MRI software and cognitive fusion biopsies in people with suspected prostate cancer: a systematic review, network meta-analysis and cost-effectiveness analysis

Article history

The research reported in this issue of the journal was commissioned and funded by the Evidence Synthesis Programme on behalf of NICE as award number NIHR135477. The protocol was agreed in May 2022. The draft manuscript began editorial review in December 2022 and was accepted for publication in June 2023. The authors have been wholly responsible for all data collection, analysis and interpretation, and for writing up their work. The HTA editors and publisher have tried to ensure the accuracy of the authors’ manuscript and would like to thank the reviewers for their constructive comments on the draft document. However, they do not accept liability for damages or losses arising from material published in this article.

Copyright statement

Copyright © 2024 Llewellyn et al. This work was produced by Llewellyn et al. under the terms of a commissioning contract issued by the Secretary of State for Health and Social Care. This is an Open Access publication distributed under the terms of the Creative Commons Attribution CC BY 4.0 licence, which permits unrestricted use, distribution, reproduction and adaptation in any medium and for any purpose provided that it is properly attributed. See: https://creativecommons.org/licenses/by/4.0/. For attribution the title, original author(s), the publication source – NIHR Journals Library, and the DOI of the publication must be cited.

2024 Llewellyn et al.

Description of health problem

Prostate cancer (PCa) is the most commonly diagnosed cancer in men in the UK; it accounts for more than a quarter (27%) of all male cancer diagnoses in 2016–8. 1 It is the second most common cause of cancer death in males in the UK, accounting for 14% of all cancer deaths. The estimated lifetime risk of a PCa diagnosis is one in eight for males born in the UK. 2,3 Over 57,000 new cases were diagnosed in 2018, with an estimated 10-year survival rate of 77.6%. Since the early 1990s, estimates of PCa incidence rates have increased by nearly half (48%) in males in the UK (2016–8) and are projected to rise by 12% between 2014 and 2035, resulting in 233 cases per 100,000 males by 2035. 3

Early-stage diagnosis is associated with improved survival outcomes compared with patients diagnosed at the latest stage of the disease. PCa primarily affects people aged 50 years or more, and the risk of developing PCa increases with age. 3 In England and Wales, 87% of people diagnosed with PCa are aged 60 years or older,4 and on average each year around a third of new cases (34%) were in males aged 75 and older. 2 People from an African family background and individuals with a family history of PCa are at higher risk of PCa. 5,6

Prostate cancer might be suspected if any of the following symptoms cannot be attributed to other health conditions: lower urinary tract symptoms, such as frequency, urgency, hesitancy, terminal dribbling and/or overactive bladder; erectile dysfunction; haematuria; lower back or bone pain; lethargy and weight loss.

The descriptor ‘clinically significant’ (CS) is widely used to differentiate PCa that may lead to morbidity or death from types of PCa that do not. This distinction is important as insignificant PCa that does not cause harm is common. 7 Autopsy studies, in men who died of causes other than PCa, indicate that there is a significant prevalence of non-CS prostate in the general male population, which increases with age. 7 PCa screening may therefore lead to overdiagnosis, by identifying cancers that are not destined to cause morbidity or mortality. Men with these cancers are at risk of being harmed by early detection and unnecessary treatment,8,9 such as radical prostatectomy or radiotherapy with no additional mortality benefit compared to an active surveillance approach, which includes regular monitoring of prostate-specific antigen (PSA) levels and digital rectal examination (DRE). On the other hand, individuals with undetected cancer or with lesions incorrectly classed as benign may miss out on relevant treatment. Clinical guidelines have focused efforts to address the risk of overtreatment and undertreatment of PCa, notably with recent updates to diagnosis pathways and refinements to risk stratification of cancer lesions. 10–12

Care pathways for the diagnosis and management of prostate cancer

Referral to suspected cancer pathway

There is no screening programme in the UK for PCa, although PSA testing is available for asymptomatic individuals above 50 years of age requesting this test. 13 For people presenting to primary care with certain clinical signs and symptoms that may indicate PCa, National Institute for Health and Care Excellence (NICE)’s guideline for suspected cancer recognition and referral advises to consider a PSA test and DRE to assess for PCa in men with: any lower urinary tract symptoms (such as nocturia, urinary frequency, hesitancy, urgency or retention) or erectile dysfunction or visible haematuria. 14 The guideline recommends men should be referred using a suspected cancer pathway (for an appointment within 2 weeks) for PCa if their PSA levels are above the age-specific reference range or if their prostate feels malignant (hard, or lumpy) on DRE. The NHS Faster Diagnosis Standard requires that patients are diagnosed or have cancer ruled out within 28 days of being referred urgently by their general practitioner (GP) for suspected cancer,15 and NICE requires that GPs should have direct access to appropriate imaging tests. 16

Figure 1 summarises the EAG’s interpretation of the pathway for the diagnosis and care of individuals with suspected PCa according to NICE guidance (NG) 131 and the NHS timed PCa pathway, which was validated by clinical advisers to the EAG. 12,17

Figure 1.

Diagnostic and care pathway for individuals with suspected prostate cancer. a, per mL of prostate volume. MDT, multidisciplinary team.

Description of technologies under assessment

This assessment will evaluate SF technologies matching the following criteria:

• intended for use in people with suspected PCa

• available in the UK

• holds a CE-mark

• compatible with MRI scanners of 1.5 tesla field strength or above.

This includes; ARTEMIS (InnoMedicus ARTEMIS), BioJet (Healthcare Supply Solutions Ltd), BiopSee (Medcom), bkFusion (BK Medical UK Ltd and MIM Software Inc), Fusion Bx 2.0 (Focal Healthcare), FusionVu (Exact Imaging), iSR’obot^TM Mona Lisa (Biobot iSR’obot), KOELIS Trinity (KOELIS and Kebomed) and UroNav Fusion Biopsy System (Phillips). Table 23, Appendix 1, presents a brief summary of the characteristics of these nine technologies.

Software fusion devices can have a variety of different features, which means they vary in the way in which they operate.

• Positioning of the ultrasound probe: An ultrasound probe can be cradled and held stationary using a device called a stepper which is attached to a workstation (also known as a stabilised approach). It can be supported by a semi-robotic arm, which allows for the ultrasound probe to be manoeuvred, while maintaining a stable pressure on the prostate. The semi-robotic arm can be used as a stepper for stabilised biopsies or can allow complete freedom of movement for use during a freehand biopsy. Finally, the ultrasound probe can be held by hand (using a freehand technique).

• Core sampling technique: Different techniques can be used to take the cores, especially in the case of transperineal biopsies. First, a grid or template can be used, which is attached to a stepper and placed in front of the perineum. The grid is marked with a number of holes, which correspond to a letter and a number to allow for multiple cores to be taken in a systematic way. Alternatively, a coaxial needle can be used. In this technique, a larger introductory needle is used to puncture the perineum before the biopsy needles is passed through. This biopsy needle can be angled to take multiple biopsies without creating multiple puncture wounds to the perineum. The coaxial needle is used with the freehand technique, where it is attached to the ultrasound probe, or in a double freehand technique, where the needle is held by hand.

• Image registration: During SF, the mpMRI images are fused with the ultrasound images during the biopsy procedure. The mpMRI image can be fixed (known as rigid registration) and will not move when the prostate is deformed, either by patient movement or by the insertion of a needle; or elastic, which means the mpMRI image adjusts to match the ultrasound image to account for prostate deformation.

A description of the principal features of the technologies is given in Appendix 1.

Place of the intervention in the diagnostic and care pathway

Software fusion targeted biopsy, for people with suspected PCa, takes place at the same two points in the diagnostic pathway as targeted CF biopsy, the current SOC.

Relevant comparator

The comparator for this assessment is targeted transperineal or transrectal prostate biopsy using CF with or without systematic biopsy, under local or general anaesthesia, in which the operator interprets the MRI imaging before the biopsy and manually targets the area of interest using TRUS as a guide. Clinical advisers to the EAG highlighted that the expertise of the person performing the biopsy may affect the accuracy and procedure time of CF.

Clinical effectiveness

• To perform a systematic review of the diagnostic accuracy and clinical efficacy of nine SF systems compared with CF targeted biopsy and with each other, in people with suspected PCa who have had a MRI scan that indicates a lesion.

• To compare the diagnostic accuracy of different SF biopsy systems with each other and with CF targeted biopsy in people with suspected PCa who have had a MRI scan that indicates a lesion using meta-analytical methods and to combine the diagnostic accuracy of different SF systems where appropriate.

• To perform a narrative systematic review of the clinical efficacy, safety and practical implementation of SF targeted biopsy. This includes assessment of intermediate outcomes, mortality and morbidity, patient-centred outcomes, adverse events (AEs), and acceptability to clinicians and patients.

Cost-effectiveness

• To conduct a systematic review and critical appraisal of relevant cost-effectiveness evidence of the use of SF biopsy systems compared to CF for targeted biopsy in people with suspected PCa who have had a MRI scan indicating a lesion.

• To develop and validate a decision-analytic model to estimate the cost-effectiveness of SF targeted biopsy systems in people with suspected PCa who have had an MRI scan indicating a lesion compared to targeted biopsy using CF. This will require linking intermediate outcomes, such as the diagnostic accuracy of SF biopsy systems to subsequent management decisions and to final health outcomes including morbidity and mortality associated with alternative treatment options. The analysis will take the perspective of the NHS and Personal Social Services (PSS), consistent with the current manual for health technology evaluations by the NICE. Final health outcomes will be evaluated in terms of quality-adjusted life-years (QALYs).

• To populate the model using the most appropriate available evidence. This evidence is likely to be identified from published literature, routine data sources and potentially using data elicited from relevant clinical experts and companies.

• To estimate the incremental cost-effectiveness of the SF biopsy systems compared to the current SOC for the population of interest (CF biopsy), based on an assessment of long-term NHS and PSS costs and quality-adjusted survival. The time horizon of the model will be sufficient to capture both the short-term and longer-term outcomes.

• To characterise the parameter uncertainty in the data used to populate the model and present the resulting uncertainty in the results to decision-makers. To this purpose, we will perform comprehensive (probabilistic and deterministic) sensitivity analyses varying parameter inputs, and structural assumptions of the model, as appropriate.

• Where possible and applicable, to assess the impact of potential sources of heterogeneity on cost-effectiveness, including subgroup analyses (e.g. patients with previous negative biopsy results within 12 months) and consideration of other factors that may affect diagnostic accuracy.

Systematic review methods (study selection, data extraction, quality assessment)

Data synthesis methods

Quantity and quality of evidence

Figure 2 presents an overview of the study selection process in a Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) flow diagram. The literature searches identified a total of 6289 unique records. After title and abstract screening, 247 references were retrieved and a total of 23 unique studies were included in the systematic review. 31,79–101 Fourteen studies were included in the quantitative synthesis,31,79,80,82,84,86–88,92–94,96,97,99 while nine studies were included in the narrative synthesis only. 81,83,85,89–91,95,98,100,101

FIGURE 2.

Study selection process (PRISMA flow diagram).

Evidence was included for all SF technologies specified in the scope and protocol (all versions) except for Fusion Bx (Focal Healthcare) and ExactVu (Exact Imaging). Report Supplementary Material 1 presents a summary of the evidence for Fusion Bx and ExactVu that was considered for inclusion and ultimately excluded, and a list of studies excluded from the systematic review, grouped by reason for exclusion.

Quality of included studies

Results of the quality and applicability assessment are reported in Figure 3, and further details on the rationale for decisions are reported in Appendix 5. All studies were at high risk of bias for at least one of the following domains: patient selection, index test, reference standard and flow and timing. Eight studies were at high risk of patient selection bias; all were non-randomised comparisons. 81,83,89,90,95,98,100,101 Three studies were at unclear risk of selection bias;31,84,85 including the two RCTs,31,84 and all other studies were at low risk of selection bias. Eight studies had a high risk of bias related to the comparison of index tests,31,81,84,88,89,93,97,101 and all other 15 studies were at low risk of bias for this domain.

FIGURE 3.

Risk of bias and applicability assessment summary of studies included in the systematic review.

Twenty studies were at high risk of bias associated with the reference standard. 31,79–87,89,90,92,94–96,98–101 For between-patient comparisons, this was primarily due to the fact that total cancer positive cases in each study arm or cohort were derived from different biopsy methods; in within-patient comparisons, as all biopsy methods were performed within the same examination, it was not feasible for studies to truly blind operators from tracks of preceding biopsy methods (true blinding would several biopsy sessions per patient, which would be unethical). Participants in all within-patient comparison studies received SF, CF and/or systematic biopsy within the same examination; the order in which the different biopsy methods were implemented varied where reported, therefore the overall direction of bias due to the lack of operator blinding could not be determined.

Of the 15 studies that compared SF with CF or with another SF device,31,81–85,88–90,93,95,97,98,100,101 7 did not use systematic biopsy or include systematic biopsy results as part of a reference standard test. 31,81,83–85,97,101 Of the studies that included systematic biopsy as part of a reference standard test, only one reported blinding the systematic biopsy operator to the MRI report. 88 This is an important design limitation, since knowledge of the MRI report may have influenced the placing of systematic biopsy cores. Clinical advisers to the EAG confirmed that lack of blinding to MRI reports may have improved the accuracy of systematic biopsies relative to targeted biopsies. Therefore, for most of the evidence for systematic biopsy included in this review, there is a risk that the detection of PCa from systematic biopsy may have been overestimated compared with true random, standard systematic biopsy. This said, the lack of blinding to MRI report when using systematic biopsy concomitant with targeted biopsy is reflective of current practice. Blinding of the histopathologists, who analysed the biopsy samples, was generally NR, and none of studies used TTMB. Two studies were at high risk of bias due to missing outcomes data (flow and timing domain),93,95 and all other studies were at low risk of bias for this domain.

Three studies raised no concerns about their applicability to the review question. 79,82,88 Five studies included a population that was deemed not applicable (NA) to the review question (high concern),31,86,90,94,99 and five included a significant proportion (approximately half) of patients undergoing repeat biopsy following a prior negative biopsy. 87,89,92,93,101 Although patients with a prior negative biopsy were eligible in this systematic review, clinical advisers to the EAG noted that they made up only a minority (approximately under 10%) of the total population undergoing targeted biopsy who are not under active surveillance. All other studies included mostly biopsy-naive patients and had a population that was considered broadly representative. Five studies used an intervention that was not considered applicable to the review question,31,84,89,95,101 primarily due to the use of general anaesthesia in all procedures. Clinical advisers to the EAG noted that general anaesthesia is normally only used in a minority of patients, although it may facilitate biopsy targeting due to the lack of patient movement. The applicability of SF was uncertain in 10 studies. 80,81,90,92–94,96,98–100 In four cases, this was due to insufficient reporting about biopsy routes and anaesthesia methods,90,96,99,100 and in six studies, a KOELIS device with no integrated ultrasound was evaluated, and the applicability of their results to KOELIS Trinity was uncertain. 80,81,92–94,98 Following request for further information from the EAG, the company did not clarify or provide evidence that the diagnostic accuracy of older versions of KOELIS was equivalent to KOELIS Trinity. Eight studies raised concerns about the applicability of the reference standard test. 31,81,83–85,95,97,101

Diagnostic accuracy results

This section presents the evidence included in the meta-analyses and structure of the networks of evidence (see Studies included in the meta-analysis and network structure), the results of the NMAs (see Meta-analysis results), and results of studies not included in the meta-analyses (see Narrative synthesis results).

Studies included in the meta-analysis and network structure

Model 1a: multinomial synthesis model (base case)

Thirteen studies, identified by the systematic review, with data suitable for inclusion in the NMA are presented in Table 24, Appendix 3 and form the network in Figure 4. Rabah et al. 84 is excluded as it compared two SF devices, assumed to have identical effects, and therefore does not contribute to the analysis. The multinomial synthesis model was used to synthesise comparative information on the probabilities of being classified at the various ISUP grades of PCa (see Meta-analysis results). Resulting estimates are then used in the base-case economic model.

FIGURE 4.

Network of biopsy types compared, under the assumption of a common effect for different SF devices. Lines represent comparisons made in studies, numbers on the lines show how many studies included that comparison and shaded areas represent multiarm studies. SB, systematic biopsy.

Due to data sparseness, we assumed that there is no difference in relative effects of the various SF biopsy devices compared to cognitive biopsy and only fixed-effect models could be fitted. This assumption is supported by the limited direct evidence comparing different fusion devices and clinical advice to the EAG. However, the different costs of each device will still be taken into account in the economic model. This assumption will be relaxed in an additional analysis (Model 1b: Multinomial synthesis model, individual device effects).

Although the network in Figure 4 is fully connected (there is a path connecting every intervention to every other), not all studies reported the breakdown of cancers detected by ISUP grade (see Appendix 3, Table 24). This resulted in a de facto disconnect in the network for comparisons of CF + SB and SF + SB for ISUP grades > 2. Relative effects comparing disconnected components of the network cannot be estimated and are reported separately.

Calculating absolute probabilities

As noted in Meta-analysis results, odds ratios estimated from this model are hard to interpret. We will therefore also present results on the absolute probability scale to aid interpretation. To calculate the absolute probabilities of being classified in each category using each intervention, we need to assume a set of underlying baseline probabilities of being classified in each category on one of the included interventions. For ease of interpretation, in this section these underlying baseline probabilities will be assumed to be fixed, that is, to have no uncertainty. All other probabilities are then obtained by applying the estimated odds ratios to these probabilities, as described in Appendix 2. These baseline probabilities should be as representative as possible of the population of interest. A targeted review was carried out to determine a good source of evidence on these probabilities (see Review of additional prevalence, test results and diagnostic accuracy evidence and Appendix 8 Distribution of test results obtained with cognitive fusion or software fusion biopsy).

The two studies with the largest sample size that were identified and deemed most representative of NHS practice were considered as a source of evidence for the baseline probabilities: Filson et al. 96 and PAIREDCAP (2019). 88 Two subgroups of patients are of interest: biopsy-naive patients and those undergoing a repeat biopsy after a negative result. Filson (2016)96 reported probabilities for these two subgroups separately and for two interventions of interest, SF using ARTEMIS and combined SF (ARTEMIS) with systematic biopsy, allowing the same source of baseline probabilities to be used for both disconnected components of the network.

However, Filson et al. 96 does not report separate data for ISUP grades 3 and 4–5, as required by the model. We approximated the probabilities of patients being in grade 3 and 4–5 by splitting the combined patients according to the proportions in each category reported in PAIREDCAP (2019)88 (approximately 60/40).

In a sensitivity analysis for the subgroup of biopsy-naive patients, the distribution of test results from PAIREDCAP (2019)88 (which only include biopsy-naive patients) was used to inform the baseline probabilities in the first part of the network. In the absence of other suitable sources of evidence, data on biopsy-naive patients from Filson et al. 96 will continue to inform the baseline probabilities in the combined biopsy (software/CF plus systematic biopsy) network.

Absolute probabilities were therefore reported for:

• subgroup of biopsy-naive patients (based on Filson et al. 96 biopsy-naive data)

• subgroup of previous negative biopsy patients (based on Filson et al. 96 previous negative biopsy data)

• a sensitivity analysis using alternative baseline probabilities for the biopsy-naive subgroup (based on biopsy-naive data from PAIREDCAP (2019)88 and Filson et al. 96).

Results will be reported separately for comparisons of CF, SF and systematic biopsy, and comparisons of combined cognitive/SF with systematic biopsy.

Model 1b: multinomial synthesis model, individual device effects

Fourteen studies identified by the systematic review with data suitable for inclusion in the NMA are presented in Table 24, Appendix 3 and form the network in Figure 5. The multinomial synthesis model was used to synthesise comparative information on the probabilities of being classified at the various ISUP grades of PCa (see Meta-analysis results).

FIGURE 5.

Network of biopsy types and devices compared. Lines represent comparisons made in studies, numbers on the lines show how many studies included that comparison and shaded areas represent multiarm studies. SB, systematic biopsy.

Although the network in Figure 5 is fully connected (there is a path connecting every intervention to every other), not all studies reported the breakdown of cancers detected by ISUP grades. This resulted in a de facto disconnect in the network for comparisons of some devices for higher ISUP grades (see Table 24, Appendix 3). Relative effects comparing disconnected components of the network cannot be estimated and are reported separately, where possible.

Calculating absolute probabilities

Absolute probabilities will be reported for:

• subgroup of biopsy-naive patients (based on Filson et al. 96 biopsy-naive data);

• subgroup of previous negative biopsy patients (based on Filson et al. 96 previous negative biopsy data).

As many network components are disconnected for high ISUP grades in this analysis, absolute probabilities are only reported where they can be reliably obtained, which limits the interpretation of results.

Model 2a: cancer detection

Data from the studies identified by the systematic review (see Figure 4) were pooled in a NMA to compare the proportion of PCas (CS and non-CS, i.e. Gleason ≥ 3 + 3) detected by the different biopsy strategies. Data were obtained by adding the relevant ISUP grades in Table 24, Appendix 3, and are presented in Table 30, Appendix 6.

In Model 2a we assumed that there is no difference in relative effects of the various SF biopsy devices compared to cognitive biopsy. This assumption relaxed in Model 2b where the individual device effects are estimated. Both fixed- and random-effects models were considered.

Model 2b: cancer detection, individual device effects

Data from the studies, identified by the systematic review (see Figure 5 and Table 30, Appendix 6), were pooled in a NMA to compare the proportion of PCas (CS and non-CS) detected by the different biopsy strategies. Both fixed- and random-effects models were considered.

Model 3a: clinically significant cancer detection

Data from the studies identified by the systematic review were pooled in aNMA to compare the proportion of CSPCas (Gleason > 3 + 3) detected by the different biopsy strategies. Only 10 studies reported the number of CS cancers detected, obtained by adding the relevant ISUP grades in Table 24 (see Appendix 3), and are presented in Table 31, Appendix 6. In addition, Rabah et al. 84 is excluded as it compared two SF devices, assumed to have identical effects in Model 3a, and therefore does not contribute to this analysis. Nine studies were included in the network (see Figure 13, Appendix 6). Both fixed- and random-effects models were considered.

Model 3b: clinically significant cancer detection, individual device effects

Data from 10 studies reporting the number of CSPCas detected by the different biopsy strategies (see Table 31 and Figure 14, Appendix 6) were pooled in a NMA. Both fixed- and random-effects models were considered.

Meta-analysis results

Model 1a: multinomial synthesis model (base-case)

Models were sampled for 100,000 iterations from 2 independent chains (50,000 iterations on each chain) after checking that convergence was achieved after a burn-in of 50,000 iterations.

Results from fitting Model 1a to the data in Table 24, Appendix 3 (network in Figure 4) are presented in Table 3. One study (Gomez-Ortiz et al. 99) had a higher than expected contribution to the mean residual deviance (15 compared to its expected contribution of 6) but overall the model fitted the data well with a posterior mean of the residual deviance of 77.4, which is close to the 75 data points included.

TABLE 3 - Odds ratios (median and 95% credible interval) of being classified as ISUP grades 1 to 4–5 compared to being categorised as having NC, for systematic biopsy and SF biopsy, compared to categorisations using CF biopsy; and for SF plus systematic biopsy, compared to CF plus systematic biopsy

SB, systematic biopsy.

ISUP grade	Compared to CF biopsy				Compared to cognitive fusion + systematic biopsy
	SB		SF		SF + SB
No cancer	Reference
1	1.57	(1.09 to 2.26)	1.98	(1.28 to 3.06)	1.20	(0.72 to 1.99)
2	2.24	(1.45 to 3.47)	1.34	(0.80 to 2.25)	2.57	(0.95 to 7.97)
3	1.40	(0.82 to 2.38)	1.25	(0.66 to 2.33)	0.66	(0.12 to 2.92)
4–5	1.54	(0.83 to 2.84)	1.58	(0.90 to 2.77)	4.33	(0.45 to 158.38)

Compared to CF biopsy, there is evidence of higher odds of being categorised in ISUP grade 1 instead of NC when using SF (OR 1.98 95% CrI 1.28 to 3.06, Table 3). There is no evidence of more patients being categorised as ISUP 2, 3 or 4–5 instead of NC for SF biopsy compared to CF biopsy (see Table 3). More patients are categorised as having non-CS cancer (ISUP grade 1) (OR 1.57 95% CrI 1.09 to 2.26) and as having a CS cancer with ISUP grade 2 (OR 2.24 95% CrI 1.45 to 3.47), instead of having NC when using systematic biopsy compared to CF biopsy. There is no clear evidence of more patients being categorised as ISUP 3 or 4–5 instead of NC for systematic biopsy compared to CF biopsy (see Table 3). However, we note the large uncertainty in all results, particularly for higher ISUP grades, due to limited data broken down by higher ISUP grades. As discussed in Quality of included studies, most of the evidence for systematic biopsy was not blinded to MRI reports. This may have inflated the accuracy of systematic biopsy compared with SF and CF.

Compared to CF plus systematic biopsy, there is no clear evidence of more patients being categorised as having cancer (ISUP grades 1 to 4–5) instead of NC for SF plus systematic biopsy. However, we note the large uncertainty in all results, particularly for the highest category. This is due to few studies reporting data broken down by higher ISUP grades and the small number of patients categorised as ISUP 4–5 using any of the two biopsy types (see Table 24, Appendix 3).

Absolute probabilities of being classified as having NC or at different ISUP grades for the two subgroups of interest: biopsy-naive patients and patients with a previous negative biopsy based on data from Filson et al. 96 are presented for ease of interpretation. A sensitivity analysis for the biopsy-naive subgroup is presented in Table 32, Appendix 6.

Absolute probabilities: biopsy-naive patients

Using baseline probabilities for SF biopsy and SF plus systematic biopsy from the biopsy-naive subgroup in Filson et al.,96 and applying the odds ratios in Table 3 the probabilities of being classified as having NC or at different ISUP grades are given in Table 4.

TABLE 4 - Probabilities (median and 95% CrI) of being classified at different ISUP grades for biopsy-naive patients

Assumed underlying baseline probabilities.

ISUP	ARTEMIS probabilities from Filson et al.96 biopsy-naive data					ARTEMIS + SB probabilities from Filson et al.96 biopsy-naive data
	Cognitive		Systematic		Softwarea	Cognitive + SB		Software + SBa
No cancer	0.55	(0.48 to 0.62)	0.42	(0.37 to 0.47)	0.47	0.41	(0.21 to 0.56)	0.36
1	0.17	(0.13 to 0.22)	0.21	(0.17 to 0.25)	0.16	0.21	(0.10 to 0.33)	0.22
2	0.12	(0.08 to 0.16)	0.20	(0.16 to 0.24)	0.20	0.10	(0.03 to 0.23)	0.22
3	0.09	(0.06 to 0.14)	0.10	(0.06 to 0.15)	0.11	0.21	(0.06 to 0.59)	0.12
4–5	0.06	(0.03 to 0.10)	0.07	(0.04 to 0.12)	0.06	0.02	(0.00 to 0.18)	0.08

For biopsy-naive patients, Model 1a suggests that compared to SF biopsy, patients undergoing CF biopsy may have (see Table 4):

1. a higher probability of being classified as not having cancer (55% vs. 47%)

2. similar probability of being classified as having non-CS cancer (ISUP 1, 17% vs. 16%)

3. lower probability of being classified at higher ISUPs, particularly ISUP 2.

Probabilities for systematic biopsy are similar to those for SF biopsy for NC and all ISUP grades. However, results are uncertain.

Results were similar when systematic biopsy was added to software and CF, although there may be a higher probability of patients being classified at ISUP grade 2 with software plus systematic biopsy compared to cognitive plus systematic biopsy, and versus lower probability for ISUP grade 3 (see Table 4), although results are imprecise due to the small number of observed events. The proportion of patients classified at ISUP grades 4–5 are similar. However, these results are very uncertain.

Absolute probabilities: previous negative biopsy patients

Using baseline probabilities for SF biopsy and SF plus systematic biopsy from the subgroup of patients with a previous negative biopsy in Filson et al.,96 and applying the odds ratios in Table 3 the probabilities of being classified as having NC or at different ISUP grades are given in Table 5.

TABLE 5 - Probabilities (median and 95% CrI) of being classified at different ISUP grades for patients with a previous negative biopsy

Assumed underlying baseline probabilities.

ISUP	ARTEMIS probabilities from Filson et al.96 previous negative biopsy data					ARTEMIS + SB probabilities from Filson et al.96 previous negative biopsy data
	Cognitive		Systematic		Softwarea	Cognitive + SB	Software + SBa
NC	0.75	(0.69 to 0.80)	0.64	(0.59 to 0.69)	0.69	0.63	(0.38 to 0.76)	0.58
1	0.08	(0.06 to 0.11)	0.11	(0.09 to 0.14)	0.09	0.13	(0.07 to 0.21)	0.15
2	0.06	(0.04 to 0.08)	0.11	(0.09 to 0.13)	0.10	0.05	(0.02 to 0.12)	0.12
3	0.06	(0.04 to 0.10)	0.08	(0.05 to 0.12)	0.08	0.14	(0.04 to 0.47)	0.09
4–5	0.04	(0.02 to 0.07)	0.05	(0.03 to 0.09)	0.05	0.01	(0.00 to 0.13)	0.06

For patients with a previous negative biopsy, given a 69% probability of being classified as not having cancer with SF biopsy,96 the probability of being classified as not having cancer is higher for patients undergoing cognitive biopsy (75% 95% CrI 69% to 80%) but lower for patients undergoing systematic biopsy (64% 95% CrI 59% to 69%). As there is high probability that patients with a prior negative biopsy will again be classified as having NC with SF, CF or systematic biopsy, the probabilities of being classified at different ISUP grades are small and similar across these biopsy strategies (see Table 5).

Similar results were obtained when adding a systematic biopsy to software and CF.

Model 1b: multinomial synthesis model, individual device effects

Models were sampled for 100,000 iterations from 2 independent chains (50,000 iterations on each chain) after checking that convergence was achieved after a burn-in of 50,000 iterations.

Results from fitting Model 1b to the data in Table 24, Appendix 3 (network in Figure 5) are presented in Table 33, Appendix 6. One study (Gomez-Ortiz et al. 99) had a higher than expected contribution to the mean residual deviance (16 compared to its expected contribution of six). Other studies had deviances in the range expected, although the posterior mean of the residual deviance was 89.2, which is higher than the number points included (79). Often a model fit can be poor when data are sparse as many parameters cannot be reliably estimated. However, more complex models, such as random-effects models, cannot be considered due to data sparseness. We would advise caution when interpreting the results from this model.

No odds ratios can be estimated for SF biopsy using UroNav or UroNav plus systematic biopsy since the only study comparing this device does not report details of classifications broken down by category (see Table 24, Appendix 3).

Compared to CF biopsy, there is only evidence of higher odds of being categorised in ISUP grade 1 instead of NC when using systematic biopsy (OR 1.54 95% CrI 1.06 to 2.24, Table 33, Appendix 6). There is some evidence that more patients are categorised as ISUP grade 2 instead of having NC when using systematic biopsy, ARTEMIS or Urostation, compared to CF biopsy. There is no clear evidence of more patients being categorised as ISUP 3 or 4–5 instead of NC for systematic biopsy or ARTEMIS compared to CF biopsy. No relative effects are estimable for the other devices and there is large uncertainty in all results.

Compared to CF plus systematic biopsy, there is no clear evidence of more patients being categorised as having cancer (ISUP grades 1 to 4–5) instead of NC for ARTEMIS or Urostation plus systematic biopsy. However, we note the large uncertainty in all results which led to some relative effects not being estimable (see Table 33, Appendix 6).

Absolute probabilities of being classified as having NC or at different ISUP grades for the two subgroups of interest can only be reported where the odds ratios are estimable for all ISUP grades. Therefore, these are only presented for CF, systematic biopsy and SF using ARTEMIS (assumed underlying baseline probabilities), and when adding systematic biopsy, for biopsy-naive patients and patients with a previous negative biopsy based on data from Filson et al. 96 (see Tables 34 and 35, Appendix 6).

Model 2a: cancer detection

Fixed- and random-effects models were fitted. Based on the model fit statistics (see Table 36, Appendix 6) both the fixed- and random-effects models fitted the data well and differences in DIC were small. Therefore, the fixed-effect model was selected. The fixed-effect unrelated mean effects model suggested no evidence of inconsistency between direct and indirect evidence based on both the model fit statistics and deviance plots (see Table 36 and Figure 15, Appendix 6).

Results from fitting Model 2a to the data in Table 30, Appendix 6 (network in Figure 4) are presented in Figure 6 and all pairwise comparisons are reported in Table 37, Appendix 6.

FIGURE 6.

Odds ratio of detection (median and 95% CrI) of cancer. OR, odds ratio; SB, systematic biopsy.

Model 2a suggests SF biopsy may classify more patients as having cancer (any ISUP), than CF biopsy (OR 1.30 95% CrI 1.06 to 1.61; Figure 6). However, note that this cannot be directly compared to the results from ISUP 1 for Model 1a, since Model 2a is comparing detection of any cancer, that is, all ISUP 1 to 5 combined, and not only the detection of non-CS cancer. The increase in the ORs for detection of any cancer is driven by the increase in the probability of categorisation at ISUP>1, which in this case is driven by increases at ISUP 2 [see Model 1a: multinomial synthesis model (base-case)]. Results for the random-effects model are presented as a sensitivity analysis in Table 37 and Figure 16, Appendix 6. As discussed in Quality of included studies, the accuracy of systematic biopsy may have been inflated due to study design limitations.

Model 2b: cancer detection, individual device effects

Fixed- and random-effects models were fitted. Based on the model fit statistics (see Table 36, Appendix 6) both the fixed- and random-effects models fitted the data well and differences in DIC were small. Therefore, the fixed-effects model was selected. The fixed-effects unrelated mean effects model suggested no evidence of inconsistency between direct and indirect evidence based on both the model fit statistics and deviance plots (see Table 36 and Figure 15, Appendix 6).

Results from fitting Model 2b to the data in Table 30, Appendix 6 (network in Figure 4) are presented in Figure 7 and all pairwise comparisons are reported in Table 38, Appendix 6.

FIGURE 7.

Odds ratio of detection (median and 95% CrI) of cancer, individual device effects. OR, odds ratio; SB, systematic biopsy.

Compared to CF biopsy, there is evidence that SF biopsy with BioJet, Urostation and ARTEMIS, and Urostation, UroNav or cognitive biopsy combined with systematic biopsy may detect more cancers. Results for the random-effects model are presented as a sensitivity analysis in Table 38 and Figure 16, Appendix 6.

Model 3a: clinically significant cancer detection

Fixed- and random-effects models were fitted. Based on the model fit statistics (see Table 36, Appendix 6) the random-effects model had a better fit to the data and the difference in DIC was > 3. Therefore, the random-effects model was selected. The random-effects unrelated mean effects model suggested no evidence of inconsistency between direct and indirect evidence based on both the model fit statistics and deviance plots (see Table 36 and Figure 15, Appendix 6).

Results from fitting Model 3a to the data in Table 31, Appendix 6 (network in Figure 13) are presented in Figure 6 and all pairwise comparisons are reported in Table 37, Appendix 6. The posterior median of the between-study heterogeneity SD was 0.313 (95% CrI 0.132 to 0.634), which is moderate on the log odds ratio scale. The full posterior distribution of the between-study SD is presented in Figure 16, Appendix 6.

While Model 2a suggested SF biopsy may classify more patients as having cancer (ISUP 1 to 5) than CF biopsy (see Figure 5), Model 3a also suggests SF biopsy may classify more patients as having CS cancer (ISUP 2, 3, 4–5), as opposed to NC or ISUP 1, than CF biopsy, but with a wider confidence interval (CI) that includes the null effect (OR 1.35 95% CrI 0.86, 2.10; Figure 6). By using odds and collapsing ISUP grades, the statistical model has higher power to detect statistically significant differences against NC than against non-CS cancer (which pools NC with ISUP 1). These results are consistent with the findings from Models 1a and 2a. There is some evidence that adding systematic biopsy to cognitive or SF increases CS cancer detection.

Model 3b: clinically significant cancer detection, individual device effects

Fixed- and random-effects models were fitted. Based on the model fit statistics (see Table 36, Appendix 6) the random-effects model had a better fit to the data and the difference in DIC was > 3. Therefore, the random-effects model was selected. The random-effects unrelated mean effects model suggested no evidence of inconsistency between direct and indirect evidence based on both the model fit statistics and deviance plots (see Table 36 and Figure 15, Appendix 6).

Results from fitting Model 3b to the data in Table 31, Appendix 6 (network in Figure 13) are presented in Figure 7 and all pairwise comparisons are reported in Table 39, Appendix 6. The posterior median of the between-study heterogeneity SD was 0.304 (95% CrI 0.048 to 0.769), which is similar to the posterior heterogeneity from Model 3a. This suggests there is moderate heterogeneity (log odds ratio scale) and that splitting the device effects did not explain the between-study variability. The full posterior distribution of the between-study SD is presented in Figure 16, Appendix 6.

Compared to CF biopsy, there is no evidence that SF with ARTEMIS, BiopSee, Urostation, or systematic biopsy detect more CS cancers. However, there is evidence that SF with BioJet or adding systematic biopsy to cognitive or SF with ARTEMIS or Urostation may increase CS cancer detection.

Clinical effectiveness results

Diagnostic accuracy and clinical effectiveness: summary and conclusions

The evidence identified by the systematic review included a total of 3733 patients who received SF and 2154 individuals with CF from 23 studies. Evidence was included for all devices specified in the protocol, except for Fusion Bx 2.0 and FusionVu. Overall, the evidence for all devices was at high risk of bias. Up to 14 studies were included in NMAs. Analyses compared the relative diagnostic accuracy of SF, CF, CF with concomitant systematic biopsies, SF with systematic biopsies, and systematic biopsies alone.

Our main NMAs looked at how CF compares to SF in classifying patients across the range of ISUP grades. Results must be cautiously interpreted due to the high risk of bias, but suggest that patients undergoing software biopsy may show: (1) a lower probability of being classified as not having cancer, (2) similar probability of being classified as having non-CS cancer (ISUP grade 1) and (3) higher probability of being classified at higher ISUP grades, particularly ISUP 2. Similar results were obtained when comparing between same biopsy methods where both were combined with systematic biopsy.

Additional meta-analyses of cancer detection rates suggest that, compared with CF biopsy, SF may identify more PCa (any grade) (OR 1.30; 95% CrI 1.06 to 1.61). Adding systematic biopsy to cognitive or SF may increase the detection of all PCa and of CS cancer, and from this evidence there is no suggestion that SF with concomitant systematic biopsy is superior to CF with systematic biopsy.

Meta-analyses by individual device showed that compared with CF biopsy, BioJet and Urostation are associated with a higher detection of PCa overall, and that BioJet is associated with more CS cancer, although only one study of BioJet was included. The evidence for all other software devices was insufficient to evaluate their accuracy compared with CF reliably, or to assess whether some SF technologies are more accurate than others.

There was large uncertainty in all estimates due to the limited evidence, particularly for higher ISUP grades and by individual device. Results from studies, excluded from the meta-analyses, broadly reflected these findings. Compared with CF, there was no evidence that the accuracy of SF may differ by lesion location, or between biopsy naive and prior negative biopsy patients, or according to operator experience.

The applicability of the evidence for KOELIS Trinity is uncertain, as it was almost entirely informed by evaluations of a previous version (KOELIS Urostation) without integrated ultrasound. The applicability of the evidence for BiopSee is also limited due to the lack of evaluations under local anaesthesia. There is no evidence comparing the accuracy of Fusion Bx 2.0 and FusionVu with CF, and no evidence for these devices was eligible for inclusion in the indirect comparisons.

Evidence for all other protocol specified outcomes was limited and inconclusive. Overall, there is no evidence that biopsy positivity rates differ significantly between SF and CF, or between SF devices. There was some evidence that systems with rigid registration (BioJet or UroNav) are easier and faster to use than elastic registration (KOELIS Trinity), although this is informed by a single, small study and is not conclusive. Overall, there is no evidence of a significant difference in safety outcomes between biopsies conducted with SF and CF or between SF devices, although the evidence is limited by poor reporting and at high risk of confounding.

No relevant evidence was found for the following outcomes: biopsy sample suitability/quality, number of repeat biopsies performed, procedure completion rates, software failure rate, time to diagnosis, length of hospital stay, time taken for MR image preparation, subsequent PCa management, re-biopsy rate, hospitalisation, OS, PFS, patient- and carer-reported outcomes (including tolerability and HRQoL), barriers and facilitators to implementations.

Additional evidence to inform model structure and parameterisation

Additional evidence was required to inform a number of economic parameters, including (1) PCa prevalence; (2) distribution of test results for cognitive and SF broken down by Gleason grade; (3) test accuracy of cognitive and SF and (4) long-term evidence on outcomes from management strategies for patients with PCa. In addition to the systematic review of diagnostic accuracy and clinical effectiveness, targeted reviews were conducted to identify the most relevant evidence to inform these parameters.

Overview

In the next sections, we provide an overview of published cost-effectiveness studies on the use of SF biopsy systems in comparison with CF for targeted prostate biopsy (see Methodology of the cost-effectiveness of software fusion biopsy for suspected prostate cancer and Results of the review of the cost-effectiveness of MRI fusion biopsy for suspected prostate cancer), to determine generalisability of the evidence to inform this assessment’s decision problem. In addition, this chapter presents a targeted review of diagnostic cost-effectiveness studies (see Methodology of the additional targeted reviews to support model conceptualisation and Results of the additional targeted reviews to support model conceptualisation), which model prostate biopsy procedures to identify aPCa (same point in the diagnostic pathway as the interventions in this assessment). This targeted review is done with the aim to support the conceptualisation and parameterisation of a de novo decision-analytic model.

Methodology of the review of economic evidence provided by the companies

We reviewed the economic evidence submitted by the companies in response to requests for information (RFIs) by NICE and the EAG. We listed this economic evidence grouped into three categories:

1. full economic evaluations that consider both costs and consequences (including cost-effectiveness, cost–utility and cost–benefit analyses)

2. resource use and cost data

3. other.

Full economic evaluations were considered for inclusion in one of the two other economic reviews (see Methodology of the cost-effectiveness of software fusion biopsy for suspected prostate cancer or Methodology of the additional targeted reviews to support model conceptualisation) as appropriate given their study characteristics.

Resource use and cost data were considered for the parametrisation of the de novo model.

Results of the review of the cost-effectiveness of magnetic resonance imaging fusion biopsy for suspected prostate cancer

Results of the additional targeted reviews to support model conceptualisation

The results of the searches are given in detail in Appendix 1. In total, 15 cost-effectiveness models116,119–133 were considered potentially relevant to inform the de novo model conceptualisation for inclusion. These studies are summarised in Table 63, Appendix 9.

Of the 15 cost-effectiveness models identified at the first stage of the review, 9 were selected for a more in-depth review, as these were identified as the most appropriate to support the conceptualisation of the de novo model given the relevance of:

• the comparisons and position in the diagnostic pathway – studies which compared biopsies conducted with MRI-influence methods (i.e. targeted and/or combined biopsies) for PCa diagnosis;119,120,124,129,130

• UK policy relevance. 116,121,123,125,126

Diagnostic strategies

The model evaluated two strategies for two alternative comparisons: (1) targeted SF biopsy versus targeted cognitive biopsy and (2) combined (targeted and systematic) SF biopsy versus combined cognitive biopsy. The four strategies could not be incrementally compared due to the mechanism of evidence generation for the diagnostic accuracy, which relied on separate evidence networks.

The test sequence and clinical management for each strategy:

1. all patients receive the index biopsy:

   1. If biopsy result suggests no PCa or ISUP grade 1, a proportion of patients undergo repeat biopsy. Patients who do not undergo repeat biopsy are managed in accordance with their diagnosed ISUP grade/CPG or discharged to routine monitoring.

   2. If biopsy result suggests ISUP grade 2 or greater, the individual receives treatment according to CPG.

2. for the patients who receive repeat biopsy:

   1. Individuals are clinically managed according to the highest ISUP grade/CPG score between the two biopsy results or discharged to routine monitoring if the biopsy suggests no PCa.

Model development

Model structure and parameterisation

Modelling of first biopsy results

Determining diagnostic accuracy

As identified above (see Conceptualisation), the fact that the evidence synthesis conducted as part of this assessment does not consider the accuracy of the different biopsy methods in relation to a reference standard has important implications for economic modelling. In the absence of a robust and representative outcomes RCT, economic modelling relies on evidence linkage facilitated by knowing the extent of misclassification of the different tests in relation to true disease status and determining its consequences to health and economic outcomes.

The extent of misclassification can, however, be made explicit by the accuracy matrix, the elements of which reflect the probabilities of obtaining a particular test result with one method conditional on a particular level of (true) disease status. Together with prevalence estimates, this matrix determines the distribution of test results, shown at the top of Figure 8.

FIGURE 8.

Illustration of the relationship between prevalence, and the accuracy and distribution of test results across five categories, for two hypothetical tests.

Note that, due to the nature of biopsy and histological examination of the biopsy specimen, it is reasonable to assume that false-positive results are not possible, that is, if cancer is histologically identified, then it is present. This implies that biopsy methods cannot identify a higher category than true disease status, and therefore zero probability is attributed to such cases in the above accuracy matrices.

Where multiple methods are of interest, the problem becomes more complex for two reasons. First, the prevalence (i.e. the true distribution across categories) is independent of test results and therefore a common prevalence estimate needs to ground all distributions of test results, and be consistent with these. Second, explicit accounts of accuracy need to respect both the prevalence estimates and the marginal distribution of test results derived from the synthesis. Therefore, a structural approach is required for determining accuracy from the marginal distributions obtained through application of the synthesis model.

Summary of approaches used in previous cost-effectiveness models

From the conceptualisation reviews (see Results of the additional targeted reviews to support model conceptualisation), two cost-effectiveness reviews have focused on a similar context where no accuracy evidence was synthesised.

A previous DAR,116 from now on referred as the Southampton DAR, conducted a meta-analysis on cancer detection rates [using relative risks (RRs)] including studies comparing the biopsy methods of interest to the decision problem (e.g. LATP vs. LATRUS), and did not include evidence comparing either method to a reference standard. In this work, the authors sourced the baseline distribution for LATRUS and its accuracy matrix, from an external diagnostic accuracy study (the PROMIS study125,126). The authors then applied the synthesised RRs of cancer detection for LATP biopsy (derived for marginal distributions) directly to both (1) the conditional probability of LATRUS identifying CS cancer conditional on true disease status, and to the (2) conditional probability of identifying CNS cancer (assumption imposed in the base case). The conditional probability of NC given true disease status was then adjusted to be one minus the remaining. The way the RRs were applied in the model is not consistent with the way in which they were derived, in that the RR derived from the synthesis model refers to the relative increase in detection rate with one method in relation to another; the RRs were therefore derived on marginal probabilities and not on conditional probabilities. Their application to conditional probabilities in such a way implies that the increase in accuracy of detecting cancer with a particular test is independent of whether the cancer was CS or non-CS, and that the increase in accuracy of detecting non-CS cancer given the patient has non-CS cancer is equal to the increase in accuracy of detecting CS cancer, given the patient has CS cancer.

An alternative study, Wilson et al. 121 also investigating LATP in relation to LATRUS, assumed no difference in the expected accuracy of the biopsy methods in the comparison of interest. Therefore, the authors sourced prevalence and accuracy estimates for LATRUS from the PROMIS study125,126 and used it to represent the expected results for both biopsy methods. In reflecting uncertainty, the authors sampled from the accuracy matrix directly, taking two independent samples to represent the two different biopsy methods, and therefore generate differences in the accuracy matrix between the methods, due to randomness only.

None of the existing approaches has direct applicability in the current assessment, where a disaggregation by ISUP grade is required.

Summary of methods

The approach used in the current assessment was designed to:

• be grounded on the results of the evidence synthesis model

• return a true distribution across ISUP grade categories (prevalence) that is internally valid, that is not lower than the estimated ISUP Grade detection rates of the different biopsy methods

• be grounded on available evidence on the likely accuracy of targeted MRI fusion conditional on ISUP grade

• define accuracy matrices for the remaining biopsy methods of interest that are consistent with both prevalence and the distributions of biopsy results from the evidence synthesis.

To achieve this, an extension to the synthesis model was developed, drawing on the broader evidence in Review of additional prevalence, test results and diagnostic accuracy evidence. To allow for an internally consistent approach, we grounded our methodology on the distribution of test results obtained with MRI-influenced methods and their accuracy. Given that disease prevalence is fully determined by these two results, the prevalence evidence identified in Review of additional prevalence, test results and diagnostic accuracy evidence will not be explicitly incorporated in our analyses but will instead be used qualitatively to put our results into context.

The methodology is summarised below. A more comprehensive description of the methods used is presented in Appendix 10.

Distribution of test results

The distributions of test results across the disease categories for the relevant biopsy methods within each disconnected component of the network in Model 1a were computed by applying network-specific baseline distributions to the results of the NMA. Building from the analyses in the evidence synthesis section, the baseline distributions were sampled from a multinomial likelihood with an uninformative Dirichlet prior distribution for its hyperparameters, to allow for uncertainty in describing the data from the empirical studies.

Accuracy matrix for software fusion

Evidence on the accuracy matrix for SF, sourced from the literature, was used to characterise the elements of the accuracy matrix probabilistically in the model. A multinomial likelihood was used to describe the distribution of test results conditional on each particular level of true disease status (each line in the matrix in Figure 8) with Dirichlet uninformative prior distributions.

Prevalence

The derivation of prevalence followed two steps, the first consisted of the analytical derivation of an initial prevalence estimate from the marginal distribution and accuracy matrix for SF. The second step entailed applying a constraint to ensure that the prevalence is always higher than the detection rates (by ISUP grade) observed across all tests.

Accuracy matrix for remaining biopsy methods

The diagonals of the accuracy matrices for the remaining biopsy methods were determined by the prevalence and the test-specific distribution of results. To define the remaining non-zero and free elements of the matrix, uninformative beta distributions were used, constrained so that their multiplication by the prevalence retrieves the test results estimated within the evidence synthesis.

Implementation

The extension to the synthesis model, developed to determine accuracy, was implemented alongside the synthesis model in a Bayesian framework estimated through Markov chain Monte Carlo methods using WinBUGS 1.4.3. 140 Due to the sparseness of evidence in other networks, this was applied to Model 1a [see Model 1a: Multinomial synthesis model (base case)] which includes SF, CF and systematic biopsy in a first connected network, and the combination of software and CF with systematic biopsy in a second connected network. As in the evidence synthesis, model convergence was assessed where possible by running two independent chains with different starting values looking at a history plot and through inspection of Gelman–Rubin diagnostic plots. Model fit was assessed by comparing the mean total residual deviance to the number of independent data points contributing to the analysis. 71

Sensitivity analysis

Given that the approach proposed here is heavily data driven, sensitivity analyses focused on varying the data sources for the baseline distributions and accuracy matrix.

Results

The extension to the synthesis model reflects the data sources described in Model 1a: Multinomial synthesis model (base-case) for the baseline distribution of test results for SF, the reference method. The extension model also required data to characterise the accuracy matrix for the reference biopsy method, two sources for these data were available (see Section Review of additional prevalence, test results and diagnostic accuracy evidence and Appendix 8, Distribution of test results obtained with cognitive fusion or software fusion biopsy) and were used here. According to the data sources used, the following analyses were conducted:

• Main analysis, for the subgroup of biopsy-naive individuals: baseline distribution of test results for SF sourced from biopsy-naive data from Filson et al. 96 relative accuracy data from the multinomial evidence synthesis model (see Multinomial synthesis model) which was incorporated in this extension, and accuracy data from Mortezavi et al. 141 Mortezavi et al. 141 was chosen for the main analysis over Zhou et al. 142 as it more closely reflects the lower accuracy observed in UK-specific evidence sources.

• Subgroup analysis for previous negative-biopsy individuals: all sources were equal to those used in the main analysis except the baseline distribution of test results for SF which was sourced from previous negative-biopsy data from Filson et al. 96

• Sensitivity analysis to data source on baseline distribution: all sources were equal to those used in the main analysis except the baseline probabilities, which were based on biopsy-naive data from PAIREDCAP (2019),88 for network 1.

• Sensitivity analysis to data source on accuracy matrix: all sources were equal to those used in the main analysis except accuracy data which was sourced from Zhou et al. 142

Note that given the two networks are disconnected, results are reported separately for comparisons of CF and SF – network 1, and for comparisons of combined cognitive/SF with systematic biopsy – network 2. Note that while network 1 includes systematic biopsy, results for this biopsy method are NR here.

Main analyses (biopsy naive)

Table 9 shows the results of the structured approach applied to the main analysis for the subgroup of biopsy-naive patients. Results are internally consistent, and consistent with the sources of evidence these drew upon. They mirror the high level of uncertainty in the evidence base.

TABLE 9 - Distribution of test results, conditional accuracy and prevalence probabilities (mean and 95% CrI) according to ISUP grade for biopsy-naive individuals

Note

Diagnostic accuracy extension to the evidence synthesis model. Results of main analysis.

Network 1		(Distribution of test results)					(Distribution of test results)
		0.516 (0.416 to 0.615)	0.186 (0.131 to 0.249)	0.136 (0.068 to 0.211)	0.098 (0.052 to 0.157)	0.064 (0.031 to 0.114)	0.457 (0.403 to 0.513)	0.173 (0.137 to 0.214)	0.196 (0.157 to 0.233)	0.108 (0.079 to 0.144)	0.066 (0.043 to 0.095)
		CF					SF
		(Accuracy matrix)					(Accuracy matrix)
(Prevalence)	ISUP	NC	1	2	3	4 or 5	NC	1	2	3	4 or 5
Network 1		(Distribution of test results)					(Distribution of test results)
		0.460 (0.335 to 0.583)	0.250 (0.152 to 0.356)	0.127 (0.034 to 0.261)	0.131 (0.046 to 0.231)	0.033 (0.001 to 0.107)	0.348 (0.273 to 0.418)	0.223 (0.179 to 0.273)	0.232 (0.168 to 0.311)	0.115 (0.081 to 0.152)	0.082 (0.054 to 0.114)
		Combined CF and systematic biopsy					Combined SF and systematic biopsy
		(Accuracy matrix)					(Accuracy matrix)
(Prevalence)	ISUP	NC	1	2	3	4 or 5	NC	1	2	3	4 or 5
0.121 (0.007 to 0.238)	NC	1	0	0	0	0	1	0	0	0	0
0.318 (0.212 to 0.452)	1	0.829 (0.529 to 0.994)	0.171 (0.006 to 0.471)	0	0	0	0.671 (0.538 to 0.796)	0.329 (0.204 to 0.462)	0	0	0
0.262 (0.193 to 0.341)	2	0.300 (0.016 to 0.64)	0.362 (0.083 to 0.674)	0.338 (0.111 to 0.55)	0	0	0.251 (0.167 to 0.347)	0.204 (0.128 to 0.288)	0.544 (0.443 to 0.64)	0	0
0.183 (0.119 to 0.265)	3	0.189 (0.006 to 0.526)	0.140 (0.005 to 0.422)	0.192 (0.008 to 0.537)	0.479 (0.213 to 0.804)	0	0.224 (0.121 to 0.343)	0.059 (0.012 to 0.138)	0.207 (0.112 to 0.322)	0.510 (0.387 to 0.65)	0
0.116 (0.077 to 0.174)	4 or 5	0.125 (0.004 to 0.389)	0.111 (0.004 to 0.357)	0.111 (0.004 to 0.362)	0.101 (0.002 to 0.332)	0.552 (0.299 to 0.882)	0.111 (0.046 to 0.199)	0.047 (0.011 to 0.112)	0.130 (0.063 to 0.217)	0.140 (0.068 to 0.226)	0.573 (0.467 to 0.687)
0.121 (0.007 to 0.238)	NC	1	0	0	0	0	1	0	0	0	0
0.318 (0.212 to 0.452)	1	0.709 (0.289 to 0.987)	0.291 (0.013 to 0.711)	0	0	0	0.528 (0.206 to 0.824)	0.472 (0.176 to 0.794)	0	0	0
0.262 (0.193 to 0.341)	2	0.249 (0.01 to 0.689)	0.437 (0.07 to 0.836)	0.314 (0.028 to 0.78)	0	0	0.078 (0.001 to 0.273)	0.152 (0.011 to 0.384)	0.770 (0.523 to 0.975)	0	0
0.183 (0.119 to 0.265)	3	0.126 (0.002 to 0.488)	0.124 (0.003 to 0.449)	0.134 (0.002 to 0.482)	0.616 (0.148 to 0.981)	0	0.132 (0.005 to 0.441)	0.135 (0.005 to 0.411)	0.130 (0.004 to 0.403)	0.603 (0.338 to 0.92)	0
0.116 (0.077 to 0.174)	4 or 5	0.195 (0.004 to 0.618)	0.187 (0.005 to 0.603)	0.173 (0.004 to 0.561)	0.163 (0.004 to 0.543)	0.281 (0.006 to 0.865)	0.069 (0.001 to 0.282)	0.070 (0.001 to 0.265)	0.066 (0.001 to 0.27)	0.071 (0.001 to 0.266)	0.724 (0.402 to 0.98)

The prevalence estimates inferred by the extended synthesis model are in line with those available in the literature (see Review of additional prevalence, test results and diagnostic accuracy evidence and Appendix Distribution of test results obtained with cognitive fusion or software fusion biopsy), perhaps closer to the lowest available estimate of cancer prevalence (i.e. low probability of NC). This is, however, expected, as the inferred prevalence in the extended model is bounded by a composite of all five tests, and is sampled from a distribution that allows for even higher cancer prevalences than those identified by the composite of all five tests.

In terms of distribution of test results, the results obtained here (presented at the top of each accuracy matrix in Table 9) are consistent with those in the synthesis section (see Tables 68–71, Appendix 10 for detailed comparisons). In summary, within network 1 (which includes cognitive and SF), the results suggest that SF may retrieve a higher detection of cancer at ISUP grade 2 and above when compared to CF, with the detection at ISUP grade 2 being highest. These results are not statistically significant, in that CrIs overlap significantly. Table 10 presents the information on distribution of test results converted onto detection rates at thresholds of categories. This information highlights that: SF presents a similar level of detection at ISUP grades 4–5, slightly increased detection of ISUP grade 3 or above of 1.3%, increased detection of ISUP grade 2 or above of 7.1% and increased detection at ISUP grade 1 or above of 5.9%.

TABLE 10 - Proportion correctly identified and detection rates (mean and 95% CrI) with the different biopsy methods for biopsy-naive individuals

Note

Diagnostic accuracy extension to the evidence synthesis model. Results of main analysis.

ISUP grade	Estimated detection rates with the different biopsy methods
	Network 1		Network 2
	CF	SF	Combined CF and systematic biopsy	Combined SF and systematic biopsy
4 or 5	0.064 (0.031 to 0.114)	0.066 (0.043 to 0.095)	0.033 (0.001 to 0.107)	0.082 (0.054 to 0.114)
3 to 5	0.162 (0.102 to 0.237)	0.175 (0.135 to 0.217)	0.164 (0.064 to 0.27)	0.197 (0.152 to 0.243)
2 to 5	0.299 (0.209 to 0.396)	0.370 (0.322 to 0.424)	0.290 (0.173 to 0.428)	0.429 (0.358 to 0.502)
1 to 5	0.484 (0.385 to 0.584)	0.543 (0.487 to 0.597)	0.540 (0.417 to 0.665)	0.652 (0.582 to 0.727)

In terms of accuracy, the results for network 1 suggest that SF is more accurate at detecting the correct category (the diagonal of the accuracy matrix is always higher for SF), with higher differences at ISUP grades 1 and 2.

The accuracy matrix results show that despite cognitive presenting a higher likelihood of an ISUP grade 1 result, there is an increased accuracy of SF at ISUP grade 1. This is due to, with CF, individuals at higher ISUP categories being misclassified as grade 1. The accuracy matrix shows increased accuracy at ISUP grade 2 for SF, but retains a significant proportion inaccuratelyclassified as ‘NC’ [with a probability of 0.25 95% CrI (0.17 to 0.34)] which is higher than the proportion innacurately classified as ISUP grade 1 [with a probability of 0.20 95% CrI (0.13 to 0.29)]. A similar effect is observed in ISUP grade 3, where the likelihood of being classified as ‘NC’ is higher for SF than for CF – probabilities 0.224 95% CrI (0.121 to 0.343) versus 0.189 95% CrI (0.006 to 0.526). This is a result of the increased detection at ISUP grade 2 not being matched by a similar level of detection at ISUP grade 1.

By multiplying the prevalence by the respective element of the accuracy matrix, the joint probability matrix is obtained (see Table 67, Appendix 10). This matrix identifies, for a cohort with the mix of ISUP grades as per the prevalence estimates, the probability of both events, that is the probability of a particular ‘true’ ISUP grade and a particular test result. This matrix identifies that, at all grades, the probability of an accurate result is 0.524 95% CrI (0.411, 0.628) for SF – 0.12 at NC, 0.10 at ISUP grade 1, 0.14 at ISUP grade 2, 0.09 at ISUP grade 3 and 0.07 at ISUP grade 4 or 5. The probability of an accurate result is 0.413 95% CrI (0.256, 0.583) for CF – 0.12 at NC, 0.05 at ISUP grade 1, 0.09 at ISUP grade 2, 0.09 at ISUP grade 3 and 0.06 at ISUP grade 4 or 5. The highest difference between software and cognitive is observed at ISUP grades 1 and 2 (approximately 5% increase in each with software). Notably, in terms of misclassification, the overall proportion of ISUP grade 3 identified as ‘no cancer’ is higher with SF 4.2% than with CF (3.5%). This implies that the key trade-offs for SF are the benefits achieved by the general increase in detection, but particularly for ISUP grades 1 and 2, at the expense of a slightly higher proportion of grade 3s that will not be detected as cancerous.

Network 2 (including software and CF combined with systematic biopsy) shows higher identification in the distribution of test results (due to the baseline used) to but qualitative results are similar to those in network 1 to noting that there is substantial uncertainty in these results. Detection rates at thresholds of categories show that cancer detection is expected to be higher with combined software to at all levels to but particularly at ISUP grade 2 or above where detection is 13.9% higher than with combined CF and at ISUP grade 1 or above where detection is 9.2% higher that with combined CF.

In terms of accuracy to at all grades to the probability of an accurate result is 0.655 95% CrI (0.471 to 0.816) for combined SF and 0.438 95% CrI (0.218 to 0.665) for combined CF. For both combined strategies to the likelihood of a ‘no cancer’ result for ISUP grades 2 and 3 is still relatively high to but this is now comparable to the likelihood of an ISUP grade 1 result.

Subgroup analysis (previous negative biopsy)

We conducted a subgroup analysis to where the baseline distribution of test results for SF was sourced from Filson et al. 96 but using the group of individuals recruited into this study that had previous negative-biopsy results. However, the diagnostic accuracy evidence synthesis and the accuracy matrix are still sourced as per the main analysis to grounded on evidence over biopsy-naive and repeat biopsy patients. Summary results of distribution of test results for the subgroup analysis are presented in Table 72 (Appendix 10) alongside their interpretation. Prevalence probabilities and results of the accuracy matrices are also presented in Tables 73 and 74 to Appendix 10.

Sensitivity analysis

Sensitivity analyses change the main sources of evidence of the main analyses (on biopsy-naive patients): a first sensitivity analysis uses an alternative baseline distribution of test results for SF [from PAIREDCAP (2019)],88 and a second analysis uses an alternative source for accuracy matrix evidence [from Zhou et al.]. 142

In both these analyses, results for the accuracy matrices could only be presented for the first network because of increased uncertainty.

The summary results in Tables 75 and 76, Appendix 10, for the first sensitivity analysis, indicate that results are sensitive to the distribution of test results. The PAIREDCAP study distribution showed a higher proportion of ‘no cancer’ identified with SF (31% vs. 46% in the main analysis grounded on Filson, Table 9), identical in ISUP grade 1, and higher proportions across all remaining ISUP categories (26%, 16% and 10%, respectively for ISUP grades 2, 3 and 4 or 5, vs. 20%, 11% and 7% in the main analysis grounded on Filson, Table 9). The distribution of test results for ISUP grade 4 or 5 are similar between software and CF, but are significantly increased for software at ISUP grade 2, slightly increased at ISUP grade 3 and slightly reduced for ISUP grade 1.

The summary results in Tables 77 and 78, Appendix 10, for the second sensitivity analysis indicate that results on the distribution of test results are only slightly sensitive to the source of evidence on the accuracy matrix in Filson (see Table 9). The main difference distribution of test results for ISUP grades 4 or 5 are slightly higher for SF in this analysis in relation to the main analysis in Table 9. The estimates of the accuracy matrices (in Appendix 10) show increased accuracy (in classifying individuals in the right category) for both technologies in relation to the main analysis in Table 9, which reflect the data from Zhou et al. 142 However, differences between the technologies in the accuracy matrices are encountered in individuals with true ISUP grade 4 or 5 where the misclassified have an equal chance across being identified across all other categories in cognitive but are slightly less likely to be identified as NC or ISUP grade 1 with SF. For those in ISUP grade 2, sensitivity analysis indicates a low likelihood of the misclassified being identified as grade 1 with SF (and therefore being more likely to be classified as ‘no cancer’), which was not observed in the main analysis.

Diagnostic pathway

The diagnostic pathway is structured as a decision tree that captures AEs, repeat biopsies and classifies individuals according to the result of the biopsy (or biopsies), and the true disease status (see Diagnostic pathway), defined as ISUP grade for those with PCa (ISUP grades 1, 2, 3, 4–5). Figure 9 shows a simplified schematic of the decision tree illustrating biopsy-related mortality, sequence of biopsies, and cost and HRQoL pay-offs which apply for each strategy. The diagram does not show the biopsy-related non-fatal events, as these do not modify the probability of moving forward in the diagnostic pathway. The probabilities of AEs are applied as weights to adjust the branch costs and HRQoL pay-offs. The diagram also does not show how the classification is established conditional on the true disease state and test accuracy at each biopsy, or how the classification conditions the probability of repeat biopsy; this is illustrated in Appendix 11, Table 79.

FIGURE 9.

Decision-tree schematics. •, probability node; , terminal node; #, complement probability (1-probability); Bx, biopsy; c_, cost; du_, disutility; p1, probability of surviving the first biopsy; p2, probability of repeat biopsy (is conditional on first biopsy result); p3, probability of surviving second biopsy.

All individuals who undergo the first biopsy are at risk of biopsy-related non-fatal and fatal AEs. The mortality risk corresponds to the complement of probability (p1). For those who survive the first biopsy, the probability of receiving a repeat biopsy (p2) is conditional on the result of the first biopsy. Individuals who test positive at first biopsy (biopsy result ISUP grade ≥ 2) and survived the first biopsy receive no further testing (p2 = 0). Those who test negative (no PCa or ISUP grade 1) and survived the first biopsy have a probability of undergoing repeat biopsy (p2), with the remaining individuals receiving no further testing. The individuals who receive a repeat biopsy are again exposed to biopsy mortality risk (p1-p3), and to a probability of having non-fatal biopsy AEs. Time is not modelled within the decision tree, so events are assumed to occur instantaneously (or in rapid succession prior to long-term model entry); this is in line with the other cohort models examined in Results of the additional targeted reviews to support model conceptualisation.

The decision-tree models repeat biopsies for a proportion of individuals who have a negative-first biopsy result. In the base case, this proportion is not conditional on whether the strategy includes a cognitive or SF component. The base-case analysis assumes that the proportion of repeat biopsy is only conditional on the result of the first biopsy (15.45% and 5%, if the result of the first biopsy indicated a lesion with ISUP grade 1 and no PCa, respectively) as per a previous DAR. 116

Similar to a previous DAR,116 we assume the same rates of biopsy complications per biopsy approach for the first and repeat biopsies. However, because we assume a different distribution between transperineal and transrectal biopsy, for the first and repeat biopsies in the diagnostic pathway, the repeat biopsy complication rates reflect a higher proportion of TP (10% GATP and 60% LATP) compared to first biopsy (65% LATP) (see Biopsy procedure costs).

In the base-case scenario, the diagnostic performance of the repeat biopsy is assumed the same as of the first biopsy. The model allows exploring a degradation in the diagnostic performance of repeat when compared to first biopsy; the impact of applying this alternative assumption is assessed through scenario analysis.

We note the (first and repeat) biopsy results are assigned in the decision tree immediately before the biopsy mortality risk is applied, meaning the proportion of individuals in each category is adjusted by the proportion who survived the biopsy procedure (assuming the same mortality risk applies to all individuals regardless of true disease category and biopsy result). Similarly, we assumed that the biopsy AEs apply to all individuals who undergo a biopsy procedure.

The costs and QALY pay-offs in the decision tree capture the short-term impacts of first and repeat biopsy. First biopsy cost pay-offs apply to all branches and include the cost of the biopsy procedure and of associated AEs. Similarly, the QALY pay-offs of the first biopsy also apply to all decision-tree branches. These QALY pay-offs aim to quantify the QALY loss associated with biopsy procedural complications. The repeat biopsy-related costs (including the same cost categories as for the first biopsy) and repeat biopsy complications QALY loss apply only to the decision-tree branches which include a repeat biopsy.

The costs of the biopsy procedure vary by strategy to reflect the differences in cost between CF and SF with each of the MRI fusion systems modelled (see Biopsy procedure costs for the estimation of biopsy procedure costs). The biopsy procedure and AEs costs are both specific to the biopsy approach (LATP, GATP or LATRUS); these costs are estimated as a weighted average of the costs by biopsy approach (where the weights correspond to the proportion of LATP, GATP and LATRUS for each biopsy in the strategy). The QALY loss from biopsy-related complications also varies by biopsy approach to reflect the different biopsy complication rates by biopsy route of access (transperineal or transrectal) and, therefore, is also estimated as a weighted average by biopsy approach.

Clinical management conditional on biopsy final classification

There are 15 possible final classifications at the end of the diagnostic pathway, which are as follows:

1. For individuals correctly classified:

   1. Diagnosed as having no PCa and without PCa;

   2. Diagnosed as ISUP grade 1 and with ISUP grade 1;

   3. Diagnosed as ISUP grade 2 and with ISUP grade 2;

   4. Diagnosed as ISUP grade 3 and with ISUP grade 3;

   5. Diagnosed as ISUP grades 4–5 and with ISUP grades 4–5;

2. For individuals misclassified:

   1. Diagnosed as having no PCa and with:

      1. ISUP grade 1;

      2. ISUP grade 2;

      3. ISUP grade 3;

      4. ISUP grades 4–5;

   2. Diagnosed as ISUP grade 1 and with:

      1. ISUP grade 2;

      2. ISUP grade 3;

      3. ISUP grades 4–5;

   3. Diagnosed as ISUP grade 2 and with:

      1. ISUP grade 3;

      2. ISUP grades 4–5;

   4. Diagnosed as ISUP grade 3 and with:

      1. ISUP grades 4–5.

The clinical management for each of these possible classifications is dependent on the diagnosed category. As detailed in Care pathways for the diagnosis and management of prostate cancer, current clinical guidance10 recommends that individuals, diagnosed as having localised or locally advanced disease (henceforth referred to as localised disease for simplicity), are involved in decisions about the management of their disease, with the range of management options offered varying as a function of their prognostic risk. Thus, patients with lower CPG scores (better prognosis) are offered more conservative management (active surveillance) with option to undergo radical treatment, while those with higher CPG scores are offered radical treatment as the preferred management option.

The diagnostic performance evidence only allows classifying patients according to their histopathological information (i.e. ISUP grade), which is only part of the prognostic information used to determine the CPG scores. Therefore, we made a simplifying assumption that ISUP grade can be used as a proxy for the individuals’ CPG score (e.g. CPG1 = ISUP grade 1), to allow establishing the evidence linkage between classification and clinical management and subsequently from this to treatment outcomes. Henceforth, we refer to the classification in the model in terms of CPG score, assuming interchangeability between ISUP grades and CPG scores. The treatment options for localised disease include active surveillance or radical treatment. Radical treatment includes radiotherapy [consisting of the model of brachytherapy or external beam radiotherapy for costing purposes (see Prostate cancer treatment costs – metastatic disease)], and radical prostatectomy.

For individuals identified as having PCa, the model assigns varying proportions of active surveillance and radical treatment, according to diagnosed CPG score (see Treatment of localised prostate cancer). All patients in the localised disease health states receive monitoring, with the set of monitoring tests and schedule varying according to whether they are receiving active surveillance or radical treatment. Individuals without a PCa diagnosis also receive monitoring, but its regime is less intensive compared to individuals diagnosed with PCa and is time limited (maximum of 10 years).

Prostate cancer treatment is associated with AEs, such as sexual, urinary and bowel dysfunction, with rates of AEs varying by treatment (see Localised treatment adverse events). AEs from PCa management are associated with disutility and costs of managing these events, which are quantified within the long-term model.

Modelling of long-term outcomes

Overview of the decision-analytic model

The long-term outcomes of the model cohort conditional on latent true disease status, the diagnosed disease category and PCa management assigned are quantified in a state transition Markov model. The model has yearly cycles (with a half-cycle correction applied) and a lifetime time horizon (40 years).

The core structure of the model is illustrated in Figure 10. Individuals who survived the biopsy procedure(s) in the diagnostic pathway can enter the model through the no PCa state if they are disease free or the localised (and locally advanced) disease state if they have PCa. Patients with PCa at model entry can remain in the localised disease health state or transition to the metastatic disease state at each yearly model cycle. The individuals who died due to the diagnostic procedure enter the ‘other cause’ death state, one of the two absorbent states in the model (highlighted in grey in Figure 10). Transitions to the other-cause death state are possible from the ‘no PCa’, localised and metastatic disease health states, with the same probability as the general population (see Other-cause mortality). The only other possible transition for the ‘no PCa’ state is to remain in the same state (i.e. the model does not consider that individuals can develop PCa, so disease progression is not modelled for those who do not have the disease at model entry). The metastatic health state is modelled as three tunnel health states (not illustrated in this diagram, Prostate cancer treatment adverse event costs – localised disease, Figure 12), where individuals can only stay in the two first tunnels states for a maximum of 1 year. Patients who transition to the metastatic health state can only remain in that health state or die. PCa mortality only applies to patients in the metastatic disease states.

FIGURE 10.

Long-term outcomes Markov model structure.

There are 15 possible localised disease health states (illustrated in the box below the model schematics), each reflecting the final classification (here expressed as CPG scores) attributed by the diagnostic pathway and the different treatments assigned conditional on the diagnosed category in the final classification.

Over the next subsections we provide details on the parameterisation of long-term transition probabilities.

Inference sub-model (disease progression by Cambridge Prognostic Group and treatment intensity)

The decision-analytic PCa model requires consideration of the impact of treatment decisions according to diagnostic accuracy. Treatment decisions are currently grounded on the identification of CPG groups, and therefore the outcomes component of the model aims to reflect: (1) differences in outcomes across the CPG risk groups that underlie treatment decisions in clinical practice and (2) the impact of different treatment intensities on each of these risk groups. Our conceptualisation review has not identified any previous cost-effectiveness model where treatment outcomes for five-level CPG groups have been considered (see Results of the additional targeted reviews to support model conceptualisation). Therefore, an estimation strategy was developed in this assessment grounded on the targeted review of evidence on the long-term outcomes of PCa (see Review of long-term evidence).

The brief overview of the wider literature highlights that, while there is evidence on the effectiveness of radical versus ‘conservative’ treatment options in delaying progression to metastatic disease, there are limited mortality benefits observed within the follow-up of clinical trials in this area. Also, we did not find evidence on treatment effectiveness stratified by CPG scores, despite the prognostic ability of the five-level score for PCa-specific death having been demonstrated in a large UK-based observational study. 112

The aim of the inference model is therefore to pull existing evidence together to predict differences in progression to metastatic disease by five-level CPG score and by treatment. Given this has not been directly observed, a calibration model was developed to infer these. The calibration model uses the structure of the decision-analytic model in Figure 10, but without considering the ‘no PCa’ health state, which has, thus, been faded out in the diagram.

The model structure is underpinned by the following assumptions. All individuals are assumed to begin with localised disease. They can continue to have localised disease, progress to metastatic disease or die from causes other than PCa. The speed of progression to metastatic disease is expected to depend on CPG group and is given by λ_i, where the index i reflects the CPG group. Other-cause mortality is age-specific and is determined by δ_age. Those with metastatic disease may (1) continue to live with metastatic disease, (2) die from PCa or (3) die from other causes. Following the NICE NG131 model,123 it was assumed that death from PCa could only occur after metastatic disease. The model was parameterised for each CPG score of interest to this assessment (CPG 1, 2, 3, and 4 and 5 combined).

The inference procedure is undertaken in two parts.

Part 1: identifying rates of progression to metastatic disease by CPG, λ_i

This part uses calibration. For any calibration process, two sets of parameters are of interest. The first concerns model parameters, some of which are unobserved and the target of inference, and others are observed and therefore evidence directly informs these. The second set concerns calibration targets, which are functions of the model parameters that have been observed and are used to identify the unobserved parameters under the model structure and other observed inputs. Table 11 lists the calibration parameters and targets and presents the results of the calibration model. A more detailed description of these parameters and their evidence sources is presented in the subsequent subsections.

TABLE 11 - Calibration model parameters and targets, and calibration results

	Description	Source	Parameter value	Results
PART 1	Calibration targets
	10-year PC death by CPG group at diagnosis	Gnanapragasam et al.112 pooled results for testing and training sets	10-year PC survival (SE) (a, b parameters of a beta distribution): G1: 0.968 (0.007) (586, 19) G2: 0.938 (0.010) (577, 38) G3: 0.871 (0.016) (356, 53) G4/5: 0.763 (0.052) (50, 16)	–
	Calibration model parameters
	Unobserved rate of progression from localised to metastatic disease, by CPG: λ1, λ2, λ3, λ4	Unobserved (calibration parameters)	NA	Rate (SE): G1: 0.0101 (0.00236) G2: 0.0229 (0.00403) G3: 0.0645 (0.01058) G4/5: 0.1788 (0.08641)
	Observed rate of PC mortality from metastatic disease, γ	STAMPEDE143	Yearly rate of PC mortality in ADT arm of 0.162 (SE 0.0073), calculated from 5-year PC mortality	–
	Observed rate of death from other causes, age-specific	ONS life tables (2000–2)144	Assumed mean age for each CPG group G1: 66.2 G2: 68.14 G3: 71.13 G4/5: 72.18	–
PART 2	Proportions under radical treatment vs. conservative management, by CPG	Gnanapragasam112	G1: 0.53 G2: 0.70 G3: 0.81 G4/5: 0.95	–
	Rate ratio for the development of metastasis of radical vs. conservative treatment	ProtecT55	Rate ratio = 0.43 95% CI (0.26 to 0.72), log rate ratio mean = –0.834, SE = 0.2545	–
	Rate of progression from localised to metastatic disease, by CPG and by treatment	Unobserved	NA	Conservative management λ1(0) = 0.0143 (0.00357) λ2(0) = 0.0380 (0.00832) λ3(0)= 0.1197 (0.02812) λ4(0) = 0.3950 (0.22287) Radical treatment: λ1(1) = 0.0063 (0.00184) λ2(1) = 0.0165 (0.00357) λ3(1)= 0.0516 (0.00964) λ4(1) = 0.1674 (0.08066)

Calibration targets

Our calibration target is 10-year PCa-specific mortality according to CPG group at diagnosis of localised disease, as reported in Gnanapragasam et al. 112 Our analysis combines groups 4 and 5, as the recommended treatment is the same for both groups. 10 We used a single data point for each CPG group of interest, at 10-year follow-up, in the calibration model. We used WebPlotDigitizer145 to extract point estimates and upper and lower CIs for PCa survival at 10 years (3652 days) from both training and validation sets in Gnanapragasam et al. 112 (from figures 1a and 2a). Standard errors (SEs) were calculated by considering the average distance between the point estimate and the upper and lower confidence limits (where both were available). The figures were then combined across data sets to derive a single estimate for each CPG, by weighting according to the inverse of their precision (analogously to a fixed-effect meta-analysis). Values for the combined CPG 4 and 5 group were derived by pooling the distributions, that is, assuming that the variance of the combined group is the weighted sum of the variances in each group. To describe the survival probabilities probabilistically, the parameters of beta distributions were specified using the method of moments. The estimates of 10-year PCa survival and the parameters of the beta distributions used to describe this in the calibration model are presented in Table 11. When simulating from the beta distribution to run the calibration, we preserved the ordering ensuring survival is highest in group 1 then group 2, group 3 and groups 4–5.

Calibration model parameters

The rates of progression from localised to metastatic disease by CPG score (λ_i, where i represent the CPG score groups of interest) were the unobserved parameters we sought to achieve inference on.

The remaining model parameters were observed. PCa specific mortality was assumed to only be possible after progression to metastatic disease, and therefore to inform this model parameter we used outcomes reported from STAMPEDE, a UK study. 143 Data from STAMPEDE’s control arm were used, as long-term hormonal treatment was the SOC at the time the study informing the calibration target was conducted. The individual patient data were reconstructed from the published Kaplan-Meier curve using the Guyot algorithm146 and a Weibull distribution was fitted using the flexsurv package in R. 147 PCa survival at 5 years, predicted by the fitted Weibull function, was 40.6% (95% CI from 43.9 to 47.0%), which was converted onto a rate assuming constant hazard. This resulted in a mean hazard of 0.162 and a 95% CI from 0.1777 to 0.1492. Assuming a symmetrical distribution, this implies a SE of 0.0073.

Office for National Statistics (ONS) life tables for men144 were used to parameterise the transitions to death from other causes (both from localised disease and from metastatic disease). Life tables were used for the years 2000–2 to approximate the mortality at the time of the Gnanapragasam study (2000–10). The average age at the start of that study differs by risk group according to data reported in the NICE model. 123 Using linear interpolation, we extended the three risk groups reported in the NICE report to the four risk groups we are considering. The ages assumed were: 66.2 years for Group 1; 68.14 years for Group 2; 71.13 years for Group 3 and 72.18 years for Groups 4–5. Due to the large sample size underlying the life tables, we did not consider this parameter uncertain.

Analysis methods

Using the parameters and targets described above, we ran the calibration analysis in the software package R, according to the algorithm below:

1. Sample a value from the uncertainty distribution for the target (PCa mortality at 10 years, for each risk group) and the known model inputs (metastatic mortality rate).

2. For each risk group, identify the value of the rate of transition from localised to metastatic disease (λ_i) that is consistent with the PCa mortality at 10 years sampled in step 1. Record the result.

3. Repeat steps 1 and 2, 10,000 times.

The optim function in R was used for the second step in this algorithm. 36 To find the rate consistent with the target, we defined a discrete time Markov model with the structure in Figure 11, and determined that its predicted 10-year survival should be compared against the target value. The loss function used was the squared distance from the proposed value to the target value. The Brent method was used with lower and upper bounds of 0 and 10, respectively. 148

FIGURE 11.

Calibration model structure and parameters.

Results

The results from the calibration procedure regarding the unobserved rate of progression from localised to metastatic disease by CPG are shown in Table 11. Comparisons of calibration parameter estimates with those from recent UK cost-effectiveness models are presented in Tables 80 and 81, Appendix 11.

Part 2: identifying the effect of treatment on the rates of metastasis

The estimated rate of progression to metastatic disease from the calibration exercise above reflect outcomes with current practice, which comprises a mix of radical and conservative treatment. In part 2, we back-calculate how these rates differ for the proportions treated with radical and conservative treatment observed in Gnanapragasam et al. 112 using an external estimate of effect for radical treatment.

Gnanapragasam et al. 112 reported the treatment mix by risk group observed in UK clinical practice during the years 2000–10. The treatment categories considered were: conservative management, brachytherapy, primary ADT, radical prostatectomy and radical radiotherapy. We further grouped treatments into two categories: conservative management and all other options, which we considered ‘radical treatment’ (see Appendix 11 for further details and comparison to NICE guidance). The split by risk group is shown in Table 11.

The rates of progression to metastatic disease inferred in the calibration step (part 1) reflect the treatment allocations in Gnanapragasam et al. 112 (see Table 11). To consider the rates of progression with and without radical treatment, we disentangle the effect of treatment by considering that the pooled estimate of the rate of progression is a weighted average of the rates under radical treatment and conservative management (weighted by the proportions treated). The rates under radical treatment are assumed to be the rates under conservative treatment multiplied by a rate ratio sourced from external evidence. For such, we use the treatment effect from ProTecT,55 the most recent UK study identified in the targeted review of the literature (see Review of long-term evidence). The rate ratio data for radical treatment pooled the PROTecT results for radical prostatectomy and radiotherapy, retrieving an estimate of 0.43 (95% CI from 0.26 to 0.72). Note that this estimate is similar to the US-based PIVOT study which estimated a HR for developing bone metastasis for radical prostatectomy of 0.40 (0.22–0.70).

Results

The results from the second part of the inference model, regarding the rate of progression from localised to metastatic disease by CPG and by treatment, are shown in Table 11.

Parameterisation of the prostate cancer health states transition probabilities

The transition probabilities from each of the 15 localised disease health states to metastatic disease were informed by calibration as described above [see Inference sub-model (disease progression by Cambridge Prognostic Group and treatment intensity)]. The calibration estimated the transition rates by true disease status and treatment assigned (active surveillance or radical treatment). Transition probabilities were subsequently estimated by weighting the annual transition rates according to the treatments assigned based on diagnosed category (see Treatment of localised prostate cancer), and then converted to annual transition probabilities assuming constant hazards over time (i.e. an exponential time to event distribution).

For patients in the metastatic disease health state transitions to PCa death were informed by PCa-specific death from the UK-STAMPEDE trial. 143 As described previously, a Weibull distribution was fitted using the flexsurv package in R data to the reconstructed individual-level PCa mortality data for the SOC (ADT) arm (metastatic patient subgroup) in Clarke et al. 143 The choice of parametric distribution was in line with a recent NICE TA evaluating enzalutamide in combination with ADT for hormone-sensitive metastatic cancer and based on a visual fit assessment conducted by the EAG (a full assessment of survival curve fit was considered out of scope for this assessment, so a targeted approach was taken). This baseline probability was parametrised in the executable model based on the flexsurv estimated Weibull coefficients (with a multivariate normal distribution fitted using the corresponding Cholesky decomposition for the PSA) and then adjusted by the effectiveness of contemporaneous combination treatments (i.e. in addition to ADT) weighted HR according to the current treatment distribution (see Treatment of localised prostate cancer). The weighted hazard ratios (HRs), for three combinations used in current clinical practice for first-line metastatic PCa compared to ADT alone, were applied to the baseline probability of PCa death to derive the metastatic to PCa death transition probability. The combination treatments considered in the model included DTX (HR vs. ADT 0.78, 95% CI: 0.66 to 0.93),143 enzalutamide (HR vs. ADT 0.66, 95% CI: 0.53 to 0.81)149 and apalutamide (HR vs. ADT 0.65, 95% CI: 0.53 to 0.79). 150 Lognormal distributions were fitted to each HR in the probabilistic model setup.

Other-cause mortality

Age-dependent other-cause mortality rates for men from Office for National Statistics (ONS) lifetables (2018–20 collection period)144 was used to estimate other-cause death probabilities in the long-term model. Parameter uncertainty was not considered for these inputs, due to the large sample size of the source data set.

Resource use and costs

The resource use and costs, considered in the diagnostic pathway, include those associated with the biopsy procedure and its adverse eves. The long-term model quantifies the costs of monitoring individuals following the diagnostic procedures in the diagnostic model, the costs of PCa treatment and end of life. Costs in the model are expressed as 2020–1 Great British pounds, estimated from a NHS and PSS perspective, and discounted at a 3.5% annual discount rate.

The resource use and cost in the long-term model (costs associated with monitoring, PCa treatment, treatment AEs and end of life) was largely informed by the Southampton DAR,116 as were the unit costs sources (updated or inflated to 2020–1 price year as appropriate). Therefore, descriptions of these categories of cost and resource use are brief and refer back to the Southampton DAR model. 116 Emphasis is put into describing elements where our assumptions and/or parameter sources differ from those of the Southampton DAR model. 116

Parameter uncertainty in resource use and costs inputs was not considered for the large majority of the inputs due to lack of information on their variance and the reliance on assumptions to define parameter quantities.

Biopsy procedure costs

This section reports the costs associated with the biopsy procedure, which include the following components:

1. Cost of the SF system – costs of the fusion software and, in some cases, a workstation (or cart). This cost only applies to the diagnostic strategies which include a SF component.

2. Cost of the ultrasound – cost of the ultrasound probe/transducer, and any required software. This cost applies to diagnostic strategies with either software or cognitive function components, but some SF systems are not compatible with third-party ultrasounds.

3. Cost of SF system installation – cost of connecting the SF system to the NHS trust IT system. This cost only applies to the diagnostic strategies which include a SF component.

4. Cost of SF system maintenance – costs of service contracts to maintain the technology and keep software up to date. This cost only applies to the diagnostic strategies which include a SF component.

5. Costs of SF system training – staff time costs required to train NHS professionals to perform biopsies. The use of SF methods requires additional training compared to CF, but the cost of training also varies across biopsy approaches (by route of access).

6. Cost of staff time to perform the biopsy procedure – cost of urologists, nurses and anaesthetist (for procedures requiring general anaesthesia). This cost varies across biopsy approaches (by route of access and type of anaesthesia), but there is also a difference in procedural time between SF and CF.

7. Cost of the biopsy setting – costs of the setting in which the biopsy procedure takes place (outpatient room, theatre session); it varies by route of access, type of anaesthesia, and MRI-influenced method.

8. Costs of other biopsy devices and consumables – cost of (a) devices and equipment (e.g. freehand needle positioning devices, lithotomy beds and biopsy guns) and (b) needles and other materials requiring replacement (immediate or after a certain number of uses). These costs are often specific to the biopsy approach [transrectal or transperineal (stabilised, freehand or double freehand)], and may differ across MRI-influenced methods and across SF systems, due to compatibility issues.

9. Cost of histopathology analysis and report – costs of processing the biopsy sample and communicating the results to the patient in a consultation. This cost applies to all strategies but may differ for strategies using different sampling methods (combined vs. targeted-only biopsy), as these may result in different number of cores being sampled.

The evidence considered to estimate these components of costs (and their calculation) is detailed in Tables 85–95, Appendix 11. Table 13 summarises the aggregated cost per biopsy for each technology and by biopsy approach, with further breakdown of costs in Tables 96–98 in Appendix 11.

TABLE 13 - Cost per biopsy by technology and biopsy approach

GATP, general anaesthesia transperineal biopsy; LATRUS, local anaesthesia transrectal ultrasound.

	Technology	Biopsy approach
		LATRUS (£)	GATP (£)	LATP (£)
Technology specific	bkFusion	147.48	380.67	231.68
	FusionVu	169.47	402.67	253.68
	KOELIS Trinity	150.37	384.86	235.87
	BiopSee	89.51	323.52	179.18
	Fusion Bx 2.0	158.10	391.72	242.73
	CF	48.44	260.00	133.07
Non-technology specific		209.95	634.15	239.25
Total cost per biopsy	bkFusion	356.53	914.82	470.93
	FusionVu	378.53	936.82	492.93
	KOELIS Trinity	359.43	919.01	475.12
	BiopSee	298.56	857.67	418.43
	Fusion Bx 2.0	367.15	925.87	481.98
	CF	257.49	794.15	372.32

As stated in Modelling of long-term outcomes, we assumed for the first biopsy in the diagnostic pathway, 65% of biopsies were conducted with LATP and the remainder with LATRUS. For the repeat biopsy, we assume that 60% are LATP, 30% are LATRUS and 10% are GATP (to reflect those individuals where there was concern that first biopsy may not have been accurate due to patient moving excessively during the procedure). We weighted the costs per biopsy approach by the corresponding proportions for first and repeat biopsy to estimate their costs in the model; these costs are reported in Table 14.

TABLE 14 - Cost of first and repeat biopsy in the model

Technology	1st biopsy cost (£)	Repeat biopsy cost (£)
bkFusion	430.89	481.00
FusionVu	452.89	503.00
KOELIS Trinity	434.62	484.80
BiopSee	376.47	426.39
Fusion Bx 2.0	441.79	491.92
Average cost SF	427.33	477.42
CF	332.13	380.05

There are a number of uncertainties in the biopsy procedure costs. These pertain to:

• the set of essential components that are integral part of each technology and the lifespan for all components

• the potential commercial discounts that may be offered by the companies, what is included in the commercial arrangements and how do these apply to each technology

• what additional costs may stem from compatibility issues with existing equipment and accessories in use in the NHS

• the additional time required to perform SF

• how training for the use of SF is delivered (to whom and for how long), and if the training requirements differ substantially between SF technologies.

Given these uncertainties and that it was not possible to calculate diagnostic performance evidence by individual SF devices at the granularity of classification (ISUP G1, ISUP G2, ISUP G3, and ISUP G4 or 5) required by the economic model, it was considered the biopsy procedure costs for each individual technology was potentially misleading to decision-makers. Thus, we apply the average biopsy cost across all SF technologies in this assessment for which cost data were submitted by the companies. Given this and the numerous uncertainties in the cost estimation of each SF technology, it was not considered appropriate to compare each SF technology against each other and CF in the model. Instead, in the base-case analysis, we apply the average cost per biopsy across all SF. Individual SF technology costs are presented alongside the base-case analysis results to illustrate how their individual costs would impact on the estimates of cost-effectiveness.

Biopsy procedure adverse events costs

The biopsy procedure-related AE costs, considered in the diagnostic pathway model, were estimated by multiplying the AE rate by the unit cost for each type of AE. The unit costs for each type of AE were derived from the Southampton DAR sources {updated for the 2020–1 price year by either using the corresponding versions of national tariffs [e.g. Personal Social Services Research Unit (PSSRU) and NHS reference costs] or inflating costs using the NHS Cost Inflation Index (NHSCII),154 as appropriate} and using the same assumptions (e.g. on resource use required to treat a mild AEs);116 further details are presented on in Table 99 in Appendix 11.

Monitoring costs

Routine monitoring costs at model entry apply to all patients who enter the long-term model. In the model, the set of monitoring tests and schedule varying according to whether the individuals have been diagnosed:

• localised and locally advanced PCa, and if so, monitoring also varies with:

  • the diagnosed CPG category (CPG 1, CPG 2–3 or CPG 4–5)

  • treatment assigned (active surveillance or radical treatment)

  • and time in the model (first, second or subsequent years).

• or not, and if so, monitoring only varies with the underlying true disease status (no PCa or CPG1–5).

Table 100 in Appendix 11 summarises the resource use and cost per year of the monitoring tests considered in the model for patients in the diagnosed as localised (and locally advanced) disease health states. These costs are applied from model entry (cycle 0) and while individuals remain in the localised disease health states.

We assumed that individuals without a PCa diagnosis would also undergo routine monitoring, regardless of whether they had PCa. In contrast, the Southampton DAR116 only attributed a cost of monitoring to those with localised PCa who had not been identified as having PCa. We changed this assumption in the York model, because in principle these two groups of individuals would be indistinguishable, as true disease status would be unknown to clinicians. We assume that in both groups individuals receive the same monitoring schedule when they are discharged to primary care: an annual PSA test (velocity test at a threshold of 75 ng/ml/year) for up to 10 years, performed at a 10-minute nurse-led appointment, and followed by a CF biopsy (costed at £477.75; assumes 35% LATRUS and 65% LATP) if the PSA test results is positive. As per the Southampton DAR,116 the probability of testing positive in the PSA test for those with PCa was assumed to be 0.69, which corresponds to the sensitivity of the corresponding PSA velocity test used in NICE NG131 model. We further assumed that the probability of testing positive in the PSA test for those without PCa corresponded to one minus the specificity of the same test (1–0.56 = 0.44). The testing schedule is similar to what was modelled in the Southampton DAR116 for those with PCa who were diagnosed as not having the disease, but the first PSA test is assumed in the York model to occur within 1 year in the long-term model (rather than 6 months). The annual cost per year of monitoring applied in the York model was £342.50 and £223.06 for those with and without PCa, respectively. These costs are applied from model entry (cycle 0) and for up to 10 years in the entry health states.

After 2 years in the model, individuals in the local disease health states are assumed to be correctly identified at their true disease status, and move to the monitoring regime that matches their true disease status.

Individuals who enter the metastatic health state incur a one-off monitoring cost of £577.83, corresponding to the same resource use as in the Southampton DAR116 (i.e. one CT and bone scan).

Prostate cancer treatment costs – localised disease

Individuals identified as having localised PCa are assumed to receive treatment at long-term model entrance according to their diagnosed CPG (see distribution of treatments by diagnosed CPG in Treatment of localised prostate cancer). Individuals who receive active surveillance is assumed to not incur any treatment costs (only monitoring costs as detailed in Prostate cancer treatment costs – localised disease), so costs of treatment are only incurred by those who undergo radical treatment.

Radical treatment resource use and costs vary according to the type of radical treatment (radical prostatectomy, external radiotherapy or brachytherapy). The cost of each type of radical treatment procedure applied in the York model is reported in Table 101, Appendix 11, alongside details on resource use and unit costs. We note that the cost of brachytherapy has increased considerably in relation to the one used in the Southampton model (£9156.96 vs. £3106.02); these differences are driven by an increase in the unit cost of delivering brachytherapy in an outpatient setting (as well as increased activity for the corresponding currency code) in 2020–1 compared to 2019–20. The costs of the radical treatment procedures were applied as one-off costs at long-term model entry (cycle 0). For those who were misdiagnosed and treated with conservative treatment, it is assumed that they receive radical treatments according to their true disease status after 2 years in the model.

In addition to the medical procedures, we also included the cost of ADT for those patients who treated with radiotherapy, according to NICE guidance. 10 ADT in the localised disease setting was assumed to consist of the same treatments as in the Southampton model, that is, bicalutamide 50 mg for 21 days followed by luteinising hormone-releasing hormone (LHRH) agonists [either leuprorelin 11.25 mg (every 3 months), triptorelin 11.25 mg (every 3 months) or goserelin 3.6 mg (every 28 days)]. Similarly, to the Southampton model, the duration of LHRH treatments was varied according to category of prognostic risk; we assumed LHRH treatment duration would be 3 and 6 months for those diagnosed in CPG 1 and CPG 2–3 categories. For those diagnosed in the CPG 4–5 category, we updated the duration of treatment for 3 years, in line with the current NICE guidance for that prognostic risk group (see Care pathways for the diagnosis and management of prostate cancer). 10 The costs of ADT included drug acquisition and administration costs and were costed as per the Southampton DAR (updated to the current price year).

Prostate cancer treatment costs – metastatic disease

Metastatic disease treatment was assumed to consist of hormone-sensitive disease treatment for the first 2 years in the metastatic health states, followed by hormone-resistant disease treatment. Costs of metastatic treatment are summarised in Table 102, Appendix 11; these include drug acquisition and administration costs.

In line with the Southampton DAR, hormone-sensitive metastatic treatment was modelled as a blended treatment consisting of ADT alone (but not identical to the regimes described for the localised disease setting, as course of bicalutamide 50 mg is longer), or in combination with either DTX, apalutamide or enzalutamide. We updated the distribution of treatments for the hormone-sensitive metastatic treatment, as described in Treatment of metastatic prostate cancer. We note that yearly costs of metastatic treatment have increased considerably in the York model compared to the Southampton model (e.g. metastatic hormone-sensitive first year cost increased to £15,603.87 from £8388.63), due to the increased proportion of individuals treated with ADT combined with enzalutamide, due to the high cost of enzalutamide. Furthermore, although we apply the same DTX treatment regimen [i.e. six cycles (delivered every 3 weeks) at a dose of 75 mg/m²; body surface area 1.91] as in the Southampton model, in the York model the 2-year DTX treatment costs are assumed to be distributed evenly between two model cycles (constant annual cost).

The Southampton DAR states that ADT alone or in combination was taken until disease progression, which was assumed to occur after 2 years. We also make their stated assumption, but we implemented it in a different way. In the Southampton model, a cost for first and second year is estimated for metastatic treatment (both treatment for hormone-sensitive and hormone-resistant disease) and applied to individuals in the metastatic disease state at first and second year (modelled in a way akin to tunnel states), respectively. Thus, in the Southampton model the cost of hormone-sensitive and hormone-resistant treatment is applied to the same set of individuals. In the York model, we explicitly model a set of three tunnel health states representing the first, second and subsequent years of metastatic disease (Figure 12). We applied the costs of hormone-sensitive metastatic treatment to individuals in the first and second year of metastatic tunnel health states, and the costs of hormone-resistant metastatic treatment costs are applied as a one-off cost to individuals who enter the ‘metastatic subsequent years’ health state.

FIGURE 12.

Diagram of metastatic tunnel health states and death states.

Another difference between models is that in the York model metastatic treatment (and monitoring) is assumed to apply to all patients with metastatic disease, as we do not distinguish between diagnosed and undiagnosed metastatic disease (the latter does not appear to incur treatment costs in the Southampton model). Thus, we implicitly assume that all individuals with metastatic disease have been diagnosed.

All metastatic treatments costs are applied as an average of the costs of the different types of treatments weighted by their treatment distribution (see treatment distribution in Table 12, Treatment of localised prostate cancer).

Prostate cancer treatment adverse event costs – localised disease

The model considers the costs of managing the AEs from active surveillance, radical prostatectomy and radiotherapy for localised PCa. These costs were estimated by multiplying the AE rates (see Localised treatment adverse events) by the unit cost of the corresponding AE (see Table 103, Appendix 11). The costs are applied in the model as a one-off at localised disease health states to the proportion of patients who receives each treatment (see treatment distribution in Table 12, Treatment of localised prostate cancer).

Prostate cancer treatment adverse event costs – metastatic disease

The costs of managing metastatic treatment-related AEs was applied in the model. Similarly, to the Southampton model only AEs of treatment for hormone-sensitive metastatic disease were included. The AE costs for androgen therapy alone and in three alternative combinations (with DTX, apalutamide or enzalutamide) were estimated by applying the unit cost per type of AE (see Table 103, Appendix 11) by the corresponding rate (see Metastatic disease treatment adverse events). The resulting costs per treatment were then applied as a one-off cost at entrance to the ‘metastatic 1st year’ health state. The one-off cost was estimated by weighing each treatment cost by the metastatic treatment distribution (see treatment distribution in Table 82 in Appendix 11).

End-of-life costs

End-of-life costs are applied to all individuals who die in the model of other-cause or PCa death, but not to those who have died of peri-procedural biopsy complications (see Biopsy procedure related adverse events). This one-off cost is applied to individuals who enter the death states at each cycle in the model, and it was sourced from Round et al. 139 and inflated to 2020–1 price year. 154

Analytic methods

Statement of principal findings

The systematic review of clinical evidence included a total of 3733 patients who received SF and 2154 individuals with CF from 23 studies. Evidence was included for all devices specified in the protocol, except for Fusion Bx 2.0 and FusionVu. Fourteen studies were included in the network meta-analyses.

Overall, the evidence for all devices was at high risk of bias and therefore the quantitative synthesis results must be interpreted with caution. Results from our main analysis (looking across ISUP grades) suggest that patients undergoing software biopsy may have: (1) a lower probability of being classified as not having cancer, (2) similar probability of being classified as having non-CS cancer (ISUP grade 1) and (3) higher probability of being classified at higher ISUP grades, particularly ISUP 2. Similar results were obtained when comparing between same biopsy methods where both were combined with systematic biopsy.

Additional meta-analyses of cancer detection rates suggest that, compared with CF biopsy, SF may identify more PCa (any grade) (OR 1.30; 95% CrI 1.06, 1.61). Adding systematic biopsy to cognitive or SF may increase the detection of all PCa and of CS cancer, and from this evidence there is no suggestion that SF with concomitant systematic biopsy is superior to CF with systematic biopsy.

Meta-analyses by individual device showed that compared with CF biopsy, BioJet and Urostation may be associated with a higher detection of PCa overall, and BioJet may be associated with a higher rate of CS cancers, although only one study of BioJet was included in the meta-analyses. Evidence for all other software devices was insufficient to reliably compare their accuracy with CF, or to determine whether some SF technologies are more accurate than others. Evidence for bkFusion, iSR’obot Mona Lisa and KOELIS Trinity was included in the systematic review but not in the meta-analyses. Compared with CF, there was no evidence that the accuracy of SF may differ by lesion location, or between biopsy-naive and prior negative-biopsy patients, or according to operator experience, although the number and quality of the studies informing the potential effect modifiers were limited.

Overall, there is no evidence that biopsy positivity rates and safety outcomes differ significantly between SF and CF, or between SF devices. There was some evidence that systems with rigid registration (BioJet or UroNav) are easier and faster to use than elastic registration (KOELIS Trinity), although this is informed by a single, small study and is not conclusive.

The base-case cost-effectiveness analysis suggests for the targeted biopsy and the combined biopsy comparisons, that SF strategy is on average costlier and yields greater QALYs than the CF strategy, resulting in a probabilistic ICER of £6197 and £2199 per additional QALY for each comparison, respectively. These ICERs are below the lower bound of the cost-effectiveness threshold range recommended by NICE, suggesting that SF may be cost-effective compared to CFs in both the targeted and the combined comparisons. However, these results should be interpreted cautiously, given the uncertainties in the relative diagnostic accuracy evidence which informs the model. The probabilistic analysis suggests a higher probability of cost-effectiveness for SF versus CF at the range of cost-effectiveness thresholds recommended by NICE (0.64 and 0.68 at £20,000 and £30,000 per additional QALY for targeted SF biopsy).

The key findings on the drivers of economic value of SF compared to CF are:

1. The costs (and harms) of SF biopsy in the diagnostic model component can only be offset in the long-term model component, which will only arise from differences in diagnostic accuracy between software and CF.

2. The value gains for SF appear to stem from increased detection at CPG ≥ 2 and, once we adjust for prevalence by CPG category, the greatest contribution to the cost-effectiveness of SF compared to cognitive results from increased correct detection at CPG 2.

3. Increased detection at CPG 1 due to reduced detection of ‘no cancer’ results in value losses at all cancer grades [i.e. there are net losses from shifting classification from ‘no cancer’ to CNS cancer (CPG 1)].

4. The magnitude of value realised for SF versus CF from the balance between different degrees of misclassification and correct classification with the two technologies also depends on the prevalence at each cancer grade.

Given the uncertainties in the costing of SF, we conducted a threshold analysis to identify the SF biopsy cost at which there would be a shift in the decision to accept SF as a good use of NHS resources. This suggested that at the cost of each the five individual technologies for which there was cost data, the recommendation decision would not change.

The base-case cost-effectiveness results were not sensitive to variations to alternative data sources and assumptions, except when no difference in diagnostic accuracy is assumed between SF and CF. Under this assumption, the ICERs for SF compared to CF (targeted and combined biopsy analysis) far exceed the upper bound of the cost-effectiveness threshold range recommended by NICE.

Strengths and limitations of the assessment

This is the first systematic review to formally compare the relative accuracy of SF and CF, with and without systematic biopsy, as well as different SF devices, using both direct and indirect evidence in a formal NMA. In order to best estimate differences between biopsy methods for each PCa grade, a multinomial logistic regression model was fitted, where the odds of being categorised in each of the different ISUP grades were allowed to vary by biopsy type.

Our findings are consistent with those of recent systematic reviews that found no significant difference between SF and CF at detecting non-CSPCas,51–53 although unlike recent evidence,51,53 our NMA found that SF increased detection of CS cancer compared with CF. This result might be explained by differences in review and synthesis methods.

Our review has a number of limitations. Despite attempts to reduce bias by excluding unpaired, non-randomised studies, the evidence included in the meta-analysis remains at high risk of bias. Although within-patient comparisons remove much of the risk of confounding from imbalances in participant characteristics, true blinding from tracks of preceding biopsy methods within the same examination is not feasible (or would require two separate biopsy sessions per patient, which would be unethical). So far, no high-quality RCTs have been published.

There was variation across the studies in patient characteristics. In particular, a number of studies included patients with prior negative-biopsy and biopsy-naive patients, who form the large majority of patients eligible for targeted biopsy, were under-represented. Some variation and gaps in reporting were observed in MRI acquisition methods, criteria for referral to biopsy, biopsy routes and anaesthesia methods. Definitions of PCa and CS cancer varied across the studies. There was insufficient evidence to explore the impact of a number of potential effect modifiers, including lesion location, operator experience, biopsy routes and anaesthesia methods.

The results of the synthesis models require careful interpretation, as they refer to comparisons between different cancer grades. The interpretation of the multinomial models on the absolute probability scale results is more intuitive and directly relevant to clinical practice. Overall, results are concordant across analyses and concordant with the data. Only the multinomial results are used in the economic model, as the value of diagnostic information provided by each test is dependent on the subsequent clinical decisions based on test results, and clinical management is conditional on cancer grades (jointly with other prognostic information).

Most estimates from the NMAs were imprecise, particularly in the multinomial synthesis and at higher ISUP grades where data were most sparse. The NMA relied on a number of assumptions. CF was assumed to be equivalent across studies. The risk and extent to which the accuracy of CF may vary by centre and operator experience are uncertain due to lack of evidence. It was also assumed that data from within-patient studies were independent. A model that accounted for the full structure of the data was not available, although it could have added precision to the estimates.

There were few studies per comparison and not all studies reported outcomes by all cancer grades. Therefore, only fixed-effect models were fit to the data. Data were sparse for most SF devices, and few studies included more than one SF technology, making it difficult to draw conclusions for relative accuracy of individual devices.

While our review identified several relevant studies, many could not be included in the synthesis due to lack of reporting of key data. For example, studies comparing software and CF to systematic biopsy reported data on both targeted technologies jointly, and few studies reported a sufficient breakdown of biopsy results by ISUP grades (or equivalent breakdown) to inform the evidence synthesis required for the economic model. In addition, where studies included a mixed population of patients, a lack of reporting of biopsy results for the relevant population led to their exclusion from the meta-analysis. We were therefore limited in the models we could consider due to data sparseness, and results are uncertain.

Studies not included in the meta-analyses mostly reported test-positive rates (positive cases as percentage of all patients). As this measure is dependent on disease prevalence rather than diagnostic accuracy, results from these studies may be influenced by differences in PCa rates between cohorts and may not be reliable.

The above-mentioned limitations in the evidence are not captured in the quantitative evidence synthesis, which is used to inform the economic analysis.

The cost-effectiveness analysis relies on the evidence informing it. Beyond the evidence sourced from the synthesis, this includes evidence on the long-term outcomes of treating PCa and the cost data on each SF technology. This evidence is limited.

Uncertainties

No evidence was found for most of this assessment’s prespecified outcomes: biopsy sample suitability/quality, number of repeat biopsies performed, procedure completion rates, software failure rate, time to diagnosis, length of hospital stay, time taken for MR image preparation, subsequent PCa management, re-biopsy rate, hospitalisation, OS, PFS, patient- and carer-reported outcomes (including tolerability and HRQoL), barriers and facilitators to implementations.

There was large uncertainty in all estimates due to the limited evidence. Meta-analyses showed moderate heterogeneity that could not be explained by differences in individual SF devices. The evidence for all SF devices was at high risk of bias, and the diagnostic accuracy of systematic biopsy relative to SF and cognitive may have been overestimated in the meta-analyses. The applicability of the evidence for KOELIS Trinity and BiopSee is uncertain. There is no evidence comparing the accuracy of Fusion Bx 2.0 and FusionVu with CF, and no evidence for these devices were eligible for inclusion in the indirect comparisons.

None of the studies included in the systematic review of diagnostic accuracy used template mapping biopsy as a reference standard, and many studies did not use standard 12-core systematic biopsy in addition to targeted biopsy methods. This means that the absolute true rate of PCa lesions was underestimated and is uncertain. However, the lack of a gold-standard test is likely to have affected comparisons between all devices similarly, and therefore is unlikely to have biased relative estimates of PCa detection.

The evidence supporting recent consensuses for classifying the clinical significance of PCa is not without limitations, despite recent improvements in imaging. 34,112,151 Trial evidence indicates that survival outcomes for some patients with more severe grades of localised PCa (CPG 3 and above) are favourable, and that the detriment associated with active monitoring may be small. 55,114 This raises further uncertainty regarding the added clinical (and economic) value of SF. Where reported, the number of targeted cores performed with software and CF were broadly comparable between the studies. However, not all studies reported data on number of targeted cores to fully assess the risk of confounding from a possible difference in number of targeted cores between software and CF. Evidence for all other protocol-specified outcomes was limited and inconclusive.

The economic value of SF seems to be driven by (1) comparative diagnostic accuracy derived where evidence is particularly sparse (cancer grades above 2), and (2) by prevalence, which is also affected by evidence sparsity.

Structural assumptions were applied to allow estimating the economic value of SF given the evidence gaps in the diagnostic accuracy of SF and in the longer-term outcomes of PCa. The structural uncertainty of these assumptions was explored in scenario analysis, but could not be jointly captured with parameter uncertainty. This means that the probabilistic results presented in this report are likely to underestimate the overall uncertainty.

Other relevant factors

Participants of studies included in the systematic review of diagnostic accuracy and clinical effectiveness had elevated PSA and/or abnormal DRE results and were referred to targeted biopsy following a PI-RADS or Likert score of three or more on MRI. This is reflective of NICE guidance, which recommends that men should be referred for mpMRI if their PSA levels are above the age-specific reference range or if their prostate feels malignant on DRE. However, other organisations have recommended that PSA levels should be used as part of a risk prediction tool, potentially leading to better targeting of patients referred to mpMRI. It is unclear how a change in referral criteria may affect the applicability of this assessment’s findings. 13

Equality, diversity and inclusion

This study was carried out by a multidisciplinary team, consisting of an information specialist, a statistician, systematic reviewers and health economists. Junior members of staff contributed according to their skills and had the opportunity to undertake tasks that furthered their training. Clinical experts that provided advice included a nurse and senior consultants.

This report was prepared to support NICE guidance to help reduce health inequalities, improve access to health care and encourage health improvement (www.nice.org.uk/about/who-we-are/policies-and-procedures/nice-equality-scheme).

Patient and public involvement

This report was prepared to support NICE guidance. Lay people, and organisations representing their interests, have opportunities to contribute to developing NICE guidance, advice and quality standards, and support their implementation (www.nice.org.uk/about/nice-communities/nice-and-the-public/public-involvement/public-involvement-programme/patient-public-involvement-policy).

Suggested research priorities

High-quality, sufficiently powered RCT evidence, comparing SF and CF with or without systematic biopsy in trained operators, is needed to address the limitations of the diagnostic accuracy evidence identified in this study. Ideally, a trial should be sufficiently powered to detect long-term oncological outcomes (PFS, PCa mortality, overall mortality), although we acknowledge that such a study may not be feasible. IP7-PACIFIC (NCT05574647),155 a large UK-based randomised trial, will aim to determine whether SF biopsy is superior to CF at detecting CSPCa in patients with suspicious MRI in patients randomised to either mpMRI or bpMRI. It is hoped that this trial will provide more precise diagnostic accuracy estimates, although it is not clear which specific SF devices will be used, and the protocol indicates that estimates of diagnostic accuracy will not be informed by a gold standard test.

Full reporting of ISUP grades for each randomised arm is recommended, and for within-patient comparison studies, full reporting of cross-tabulation tables, where the classification of patients’ cancer by ISUP grade for each biopsy type is described and the relative accuracy of the interventions can be derived. In mixed population studies, reporting by key patient characteristics, such as PI-RADS score, whether biopsy naive or experienced, and route of referral for MRI (e.g. following clinical concerns, routine surveillance, screening etc.) are required to inform decision-making. Availability of more granular data, from already published studies, would enable future syntheses to make use of a larger body of evidence. Qualitative evidence on the acceptability of SF to patients, notably where biopsy procedure time might be significantly increased, is needed.

Contributions of authors

Alexis Llewellyn (https://orcid.org/0000-0003-4569-5136) (Systematic Reviewer) contributed to the protocol, performed the systematic review, wrote the background and most of the sections on clinical effectiveness.

Thai Han Phung (https://orcid.org/0000-0001-6193-4673) (Health Economist) assisted with the economic evidence reviews and with the development of the economic model, its analyses and validation, and contributed to the writing of the cost-effectiveness sections.

Marta O Soares (https://orcid.org/0000-0003-1579-8513) (Health Economist) developed the methods to extend the evidence synthesis for use in the economic model and conducted the required analyses over the extended synthesis model. She provided oversight and contributed to the development of the inference sub model to inform the economic model. She wrote the cost-effectiveness sections of the protocol and main report, and provided leadership support to the economic sections.

Lucy Shepherd (https://orcid.org/0000-0001-6803-5566) (Systematic Reviewer) contributed to the protocol, performed the systematic review and wrote parts of the clinical sections.

David Glynn (https://orcid.org/0000-0002-0989-1984) (Health Economist) reviewed the prostate cancer long-term outcome evidence and developed the inference sub model to inform the economic model.

Melissa Harden (https://orcid.org/0000-0003-2338-6869) (Information Specialist) designed and ran all searches for the study, managed the library of references, and wrote the search sections of the report.

Ruth Walker (https://orcid.org/0000-0003-2765-7363) (Systematic Reviewer) contributed to the protocol and background materials.

Ana Duarte (https://orcid.org/0000-0002-0528-4773) (Health Economist) developed the economic model, conducted analyses over this model and validated it. Ana had overall responsibility for the cost-effectiveness sections of the report. She wrote the cost-effectiveness sections of the protocol and of the report.

Sofia Dias (https://orcid.org/0000-0002-2172-0221) (Statistician) contributed to the protocol, adapted the network meta-analysis model and performed the meta-analyses, wrote up the meta-analysis section results, and oversaw the conduct and writing of the clinical effectiveness sections and the report as a whole. Sofia has overall responsibility for the clinical effectiveness sections of the report.

Acknowledgements

We would like to thank specialist DAC members Dr Oliver Hulson, Consultant Radiologist at Leeds Teaching Hospital Trust, Ms Lillian White, Clinical Nurse Specialist at NHS Ayshire and Arran, and Mr Hide Yamamoto, Consultant Urological Surgeon at Maidstone and Tunbridge Wells Trust for providing helpful advice during this study. We also thank Ms Lillian White for commenting on an early draft of the report. We would also like to thank Prof Hashim Ahmed for providing helpful advice during this study; Mr Connor Evans, Research Trainee at CRD, for helping to edit the final version of this report; and Prof Stephen Palmer for advice on the conceptualisation of the economic model.

We thank Dr Bhash Naidoo and the NICE Centre for Guidelines who developed the NICE NG131 health economics model and for sharing it. The review of the electronic version of the model contributed to inform the York-model conceptualisation. Similarly, we would like to thank Dr Joanne Lord and the Southampton TAR group who developed the Southampton model and for sharing it. The electronic version of the Southampton model was used to support both model conceptualisation and parameterisation of the York model.

Data-sharing statement

All data are provided in appendices to this report. Additional requests for access to data should be addressed to the corresponding author.

Ethics statement

Due to the nature of this study (systematic review of aggregate data and economic analysis), no ethical approval was required.

Information governance statement

This study did not handle any personal information.

Disclosure of interests

Full disclosure of interests: Completed ICMJE forms for all authors, including all related interests, are available in the toolkit on the NIHR Journals Library report publication page at https://doi.org/10.3310/PLFG4210.

Primary conflict of interest: Marta O Soares is a member of the NIHR Health Technology Assessment (HTA) General Committee.

Disclaimers

This article presents independent research funded by the National Institute for Health and Care Research (NIHR). The views and opinions expressed by authors in this publication are those of the authors and do not necessarily reflect those of the NHS, the NIHR, the HTA programme or the Department of Health and Social Care. If there are verbatim quotations included in this publication the views and opinions expressed by the interviewees are those of the interviewees and do not necessarily reflect those of the authors, those of the NHS, the NIHR, the HTA programme or the Department of Health and Social Care.

ARTEMIS (InnoMedicus ARTEMIS)

The ARTEMIS fusion biopsy system comprises a semi-robotic mechanical arm and a mobile workstation. The system includes the ProFuse radiology software for preparation of MRI data for fusion and reporting findings on the ARTEMIS biopsy system. The system allows both elastic and rigid estimation to account for prostate deformation, and supports both transrectal and transperineal biopsies. The mechanical arm is used to track the prostate in real time and guide the biopsy needle.

At the time of writing the EAG report, the company had not registered with NICE, and therefore did not provide information on this technology’s compatibility with a picture archiving and communication system (PACS), image measurement capabilities and ability to produce archivable cartograms.

BioJet (Healthcare Supply Solutions Ltd)

The BioJet MR Fusion system comprises MRI fusion software, a mobile workstation and is compatible with third-party ultrasounds. The system uses elastic estimations and is compatible with both transrectal and transperineal biopsies and supports both stabilised and freehand biopsy approaches.

The software enables image measurements and generates reports displaying the location of sampled areas. BioJet can be connected to a local PACS.

BiopSee (Medcom)

The BiopSee consists of the BiopSee software and the MedSta cart (workstation) and is compatible with third-party ultrasounds. The system supports both elastic and rigid estimation to account for prostate deformation, and allows both transrectal and transperineal biopsies. The system can be used for stabilised and freehand biopsy approaches. A stabilising arm is available for transperineal stabilised biopsies. Patient movement is tracked through the stepper during stabilised biopsies, or through a magnetic tracker, which is attached to the probe during freehand biopsies. The system can automatically adjust for patient movement, or the user can manually adjust the contours when a patient moves.

The BiopSee records all positions of the needle and shows the coverage of the prostate. Image measurements such as prostate and lesion volumes are also possible. The data are stored locally and can be connected to a PACS for import and export of images.

bkFusion (BK Medical UK Ltd and MIM Software Inc.)

BK Medical UK Ltd offers three versions of bkFusion software: one for transrectal, one for freehand transperineal and one for stabilised transperineal biopsies. The software can be integrated into either the bk3000 or bk5000 ultrasounds. The bkFusion system uses rigid estimation to account for prostate deformation. Predictive Fusion software re-orientates the MRI image before the biopsy. The transrectal and freehand transperineal fusion systems comprise a magnetic field generator and sensor to track the probe position.

Image measurements such as prostate volume are possible. A detailed report of the biopsy can be saved locally, or transferred to a PACS.

Fusion Bx 2.0 (Focal Healthcare)

The Fusion Bx 2.0 is a biopsy device that includes a counter balanced, semi-robotic arm that is mounted to a mobile cart. The Fusion Bx 2.0 comprises Fusion MR software which is compatible with third-party ultrasounds. The system uses both elastic and rigid estimation to account for prostate deformation, and supports both transrectal and transperineal biopsies. Patient movements are tracked with sensors inside the semi-robotic arm.

The software allows image measurements such as prostate volume and distances can be calculated. Data on the biopsied samples and the regions of interest are recorded on a 3D image of the prostate. The system can connect to PACS using a wired Ethernet or Wi-Fi connection.

FusionVu (Exact Imaging)

The ExactVu device includes a micro-ultrasound (high-resolution ultrasound at > 20MHz) and a FusionVu feature that enables SF biopsy. A stabiliser arm or stepper is available for stabilised biopsies, and freehand biopsies are also possible. The system uses rigid estimation followed by real-time visualisation of the lesions using micro-ultrasound, and supports both transperineal and transrectal biopsies. The system tracks and adjusts for patient movement using data from a movement sensor together with the live ultrasound images.

The software provides image measurements such as prostate volume and lesion size. Information on the orientation of all images and video frames are recorded so that the same position can be found if a repeat biopsy is performed. The system is PACS compatible, but a separate software (Weasis DICOM viewer) is available in the case that a PACS is not available.

iSR’obot Mona Lisa (Biobot iSR’obot)

The iSR’obot Mona Lisa is a robotic transperineal prostate biopsy system with MRI-ultrasound fusion capability. The system uses UroFusion software to highlight regions of interest on MR images and fuses the MRI model with the ultrasound model. The robotic needle guide allows automated positioning and depth control of the biopsy needle to the targeted biopsy core. The system uses elastic estimation to account for prostate deformation.

Reports are generated with 3D-images and co-ordinates are recorded of each biopsy sample. At the time of writing the EAG report, the company had not registered with NICE, and therefore did not provide information on the tracking of patient movement, whether freehand biopsies can be done, PACS compatibility and image measurement capabilities of this system.

KOELIS Trinity (KOELIS and Kebomed)

The KOELIS Trinity is a mobile ultrasound system with mapping fusion software, which comprises PROMAP 3D-Prostate Suite software and the TRINITY ultrasound system (workstation, RECFIRE ultrasound probes, guides specific to transperineal or transrectal biopsies and a Steady Pro probe holder). The system uses elastic estimation to account for prostate deformation, and supports both transrectal and transperineal biopsies. It enables both stabilised and freehand probe biopsies. The Organ-Based Tracking Fusion software identifies and compensates for patient movements and prostate deformations to record each core location.

The PROMAP software produces a 3D map of the prostate recording the position of MRI lesion targets and location of biopsy samples. The KOELIS Trinity provides image measurements such as prostate volume, exact measurements of the regions of interest and other quantitative measurements of the image. Data can be transferred to a PACS.

UroNav Fusion Biopsy System (Phillips)

The UroNav Fusion Biopsy System includes an electromagnetic tracking system, a mobile workstation and DynaCAD Prostate fusion software. The system is compatible with third-party ultrasounds. It supports both transperineal and transrectal biopsies, with stabilised or freehand approaches. UroNav uses both rigid and elastic registration methods to create and maintain 3D registration of MR/US images and compensate for patient movement. The system can be used with the UroNav mobile stepper system and the two navigation sensors to track patient movement.

The UroNav Fusion Biopsy system provides the core location data, images and videos. At the time of writing the EAG report, the company had not registered with NICE, and therefore did not submit any information on image estimation methods for prostate deformation, patient movement tracking feasibility for freehand biopsies, PACS compatibility and image measurement capabilities of this system.

Database search strategies

MEDLINE ALL

(includes: Epub Ahead of Print, In-process and Other Non-Indexed Citations, Ovid MEDLINE Daily and Ovid MEDLINE)

Via Ovid http://ovidsp.ovid.com/

Date range: 1946 to 13 May, 2022

Date searched: 16 May 2022

Records retrieved: 3129

MEDLINE ALL was searched again on 2 August 2022. 3218 studies were retrieved.

1. exp Prostatic Neoplasms/ (142378)

2. Prostatic Intraepithelial Neoplasia/ (1399)

3. ((prostate$ or prostatic or intraprostatic) adj4 (cancer$ or neoplas$ or tumour$ or tumor$ or malignan$ or metasta$ or carcinoma$ or adenocarcinoma$ or lesion$ or nodul$ or sarcoma$ or lymphoma$)).ti,ab. (165600)

4. (PCa or sPCa or csPCa or PrCa).ti,ab. (52571)

5. (((atypical adj3 proliferation) or ASAP) and prostat$).mp. (292).

6. or/1-5 (224791)

7. Magnetic Resonance Imaging/ (453356)

8. Multiparametric Magnetic Resonance Imaging/ (961)

9. (magnetic resonance or MRI or MR imag$ or MR scan$).ti,ab. (560471)

10. (mpMRI or mp-MRI or mpMR imag$ or mpMR scan$ or mp-MR imag$ or mp-MR scan$ or bpMRI or bp-MRI or bpMR imag$ or bpMR scan$ or bp-MR imag$ or bp-MR scan$).ti,ab. (2060)

11. or/7-10 (721668)

12. Image Interpretation, Computer-Assisted/ (47627)

13. (fusion$ or fuse$ or fusing$).ti,ab. (299284)

14. cognitive$.ti,ab. (424900)

15. (visual$ adj3 (estimat$ or direct$ or align$ or guid$ or influenc$)).ti,ab. (28436)

16. registration$.ti,ab. (161125)

17. (elastic or rigid or nonrigid).ti,ab. (138219)

18. Software/ (120348)

19. (software or hardware).ti,ab. (224399)

20. or/12-19 (1355053)

21. Prostate/ (39209)

22. (prostate$ or prostatic).ti,ab. (234214)

23. 21 or 22 (238231)

24. Biopsy/ (185156)

25. Image-Guided Biopsy/ (5020)

26. Endoscopic Ultrasound-Guided Fine Needle Aspiration/ (3254)

27. Biopsy, Fine-Needle/ (14970)

28. Biopsy, Large-Core Needle/ (2307)

29. Biopsy, Needle/ (49647)

30. (biopsy or biopsie$ or rebiopsy or rebiopsie$).ti,ab. (427177)

31. or/24-30 (548867)

32. 23 and 31 (26179)

33. 6 and 11 and 20 and 32 (1621)

34. ((MRI or MR or magnetic resonance or mpMRI or mp-MRI or bpMRI or bp-MRI) adj6 (prior or previous$ or preced$ or before$ or earlier or first or initial$) adj6 (biopsy or biopsie$)).ti,ab. (860)

35. ((MRI or MR or magnetic resonance or mpMRI or mp-MRI or bpMRI or bp-MRI) adj6 prebiops$)).ti,ab. (160)

36. ((MRI or MR or magnetic resonance or mpMRI or mp-MRI or bpMRI or bp-MRI) adj6 (prior or previous$ or preced$ or before$ or earlier or first or initial$) adj6 (ultrasound$ or ultrasonic$ or ultrasonograph$ or TRUS or transperineal$ or transrectal$)).ti,ab. (773)

37. or/34-36 (1626)

38. 6 and 37 (662)

39. (target$ adj4 (biopsy or biopsie$ or rebiopsy or rebiopsie$)).ti,ab. (3800)

40. (focal adj2 (biopsy or biopsie$ or rebiopsy or rebiopsie$)).ti,ab. (636)

41. 39 or 40 (4405)

42. 6 and 41 (1842)

43. (target$ adj4 (MRI or MR or magnetic resonance or mpMRI or mp-MRI or bpMRI or bp-MRI)).ti,ab. (4003)

44. (focal adj2 (MRI or MR or magnetic resonance or mpMRI or mp-MRI or bpMRI or bp-MRI)).ti,ab. (546)

45. 43 or 44 (4534)

46. 6 and 32 and 45 (1125)

47. ((MRI or MR or magnetic resonance or mpMRI or mp-MRI or bpMRI or bp-MRI) adj3 (guid$ or influenc$ or direct$ or align$)).ti,ab. (11942)

48. 6 and 32 and 47 (951)

49. ((MRI stratified or magnetic resonance imaging stratified) adj3 pathway$)).ti,ab. (3)

50. 33 or 38 or 42 or 46 or 48 or 49 (3265)

51. (MRGB or MR-GB or MRIGB or MRI-GB).ti,ab. (75)

52. (MRIFB or MRI-FB).ti,ab. (3)

53. (MRFTB or MRF-TB).ti,ab. (9)

54. (MRTB or MR-TB or MRITB or MRI-TB or MRTBx or MR-TBx or MRITBx or MRI-TBx).ti,ab. (96)

55. FBx.ti,ab. (94)

56. (FUSTB or FUS-TB or TB-FUS).ti,ab. (9)

57. Fusion TB.ti,ab. (21)

58. (MRI-TRUS or MRI-TRUSB or MRI-TPB).ti,ab. (189)

59. (COG-TB or TB-COG or CBx).ti,ab. (530)

60. TRUS-TB.ti,ab. (3)

61. (‘MRI/TRUS’ or ‘mpMRI/TRUS’ or ‘MR/US’ or ‘MRI/TRUS-TB’).ti,ab. (306)

62. or/51-61 (1105)

63. 6 and 62 (437)

64. 50 or 63 (3292)

65. (fusion$ adj3 (software or hardware or computer$ or device$ or system$ or technolog$ or machine$ or platform$)).ti,ab. (5680)

66. 6 and 65 (294)

67. 64 or 66 (3331)

68. KOELIS.ti,ab. (23)

69. Fusion Bx.ti,ab. (1)

70. BioJet.ti,ab. (28)

71. (Trinity or PROMAP).ti,ab. (1329)

72. Fusion MR.ti,ab. (8)

73. (bkFusion or bk Fusion or BK3000 or BK 3000 or BK5000 or BK 5000 or Predictive Fusion).ti,ab. (7)

74. or/70-73 (1371)

75. 6 and 74 (20)

76. 68 or 69 or 75 (38)

77. BiopSee .ti,ab. (6)

78. UroNav.ti,ab. (17)

79. (‘iSR’obot’ or iSRobot or iSR obot or UroFusion or UroBiopsy).ti,ab. (2)

80. (FusionVu$ or ExactVu$).ti,ab. (12)

81. DynaCAD.ti,ab. (9)

82. (ARTEMIS or ProFuse).ti,ab. (4760)

83. Mona Lisa.ti,ab. (106)

84. or/81-83 (4874)

85. 6 and 84 (54)

86. or/77-80 (34)

87. 85 or 86 (81)

88. 67 or 76 or 87 (3362)

89. exp animals/not humans.sh. (5007245)

90. 88 not 89 (3357)

91. limit 90 to yr=‘2008 -Current’ (3129)

Key

1. / = subject heading (MeSH heading)

2. sh = subject heading (MeSH heading)

3. exp = exploded subject heading (MeSH heading)

4. $ = truncation

5. ti,ab = terms in title or abstract fields

6. mp = multi-purpose field search – terms in title, original title, abstract, name of substance word, or subject heading word

7. adj3 = terms within three words of each other (any order)

Cochrane Controlled Register of Trials (CENTRAL)

Via Wiley http://onlinelibrary.wiley.com/

Issue: Issue 4 of 12, April 2022

Date searched: 16 May 2022

Records retrieved: 425

CENTRAL was searched again on 2 August 2022. 434 studies were retrieved.

1. MeSH descriptor: [Prostatic Neoplasms] explode all trees 6115

2. MeSH descriptor: [Prostatic Intraepithelial Neoplasia] this term only 47

3. ((prostate* or prostatic or intraprostatic) near/4 (cancer* or neoplas* or tumour* or tumor* or malignan* or metasta* or carcinoma* or adenocarcinoma* or lesion* or nodul* or sarcoma* or lymphoma*)):ti,ab,kw 15719

4. (PCa or sPCa or csPCa or PrCa):ti,ab,kw 5554

5. ((atypical near/3 proliferation) or ASAP) and prostat*)):ti,ab,kw 21

6. #1 or #2 or #3 or #4 or #5 20099

7. MeSH descriptor: [Magnetic Resonance Imaging] this term only 7831

8. MeSH descriptor: [Multiparametric Magnetic Resonance Imaging] this term only 11

9. (‘magnetic resonance’ or MRI or (MR next imag*) or (MR next scan*)):ti,ab,kw 41256

10. (mpMRI or mp-MRI or (mpMR next imag*) or (mpMR next scan*) or (mp-MR next imag*) or mp-MR scan* or bpMRI or bp-MRI or (bpMR next imag*) or (bpMR next scan*) or bp-MR imag* or bp-MR scan*):ti,ab,kw 260

11. #7 or #8 or #9 or #10 41264

12. MeSH descriptor: [Image Interpretation, Computer-Assisted] this term only 875

13. (fusion* or fuse* or fusing*):ti,ab,kw 8635

14. cognitive*:ti,ab,kw 80126

15. (visual* near/3 (estimat* or direct* or align* or guid* or influenc*)):ti,ab,kw 2089

16. registration*:ti,ab,kw 66768

17. (elastic or rigid or nonrigid):ti,ab,kw 6102

18. MeSH descriptor: [Software] this term only 1008

19. (software or hardware):ti,ab,kw 26282

20. #12 or #13 or #14 or #15 or #16 or #17 or #18 or #19 180581

21. MeSH descriptor: [Prostate] this term only 975

22. (prostate* or prostatic):ti,ab,kw 23298

23. #21 or #22 23298

24. MeSH descriptor: [Biopsy] this term only 3365

25. MeSH descriptor: [Image-Guided Biopsy] this term only 119

26. MeSH descriptor: [Endoscopic Ultrasound-Guided Fine Needle Aspiration] this term only 156

27. MeSH descriptor: [Biopsy, Needle] explode all trees 1270

28. (biopsy or biopsie* or rebiopsy or rebiopsie*):ti,ab,kw 32970

29. #24 or #25 or #26 or #27 or #28 33007

30. #23 and #29 2832

31. #6 and #11 and #20 and #30 211

32. ((MRI or MR or ‘magnetic resonance’ or mpMRI or mp-MRI or bpMRI or bp-MRI) near/6 (prior or previous* or preced* or before* or earlier or first or initial*) near/6 (biopsy or biopsie*)):ti,ab,kw 95

33. ((MRI or MR or ‘magnetic resonance’ or mpMRI or mp-MRI or bpMRI or bp-MRI) near/6 prebiops*):ti,ab,kw 34

34. ((MRI or MR or ‘magnetic resonance’ or mpMRI or mp-MRI or bpMRI or bp-MRI) near/6 (prior or previous* or preced* or before* or earlier or first or initial*) near/6 (ultrasound* or ultrasonic* or ultrasonograph* or TRUS or transperineal* or transrectal*)):ti,ab,kw 84

35. (target* near/4 (biopsy or biopsie* or rebiopsy or rebiopsie*)):ti,ab,kw 573

36. (focal near/2 (biopsy or biopsie* or rebiopsy or rebiopsie*)):ti,ab,kw 22

37. #32 or #33 or #34 or #35 or #36 715

38. #6 and #37 324

39. (target* near/4 (MRI or MR or ‘magnetic resonance’ or mpMRI or mp-MRI or bpMRI or bp-MRI)):ti,ab,kw 453

40. (focal near/2 (MRI or MR or ‘magnetic resonance’ or mpMRI or mp-MRI or bpMRI or bp-MRI)):ti,ab,kw 38

41. ((MRI or MR or ‘magnetic resonance’ or mpMRI or mp-MRI or bpMRI or bp-MRI) near/3 (guid* or influenc* or direct* or align*)):ti,ab,kw 923

42. #39 or #40 or #41 1299

43. #6 and #30 and #42 279

44. ((‘MRI stratified’ or ‘magnetic resonance imaging stratified’) near/3 pathway*):ti,ab,kw 0

45. #31 or #38 or #43 or #44 430

46. (MRGB or MR-GB or MRIGB or MRI-GB or MRIFB or MRI-FB or MRFTB or MRF-TB or MRFTB or MRF-TB or MRTB or MR-TB or MRITB or MRI-TB or MRTBx or MR-TBx or MRITBx or MRI-TBx or FBx or FUSTB or FUS-TB or TB-FUS or ‘Fusion TB’ or MRI-TRUS or MRI-TRUSB or MRI-TPB or COG-TB or TB-COG or CBx or TRUS-TB or ‘MRI/TRUS’ or ‘mpMRI/TRUS’ or ‘MR/US’ or ‘MRI/TRUS-TB’):ti,ab,kw 136

47. #6 and #46 82

48. #45 or #47 431

49. (fusion* near/3 (software or hardware or computer* or device* or system* or technolog* or machine* or platform*)):ti,ab,kw 267

50. #6 and #49 57

51. #48 or #50 434

52. (KOELIS or ‘Fusion Bx’ or BioJet):ti,ab,kw 18

53. (Trinity or PROMAP or ‘Fusion MR’ or bkFusion or ‘bk Fusion’ or BK3000 or ‘BK 3000’ or BK5000 or ‘BK 5000’ or ‘Predictive Fusion’):ti,ab,kw 161

54. #6 and #53 3

55. #52 or #54 19

56. (BiopSee or UroNav or ‘iSR’obot’ or iSRobot or ‘iSR obot’ or UroFusion or UroBiopsy or FusionVu* or ExactVu*):ti,ab,kw 19

57. (DynaCAD or ARTEMIS or ProFuse or ‘Mona Lisa’):ti,ab,kw 283

58. #6 and #57 9

59. #56 or #58 27

60. #51 or #55 or #59 with Publication Year from 2008 to 2022, in Trials 425

61. #51 or #55 or #59 in Cochrane Reviews, Cochrane Protocols 1

Key

1. MeSH descriptor = subject heading (MeSH heading)

2. * = truncation

3. ti,ab,kw = terms in title, abstract or keyword fields

4. near/3 = terms within three words of each other (any order)

5. next = terms are next to each other

Cochrane Database of Systematic Reviews (CDSR)

Via Wiley http://onlinelibrary.wiley.com/

Issue: Issue 5 of 12, May 2022

Date searched: 16 May 2022

Records retrieved: 1

See above under CENTRAL for search strategy.

Cumulative Index to Nursing and Allied Health (CINAHL Plus)

Via Ebsco http://onlinelibrary.wiley.com/

Date range: Inception to 20220516

Date searched: 16 May 2022

Records retrieved: 916

1. (MH ‘Prostatic Neoplasms+’) 34,206

2. TI ((prostate* or prostatic or intraprostatic) N4 (cancer* or neoplas* or tumour* or tumor* or malignan* or metasta* or carcinoma* or adenocarcinoma* or lesion* or nodul* or sarcoma* or lymphoma*)) OR AB ((prostate* or prostatic or intraprostatic) N4 (cancer* or neoplas* or tumour* or tumor* or malignan* or metasta* or carcinoma* or adenocarcinoma* or lesion* or nodul* or sarcoma* or lymphoma*)) 35,654

3. TI ((PCa or sPCa or csPCa or PrCa)) OR AB ((PCa or sPCa or csPCa or PrCa)) 7320

4. TI (((atypical N3 proliferation) or ASAP) and prostat*) OR AB (((atypical N3 proliferation) or ASAP) and prostat*) 29

5. S1 OR S2 OR S3 OR S4 48,489

6. (MH ‘Magnetic Resonance Imaging’) 136,332

7. TI ((‘magnetic resonance’ or MRI or (MR N1 imag*) or (MR N1 scan*))) OR AB ((‘magnetic resonance’ or MRI or (MR N1 imag*) or (MR N1 scan*))) 123,908

8. TI (((mpMRI or mp-MRI or (mpMR N1 imag*) or (mpMR N1 scan*) or (mp-MR N1 imag*) or (mp-MR N1 scan*) or bpMRI or bp-MRI or (bpMR N1 imag*) or (bpMR N1 scan*) or (bp-MR N1 imag*) or (bp-MR N1 scan*))) OR AB (((mpMRI or mp-MRI or (mpMR N1 imag*) or (mpMR N1 scan*) or (mp-MR N1 imag*) or (mp-MR N1 scan*) or bpMRI or bp-MRI or (bpMR N1 imag*) or (bpMR N1 scan*) or (bp-MR N1 imag*) or (bp-MR N1 scan*))) 631

9. S6 OR S7 OR S8 181,020

10. (MH ‘Image Interpretation, Computer Assisted’) 9454

11. TI (fusion* or fuse* or fusing*) OR AB (fusion* or fuse* or fusing*) 26,160

12. TI cognitive* OR AB cognitive* 154,740

13. TI (visual* N3 (estimat* or direct* or align* or guid* or influenc*)) OR AB (visual* N3 (estimat* or direct* or align* or guid* or influenc*)) 4578

14. TI registration* OR AB registration* 64,987

15. TI (elastic or rigid or nonrigid) OR AB (elastic or rigid or nonrigid) 12,473

16. (MH ‘Software’) 31,273

17. TI (software or hardware) OR AB (software or hardware) 59,300

18. S10 OR S11 OR S12 OR S13 OR S14 OR S15 OR S16 OR S17 341,260

19. (MH ‘Prostate’) 3816

20. TI (prostate* or prostatic) OR AB (prostate* or prostatic) 45,719

21. S19 OR S20 46,101

22. (MH ‘Biopsy’) 35,975

23. (MH ‘Biopsy, Needle’) 11,989

24. TI (biopsy or biopsie* or rebiopsy or rebiopsie*) OR AB (biopsy or biopsie* or rebiopsy or rebiopsie*) 59,743

25. S22 OR S23 OR S24 84,744

26. S21 AND S25 4603

27. S5 AND S9 AND S18 AND S26 463

28. TI ((MRI or MR or ‘magnetic resonance’ or mpMRI or mp-MRI or bpMRI or bp-MRI) N6 (prior or previous* or preced* or before* or earlier or first or initial*) N6 (biopsy or biopsie*)) OR AB ((MRI or MR or ‘magnetic resonance’ or mpMRI or mp-MRI or bpMRI or bp-MRI) N6 (prior or previous* or preced* or before* or earlier or first or initial*) N6 (biopsy or biopsie*)) 254

29. TI ((MRI or MR or ‘magnetic resonance’ or mpMRI or mp-MRI or bpMRI or bp-MRI) N6 prebiops*) OR AB ((MRI or MR or ‘magnetic resonance’ or mpMRI or mp-MRI or bpMRI or bp-MRI) N6 prebiops*) 45

30. TI ((MRI or MR or ‘magnetic resonance’ or mpMRI or mp-MRI or bpMRI or bp-MRI) N6 (prior or previous* or preced* or before* or earlier or first or initial*) N6 (ultrasound* or ultrasonic* or ultrasonograph* or TRUS or transperineal* or transrectal*)) OR AB ((MRI or MR or ‘magnetic resonance’ or mpMRI or mp-MRI or bpMRI or bp-MRI) N6 (prior or previous* or preced* or before* or earlier or first or initial*) N6 (ultrasound* or ultrasonic* or ultrasonograph* or TRUS or transperineal* or transrectal*)) 289

31. TI (target* N4 (biopsy or biopsie* or rebiopsy or rebiopsie*)) OR AB (target* N4 (biopsy or biopsie* or rebiopsy or rebiopsie*)) 961

32. TI (focal N2 (biopsy or biopsie* or rebiopsy or rebiopsie*)) OR AB (focal N2 (biopsy or biopsie* or rebiopsy or rebiopsie*)) 136

33. S28 OR S29 OR S30 OR S31 OR S32 1512

34. S5 AND S33 591

35. TI (target* N4 (MRI or MR or magnetic resonance or mpMRI or mp-MRI or bpMRI or bp-MRI)) OR AB (target* N4 (MRI or MR or magnetic resonance or mpMRI or mp-MRI or bpMRI or bp-MRI)) 880

36. TI (focal N2 (MRI or MR or magnetic resonance or mpMRI or mp-MRI or bpMRI or bp-MRI)) OR AB (focal N2 (MRI or MR or magnetic resonance or mpMRI or mp-MRI or bpMRI or bp-MRI)) 257

37. TI ((MRI or MR or ‘magnetic resonance’ or mpMRI or mp-MRI or bpMRI or bp-MRI) N3 (guid* or influenc* or direct* or align*)) OR AB ((MRI or MR or ‘magnetic resonance’ or mpMRI or mp-MRI or bpMRI or bp-MRI) N3 (guid* or influenc* or direct* or align*)) 3420

38. S35 OR S36 OR S37 4296

39. S5 AND S26 AND S38 533

40. TI ((‘MRI stratified’ or ‘magnetic resonance imaging stratified’) N3 pathway*) OR AB ((‘MRI stratified’ or ‘magnetic resonance imaging stratified’) N3 pathway*) 2

41. S27 OR S34 OR S39 OR S40 909

42. TI (MRGB or MR-GB or MRIGB or MRI-GB or MRIFB or MRI-FB or MRFTB or MRF-TB or MRFTB or MRF-TB or MRTB or MR-TB or MRITB or MRI-TB or MRTBx or MR-TBx or MRITBx or MRI-TBx or FBx or FUSTB or FUS-TB or TB-FUS or ‘Fusion TB’ or MRI-TRUS or MRI-TRUSB or MRI-TPB or COG-TB or TB-COG or CBx or TRUS-TB or ‘MRI/TRUS’ or ‘mpMRI/TRUS’ or ‘MR/US’ or ‘MRI/TRUS-TB’) OR AB (MRGB or MR-GB or MRIGB or MRI-GB or MRIFB or MRI-FB or MRFTB or MRF-TB or MRFTB or MRF-TB or MRTB or MR-TB or MRITB or MRI-TB or MRTBx or MR-TBx or MRITBx or MRI-TBx or FBx or FUSTB or FUS-TB or TB-FUS or ‘Fusion TB’ or MRI-TRUS or MRI-TRUSB or MRI-TPB or COG-TB or TB-COG or CBx or TRUS-TB or ‘MRI/TRUS’ or ‘mpMRI/TRUS’ or ‘MR/US’ or ‘MRI/TRUS-TB’) 185

43. S5 AND S42 126

44. S41 OR S43 915

45. TI (fusion* N3 (software or hardware or computer* or device* or system* or technolog* or machine* or platform*)) OR AB (fusion* N3 (software or hardware or computer* or device* or system* or technolog* or machine* or platform*)) 832

46. S5 AND S45 86

47. S44 OR S46 922

48. TI (KOELIS or ‘Fusion Bx’ or BioJet) OR AB (KOELIS or ‘Fusion Bx’ or BioJet) 17

49. TI (Trinity or PROMAP or ‘Fusion MR’ or bkFusion or ‘bk Fusion’ or BK3000 or ‘BK 3000’ or BK5000 or ‘BK 5000’ or ‘Predictive Fusion’) OR AB (Trinity or PROMAP or ‘Fusion MR’ or bkFusion or ‘bk Fusion’ or BK3000 or ‘BK 3000’ or BK5000 or ‘BK 5000’ or ‘Predictive Fusion’) 482

50. S5 AND S49 2

51. S48 OR S50 18

52. TI (BiopSee or UroNav or ‘iSR’obot’ or iSRobot or ‘iSR obot’ or UroFusion or UroBiopsy or FusionVu* or ExactVu*) OR AB (BiopSee or UroNav or ‘iSR’obot’ or iSRobot or ‘iSR obot’ or UroFusion or UroBiopsy or FusionVu* or ExactVu*) 11

53. TI (DynaCAD or ARTEMIS or ProFuse or ‘Mona Lisa’) AND AB (DynaCAD or ARTEMIS or ProFuse or ‘Mona Lisa’) 32

54. S5 AND S53 0

55. S52 OR S54 11

56. S47 OR S51 OR S55 925

57. S47 OR S51 OR S55 Limiters – Published Date: 20080101-20221231 916

Key

1. MH = CINAHL subject heading

2. + = exploded CINAHL subject heading

3. * = truncation

4. TI = terms in the title

5. AB = terms in the abstract

6. N3 = terms within three words of each other (any order)

Database of Abstracts of Reviews of Effects (DARE)

Via http://onlinelibrary.wiley.com/

Date range: Inception – 31 March 2015

Date searched: 16 May 2022

Records retrieved: 7

1. MeSH DESCRIPTOR Prostatic Neoplasms EXPLODE ALL TREES 709

2. MeSH DESCRIPTOR Prostatic Intraepithelial Neoplasia 2

3. ((prostate* or prostatic or intraprostatic) NEAR4 (cancer* or neoplas* or tumour* or tumor* or malignan* or metasta* or carcinoma* or adenocarcinoma* or lesion* or nodul* or sarcoma* or lymphoma*)) 891

4. (PCa or sPCa or csPCa or PrCa) 44

5. ((atypical NEAR3 proliferation) or ASAP) AND (prostat*) 1

6. #1 OR #2 OR #3 OR #4 OR #5 935

7. MeSH DESCRIPTOR Magnetic Resonance Imaging 693

8. MeSH DESCRIPTOR Multiparametric Magnetic Resonance Imaging 0

9. (‘magnetic resonance’ or MRI or MR imag* or MR scan*) 1337

10. (mpMRI or mp-MRI or mpMR imag* or mpMR scan* or mp-MR imag* or mp-MR scan* or bpMRI or bp-MRI or bpMR imag* or bpMR scan* or bp-MR imag* or bp-MR scan*) 2

11. #7 OR #8 OR #9 OR #10 1337

12. MeSH DESCRIPTOR Image Interpretation, Computer-Assisted 27

13. (fusion* or fuse* or fusing* or cognitive or registration* or elastic or rigid or nonrigid) 3376

14. (visual* NEAR3 (estimat* or direct* or align* or guid* or influenc*)) 23

15. MeSH DESCRIPTOR Software 76

16. (software or hardware) 812

17. #12 OR #13 OR #14 OR #15 OR #16 4163

18. MeSH DESCRIPTOR Prostate 82

19. (prostate* or prostatic) 1283

20. #18 OR #19 1283

21. MeSH DESCRIPTOR Biopsy 248

22. MeSH DESCRIPTOR Image-Guided Biopsy 11

23. MeSH DESCRIPTOR Endoscopic Ultrasound-Guided Fine Needle Aspiration 19

24. MeSH DESCRIPTOR Biopsy, Fine-Needle 83

25. MeSH DESCRIPTOR Biopsy, Large-Core Needle 8

26. MeSH DESCRIPTOR Biopsy, Needle 164

27. (biopsy or biopsie* or rebiopsy or rebiopsie*) 1457

28. #21 OR #22 OR #23 OR #24 OR #25 OR #26 OR #27 1473

29. #20 AND #28 137

30. #6 AND #11 AND #17 AND #29 4

31. ((MRI or MR or ‘magnetic resonance’ or mpMRI or mp-MRI or bpMRI or bp-MRI) NEAR6 (biopsy or biopsie*)) 39

32. ((MRI or MR or ‘magnetic resonance’ or mpMRI or mp-MRI or bpMRI or bp-MRI) NEAR6 prebiops*) 0

33. ((MRI or MR or ‘magnetic resonance’ or mpMRI or mp-MRI or bpMRI or bp-MRI) NEAR6 (ultrasound* or ultrasonic* or ultrasonograph* or TRUS or transperineal* or transrectal*)) 121

34. (target* NEAR4 (biopsy or biopsie* or rebiopsy or rebiopsie*)) 10

35. (focal NEAR2 (biopsy or biopsie* or rebiopsy or rebiopsie*)) 0

36. #31 OR #32 OR #33 OR #34 OR #35 155

37. #6 AND #36 10

38. (target* NEAR4 (MRI or MR or ‘magnetic resonance’ or mpMRI or mp-MRI or bpMRI or bp-MRI)) 3

39. (focal* NEAR2 (MRI or MR or ‘magnetic resonance’ or mpMRI or mp-MRI or bpMRI or bp-MRI)) 1

40. ((MRI or MR or ‘magnetic resonance’ or mpMRI or mp-MRI or bpMRI or bp-MRI) NEAR3 (guid* or influenc* or direct* or align*)) 65

41. #38 OR #39 OR #40 67

42. #6 AND #29 AND #41 5

43. ((‘MRI stratified’ or ‘magnetic resonance imaging stratified’) NEAR3 pathway*) 0

44. (MRGB or MR-GB or MRIGB or MRI-GB or MRIFB or MRI-FB or MRFTB or MRF-TB or MRFTB or MRF-TB or MRTB or MR-TB or MRITB or MRI-TB or MRTBx or MR-TBx or MRITBx or MRI-TBx or FBx or FUSTB or FUS-TB or TB-FUS or ‘Fusion TB’ or MRI-TRUS or MRI-TRUSB or MRI-TPB or COG-TB or TB-COG or CBx or TRUS-TB or ‘MRI/TRUS’ or ‘mpMRI/TRUS’ or ‘MR/US’ or ‘MRI/TRUS-TB’) 1

45. (fusion* NEAR3 (software or hardware or computer* or device* or system* or technolog* or machine* or platform*)) 34

46. #44 OR #45 35

47. #6 AND #46 2

48. #30 OR #37 OR #42 OR #43 OR #47 11

49. (Trinity or PROMAP or ‘Fusion MR’ or bkFusion or ‘bk Fusion’ or BK3000 or ‘BK 3000’ or BK5000 or ‘BK 5000’ or ‘Predictive Fusion’) 2

50. (DynaCAD or ARTEMIS or ProFuse or ‘Mona Lisa’) 8

51. #49 OR #50 10

52. #6 AND #51 0

53. (KOELIS or ‘Fusion Bx’ or BioJet or BiopSee or UroNav or ‘iSR’obot’ or iSRobot or ‘iSR obot’ or UroFusion or UroBiopsy or FusionVu* or ExactVu*) 0

54. #48 OR #52 OR #53 11

Key

1. MeSH DESCRIPTOR = subject heading (MeSH heading)

2. * = truncation

3. NEAR3 = terms within three words of each other (order specified)

EconLit

Via Ovid http://ovidsp.ovid.com/

Date range: 1886 to 5 May, 2022

Date searched: 16 May 2022

Records retrieved: 0

1. ((prostate$ or prostatic or intraprostatic) adj4 (cancer$ or neoplas$ or tumour$ or tumor$ or malignan$ or metasta$ or carcinoma$ or adenocarcinoma$ or lesion$ or nodul$ or sarcoma$ or lymphoma$)).mp. (114)

2. (PCa or sPCa or csPCa or PrCa).mp. (541)

3. (((atypical adj3 proliferation) or ASAP) and prostat$).mp. (0)

4. or/1-3 (651)

5. (magnetic resonance or MRI or MR imag$ or MR scan$).mp. (188)

6. (mpMRI or mp-MRI or mpMR imag$ or mpMR scan$ or mp-MR imag$ or mp-MR scan$ or bpMRI or bp-MRI or bpMR imag$ or bpMR scan$ or bp-MR imag$ or bp-MR scan$).mp. (0)

7. 5 or 6 (188)

8. (fusion$ or fuse$ or fusing$).mp. (643)

9. cognitive$.mp. (17030)

10. (visual$ adj3 (estimat$ or direct$ or align$ or guid$ or influenc$)).mp. (75)

11. registration$.mp. (1925)

12. (elastic or rigid or nonrigid).mp. (4352)

13. (software or hardware).mp. (15832)

14. or/8-13 (39541)

15. (prostate$ or prostatic).mp. (141)

16. (biopsy or biopsie$ or rebiopsy or rebiopsie$).mp. (17)

17. 15 and 16 (4)

18. 4 and 7 and 14 and 17 (0)

19. ((MRI or MR or magnetic resonance or mpMRI or mp-MRI or bpMRI or bp-MRI) adj6 (prior or previous$ or preced$ or before$ or earlier or first or initial$) adj6 (biopsy or biopsie$)).mp. (0)

20. ((MRI or MR or magnetic resonance or mpMRI or mp-MRI or bpMRI or bp-MRI) adj6 prebiops$).mp. (0)

21. ((MRI or MR or magnetic resonance or mpMRI or mp-MRI or bpMRI or bp-MRI) adj6 (prior or previous$ or preced$ or before$ or earlier or first or initial$) adj6 (ultrasound$ or ultrasonic$ or ultrasonograph$ or TRUS or transperineal$ or transrectal$)).mp. (0)

22. (target$ adj4 (biopsy or biopsie$ or rebiopsy or rebiopsie$)).mp. (0)

23. (focal adj2 (biopsy or biopsie$ or rebiopsy or rebiopsie$)).mp. (0)

24. (target$ adj4 (MRI or MR or magnetic resonance or mpMRI or mp-MRI or bpMRI or bp-MRI)).mp. (2)

25. 4 and 24 (0)

26. (focal adj2 (MRI or MR or magnetic resonance or mpMRI or mp-MRI or bpMRI or bp-MRI)).mp. (0)

27. ((MRI or MR or magnetic resonance or mpMRI or mp-MRI or bpMRI or bp-MRI) adj3 (guid$ or influenc$ or direct$ or align$)).mp. (9)

28. 4 and 27 (0)

29. ((MRI stratified or magnetic resonance imaging stratified) adj3 pathway$).mp. (0)

30. (MRGB or MR-GB or MRIGB or MRI-GB).mp. (0)

31. (MRIFB or MRI-FB).mp. (0)

32. (MRFTB or MRF-TB).mp. (0)

33. (MRFTB or MRF-TB).mp. (0)

34. (MRTB or MR-TB or MRITB or MRI-TB or MRTBx or MR-TBx or MRITBx or MRI-TBx).mp. (0)

35. FBx.mp. (0)

36. (FUSTB or FUS-TB or TB-FUS).mp. (0)

37. Fusion TB.mp. (0)

38. (MRI-TRUS or MRI-TRUSB or MRI-TPB).mp. (0)

39. (COG-TB or TB-COG or CBx).mp. (1)

40. 4 and 39 (0)

41. TRUS-TB.mp. (0)

42. (‘MRI/TRUS’ or ‘mpMRI/TRUS’ or ‘MR/US’ or ‘MRI/TRUS-TB’).mp. (0)

43. (fusion$ adj3 (software or hardware or computer$ or device$ or system$ or technolog$ or machine$ or platform$)).mp. (26)

44. 4 and 43 (0)

45. (KOELIS or Fusion Bx).mp. (0)

46. (BioJet or Trinity or PROMAP or Fusion MR or bkFusion or bk Fusion or BK3000 or BK 3000 or BK5000 or BK 5000 or Predictive Fusion).mp. (356)

47. 4 and 46 (0)

48. (BiopSee or UroNav or ‘iSR’obot’ or iSRobot or iSR obot or UroFusion or UroBiopsy or FusionVu$ or ExactVu$).mp. (0)

49. (DynaCAD or ARTEMIS or ProFuse or Mona Lisa).mp. (24)

50. 4 and 49 (0)

51. 18 or 19 or 20 or 21 or 22 or 23 or 25 or 26 or 28 or 29 or 30 or 31 or 32 or 33 or 34 or 35 or 36 or 37 or 38 or 40 or 41 or 42 or 44 or 45 or 47 or 48 or 50 (0)

Key

1. $ = truncation

2. mp = multi-purpose field search – terms in title, original title, abstract, name of substance word, or subject heading word

3. adj3 = terms within three words of each other (any order)

EMBASE

Via Ovid http://onlinelibrary.wiley.com/

Date range: 1974 to 13 May 2022

Date searched: 16 May 2022

Records retrieved: 6221

Embase was searched again on 2 August 2022. After conference abstracts were removed, 3318 studies were retrieved.

1. exp prostate tumor/ (271321)

2. ((prostate$ or prostatic or intraprostatic) adj4 (cancer$ or neoplas$ or tumour$ or tumor$ or malignan$ or metasta$ or carcinoma$ or adenocarcinoma$ or lesion$ or nodul$ or sarcoma$ or lymphoma$)).ti,ab. (244110)

3. (PCa or sPCa or csPCa or PrCa).ti,ab. (77312)

4. (((atypical adj3 proliferation) or ASAP) and prostat$).mp. (644)

5. or/1-4 (351675)

6. nuclear magnetic resonance imaging/ (903701)

7. multiparametric magnetic resonance imaging/ (6477)

8. (magnetic resonance or MRI or MR imag$ or MR scan$).ti,ab. (819806)

9. (mpMRI or mp-MRI or mpMR imag$ or mpMR scan$ or mp-MR imag$ or mp-MR scan$ or bpMRI or bp-MRI or bpMR imag$ or bpMR scan$ or bp-MR imag$ or bp-MR scan$).ti,ab. (4373)

10. or/6-9 (1163227)

11. computer assisted diagnosis/ (41296)

12. (fusion$ or fuse$ or fusing$).ti,ab. (361890)

13. cognitive$.ti,ab. (585952)

14. (visual$ adj3 (estimat$ or direct$ or align$ or guid$ or influenc$)).ti,ab. (35918)

15. registration$.ti,ab. (163670)

16. (elastic or rigid or nonrigid).ti,ab. (152621)

17. software/ or imaging software/ or nuclear magnetic resonance scanner software/ or ultrasound imaging system software/ (139562)

18. (software or hardware).ti,ab. (363110)

19. or/11-18 (1695591)

20. exp prostate/ (54557)

21. (prostate$ or prostatic).ti,ab. (336120)

22. 20 or 21 (339264)

23. biopsy/ (174400)

24. image guided biopsy/ (6935)

25. endoscopic ultrasound guided fine needle biopsy/ (5968)

26. exp needle biopsy/ (79356)

27. biopsy technique/ (7739)

28. tumor biopsy/ (43525)

29. (biopsy or biopsie$ or rebiopsy or rebiopsie$).ti,ab. (685526)

30. or/23-29 (782015)

31. 22 and 30 (43352)

32. prostate biopsy/ or exp transperineal biopsy/ or exp transrectal biopsy/ (24654)

33. 31 or 32 (48987)

34. 5 and 10 and 19 and 33 (3137)

35. ((MRI or MR or magnetic resonance or mpMRI or mp-MRI or bpMRI or bp-MRI) adj6 (prior or previous$ or preced$ or before$ or earlier or first or initial$) adj6 (biopsy or biopsie$)).ti,ab. (1707)

36. ((MRI or MR or magnetic resonance or mpMRI or mp-MRI or bpMRI or bp-MRI) adj6 prebiops$).ti,ab. (248)

37. ((MRI or MR or magnetic resonance or mpMRI or mp-MRI or bpMRI or bp-MRI) adj6 (prior or previous$ or preced$ or before$ or earlier or first or initial$) adj6 (ultrasound$ or ultrasonic$ or ultrasonograph$ or TRUS or transperineal$ or transrectal$)).ti,ab. (1370)

38. or/35-37 (2954)

39. 5 and 38 (1359)

40. (target$ adj4 (biopsy or biopsie$ or rebiopsy or rebiopsie$)).ti,ab. (7633)

41. (focal adj2 (biopsy or biopsie$ or rebiopsy or rebiopsie$)).ti,ab. (1195)

42. 40 or 41 (8750)

43. 5 and 40 (3525)

44. (target$ adj4 (MRI or MR or magnetic resonance or mpMRI or mp-MRI or bpMRI or bp-MRI)).ti,ab. (6907)

45. (focal adj2 (MRI or MR or magnetic resonance or mpMRI or mp-MRI or bpMRI or bp-MRI)).ti,ab. (959)

46. 44 or 45 (7838)

47. 5 and 31 and 46 (2297)

48. mri guided biopsy/ (246)

49. ((MRI or MR or magnetic resonance or mpMRI or mp-MRI or bpMRI or bp-MRI) adj3 (guid$ or influenc$ or direct$ or align$)).ti,ab. (18601)

50. 48 or 49 (18743)

51. 5 and 31 and 50 (1937)

52. ((MRI stratified or magnetic resonance imaging stratified) adj3 pathway$).ti,ab. (3)

53. magnetic resonance imaging ultrasound fusion biopsy/ (128)

54. image guided noninferiority targeted biopsy/ (1)

55. cognitive biopsy/ (4)

56. software based targeted biopsy/ (1)

57. visually directed targeted biopsy/ (1)

58. ultrasound fusion targeted biopsy/ (3)

59. or/52-58 (140)

60. 34 or 39 or 43 or 47 or 51 or 59 (6166)

61. (MRGB or MR-GB or MRIGB or MRI-GB).ti,ab. (132)

62. (MRIFB or MRI-FB).ti,ab. (8)

63. (MRFTB or MRF-TB).ti,ab. (36)

64. (MRTB or MR-TB or MRITB or MRI-TB or MRTBx or MR-TBx or MRITBx or MRI-TBx).ti,ab. (168)

65. FBx.ti,ab. (226)

66. (FUSTB or FUS-TB or TB-FUS).ti,ab. (11)

67. Fusion TB.ti,ab. (29)

68. (MRI-TRUS or MRI-TRUSB or MRI-TPB).ti,ab. (485)

69. (COG-TB or TB-COG or CBx).ti,ab. (829)

70. TRUS-TB.ti,ab. (8)

71. (‘MRI/TRUS’ or ‘mpMRI/TRUS’ or ‘MR/US’ or ‘MRI/TRUS-TB’).ti,ab. (777)

72. or/61-71 (2124)

73. 5 and 72 (1009)

74. 60 or 73 (6215)

75. (fusion$ adj3 (software or hardware or computer$ or device$ or system$ or technolog$ or machine$ or platform$)).ti,ab. (7446)

76. 5 and 75 (707)

77. magnetic resonance imaging-ultrasound fusion-guided prostate biopsy device/ (245)

78. 74 or 76 or 77 (6346)

79. KOELIS.ti,ab,dv. (180)

80. Fusion Bx.ti,ab,dv. (16)

81. BioJet.ti,ab,dv. (105)

82. (Trinity or PROMAP).ti,ab,dv. (2121)

83. Fusion MR.ti,ab,dv. (13)

84. (bkFusion or bk Fusion or BK3000 or BK 3000 or BK5000 or BK 5000 or Predictive Fusion).ti,ab,dv. (60)

85. or/81-84 (2295)

86. 5 and 85 (148)

87. 79 or 80 or 86 (307)

88. BiopSee .ti,ab,dv. (52)

89. UroNav.ti,ab,dv. (163)

90. (‘iSR’obot’ or iSRobot or iSR obot or UroFusion or UroBiopsy).ti,ab,dv. (31)

91. (FusionVu$ or ExactVu$).ti,ab,dv. (84)

92. DynaCAD.ti,ab,dv. (73)

93. (ARTEMIS or ProFuse).ti,ab,dv. (6586)

94. Mona Lisa.ti,ab,dv. (162)

95. or/92-94 (6817)

96. 5 and 95 (247)

97. 88 or 89 or 90 or 91 or 96 (506)

98. 78 or 87 or 97 (6483)

99. (animal/ or animal experiment/ or animal model/ or animal tissue/ or nonhuman/) not exp human/ (6457016)

100. 98 not 99 (6455)

101. limit 100 to yr=‘2008 -Current’ (6221)

Key

1. / = subject heading (Emtree heading)

2. exp = exploded subject heading (Emtree heading)

3. $ = truncation

4. ti,ab = terms in title or abstract fields

5. mp = multi-purpose field search – terms in title, original title, abstract, name of substance word, or subject heading word

6. dv = terms in the device trade name field

7. adj3 = terms within three words of each other (any order)

Health Management and Information Consortium (HMIC)

Via Ovid http://onlinelibrary.wiley.com/

Date range: 1979 to March 2022

Date searched: 16 May 2022

Records retrieved: 0

1. ((prostate$ or prostatic or intraprostatic) adj4 (cancer$ or neoplas$ or tumour$ or tumor$ or malignan$ or metasta$ or carcinoma$ or adenocarcinoma$ or lesion$ or nodul$ or sarcoma$ or lymphoma$)).mp. (736)

2. (PCa or sPCa or csPCa or PrCa).mp. (74)

3. (((atypical adj3 proliferation) or ASAP) and prostat$).mp. (0)

4. or/1-3 (792)

5. (magnetic resonance or MRI or MR imag$ or MR scan$).mp. (483)

6. (mpMRI or mp-MRI or mpMR imag$ or mpMR scan$ or mp-MR imag$ or mp-MR scan$ or bpMRI or bp-MRI or bpMR imag$ or bpMR scan$ or bp-MR imag$ or bp-MR scan$).mp. (0)

7. 5 or 6 (483)

8. (fusion$ or fuse$ or fusing$).mp. (94)

9. cognitive$.mp. (2602)

10. (visual$ adj3 (estimat$ or direct$ or align$ or guid$ or influenc$)).mp. (23)

11. registration$.mp. (4038)

12. (elastic or rigid or nonrigid).mp. (258)

13. (software or hardware).mp. (1828)

14. or/8-13 (8757)

15. (prostate$ or prostatic).mp. (914)

16. (biopsy or biopsie$ or rebiopsy or rebiopsie$).mp. (303)

17. 15 and 16 (36)

18. 4 and 7 and 14 and 17 (0)

19. ((MRI or MR or magnetic resonance or mpMRI or mp-MRI or bpMRI or bp-MRI) adj6 (prior or previous$ or preced$ or before$ or earlier or first or initial$) adj6 (biopsy or biopsie$)).mp. (0)

20. ((MRI or MR or magnetic resonance or mpMRI or mp-MRI or bpMRI or bp-MRI) adj6 prebiops$).mp. (0)

21. ((MRI or MR or magnetic resonance or mpMRI or mp-MRI or bpMRI or bp-MRI) adj6 (prior or previous$ or preced$ or before$ or earlier or first or initial$) adj6 (ultrasound$ or ultrasonic$ or ultrasonograph$ or TRUS or transperineal$ or transrectal$)).mp. (0)

22. (target$ adj4 (biopsy or biopsie$ or rebiopsy or rebiopsie$)).mp. (0)

23. (focal adj2 (biopsy or biopsie$ or rebiopsy or rebiopsie$)).mp. (0)

24. (target$ adj4 (MRI or MR or magnetic resonance or mpMRI or mp-MRI or bpMRI or bp-MRI)).mp. (1)

25. 4 and 24 (0)

26. (focal adj2 (MRI or MR or magnetic resonance or mpMRI or mp-MRI or bpMRI or bp-MRI)).mp. (0)

27. ((MRI or MR or magnetic resonance or mpMRI or mp-MRI or bpMRI or bp-MRI) adj3 (guid$ or influenc$ or direct$ or align$)).mp. (22)

28. 4 and 27 (0)

29. ((MRI stratified or magnetic resonance imaging stratified) adj3 pathway$).mp. (0)

30. (MRGB or MR-GB or MRIGB or MRI-GB).mp. (0)

31. (MRIFB or MRI-FB).mp. (0)

32. (MRFTB or MRF-TB).mp. (0)

33. (MRFTB or MRF-TB).mp. (0)

34. (MRTB or MR-TB or MRITB or MRI-TB or MRTBx or MR-TBx or MRITBx or MRI-TBx).mp. (0)

35. FBx.mp. (0)

36. (FUSTB or FUS-TB or TB-FUS).mp. (0)

37. Fusion TB.mp. (0)

38. (MRI-TRUS or MRI-TRUSB or MRI-TPB).mp. (0)

39. (COG-TB or TB-COG or CBx).mp. (0)

40. TRUS-TB.mp. (0)

41. (‘MRI/TRUS’ or ‘mpMRI/TRUS’ or ‘MR/US’ or ‘MRI/TRUS-TB’).mp. (0)

42. (fusion$ adj3 (software or hardware or computer$ or device$ or system$ or technolog$ or machine$ or platform$)).mp. (1)

43. 4 and 42 (0)

44. (KOELIS or Fusion Bx or BioJet).mp. (0)

45. (Trinity or PROMAP or Fusion MR or bkFusion or bk Fusion or BK3000 or BK 3000 or BK5000 or BK 5000 or Predictive Fusion).mp. (12)

46. 4 and 45 (0)

47. (BiopSee or UroNav or ‘iSR’obot’ or iSRobot or iSR obot or UroFusion or UroBiopsy or FusionVu$ or ExactVu$).mp. (0)

48. (DynaCAD or ARTEMIS or ProFuse or Mona Lisa).mp. (12)

49. 4 and 48 (0)

50. 18 or 19 or 20 or 21 or 22 or 23 or 25 or 26 or 28 or 29 or 30 or 31 or 32 or 33 or 34 or 35 or 36 or 37 or 38 or 39 or 40 or 41 or 43 or 44 or 46 or 47 or 49 (0)

Key

1. $ = truncation

2. mp = multi-purpose field search – terms in title, original title, abstract, name of substance word, or subject heading word

3. adj3 = terms within three words of each other (any order)

Health Technology Assessment (HTA) database

Via www.crd.york.ac.uk/CRDWeb/

Date range: Inception – 31 March 2018

Date searched: 16 May 2022

Records retrieved: 2

See under DARE for search strategy used.

International Health Technology Assessment (INAHTA) database

Via http://onlinelibrary.wiley.com/

Date searched: 16 May 2022

Records retrieved: 38

1. ((((biopsy OR biopsie* OR rebiopsy OR rebiopsie*)[Title] OR (biopsy OR biopsie* OR rebiopsy OR rebiopsie*)[abs] OR (biopsy OR biopsie* OR rebiopsy OR rebiopsie*)[Keywords]) OR (‘Biopsy, Needle’[mh]) OR (‘Biopsy, Large-Core Needle’[mh]) OR (‘Biopsy, Fine-Needle’[mh]) OR (‘Endoscopic Ultrasound-Guided Fine Needle Aspiration’[mh]) OR (‘Image-Guided Biopsy’[mh]) OR (‘Biopsy’[mh])) AND (((prostate* OR prostatic)[Title] OR (prostate* OR prostatic)[abs] OR (prostate* OR prostatic)[Keywords]) OR (‘Prostate’[mh]))) AND (((software OR hardware)[Title] OR (software OR hardware)[abs] OR (software OR hardware)[Keywords]) OR (‘Software’[mh]) OR ((visual* AND (estimat* OR direct* OR align* OR guid* OR influenc*))[Title] OR (visual* AND (estimat* OR direct* OR align* OR guid* OR influenc*))[abs] OR (visual* AND (estimat* OR direct* OR align* OR guid* OR influenc*))[Keywords]) OR ((fusion* OR fuse* OR fusing* OR cognitive OR registration* OR elastic OR rigid OR nonrigid)[Title] OR (fusion* OR fuse* OR fusing* OR cognitive OR registration* OR elastic OR rigid OR nonrigid)[abs] OR (fusion* OR fuse* OR fusing* OR cognitive OR registration* OR elastic OR rigid OR nonrigid)[Keywords]) OR (‘1mage 1nterpretation, Computer-Assisted’[mh])) AND (((mpMRI OR mp-MRI OR mpMR imag* OR mpMR scan* OR mp-MR imag* OR mp-MR scan* OR bpMRI OR bp-MRI OR bpMR imag* OR bpMR scan* OR bp-MR imag* OR bp-MR scan*)[Title] OR (mpMRI OR mp-MRI OR mpMR imag* OR mpMR scan* OR mp-MR imag* OR mp-MR scan* OR bpMRI OR bp-MRI OR bpMR imag* OR bpMR scan* OR bp-MR imag* OR bp-MR scan*)[abs] OR (mpMRI OR mp-MRI OR mpMR imag* OR mpMR scan* OR mp-MR imag* OR mp-MR scan* OR bpMRI OR bp-MRI OR bpMR imag* OR bpMR scan* OR bp-MR imag* OR bp-MR scan*)[Keywords]) OR ((‘magnetic resonance’ OR MRI OR MR imag* OR MR scan*)[Title] OR (‘magnetic resonance’ OR MRI OR MR imag* OR MR scan*)[abs] OR (‘magnetic resonance’ OR MRI OR MR imag* OR MR scan*)[Keywords]) OR (‘Multiparametric Magnetic Resonance Imaging’[mh]) OR (‘Magnetic Resonance Imaging’[mh])) AND (((ASAP AND prostat*)[Title] OR (ASAP AND prostat*)[abs] OR (ASAP AND prostat*)[Keywords]) OR ((atypical AND proliferation AND prostat*)[Title] OR (atypical AND proliferation AND prostat*)[abs] OR (atypical AND proliferation AND prostat*)[Keywords]) OR ((PCa OR sPCa OR csPCa OR PrCa)[Title] OR (PCa OR sPCa OR csPCa OR PrCa)[abs] OR (PCa OR sPCa OR csPCa OR PrCa)[Keywords]) OR (((cancer* OR neoplas* OR tumour* OR tumor* OR malignan* OR metasta* OR carcinoma* OR adenocarcinoma* OR lesion* OR nodul* OR sarcoma* OR lymphoma*)[Title] OR (cancer* OR neoplas* OR tumour* OR tumor* OR malignan* OR metasta* OR carcinoma* OR adenocarcinoma* OR lesion* OR nodul* OR sarcoma* OR lymphoma*)[abs] OR (cancer* OR neoplas* OR tumour* OR tumor* OR malignan* OR metasta* OR carcinoma* OR adenocarcinoma* OR lesion* OR nodul* OR sarcoma* OR lymphoma*)[Keywords]) AND (((prostate* OR prostatic OR intraprostatic))[Title] OR ((prostate* OR prostatic OR intraprostatic))[abs] OR ((prostate* OR prostatic OR intraprostatic))[Keywords])) OR (‘Prostatic Intraepithelial Neoplasia’[mh]) OR (‘Prostatic Neoplasms’[mhe])) 4 hits

2. ((((target* OR focal) AND (biopsy OR biopsie* OR rebiopsy OR rebiopsie*))[Title] OR ((target* OR focal) AND (biopsy OR biopsie* OR rebiopsy OR rebiopsie*))[abs] OR ((target* OR focal) AND (biopsy OR biopsie* OR rebiopsy OR rebiopsie*))[Keywords]) OR (((MRI OR MR OR ‘magnetic resonance’ OR mpMRI OR mp-MRI OR bpMRI OR bp-MRI) AND (ultrasound* OR ultrasonic* OR ultrasonograph* OR TRUS OR transperineal* OR transrectal*))[Title] OR ((MRI OR MR OR ‘magnetic resonance’ OR mpMRI OR mp-MRI OR bpMRI OR bp-MRI) AND (ultrasound* OR ultrasonic* OR ultrasonograph* OR TRUS OR transperineal* OR transrectal*))[abs] OR ((MRI OR MR OR ‘magnetic resonance’ OR mpMRI OR mp-MRI OR bpMRI OR bp-MRI) AND (ultrasound* OR ultrasonic* OR ultrasonograph* OR TRUS OR transperineal* OR transrectal*))[Keywords]) OR ((((MRI OR MR OR ‘magnetic resonance’ OR mpMRI OR mp-MRI OR bpMRI OR bp-MRI) AND prebiops*)[Title] OR ((MRI OR MR OR ‘magnetic resonance’ OR mpMRI OR mp-MRI OR bpMRI OR bp-MRI) AND prebiops*)[abs] OR ((MRI OR MR OR ‘magnetic resonance’ OR mpMRI OR mp-MRI OR bpMRI OR bp-MRI) AND prebiops*)[Keywords])) OR (((MRI OR MR OR ‘magnetic resonance’ OR mpMRI OR mp-MRI OR bpMRI OR bp-MRI) AND biops*)[Title] OR ((MRI OR MR OR ‘magnetic resonance’ OR mpMRI OR mp-MRI OR bpMRI OR bp-MRI) AND biops*)[abs] OR ((MRI OR MR OR ‘magnetic resonance’ OR mpMRI OR mp-MRI OR bpMRI OR bp-MRI) AND biops*)[Keywords])) AND (((ASAP AND prostat*)[Title] OR (ASAP AND prostat*)[abs] OR (ASAP AND prostat*)[Keywords]) OR ((atypical AND proliferation AND prostat*)[Title] OR (atypical AND proliferation AND prostat*)[abs] OR (atypical AND proliferation AND prostat*)[Keywords]) OR ((PCa OR sPCa OR csPCa OR PrCa)[Title] OR (PCa OR sPCa OR csPCa OR PrCa)[abs] OR (PCa OR sPCa OR csPCa OR PrCa)[Keywords]) OR (((cancer* OR neoplas* OR tumour* OR tumor* OR malignan* OR metasta* OR carcinoma* OR adenocarcinoma* OR lesion* OR nodul* OR sarcoma* OR lymphoma*)[Title] OR (cancer* OR neoplas* OR tumour* OR tumor* OR malignan* OR metasta* OR carcinoma* OR adenocarcinoma* OR lesion* OR nodul* OR sarcoma* OR lymphoma*)[abs] OR (cancer* OR neoplas* OR tumour* OR tumor* OR malignan* OR metasta* OR carcinoma* OR adenocarcinoma* OR lesion* OR nodul* OR sarcoma* OR lymphoma*)[Keywords]) AND (((prostate* OR prostatic OR intraprostatic))[Title] OR ((prostate* OR prostatic OR intraprostatic))[abs] OR ((prostate* OR prostatic OR intraprostatic))[Keywords])) OR (‘Prostatic Intraepithelial Neoplasia’[mh]) OR (‘Prostatic Neoplasms’[mhe])) 9 hits

3. ((((MRI OR MR OR ‘magnetic resonance’ OR mpMRI OR mp-MRI OR bpMRI OR bp-MRI) AND (guid* OR influenc* OR direct* OR align*))[Title] OR ((MRI OR MR OR ‘magnetic resonance’ OR mpMRI OR mp-MRI OR bpMRI OR bp-MRI) AND (guid* OR influenc* OR direct* OR align*))[abs] OR ((MRI OR MR OR ‘magnetic resonance’ OR mpMRI OR mp-MRI OR bpMRI OR bp-MRI) AND (guid* OR influenc* OR direct* OR align*))[Keywords]) OR (((target* OR focal) AND (MRI OR MR OR ‘magnetic resonance’ OR mpMRI OR mp-MRI OR bpMRI OR bp-MRI))[Title] OR ((target* OR focal) AND (MRI OR MR OR ‘magnetic resonance’ OR mpMRI OR mp-MRI OR bpMRI OR bp-MRI))[abs] OR ((target* OR focal) AND (MRI OR MR OR ‘magnetic resonance’ OR mpMRI OR mp-MRI OR bpMRI OR bp-MRI))[Keywords])) AND ((((biopsy OR biopsie* OR rebiopsy OR rebiopsie*)[Title] OR (biopsy OR biopsie* OR rebiopsy OR rebiopsie*)[abs] OR (biopsy OR biopsie* OR rebiopsy OR rebiopsie*)[Keywords]) OR (‘Biopsy, Needle’[mh]) OR (‘Biopsy, Large-Core Needle’[mh]) OR (‘Biopsy, Fine-Needle’[mh]) OR (‘Endoscopic Ultrasound-Guided Fine Needle Aspiration’[mh]) OR (‘Image-Guided Biopsy’[mh]) OR (‘Biopsy’[mh])) AND (((prostate* OR prostatic)[Title] OR (prostate* OR prostatic)[abs] OR (prostate* OR prostatic)[Keywords]) OR (‘Prostate’[mh]))) AND (((ASAP AND prostat*)[Title] OR (ASAP AND prostat*)[abs] OR (ASAP AND prostat*)[Keywords]) OR ((atypical AND proliferation AND prostat*)[Title] OR (atypical AND proliferation AND prostat*)[abs] OR (atypical AND proliferation AND prostat*)[Keywords]) OR ((PCa OR sPCa OR csPCa OR PrCa)[Title] OR (PCa OR sPCa OR csPCa OR PrCa)[abs] OR (PCa OR sPCa OR csPCa OR PrCa)[Keywords]) OR (((cancer* OR neoplas* OR tumour* OR tumor* OR malignan* OR metasta* OR carcinoma* OR adenocarcinoma* OR lesion* OR nodul* OR sarcoma* OR lymphoma*)[Title] OR (cancer* OR neoplas* OR tumour* OR tumor* OR malignan* OR metasta* OR carcinoma* OR adenocarcinoma* OR lesion* OR nodul* OR sarcoma* OR lymphoma*)[abs] OR (cancer* OR neoplas* OR tumour* OR tumor* OR malignan* OR metasta* OR carcinoma* OR adenocarcinoma* OR lesion* OR nodul* OR sarcoma* OR lymphoma*)[Keywords]) AND (((prostate* OR prostatic OR intraprostatic))[Title] OR ((prostate* OR prostatic OR intraprostatic))[abs] OR ((prostate* OR prostatic OR intraprostatic))[Keywords])) OR (‘Prostatic Intraepithelial Neoplasia’[mh]) OR (‘Prostatic Neoplasms’[mhe])) 5 hits

4. ((MRGB OR MR-GB OR MRIGB OR MRI-GB OR MRIFB OR MRI-FB OR MRFTB OR MRF-TB OR MRFTB OR MRF-TB OR MRTB OR MR-TB OR MRITB OR MRI-TB OR MRTBx OR MR-TBx OR MRITBx OR MRI-TBx OR FBx OR FUSTB OR FUS-TB OR TB-FUS OR ‘Fusion TB’ OR MRI-TRUS OR MRI-TRUSB OR MRI-TPB OR COG-TB OR TB-COG OR CBx OR TRUS-TB OR ‘MRI/TRUS’ OR ‘mpMRI/TRUS’ OR ‘MR/US’ OR ‘MRI/TRUS-TB’)[Title] OR (MRGB OR MR-GB OR MRIGB OR MRI-GB OR MRIFB OR MRI-FB OR MRFTB OR MRF-TB OR MRFTB OR MRF-TB OR MRTB OR MR-TB OR MRITB OR MRI-TB OR MRTBx OR MR-TBx OR MRITBx OR MRI-TBx OR FBx OR FUSTB OR FUS-TB OR TB-FUS OR ‘Fusion TB’ OR MRI-TRUS OR MRI-TRUSB OR MRI-TPB OR COG-TB OR TB-COG OR CBx OR TRUS-TB OR ‘MRI/TRUS’ OR ‘mpMRI/TRUS’ OR ‘MR/US’ OR ‘MRI/TRUS-TB’)[abs] OR (MRGB OR MR-GB OR MRIGB OR MRI-GB OR MRIFB OR MRI-FB OR MRFTB OR MRF-TB OR MRFTB OR MRF-TB OR MRTB OR MR-TB OR MRITB OR MRI-TB OR MRTBx OR MR-TBx OR MRITBx OR MRI-TBx OR FBx OR FUSTB OR FUS-TB OR TB-FUS OR ‘Fusion TB’ OR MRI-TRUS OR MRI-TRUSB OR MRI-TPB OR COG-TB OR TB-COG OR CBx OR TRUS-TB OR ‘MRI/TRUS’ OR ‘mpMRI/TRUS’ OR ‘MR/US’ OR ‘MRI/TRUS-TB’)[Keywords]) AND (((ASAP AND prostat*)[Title] OR (ASAP AND prostat*)[abs] OR (ASAP AND prostat*)[Keywords]) OR ((atypical AND proliferation AND prostat*)[Title] OR (atypical AND proliferation AND prostat*)[abs] OR (atypical AND proliferation AND prostat*)[Keywords]) OR ((PCa OR sPCa OR csPCa OR PrCa)[Title] OR (PCa OR sPCa OR csPCa OR PrCa)[abs] OR (PCa OR sPCa OR csPCa OR PrCa)[Keywords]) OR (((cancer* OR neoplas* OR tumour* OR tumor* OR malignan* OR metasta* OR carcinoma* OR adenocarcinoma* OR lesion* OR nodul* OR sarcoma* OR lymphoma*)[Title] OR (cancer* OR neoplas* OR tumour* OR tumor* OR malignan* OR metasta* OR carcinoma* OR adenocarcinoma* OR lesion* OR nodul* OR sarcoma* OR lymphoma*)[abs] OR (cancer* OR neoplas* OR tumour* OR tumor* OR malignan* OR metasta* OR carcinoma* OR adenocarcinoma* OR lesion* OR nodul* OR sarcoma* OR lymphoma*)[Keywords]) AND (((prostate* OR prostatic OR intraprostatic))[Title] OR ((prostate* OR prostatic OR intraprostatic))[abs] OR ((prostate* OR prostatic OR intraprostatic))[Keywords])) OR (‘Prostatic Intraepithelial Neoplasia’[mh]) OR (‘Prostatic Neoplasms’[mhe])) 14 hits

5. ((Trinity OR PROMAP OR ‘Fusion MR’ OR bkFusion OR ‘bk Fusion’ OR BK3000 OR ‘BK 3000’ OR BK5000 OR ‘BK 5000’ OR ‘Predictive Fusion’ OR DynaCAD OR ARTEMIS OR ProFuse OR ‘Mona Lisa’)[Title] OR (Trinity OR PROMAP OR ‘Fusion MR’ OR bkFusion OR ‘bk Fusion’ OR BK3000 OR ‘BK 3000’ OR BK5000 OR ‘BK 5000’ OR ‘Predictive Fusion’ OR DynaCAD OR ARTEMIS OR ProFuse OR ‘Mona Lisa’)[abs] OR (Trinity OR PROMAP OR ‘Fusion MR’ OR bkFusion OR ‘bk Fusion’ OR BK3000 OR ‘BK 3000’ OR BK5000 OR ‘BK 5000’ OR ‘Predictive Fusion’ OR DynaCAD OR ARTEMIS OR ProFuse OR ‘Mona Lisa’)[Keywords]) AND (((ASAP AND prostat*)[Title] OR (ASAP AND prostat*)[abs] OR (ASAP AND prostat*)[Keywords]) OR ((atypical AND proliferation AND prostat*)[Title] OR (atypical AND proliferation AND prostat*)[abs] OR (atypical AND proliferation AND prostat*)[Keywords]) OR ((PCa OR sPCa OR csPCa OR PrCa)[Title] OR (PCa OR sPCa OR csPCa OR PrCa)[abs] OR (PCa OR sPCa OR csPCa OR PrCa)[Keywords]) OR (((cancer* OR neoplas* OR tumour* OR tumor* OR malignan* OR metasta* OR carcinoma* OR adenocarcinoma* OR lesion* OR nodul* OR sarcoma* OR lymphoma*)[Title] OR (cancer* OR neoplas* OR tumour* OR tumor* OR malignan* OR metasta* OR carcinoma* OR adenocarcinoma* OR lesion* OR nodul* OR sarcoma* OR lymphoma*)[abs] OR (cancer* OR neoplas* OR tumour* OR tumor* OR malignan* OR metasta* OR carcinoma* OR adenocarcinoma* OR lesion* OR nodul* OR sarcoma* OR lymphoma*)[Keywords]) AND (((prostate* OR prostatic OR intraprostatic))[Title] OR ((prostate* OR prostatic OR intraprostatic))[abs] OR ((prostate* OR prostatic OR intraprostatic))[Keywords])) OR (‘Prostatic Intraepithelial Neoplasia’[mh]) OR (‘Prostatic Neoplasms’[mhe]))

6. ((fusion* AND (software OR hardware OR computer* OR device* OR system* OR technolog* OR machine* OR platform*))[Title] OR (fusion* AND (software OR hardware OR computer* OR device* OR system* OR technolog* OR machine* OR platform*))[abs] OR (fusion* AND (software OR hardware OR computer* OR device* OR system* OR technolog* OR machine* OR platform*))[Keywords]) AND (((ASAP AND prostat*)[Title] OR (ASAP AND prostat*)[abs] OR (ASAP AND prostat*)[Keywords]) OR ((atypical AND proliferation AND prostat*)[Title] OR (atypical AND proliferation AND prostat*)[abs] OR (atypical AND proliferation AND prostat*)[Keywords]) OR ((PCa OR sPCa OR csPCa OR PrCa)[Title] OR (PCa OR sPCa OR csPCa OR PrCa)[abs] OR (PCa OR sPCa OR csPCa OR PrCa)[Keywords]) OR (((cancer* OR neoplas* OR tumour* OR tumor* OR malignan* OR metasta* OR carcinoma* OR adenocarcinoma* OR lesion* OR nodul* OR sarcoma* OR lymphoma*)[Title] OR (cancer* OR neoplas* OR tumour* OR tumor* OR malignan* OR metasta* OR carcinoma* OR adenocarcinoma* OR lesion* OR nodul* OR sarcoma* OR lymphoma*)[abs] OR (cancer* OR neoplas* OR tumour* OR tumor* OR malignan* OR metasta* OR carcinoma* OR adenocarcinoma* OR lesion* OR nodul* OR sarcoma* OR lymphoma*)[Keywords]) AND (((prostate* OR prostatic OR intraprostatic))[Title] OR ((prostate* OR prostatic OR intraprostatic))[abs] OR ((prostate* OR prostatic OR intraprostatic))[Keywords])) OR (‘Prostatic Intraepithelial Neoplasia’[mh]) OR (‘Prostatic Neoplasms’[mhe])) 2 hits

key

1. [abs] = abstract

2. [mh] = subject heading (MeSH heading)

3. [mhe] = exploded subject heading (MeSH heading)

4. * = truncation

Latin American and Caribbean Health Sciences Literature (LILACS)

Via https://pesquisa.bvsalud.org/portal/advanced/?lang=en

Date searched: 16 May 2022

Records retrieved: 98

1. 1. Search of title, abstract, subject heading fields: (prostat$) AND (cancer$ OR neoplas$ OR tumour$ OR tumor$ OR malignan$ OR metasta$ OR carcinoma$ OR adenocarcinoma$ OR lesion$ OR nodul$ OR sarcoma$ OR lymphoma$) AND (‘magnetic resonance’ OR MRI OR MR imag$ OR MR scan$ OR mpMRI OR mp-MRI OR mpMR imag$ OR mpMR scan$ OR mp-MR imag$ OR mp-MR scan$ OR bpMRI OR bp-MRI OR bpMR imag$ OR bpMR scan$ OR bp-MR imag$ OR bp-MR scan$) AND (fusion$ OR fuse$ OR fusing$ OR cognitive$ OR visual$ OR registration$ OR elastic OR rigid OR nonrigid OR software OR hardware OR target$ OR focal OR guid$ OR influenc$ OR direct$ OR align$) AND (biopsy OR biopsie$ OR rebiopsy OR rebiopsie$ OR prebiopsy$)

2. Limit: 2008–2022

3. 35 hits

4. 2. Search of title, abstract, subject heading fields: (prostat$) AND (cancer$ OR neoplas$ OR tumour$ OR tumor$ OR malignan$ OR metasta$ OR carcinoma$ OR adenocarcinoma$ OR lesion$ OR nodul$ OR sarcoma$ OR lymphoma$) AND (MRI OR MR OR magnetic resonance OR mpMRI OR mp-MRI OR bpMRI OR bp-MRI) AND (ultrasound$ OR ultrasonic$ OR ultrasonograph$ OR TRUS OR transperineal$ OR transrectal$)

5. Limit: 200–2022

6. 53 hits

7. 3. Search of title, abstract, subject heading fields: (prostat$) AND (cancer$ OR neoplas$ OR tumour$ OR tumor$ OR malignan$ OR metasta$ OR carcinoma$ OR adenocarcinoma$ OR lesion$ OR nodul$ OR sarcoma$ OR lymphoma$) AND (MRGB OR MR-GB OR MRIGB OR MRI-GB OR MRIFB OR MRI-FB OR MRFTB OR MRF-TB OR MRFTB OR MRF-TB OR MRTB OR MR-TB OR MRITB OR MRI-TB OR MRTBx OR MR-TBx OR MRITBx OR MRI-TBx OR FBx OR FUSTB OR FUS-TB OR TB-FUS OR ‘Fusion TB’ OR MRI-TRUS OR MRI-TRUSB OR MRI-TPB OR COG-TB OR TB-COG OR CBx OR TRUS-TB OR ‘MRI/TRUS’ OR ‘mpMRI/TRUS’ OR ‘MR/US’ OR ‘MRI/TRUS-TB’)

8. Limit: 2008–2022

9. 9 hits

10. 4. Search of title, abstract, subject heading fields: (KOELIS OR ‘Fusion Bx’ OR BioJet OR BiopSee OR UroNav OR ‘iSR’obot’ OR iSRobot OR ‘iSR obot’ OR UroFusion OR UroBiopsy OR FusionVu$ OR ExactVu$)

11. Limit: 2008–2022

12. 0 hits

13. 5. Search of title, abstract, subject heading fields: (Trinity OR PROMAP OR ‘Fusion MR’ OR bkFusion OR ‘bk Fusion’ OR BK3000 OR ‘BK 3000’ OR BK5000 OR ‘BK 5000’ OR ‘Predictive Fusion’ OR DynaCAD OR ARTEMIS OR ProFuse OR ‘Mona Lisa’) AND (prostat$)

14. Limit: 2008–2022

15. 1 hit

NHS Economic Evaluations Database (NHS EED)

Via www.crd.york.ac.uk/CRDWeb/

Date range: Inception – 31 March 2015

Date searched: 16 May 2022

Records retrieved: 2

See under DARE for search strategy used.

Science Citation Index

Va Web of Science, Clarivate Analytics https://clarivate.com/

Date range: 1900 – present

Date searched: 16 May 2022

Records retrieved: 3616

The Science Citation Index and the Conference Proceedings Citation Index-Science were both searched using the strategy below. Numbers of records retrieved are therefore the total number from searching both databases.

The Science Citation Index only was searched again on 2 August 2022. An amount of 3561 studies were retrieved.

1. 48 #45 OR #41 OR #37 3616

2. 47 #45 or #41 or #37 3857

3. 46 #45 OR #41 OR #37 3857

4. 45 #42 OR #44 69

5. 44 #43 AND #4 42

6. 43 TS = (DynaCAD or ARTEMIS or ProFuse or ‘Mona Lisa’) 5737

7. 42 TS = (BiopSee or UroNav or ‘iSR’obot’ or iSRobot or ‘iSR obot’ or UroFusion or UroBiopsy or FusionVu* or ExactVu*) 34

8. 41 #40 OR #38 41

9. 40 #39 AND #4 19

10. 39 TS = (BioJet or Trinity or PROMAP or ‘Fusion MR’ or bkFusion or ‘bk Fusion’ or BK3000 or ‘BK 3000’ or BK5000 or ‘BK 5000’ or ‘Predictive Fusion’) 2748

11. 38 TS = (KOELIS or ‘Fusion Bx’) 25

12. 37 #36 OR #34 OR #32 3825

13. 36 #35 AND #4 471

14. 35 TS = (fusion* NEAR/3 (software or hardware or computer* or device* or system* or technolog* or machine* or platform*)) 24,330

15. 34 #33 AND #4 451

16. 33 TS = (MRGB or MR-GB or MRIGB or MRI-GB or MRIFB or MRI-FB or MRFTB or MRF-TB or MRFTB or MRF-TB or MRTB or MR-TB or MRITB or MRI-TB or MRTBx or MR-TBx or MRITBx or MRI-TBx or FBx or FUSTB or FUS-TB or TB-FUS or ‘Fusion TB’ or MRI-TRUS or MRI-TRUSB or MRI-TPB or COG-TB or TB-COG or CBx or TRUS-TB or ‘MRI/TRUS’ or ‘mpMRI/TRUS’ or ‘MR/US’ or ‘MRI/TRUS-TB’) 1351

17. 32 #31 OR #30 OR #25 OR #18 3620

18. 31 TS=((‘MRI stratified’ or ‘magnetic resonance imaging stratified’) NEAR/3 pathway*) 3

19. 30 #29 AND #17 AND #4 2,016

20. 29 #26 OR #27 OR #28 22800

21. 28 TS=((MRI or MR or ‘magnetic resonance’ or mpMRI or mp-MRI or bpMRI or bp-MRI) NEAR/3 (guid* or influenc* or direct* or align*)) 17,122

22. 27 TS = (focal NEAR/2 (MRI or MR or ‘magnetic resonance’ or mpMRI or mp-MRI or bpMRI or bp-MRI)) 1243

23. 26 TS = (target* NEAR/4 (MRI or MR or ‘magnetic resonance’ or mpMRI or mp-MRI or bpMRI or bp-MRI)) 5682

24. 25 #24 AND #4 2567

25. 24 #23 OR #22 OR #21 OR #20 OR #19 6484

26. 23 TS = (focal NEAR/2 (biopsy or biopsie* or rebiopsy or rebiopsie*)) 666

27. 22 TS = (target* NEAR/4 (biopsy or biopsie* or rebiopsy or rebiopsie*)) 4437

28. 21 TS=((MRI or MR or ‘magnetic resonance’ or mpMRI or mp-MRI or bpMRI or bp-MRI) NEAR/6 (prior or previous* or preced* or before* or earlier or first or initial*) NEAR/6 (ultrasound* or ultrasonic* or ultrasonograph* or TRUS or transperineal* or transrectal*)) 858

29. 20 TS=((MRI or MR or ‘magnetic resonance’ or mpMRI or mp-MRI or bpMRI or bp-MRI) NEAR/6 prebiops*) 179

30. 19 TS=((MRI or MR or ‘magnetic resonance’ or mpMRI or mp-MRI or bpMRI or bp-MRI) NEAR/6 (prior or previous* or preced* or before* or earlier or first or initial*) NEAR/6 (biopsy or biopsie*)) 963

31. 18 #17 AND #14 AND #7 AND #4 1832

32. 17 #15 AND #16 28,427

33. 16 TS = (biopsy or biopsie* or rebiopsy or rebiopsie*) 379,853

34. 15 TS = (prostate* or prostatic) 336,855

35. 14 #13 OR #12 OR #11 OR #10 OR #9 OR #8 2,737,360

36. 13 TS = (software or hardware) 906,626

37. 12 TS = (elastic or rigid or nonrigid) 624,588

38. 11 TS = (registration*) 134,030

39. 10 TS = (visual* NEAR/3 (estimat* or direct* or align* or guid* or influenc*)) 43,631

40. 9 TS = cognitive* 514,118

41. 8 TS = (fusion* or fuse* or fusing*) 589,649

42. 7 #5 OR #6 757,071

43. 6 TS = (mpMRI or mp-MRI or ‘mpMR imag*’ or ‘mpMR scan*’ or ‘mp-MR imag*’ or ‘mp-MR scan*’ or bpMRI or bp-MRI or ‘bpMR imag*’ or ‘bpMR scan*’ or ‘bp-MR imag*’ or ‘bp-MR scan*’) 2175

44. 5 TS=(‘magnetic resonance’ or MRI or ‘MR imag*’ or ‘MR scan*’) 756,868

45. 4 #1 OR #2 OR #3 332,891

46. 3 TS=(((atypical NEAR/3 proliferation) or ASAP) and prostat*) 317

47. 2 TS = (PCa or sPCa or csPCa or PrCa) 101,467

48. 1 TS=((prostate* or prostatic or intraprostatic) NEAR/4 (cancer* or neoplas* or tumour* or tumor* or malignan* or metasta* or carcinoma* or adenocarcinoma* or lesion* or nodul* or sarcoma* or lymphoma*)) 246,739

Key

1. TS = topic tag; searches in title, abstract, author keywords and keywords plus fields

2. * = truncation

3. NEAR/3 = terms within three words of each other (any order)

On-going, unpublished or grey literature search strategies

Conference proceedings citation index – Science (CPCI-Science)

Via Web of Science, Clarivate Analytics https://clarivate.com/

Date range: 1990–present (CPCI-Science)

Date searched: 16 May 2022

See above under Science Citation Index for search strategy used. The number of records retrieved from CPCI-Science is not available as both Science Citation Index and CPCI-Science were searched together retrieving 3616 records in total from both databases.

EU Clinical Trials Register

via https://www.clinicaltrialsregister.eu/ctr-search/search

Search date: 15 June 2022

Records retrieved: 86

1. 68 result(s) found for: (prostate OR prostatic OR intraprostatic) AND (biopsy OR rebiopsy OR re-biopsy) AND (neoplasm OR cancer OR tumour OR tumor OR lesion OR nodule OR adenocarcinoma) AND (MRI OR MR OR ‘magnetic resonance’ OR biparametric OR multiparametric OR bpMRI OR mpMRI OR bp-MRI OR mp-MRI) date range: 2015-01-01 to 2022-06-15

2. 18 result(s) found for: (prostate OR prostatic OR intraprostatic) AND (biopsy OR rebiopsy OR re-biopsy) AND (neoplasm OR cancer OR tumour OR tumor OR lesion OR nodule OR adenocarcinoma) AND targeted date range: 2015-01-01 to 2022-06-15

3. 0 result(s) found for: (KOELIS OR ‘Fusion Bx’ OR BioJet OR BiopSee OR UroNav OR ‘iSR’obot’ OR iSRobot OR ‘iSR obot’ OR UroFusion OR UroBiopsy OR FusionVu OR ExactVu) date range: 2015-01-01 to 2022-06-15

4. 0 result(s) found for: (Trinity OR PROMAP OR ‘Fusion MR’ OR bkFusion OR ‘bk Fusion’ OR BK3000 OR ‘BK 3000’ OR BK5000 OR ‘BK 5000’ OR ‘Predictive Fusion’ OR DynaCAD OR ARTEMIS OR ProFuse OR ‘Mona Lisa’) AND (prostate OR prostatic OR intraprostatic) date range: 2015-01-01 to 2022-06-15

Open Access Theses and Dissertations (OATD)

Via https://oatd.org/

Date searched: 16 May 2022

Records retrieved: 74

3 search queries used:

Query 1

(Prostat* AND biops*) AND (fusion* OR cognitive* OR software) AND (cancer* OR neoplas* OR tumour* OR tumor* OR malignan* OR metasta* OR carcinoma* OR adenocarcinoma* OR lesion* OR nodul* OR sarcoma* OR lymphoma*) AND (‘magnetic resonance’ OR MRI OR biparametric OR multiparametric)

50 hits

Query 2

(cancer* OR neoplas* OR tumour* OR tumor* OR malignan* OR metasta* OR carcinoma* OR adenocarcinoma* OR lesion* OR nodul* OR sarcoma* OR lymphoma*) AND (prostat*) AND (BioJet OR Trinity OR PROMAP OR ‘Fusion MR’ OR bkFusion OR ‘bk Fusion’ OR BK3000 OR ‘BK 3000’ OR BK5000 OR ‘BK 5000’ OR ‘Predictive Fusion’ OR DynaCAD OR ARTEMIS OR ProFuse OR ‘Mona Lisa’)

23 hits

Query 3

KOELIS OR ‘Fusion Bx’ OR BiopSee OR UroNav OR ‘iSR’obot’ OR iSRobot OR ‘iSR obot’ OR UroFusion OR UroBiopsy OR FusionVu* OR ExactVu*

1 hit

Key

1. * = truncation

ProQuest Dissertations and Theses A&I

Via https://www.proquest.com

Date searched: 16 May 2022

Records retrieved: 207

1. ((TI,AB,SU,IF((prostate* OR prostatic OR intraprostatic) NEAR/4 (cancer* OR neoplas* OR tumour* OR tumor* OR malignan* OR metasta* OR carcinoma* OR adenocarcinoma* OR lesion* OR nodul* OR sarcoma* OR lymphoma*)) OR TI,AB,SU,IF(PCa OR sPCa OR csPCa OR PrCa)) OR TI,AB,SU,IF(((atypical NEAR/3 proliferation) OR ASAP) AND prostat*)) AND (TI,AB,SU,IF(‘magnetic resonance’ OR MRI OR MR imag* OR MR scan*) OR TI,AB,SU,IF(mpMRI OR mp-MRI OR mpMR imag* OR mpMR scan* OR mp-MR imag* OR mp-MR scan* OR bpMRI OR bp-MRI OR bpMR imag* OR bpMR scan* OR bp-MR imag* OR bp-MR scan*)) AND (TI,AB,SU,IF(prostate* OR prostatic) AND TI,AB,SU,IF(biopsy OR biopsie* OR rebiopsy OR rebiopsie*)) AND (TI,AB,SU,IF(fusion* OR fuse* OR fusing* OR cognitive* OR registration* OR elastic OR rigid OR nonrigid OR software OR hardware) OR TI,AB,SU,IF(visual* NEAR/3 (estimat* OR direct* OR align* OR guid* OR influenc*))) limit: 2008-01-01 to 2022-05-16 33 Hits

2. (TI,AB,SU,IF((prostate* OR prostatic OR intraprostatic) NEAR/4 (cancer* OR neoplas* OR tumour* OR tumor* OR malignan* OR metasta* OR carcinoma* OR adenocarcinoma* OR lesion* OR nodul* OR sarcoma* OR lymphoma*)) OR TI,AB,SU,IF(PCa OR sPCa OR csPCa OR PrCa)) AND TI,AB,SU,IF((MRI OR MR OR ‘magnetic resonance’ OR mpMRI OR mp-MRI OR bpMRI OR bp-MRI) NEAR/6 (biopsy OR biopsie* OR prebiops* OR ultrasound* OR ultrasonic* OR ultrasonograph* OR TRUS OR transperineal* OR transrectal*)) limit: 2008-01-01 to 2022-05-16 67 hits

3. (TI,AB,SU,IF((prostate* OR prostatic OR intraprostatic) NEAR/4 (cancer* OR neoplas* OR tumour* OR tumor* OR malignan* OR metasta* OR carcinoma* OR adenocarcinoma* OR lesion* OR nodul* OR sarcoma* OR lymphoma*)) OR TI,AB,SU,IF(PCa OR sPCa OR csPCa OR PrCa)) AND (TI,AB,SU,IF((target* OR focal) NEAR/4 (biopsy OR biopsie* OR rebiopsy OR rebiopsie*)) OR TI,AB,SU,IF((target* OR focal) NEAR/4 (MRI OR MR OR ‘magnetic resonance’ OR mpMRI OR mp-MRI OR bpMRI OR bp-MRI))) limit: 2008-01-01 to 2022-05-16 53 hits

4. ((TI,AB,SU,IF((prostate* OR prostatic OR intraprostatic) NEAR/4 (cancer* OR neoplas* OR tumour* OR tumor* OR malignan* OR metasta* OR carcinoma* OR adenocarcinoma* OR lesion* OR nodul* OR sarcoma* OR lymphoma*)) OR TI,AB,SU,IF(PCa OR sPCa OR csPCa OR PrCa)) OR TI,AB,SU,IF(((atypical NEAR/3 proliferation) OR ASAP) AND prostat*)) AND (TI,AB,SU,IF(prostate* OR prostatic) AND TI,AB,SU,IF(biopsy OR biopsie* OR rebiopsy OR rebiopsie*)) AND TI,AB,SU,IF((MRI OR MR OR ‘magnetic resonance’ OR mpMRI OR mp-MRI OR bpMRI OR bp-MRI) NEAR/3 (guid* OR influenc* OR direct* OR align*)) limit: 2008-01-01 to 2022-05-16 20 hits

5. ((TI,AB,SU,IF((prostate* OR prostatic OR intraprostatic) NEAR/4 (cancer* OR neoplas* OR tumour* OR tumor* OR malignan* OR metasta* OR carcinoma* OR adenocarcinoma* OR lesion* OR nodul* OR sarcoma* OR lymphoma*)) OR TI,AB,SU,IF(PCa OR sPCa OR csPCa OR PrCa)) OR TI,AB,SU,IF(((atypical NEAR/3 proliferation) OR ASAP) AND prostat*)) AND TI,AB,SU,IF(MRGB OR MR-GB OR MRIGB OR MRI-GB OR MRIFB OR MRI-FB OR MRFTB OR MRF-TB OR MRFTB OR MRF-TB OR MRTB OR MR-TB OR MRITB OR MRI-TB OR MRTBx OR MR-TBx OR MRITBx OR MRI-TBx OR FBx OR FUSTB OR FUS-TB OR TB-FUS OR ‘Fusion TB’ OR MRI-TRUS OR MRI-TRUSB OR MRI-TPB OR COG-TB OR TB-COG OR CBx OR TRUS-TB OR ‘MRI/TRUS’ OR ‘mpMRI/TRUS’ OR ‘MR/US’ OR ‘MRI/TRUS-TB’) limit: 2008-01-01 to 2022-05-16 6 hits

6. ((TI,AB,SU,IF((prostate* OR prostatic OR intraprostatic) NEAR/4 (cancer* OR neoplas* OR tumour* OR tumor* OR malignan* OR metasta* OR carcinoma* OR adenocarcinoma* OR lesion* OR nodul* OR sarcoma* OR lymphoma*)) OR TI,AB,SU,IF(PCa OR sPCa OR csPCa OR PrCa)) OR TI,AB,SU,IF(((atypical NEAR/3 proliferation) OR ASAP) AND prostat*)) AND TI,AB,SU,IF(fusion* NEAR/3 (software OR hardware OR computer* OR device* OR system* OR technolog* OR machine* OR platform*)) limit: 2008-01-01 to 2022-05-16 26 hits

7. TI,AB,SU,IF(KOELIS OR ‘Fusion Bx’ OR BiopSee OR UroNav OR ‘iSR’obot’ OR iSRobot OR ‘iSR obot’ OR UroFusion OR UroBiopsy OR FusionVu* OR ExactVu*) 0 hits

8. ((TI,AB,SU,IF((prostate* OR prostatic OR intraprostatic) NEAR/4 (cancer* OR neoplas* OR tumour* OR tumor* OR malignan* OR metasta* OR carcinoma* OR adenocarcinoma* OR lesion* OR nodul* OR sarcoma* OR lymphoma*)) OR TI,AB,SU,IF(PCa OR sPCa OR csPCa OR PrCa)) OR TI,AB,SU,IF(((atypical NEAR/3 proliferation) OR ASAP) AND prostat*)) AND TI,AB,SU,IF(BioJet OR Trinity OR PROMAP OR ‘Fusion MR’ OR bkFusion OR ‘bk Fusion’ OR BK3000 OR ‘BK 3000’ OR BK5000 OR ‘BK 5000’ OR ‘Predictive Fusion’) limit: 2008-01-01 to 2022-05-16 2 hits

Key

1. TI,AB,SU,IF = search of title, abstract, subject heading and keyword fields

2. * = truncation

3. NEAR/3 = terms within three words of each other (any order)

PROSPERO

Via https://www.crd.york.ac.uk/prospero/

Date searched: 23 May 2022

Records retrieved: 78

1. MeSH DESCRIPTOR Prostatic Intraepithelial Neoplasia 0

2. MeSH DESCRIPTOR Prostatic Neoplasms EXPLODE ALL TREES 406

3. (prostate* or prostatic or intraprostatic) adj4 (cancer* or neoplas* or tumour* or tumor* or malignan* or metasta* or carcinoma* or adenocarcinoma* or lesion* or nodul* or sarcoma* or lymphoma*) 1351

4. ((prostate* or prostatic or intraprostatic) adj4 (cancer* or neoplas* or tumour* or tumor* or malignan* or metasta* or carcinoma* or adenocarcinoma* or lesion* or nodul* or sarcoma* or lymphoma*)):TI 740

5. ((prostate* or prostatic or intraprostatic)):TI 1080

6. PCa or sPCa or csPCa or PrCa 335

7. #1 OR #2 OR #4 OR #6 951

8. #1 OR #2 OR #3 OR #6 1509

9. #5 OR #2 OR #1 1092

10. MeSH DESCRIPTOR Magnetic Resonance Imaging 458

11. MeSH DESCRIPTOR Multiparametric Magnetic Resonance Imaging 6

12. ‘magnetic resonance’ or MRI or MR imag* or MR scan* 5234

13. (‘magnetic resonance’ or MRI or MR imag* or MR scan*):TI 773

14. ((mpMRI or mp-MRI or mpMR imag* or mpMR scan* or mp-MR imag* or mp-MR scan* or bpMRI or bp-MRI or bpMR imag* or bpMR scan* or bp-MR imag* or bp-MR scan*)):TI 8

15. (mpMRI or mp-MRI or mpMR imag* or mpMR scan* or mp-MR imag* or mp-MR scan* or bpMRI or bp-MRI or bpMR imag* or bpMR scan* or bp-MR imag* or bp-MR scan*) 59

16. #10 OR #11 OR #12 OR #15 5259

17. #10 OR #11 OR #13 OR #14 887

18. MeSH DESCRIPTOR Image Interpretation, Computer-Assisted 4

19. MeSH DESCRIPTOR Software 31

20. fusion* or fuse* or fusing* or cognitive or registration* or elastic or rigid or nonrigid 17958

21. visual* adj3 (estimat* or direct* or align* or guid* or influenc*) 274

22. software or hardware 48745

23. #18 OR #19 OR #20 OR #21 OR #22 60890

24. MeSH DESCRIPTOR Prostate 102

25. prostate* or prostatic 1862

26. (prostate* or prostatic):TI 1080

27. #24 OR #25 1881

28. #24 OR #26 1102

29. (MeSH DESCRIPTOR Biopsy):TI 0

30. (MeSH DESCRIPTOR Image-Guided Biopsy EXPLODE ALL TREES):TI 0

31. (MeSH DESCRIPTOR Biopsy, Needle EXPLODE ALL TREES):TI 0

32. MeSH DESCRIPTOR Biopsy, Needle EXPLODE ALL TREES 50

33. MeSH DESCRIPTOR Biopsy, Needle EXPLODE ALL TREES 50

34. MeSH DESCRIPTOR Biopsy 103

35. MeSH DESCRIPTOR Image-Guided Biopsy EXPLODE ALL TREES 27

36. biopsy or biopsie* or rebiopsy or rebiopsie* 2655

37. (biopsy or biopsie* or rebiopsy or rebiopsie*):TI 251

38. #32 OR #34 OR #35 OR #36 2678

39. #32 OR #34 OR #35 OR #37 295

40. #8 AND #16 AND #23 AND #27 AND #38 54

41. ((MRI or MR or ‘magnetic resonance’ or mpMRI or mp-MRI or bpMRI or bp-MRI) adj6 (prior or previous* or preced* or before* or earlier or first or initial*) adj6 (biopsy or biopsie*)):TI 1

42. ((MRI or MR or magnetic resonance or mpMRI or mp-MRI or bpMRI or bp-MRI) adj6 prebiops*):TI 0

43. ((MRI or MR or ‘magnetic resonance’ or mpMRI or mp-MRI or bpMRI or bp-MRI) adj6 (prior or previous* or preced* or before* or earlier or first or initial*) adj6 (ultrasound* or ultrasonic* or ultrasonograph* or TRUS or transperineal* or transrectal*)):TI 0

44. (target* adj4 (biopsy or biopsie* or rebiopsy or rebiopsie*)):TI 15

45. (focal* adj2 (biopsy or biopsie* or rebiopsy or rebiopsie*)):TI 1

46. (MRI or MR or ‘magnetic resonance’ or mpMRI or mp-MRI or bpMRI or bp-MRI) adj6 (prior or previous* or preced* or before* or earlier or first or initial*) adj6 (biopsy or biopsie*) 9

47. (MRI or MR or magnetic resonance or mpMRI or mp-MRI or bpMRI or bp-MRI) adj6 prebiops* 0

48. (MRI or MR or magnetic resonance or mpMRI or mp-MRI or bpMRI or bp-MRI) adj6 (prior or previous* or preced* or before* or earlier or first or initial*) adj6 (ultrasound* or ultrasonic* or ultrasonograph* or TRUS or transperineal* or transrectal*) 0

49. target* adj4 (biopsy or biopsie* or rebiopsy or rebiopsie*) 48

50. focal* adj2 (biopsy or biopsie* or rebiopsy or rebiopsie*) 1

51. #46 OR #47 OR #48 OR #49 OR #50 55

52. #8 AND #51 44

53. #52 OR #40 66

54. target* adj4 (MRI or MR or ‘magnetic resonance’ or mpMRI or mp-MRI or bpMRI or bp-MRI) 22

55. focal* adj2 (MRI or MR or ‘magnetic resonance’ or mpMRI or mp-MRI or bpMRI or bp-MRI) 2

56. (MRI or MR or ‘magnetic resonance’ or mpMRI or mp-MRI or bpMRI or bp-MRI) adj3 (guid* or influenc* or direct* or align*) 139

57. #54 OR #55 OR #56 154

58. #8 AND #27 AND #38 AND #57 38

59. #53 OR #58 76

60. MRGB or MR-GB or MRIGB or MRI-GB or MRIFB or MRI-FB or MRFTB or MRF-TB or MRFTB or MRF-TB or MRTB or MR-TB or MRITB or MRI-TB or MRTBx or MR-TBx or MRITBx or MRI-TBx or FBx or FUSTB or FUS-TB or TB-FUS or ‘Fusion TB’ or MRI-TRUS or MRI-TRUSB or MRI-TPB or COG-TB or TB-COG or CBx or TRUS-TB or ‘MRI/TRUS’ or ‘mpMRI/TRUS’ or ‘MR/US’ or ‘MRI/TRUS-TB’ 17

61. fusion* adj3 (software or hardware or computer* or device* or system* or technolog* or machine* or platform*) 75

62. #60 OR #61 91

63. #8 AND #62 16

64. #63 OR #59 76

65. KOELIS OR ‘Fusion Bx’ OR BioJet OR BiopSee OR UroNav OR ‘iSR’obot’ OR iSRobot OR ‘iSR obot’ OR UroFusion OR UroBiopsy OR FusionVu* OR ExactVu* 7

66. Trinity OR PROMAP OR ‘Fusion MR’ OR bkFusion OR ‘bk Fusion’ OR BK3000 OR ‘BK 3000’ OR BK5000 OR ‘BK 5000’ OR ‘Predictive Fusion’ OR DynaCAD OR ARTEMIS OR ProFuse OR ‘Mona Lisa’ 489

67. #66 AND #8 4

68. #64 OR #65 OR #67 78