Multiparametric MRI to improve detection of prostate cancer compared with transrectal ultrasound-guided prostate biopsy alone: the PROMIS study

Article history

The research reported in this issue of the journal was funded by the HTA programme as project number 09/22/67. The contractual start date was in April 2016. The draft report began editorial review in November 2016 and was accepted for publication in July 2017. 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’ report 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 report.

Declared competing interests of authors

Hashim U Ahmed reports grants and personal fees from SonaCare Medical, Sophiris Bio Inc. and Galil Medical and grants from TROD Medical outside the submitted work. He conducts private practice in London for the diagnosis and treatment of patients with potential prostate cancer through insurance schemes and self-paying patients in addition to his NHS practice. Ahmed El-Shater Bosaily reports grants and non-financial support from University College London Hospitals/University College London Biomedical Research Centre, grants and non-financial support from The Royal Marsden NHS Foundation Trust and the Institute for Cancer Research Biomedical Research Centre and non-financial support from the Medical Research Council’s Clinical Trials Unit during the conduct of the study. Mark Emberton reports grants from Sophiris Bio Inc., grants from TROD Medical, other support from STEBA Biotech, grants and other support from SonaCare Medical and other support from Nuada Medical and London Urology Associates, outside the submitted work. Rita Faria reports grants from the National Institute for Health Research during the conduct of the study. Richard Graham Hindley reports payment as a Clinical Director from Nuada Medical during the period of recruitment. Alexander Kirkham reports other support from Nuada Medical outside the submitted work. Derek J Rosario reports personal fees from Ferring Pharmaceuticals Ltd and grants from Bayer Pharmaceuticals Division, outside the submitted work. Iqbal Shergill reports grants from Ipsen and Astellas Pharma Inc. and non-financial support from Boston Scientific and Olympus, outside the submitted work. Eldon Spackman reports personal fees from Astellas Pharma Canada, Inc., outside the submitted work. Mathias Winkler reports grants and personal fees from Zicom-Biobot (Singapore), outside the submitted work.

Copyright statement

© Queen’s Printer and Controller of HMSO 2018. This work was produced by Brown et al. under the terms of a commissioning contract issued by the Secretary of State for Health and Social Care. This issue may be freely reproduced for the purposes of private research and study and extracts (or indeed, the full report) may be included in professional journals provided that suitable acknowledgement is made and the reproduction is not associated with any form of advertising. Applications for commercial reproduction should be addressed to: NIHR Journals Library, National Institute for Health Research, Evaluation, Trials and Studies Coordinating Centre, Alpha House, University of Southampton Science Park, Southampton SO16 7NS, UK.

2018 Queen’s Printer and Controller of HMSO

Rationale

There are approximately 40,000 new cases of prostate cancer annually in the UK. The reported incidence has doubled in the last 15 years, mainly as a result of the increased use of serum prostate-specific antigen (PSA) testing in healthy men. As a result, prostate cancer has become the most commonly diagnosed cancer in men. 1 Many, if not most, prostate cancers that are currently detected are clinically non-significant (CNS) (i.e. would have no clinical impact during the man’s remaining life expectancy). Although the UK has no formal population-based PSA screening programme, the rising numbers of cases in the UK are primarily the result of increased awareness and incidental PSA testing. The existing diagnostic pathway will, if left unchecked, result in a further rise in the number of CNS cases identified and in the associated costs and harms of treatment, without necessarily reducing the risk of dying from the disease.

This position is exemplified in two large randomised controlled trials (RCTs) of population-based prostate cancer screening using serum PSA testing compared with standard practice. First, a large RCT of prostate cancer screening in the USA2 showed no evidence of a survival benefit when comparing an annual screening strategy (with an approximately 85% rate of screening) with usual care, which included quite considerable rates of PSA testing (approximately 46–52%). Second, the European Randomized Study of Screening for Prostate Cancer3 showed a modest reduction in the risk of death from prostate cancer in men who were screened every 4 years, from 8.2% to 4.8% [risk ratio 0.8, 95% confidence interval (CI) 0.65 to 0.98] at 9 years’ follow-up. The number of participants needed to screen was 1410 and the number needed to treat was 48 to prevent one death from prostate cancer over a 10-year period. 4 This benefit was maintained up to 13 years’ follow-up. 5

Once diagnosed, the lack of benefit for radical treatment of CNS cancer compared with a strategy of monitoring has been exemplified in three RCTs. First, the Scandinavian Prostate Cancer Group – 4 (SPCG-4) trial6 randomised men who were clinically diagnosed (not from screening). Most of these men had clinically significant (CS) medium- to high-risk cancers and radical prostatectomy (RP) showed a survival benefit at long-term follow-up compared with watchful waiting (a less active form of surveillance than that used in the modern era). 6 Treatment conferred significant harms for patients. 7 Second, the Prostate Cancer Intervention versus Observation Trial (PIVOT)8 randomised men diagnosed early through PSA screening in the USA between watchful waiting and RP and found no cancer-specific survival benefit, although subgroup analyses showed that men with high-risk disease did survive longer with treatment and there was a possible benefit, albeit marginal, for those with intermediate-risk cancers. Most recently, the Prostate Testing for Cancer and Treatment (ProtecT) trial in the UK9 found that, in 1643 men with localised prostate cancer detected as a result of PSA-based screening, there was no statistically significant difference in prostate-specific or all-cause mortality at 10 years in men assigned to active surveillance, surgery or radiotherapy; however, rates of disease progression and metastasis were higher with active surveillance. Commentators were quick to voice their concern that overdiagnosis, and hence overtreatment and associated morbidity, would increase further if PSA screening was adopted more widely. 10,11 However, if a diagnostic method was available that was more specific to clinically significant prostate cancer, the beneficial effect of screening and subsequent treatment on mortality could be retained, but overdiagnosis and overtreatment would be minimised.

Current clinical pathway

At present, a man is judged to be at risk of harbouring prostate cancer if he has any of the following: a raised serum PSA level (present in the majority of cases), an abnormal digital rectal examination (DRE), a positive family history of prostate cancer or a specific ethnic risk profile. 12 The diagnostic pathway of interest to this work relates to men with signs and symptoms that are suspicious of prostate cancer who have been referred for confirmatory diagnostic testing. In these men, prostate cancer, if present, is typically localised. Men with signs and symptoms of locally advanced cancer or metastatic cancer are often managed in a different manner with staging whole-body scans and the institution of systemic therapy with or without confirmatory prostate biopsy.

Men in whom there is diagnostic uncertainty are currently advised to have a transrectal ultrasound (TRUS)-guided prostate biopsy. 13 Men with a negative TRUS-guided biopsy result are advised to consider multiparametric magnetic resonance imaging (mpMRI) (using T2- and diffusion-weighted imaging) to determine whether or not another biopsy is needed. Between 59,000 and 80,000 men have a TRUS-guided biopsy in the UK each year,14 although this is likely to be an underestimate considering that three to four TRUS-guided biopsies are usually carried out to diagnose one man with cancer.

Men diagnosed with localised prostate cancer in England are managed according to their risk of progression (Table 1). The risk of progression is assessed on the basis of their serum PSA concentration, Gleason score and clinical stage. PSA is a protein produced by the prostate gland; an elevated PSA level is a sign that cancer may be present. The Gleason pattern classifies prostate cells obtained in the TRUS-guided biopsy by level of differentiation, which is related to the degree of aggressiveness of prostate cancer. Cancer cells can be classified from Gleason grade 1 (well differentiated and lower risk) to Gleason grade 5 (poorly differentiated and higher risk). 13 The Gleason score is the sum of the grades of the two most common types of cells. For example, if the man’s most common cells are Gleason grade 3 and Gleason grade 4, the Gleason pattern is 3 + 4 = 7. The clinical stage is assessed by the urologist by DRE with or without imaging scans, although the majority of men diagnosed with prostate cancer also undergo pelvic and prostate magnetic resonance imaging (MRI) after the biopsy.

The 2014 National Institute for Health and Care Excellence (NICE) clinical guideline13 recommends that men with low-risk prostate cancer are offered active surveillance, as described in Table 2. Men with intermediate- or high-risk cancer should be offered active treatment. In men with intermediate-risk cancer, the 2014 NICE clinical guideline recommends RP or radical radiotherapy but considers active surveillance or high-dose-rate brachytherapy with external-beam radiotherapy as possible options. Men with high-risk cancer should be offered RP, radical radiotherapy or high-dose-rate brachytherapy with external-beam radiotherapy. Interestingly, the NICE guidance also recommends prostate MRI as a baseline test at the commencement of active surveillance. Under the current NICE guidance and clinical practice, therefore, almost all men who have a TRUS-guided biopsy are recommended to undergo MRI after biopsy.

The current clinical pathway and potential for imaging

Men undergo prostate biopsy in the absence of accurate imaging that can visualise a suspicious lesion, as ultrasonography is used to identify the prostate, not the suspect lesion. The result is that biopsies are taken ‘blindly’ from areas of the gland. Although protocols stipulate that the biopsies should sample certain regions in a fanlike manner, studies have shown that this is often not the case and that the biopsies are clustered. 15 This results in a random, rather than systematic, deployment of the needle. This approach (Figure 1) contrasts markedly with that used for other cancers, in which the physician either visualises (e.g. using endoscopy) or images (e.g. using mammography) a suspect lesion in order to guide a biopsy needle to it.

FIGURE 1.

Current and proposed diagnostic pathway for prostate cancer. Note that the proposed future pathway is not the pathway taken by patients in the Prostate Magnetic Resonance Imaging Study. Reproduced from El-Shater Bosaily et al. 16 © 2015 The Authors. Published by Elsevier Inc. This is an Open Access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.

Reproduced from El-Shater Bosaily et al. 16 © 2015 The Authors. Published by Elsevier Inc. This is an Open Access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.

The use of mpMRI prior to TRUS-guided biopsy could offer several important advantages:

• less overdiagnosis – in that fewer CNS prostate cancers are detected by avoiding unnecessary biopsy of men who do not have CS cancer

• less overtreatment – as fewer CNS prostate cancers are detected

• increased detection of CS prostate cancers – by directing biopsies to areas of the prostate that appear abnormal on mpMRI

• improved characterisation of individual cancers – as a result of more representative biopsy sampling

• reduced complications (including sepsis and bleeding) – as fewer men are biopsied and fewer biopsies are taken in men who are biopsied.

In addition, a revised diagnostic pathway based on the findings of the PROstate Magnetic resonance Imaging Study (PROMIS) also has the potential to offer a more cost-effective use of NHS resources. This is explored in the accompanying economic evaluation.

Limitations of transrectal ultrasound-guided biopsy

• Overdetection of CNS prostate cancer: men who undergo TRUS-guided biopsy have a 25% chance of being diagnosed with prostate cancer. 17 This compares with a lifetime risk of 6–8% for symptomatic prostate cancer, illustrating the overdiagnosis of harmless cancers in men who undergo TRUS-guided biopsies (Figure 2). 18

• Underdetection of CS prostate cancer: TRUS-guided biopsies have an estimated false-negative rate of 30–45%, although the true false-negative rate is unknown as few studies have applied a detailed reference test to those men who have a negative TRUS-guided biopsy. 19,20 Although the clinician takes 10–12 biopsies in a manner that attempts to obtain representative tissue within the peripheral zone (Figure 3, left-hand image), several parts of the gland are not well sampled (a systematic error). The anterior part of the gland may be missed as a result of its greater distance from the rectum (Figure 3, middle and right-hand image). Tissue in the midline is missed as a result of efforts to avoid the urethra, whereas the apex of the prostate is often difficult to access by the transrectal route.

• Inaccurate risk stratification: TRUS-guided biopsies can be unrepresentative of the true burden of cancer as a result of random sampling error (Figure 4). Either the size or the grade of cancer may be underestimated if the cancer tissue obtained in a TRUS-guided biopsy is not representative. 21 Figure 4 illustrates how accurate estimation of tumour size will depend on hitting the centre of a lesion. At present, because these lesions are not visualised, this relies purely on chance. However, improved risk stratification is likely if MRI results can be used to guide deployment of needles in biopsies.

  Equally, the pathological status derived from TRUS-guided biopsies can be unreliable if the test is reapplied, for instance in active surveillance, not only in discriminating clinically important cancer from clinically unimportant prostate cancer but also in attributing a non-cancer status from a cancer status, in approximately one-quarter of men subject to serial testing. 22

• Side effects: TRUS-guided biopsy is associated with a number of complications, the most important being urinary tract infection (in 1–8% of cases), which can result in life-threatening sepsis (in 1–4% of cases). Haematuria (in 50% of cases), haematospermia (in 30% of cases), pain/discomfort (in most cases), dysuria (in most cases) and urinary retention (in 1% of cases) can also be expected. 23–27

FIGURE 2.

Overdetection of CNS prostate cancer.

FIGURE 3.

Underdetection of CS prostate cancer.

FIGURE 4.

Inaccurate risk stratification.

Multiparametric magnetic resonance imaging in diagnosing prostate cancer

The available evidence suggests that MRI can achieve both sensitivity and specificity of between 70% and 90% for the detection of CS prostate cancer. 28 However, a systematic review of the literature29 found the quality of the initial studies evaluating MRI to be disappointing. 30 They repeatedly showed low sensitivity and specificity as well as high interobserver variability, even when using high-resolution endorectal MRI. 31–37 Much has changed since these early reports, including an appreciation of the impact of post-biopsy changes on MRI; technological improvements such as increasing magnetic field strength (from 0.5 T to 1.5 T and 3.0 T); shorter pulse sequences enabling faster image acquisition; and the introduction of functional imaging in the form of diffusion weighting (DW) and dynamic contrast enhancement (DCE).

The main types of magnetic resonance images available are those produced by T2 weighting, DW and DCE. Multiparametric approaches (mpMRI, combining these three sequences together) have also been investigated. 38–48 Although their samples are small, single-centre case series have found an advantage of using two or three magnetic resonance sequences rather than just one. A recent systematic review49 incorporating studies that reported after PROMIS started has demonstrated the following ranges in accuracy metrics in studies using mapping biopsies and radical whole-mount prostatectomy specimens (often in retrospective case series): sensitivity, 73–95%; specificity, 28–87%; positive predictive value (PPV), 34–93%; and negative predictive value (NPV), 3–98%. However, no studies prior to PROMIS have prospectively evaluated the clinical validity of mpMRI in the population of interest (men at risk of harbouring prostate cancer who have not had a prior prostate biopsy) against an accurate and appropriate reference standard within a multicentre study in a double-blinded fashion.

Limitations of the magnetic resonance imaging literature

There are important limitations of previous studies investigating the diagnostic accuracy of MRI for prostate cancer:

• Biopsy artefact: studies mostly evaluate MRI after biopsy; however, the biopsy procedure can affect what is seen on MRI, which can result in an increase in false-positive or false-negative results.

• Limited application: studies mostly evaluate only the peripheral zone of the prostate, ignoring up to one-third of prostate cancers.

• Segmentation: when each region of interest is segmented to achieve a sufficient number of data sets, increasing the power of the analysis and accuracy by incorrectly treating each region of interest as independent.

• Poor reference standard: most studies use RP, leading to selection bias as those undergoing surgery tend to have burdens of cancer that are distinct from men with an abnormal PSA level, and patients choosing other treatments can never be evaluated. 50 Co-registration of a magnetic resonance image to a RP specimen is challenging because of shrinkage (10–20%), distortion, tissue loss as a result of ‘trimming’ (10%), orientation and absent perfusion.

These deficiencies, which mean that there is a discernible lack of level 1 evidence on the accuracy of mpMRI, probably account for the limited acceptance of MRI in contemporary prostate cancer diagnostic pathways. 51

The optimal reference standard: template prostate mapping

Template prostate mapping (TPM) biopsy was selected as the reference test for this study as it meets the required specification for our defined population when using 5-mm sampling52 (Figure 5). TPM-biopsy:

• Produces a histological map of the entire prostate in three dimensions. 52–55

• Has an estimated sensitivity and a NPV in the order of 95% for CS cancers when assessed against RP. 56–58

• Avoids selection bias because all men exposed to the index test can be exposed to the reference standard.

• Has a similar side-effect profile to that of TRUS-guided biopsy, with three important differences. TPM-biopsy:

  • carries a significantly lower risk of urosepsis (< 0.5%), as the needles do not traverse the rectal mucosa23 (urosepsis is the most serious complication of TRUS-guided biopsy)

  • confers a higher risk of self-limiting failure to void urine (5%) as a result of greater gland swelling; the risk is 5% compared with 1–2% associated with TRUS-guided biopsy55,59,60

  • requires general or spinal anaesthesia.

FIGURE 5.

Illustration of a TPM-biopsy procedure. Reproduced from El-Shater Bosaily et al. 16 © 2015 The Authors. Published by Elsevier Inc. This is an Open Access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.

Reproduced from El-Shater Bosaily et al. 16 © 2015 The Authors. Published by Elsevier Inc. This is an Open Access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.

Although the accuracy of TPM-biopsy is high in the diagnosis of prostate cancer, it is not currently recommended as standard practice as it requires general anaesthesia and theatre time and so cannot be used routinely in the tens of thousands of men undergoing TRUS-guided biopsy. 61

Template prostate mapping biopsy is carried out in accordance with a set protocol and so can be conducted blind to any clinical or imaging data. Mapping using 5-mm sampling is obtained using core needles inserted via a brachytherapy grid fixed on a stepper placed against the perineum with the patient under general anaesthesia in the lithotomy position. In most prostates, two biopsies at each grid point are required to sample the full craniocaudal gland length.

Aims and objectives

The purpose of PROMIS was to assess the clinical validity (sensitivity, specificity, PPV and NPV) of mpMRI for the detection of CS prostate cancer and compare these accuracy metrics with those of TRUS-guided biopsy.

In particular, PROMIS aimed to evaluate whether or not mpMRI improves the ability to detect, as well as rule out, CS prostate cancer in a group of men at risk of harbouring prostate cancer, who have not had a previous biopsy and who would normally be advised to undergo a prostate biopsy as part of standard care. PROMIS was designed to determine whether or not it is appropriate for men with a risk factor for harbouring CS cancer to first undergo mpMRI to select who should have a prostate biopsy and who might safely forgo a first prostate biopsy. In other words, we sought to determine whether or not mpMRI could be used as a triage test prior to a first biopsy (see Figure 1). 62

The main objectives of the trial were to:

• Assess the ability of mpMRI to identify men who can safely avoid unnecessary biopsy.

• Assess the ability of the mpMRI-based pathway to improve the rate of detection of CS cancers compared with TRUS-guided biopsy, by evaluating the diagnostic accuracy of both mpMRI (the index test) and TRUS-guided biopsy (standard test) against an accurate reference standard, TPM-biopsy.

• Estimate the cost-effectiveness of a mpMRI-based diagnostic pathway. Using data from the main study and the wider literature, the study considered the implications of alternative diagnostic strategies for NHS costs and men’s quality-adjusted survival duration.

Study design

This chapter is partially reproduced from El-Shater Bosaily et al. 16 © 2015 The Authors. Published by Elsevier Inc. This is an Open Access article distributed in accordance with the terms of the Creative Commons Attribution (CC BY 4.0) license, which permits others to distribute, remix, adapt and build upon this work, for commercial use, provided the original work is properly cited. See: http://creativecommons.org/licenses/by/4.0/. The protocol is also available at www.ctu.mrc.ac.uk/research/documents/cancer_protocols/promis_protocol_v4 (accessed 12 December 2017).

The PROMIS study was a prospective validating paired-cohort study representing level 1b evidence for diagnostic studies. 63 The population of interest was men at risk of prostate cancer who are usually recommended to undergo a first prostate biopsy within standard care. The study was conducted at 11 NHS hospitals in England. To compare the diagnostic accuracy of mpMRI (the index test) and TRUS-guided biopsy (the current standard), both must be individually compared with a reference standard, TPM-biopsy. Therefore, all participants in the study underwent all three tests (mpMRI, TPM-biopsy and TRUS-guided biopsy), with TPM-biopsy followed by TRUS-guided biopsy carried out as a combined biopsy procedure (CBP). The trial schema is shown in Figure 6. Each test was conducted blind to all of the other test results and reported independently of the other tests.

FIGURE 6.

Trial schema. Reproduced from El-Shater Bosaily et al. 16 © 2015 The Authors. Published by Elsevier Inc. This is an Open Access article distributed in accordance with the terms of the Creative Commons Attribution (CC BY 4.0) license, which permits others to distribute, remix, adapt and build upon this work, for commercial use, provided the original work is properly cited. See: http://creativecommons.org/licenses/by/4.0/.

Reproduced from El-Shater Bosaily et al. 16 © 2015 The Authors. Published by Elsevier Inc. This is an Open Access article distributed in accordance with the terms of the Creative Commons Attribution (CC BY 4.0) license, which permits others to distribute, remix, adapt and build upon this work, for commercial use, provided the original work is properly cited. See: http://creativecommons.org/licenses/by/4.0/.

A patient and public representative (Robert Oldroyd) was involved in the study, in all aspects relating to the design, the writing of the patient information and consent process, the conduct throughout the study as a member of the Trial Management Group, the critical review of the results and the dissemination of the results to both clinician and patient audiences. He contributed to the Trial Management Group deliberations on the study concept and design for the duration of the trial, at both face-to-face meetings and during teleconferences. As a lay member of the Nottingham 1 Research Ethics Committee (REC) from 2006 until 2015, he contributed expertise on ethics issues, including elements of the project proposal presented to the London–Hampstead REC, which he attended with Mark Emberton in March 2011. As a former prostate cancer patient, he voiced initial concerns about the patient volunteer burden in the study. He rewrote the patient information sheet (reducing the initial 28 pages to 14 pages) to make it more acceptable to the REC and revised the consent form. He commented fully on aspects of the study report, clarifying some sections to make them more accessible to lay readers. He spoke about PROMIS at the Nottingham Prostate Cancer UK conference in March 2017.

PROMIS was designed to overcome several shortcomings that are associated with other studies reported in the literature. These shortcomings prompted the following decisions to be made around the design of PROMIS:

• All patients were biopsy naive and underwent mpMRI prior to any biopsy, so there was no biopsy artefact.

• mpMRI was evaluated for all anatomical zones of the prostate including peripheral and transition zones.

• The study was powered so that the primary outcome was derived using the whole prostate as the sector of analysis rather than segmented sectors of the prostate.

• The study used an accurate reference test that can be applied to all men at risk.

The study was also designed to avoid or minimise a number of biases that are inherent in the current literature:

• Spectrum and selection biases were avoided by recruiting men at risk of prostate cancer and applying all tests to all of the men.

• Work-up bias was eliminated by ensuring that patients and clinicians remained blinded to all imaging test results until the biopsies were carried out and reported.

• Reviewer/reporter bias was avoided by ensuring that the radiologist was blinded to the reference test and the pathologist was blinded to the imaging. The radiology report was submitted prior to the biopsies.

• Incorporation bias was minimised by ensuring that TPM-biopsies and TRUS-guided biopsies followed a standard accepted protocol.

Patient eligibility

Patients were eligible for registration into the study if they fulfilled all of the inclusion criteria and none of the exclusion criteria (Box 1). Men with a PSA level of > 15 ng/ml were excluded as the higher prevalence of prostate cancer in this subgroup means that mpMRI is unlikely to be used as a triage test in these men. Men whose prostate volume on mpMRI was found to be > 100 ml were withdrawn from the study. Men were required to give written informed consent before participating.

Ethics considerations

The study was carried out in accordance with the principles of the Declaration of Helsinki64 and the UK Research Governance Framework version 265 and received UK REC approval on 16 March 2011 from the London–Hampstead REC. PROMIS has been registered on ClinicalTrials.gov (NCT01292291).

Test procedures

The index test: multiparametric magnetic resonance imaging

In PROMIS, mpMRI was standardised to the minimal requirements advised by a European consensus meeting,66 the European Society of Urogenital Radiology67 and the British Society of Urogenital Radiology guidelines. 68 T1-weighted, T2-weighted, diffusion-weighted [apparent diffusion coefficient (ADC) maps and long-b scan] and dynamic gadolinium contrast-enhanced imaging was acquired using a 1.5-T scanner and a pelvic phased array (Table 3).

TABLE 3 - Multiparametric magnetic resonance imaging scan specification

VIBE, volumetric interpolated breath-hold examination.

Note

Reproduced from El-Shater Bosaily et al. 16 © 2015 The Authors. Published by Elsevier Inc. This is an Open Access article distributed in accordance with the terms of the Creative Commons Attribution (CC BY 4.0) license, which permits others to distribute, remix, adapt and build upon this work, for commercial use, provided the original work is properly cited. See: http://creativecommons.org/licenses/by/4.0/.

	TR	TE	Flip angle, °	Plane	Slice thickness [mm (gap)]	Matrix size	Field of view (mm)	Time from scan
T2 TSE	5170	92	180	Axial, coronal, sagittal	3 (10%)	256 × 256	180 × 180	3 minutes 54 seconds (ax)
VIBE at multiple flip angles for T1 calculation (optional)
VIBE fat sat	5.61	2.52	15	Axial	3	192 × 192	260 × 260	Continue for ≥ 5 minutes 30 seconds after contrast
Diffusion (b = 0, 150, 500, 1000)	2200	Minimum (< 98)		Axial	5	172 × 172	260 × 260	5 minutes 44 seconds (16 averages)
Diffusion (b = 1400)	2200	Minimum (< 98)		Axial	5	172 × 172	320 × 320	3 minutes 39 seconds (32 averages)

Endorectal coils were not used as there is no consensus on their role in minimal scanning requirements. 66 Magnetic resonance spectroscopy was not included because evidence from a large prospective multicentre study at the time showed no benefit of spectroscopy for localisation of prostate cancer compared with T2-weighted imaging alone. 69 We decided to use only 1.5-T scanners as these were more widely available in the UK health-care setting and most studies in the literature had reported the accuracy of mpMRI results based on 1.5-T scanners alone.

Quality assurance and reporting

A robust quality control process was used to maintain the quality of scans and ensure uniformity across all centres. All MRI scanners and individual mpMRI scans underwent quality control checks by an independent commercial imaging clinical research organisation appointed through open tender (IXICO plc, London, UK). Prior to site initiation, the lead radiologist (Alexander Kirkham) reviewed a number of prostate MRI scans from each centre and gave iterative feedback on improving scan quality.

During the study, scans deemed to be of insufficient quality were repeated prior to biopsy. In addition, a standardised operating procedure for mpMRI reporting was adopted in line with the recommendations of the European consensus meeting66 and the European Society of Urogenital-Radiology prostate MRI guidelines. 67 This was convened before publication of the more recent Prostate Imaging Reporting and Data System (PI-RADS) mpMRI reporting consensus. 67 Subsequent comparisons of the Likert and PI-RADS reporting schemes have yielded similar results. 70,71

At each centre, mpMRI scans were reported by dedicated urological radiologists who had undergone centralised training provided by the lead centre [University College London Hospital (UCLH)]. Radiologists were provided with clinical details including PSA levels, DRE findings and any other risk factors such as a family history of prostate cancer.

Images were reported in sequence, with T2-weighted images reported first, T2-weighted and diffusion-weighted images reported together and then a third report issued for T2-weighted with diffusion-weighted and dynamic contrast-enhanced scans together. The reporting form is shown in Figure 7. Future analyses will investigate whether or not both diffusion-weighted and dynamic contrast-enhanced images are required. As DCE requires a contrast agent (with its need for intravenous access, medical supervision and contrast-related risks) and an additional 10–15 minutes of scan time, it will be useful to determine whether or not this additional resource use and cost is necessary.

FIGURE 7.

Multiparametric magnetic resonance imaging reporting form.

A five-point Likert scoring system66,67,72 was used to indicate the probability of cancer (1, highly likely to be benign; 2, likely to be benign; 3, equivocal; 4, likely to be malignant; 5, highly likely to be malignant). The prostate was divided into 12 regions of interest and each region was scored from 1 to 5. In addition, each lesion was identified and scored separately and the longest axial diameter, lesion volume, ADC value and contrast enhancement curve type were recorded. 73–76 From these observations, an overall score of 1–5 (using the same definitions) was assigned to the whole prostate. This was carried out for ‘any cancer’ and for definitions 1 and 2 of CS cancer (see Definitions of clinically significant prostate cancer).

For the primary outcome, an overall score of ≥ 3 was used to indicate a suspicious scan in relation to CS cancer (i.e. a positive mpMRI score). This reflects the level at which further tests (e.g. biopsy) would be considered if mpMRI were to be introduced into the diagnostic pathway in the future. To assess interobserver agreement, 132 scans from the lead site were rereported by a blinded second radiologist based at that site. The values used in the analyses were those from the original reporter.

The results of the mpMRI could be unblinded by the radiologist if the mpMRI revealed an enlarged prostate volume of > 100 ml, which would be impossible to sample every 5 mm because of the interference of the bony pelvic arch. This might have reduced the number of ‘negative’ prostate results in the final analyses. The mpMRI could also be unblinded if there was evidence of stage T4 prostate cancer or involved lymph nodes or colorectal/bladder invasion. This might have had a detrimental impact on the performance characteristic of mpMRI, as such tumours were more likely to be detected. The presence of other cancers such as bladder or colorectal cancers was also a criterion for withdrawal. Withdrawal was deemed appropriate in these men as expedited referral for biopsy and treatment was required. These withdrawals were unlikely to have an impact on the primary or secondary outcomes.

Definitions of clinically significant prostate cancer

Disease significance was defined using criteria that have been previously developed and validated for use with TPM-biopsy for the detection of primary Gleason grades of ≥ 478 and of a cancer core length that is predictive of the presence of lesions of ≥ 0.5 ml. 79–82 Gleason scoring was based on the most frequent pattern and not on the highest grade detected on histological analysis.

The primary definition of CS prostate cancer used a histological target condition on TPM-biopsy that incorporated the presence of a Gleason score of ≥ 4 + 3 and/or a cancer core length of ≥ 6 mm of any Gleason score, in any location (definition 1). This was chosen as the primary outcome on the basis that few physicians would disagree that any man with this burden of cancer would require treatment. A secondary definition of CS disease was also used (a cancer core length of ≥ 4 mm and/or a Gleason score of ≥ 3 + 4; definition 2). A further definition of clinical significance (any Gleason pattern of ≥ 7) was added at the request of reviewers at The Lancet and this was treated as an exploratory definition (definition 3).

Monitoring of adverse events

The expected side effects of combining the TPM-biopsy and TRUS-guided biopsy procedures were discussed with patients before registration. The rate of serious adverse events (SAEs) was monitored weekly by an independent Trial Steering Committee.

Outcome measures

The primary outcome measures were the diagnostic accuracy of mpMRI, TRUS-guided biopsy and TPM-biopsy using the primary mpMRI and histology definitions, as measured by sensitivity, specificity, PPV and NPV. The same parameters using combinations of alternative definitions of disease for each test were investigated as secondary outcomes. The other outcomes reported were inter-rater agreement between tests and adverse events (AEs).

Translational research objectives

PROMIS was an ideal setting for assessing the utility of biomarkers (from urine and blood) to identify men with CS prostate cancer. It was, to our knowledge, the first time that a broad spectrum of men at risk have been evaluated using an optimal biopsy technique that accurately characterises the presence, size and grade of prostate cancer. A comprehensive bank of tissue samples (serum, plasma, germline DNA, urine) was collected from men prior to biopsy, to enable urinary and serum biomarkers to be analysed with respect to the detection of CS prostate cancer on TPM-biopsy. The results of these analyses will be reported at a later date.

Sample size

Power calculations were performed in relation to (1) the precision of the estimates for the accuracy of mpMRI relative to TPM-biopsy in terms of the primary definition of CS cancer, (2) a head-to-head comparison of mpMRI with TRUS-guided biopsy and (3) an assumed underlying prevalence of CS cancer of 15% according to the primary definition. 19,20,83,84 For the less stringent definition (definition 2), it was assumed that 25% of participants would have CS prostate cancer as detected by the reference standard.

The largest sample size obtained from the power calculations around (1) and (2) in the previous paragraph was 714 (see protocol16); this was considered to be the maximum number of men required to have all three tests (mpMRI, TPM-biopsy and TRUS-guided biopsy), based on the assumption that mpMRI and TRUS-guided biopsy are uncorrelated.

Assuming a specificity of 77%, in order to demonstrate that the lower 95% CI is ≥ 70%, we would require 407 cases of CNS prostate cancer. This is equivalent to a total of 479 men for definition 1 and 543 men for definition 2. Assuming a sensitivity of 75%, in order to demonstrate that the lower 95% CI for sensitivity is ≥ 60%, we would require 97 cases of CS prostate cancer. This is equivalent to a total of 647 men for definition 1 and 388 for definition 2. These estimates of sensitivity and specificity were considered realistic based on current unpublished and published literature. 85,86

It was assumed that TRUS-guided biopsy detects 48% of CS prostate cancers83,87 and that mpMRI would detect ≥ 70%; these were conservative estimates. Using McNemar’s test for paired binary observations,88 in order to show an absolute increase in the proportion of CS cancers detected of ≥ 22% (from 48% to 70%) with a power of 90% and a two-sided alpha of 5%, a total of 107 cases are required. This is equivalent to a total study population of 714 men for definition 1 and 428 men for definition 2. Table 4 shows the required sample size for McNemar’s test for different levels of agreement between mpMRI and TRUS-guided biopsy. The shaded regions reflect the scenario in which virtually all cancers are detected by either mpMRI or TRUS-guided biopsy, in which there is extremely low agreement between mpMRI and TRUS-guided biopsy. This is very unlikely but is included for completeness.

The independent Trial Steering Committee carried out an a priori interim review after 50 participants had received all three tests. Although a higher than anticipated prevalence of any cancer was observed, no changes to the target sample size were recommended.

Statistical analysis

All statistical analyses were carried out in accordance with a statistical analysis plan agreed before the data were inspected. Stata^® version 13.0 (StataCorp LP, College Station, TX, USA) was used to carry out the analyses. The primary analysis was based on all evaluable data, excluding men without all three test results and any data that were rejected as part of the external mpMRI quality control/quality assurance process.

For each comparison, 2 × 2 contingency tables were used to present the results and calculate the diagnostic accuracy estimates with 95% CIs. Given the paired nature of the test results, McNemar’s tests were used for the head-to-head comparisons of sensitivity and specificity between mpMRI and TRUS-guided biopsy. Because the PPV and NPV are dependent on disease prevalence, a generalised estimating equation (GEE) logistic regression model was used to compare the PPV and NPV for mpMRI and TRUS-guided biopsy against those for TPM-biopsy. 89,90

The sensitivities, specificities and predictive values were calculated for mpMRI based on the overall radiological score for mpMRI [assigned to definition 2 on the mpMRI case report form (CRF)] and definition 1 for CS cancer on TPM-guided biopsy.

The format of the 2 × 2 table is shown in Table 5. Specificity = d/(c + d), where d = the number of men testing negative on mpMRI and negative for CS cancer on TPM-biopsy and c = the number of men testing positive on mpMRI and negative for CS cancer on TPM-biopsy. The NPV = d/(b + d), where d = the number of men testing negative on mpMRI and negative for CS cancer on TPM-biopsy and b = the number of men testing negative on mpMRI who test positive for CS cancer on TPM-biopsy. Sensitivity = a/(a + b), where a = the number of men testing positive on mpMRI and positive for CS on TPM-biopsy and b = the number of men testing negative on mpMRI who have CS cancer on TPM-biopsy. The PPV = a/(a + c), where a = the number of men testing positive on mpMRI and positive for CS on TPM-biopsy and c = the number of men testing positive on mpMRI who do not have CS cancer on TPM-biopsy.

For the comparison of TRUS-guided biopsy and mpMRI, McNemar’s test was used to compare the agreement between mpMRI (radiological score of ≥ 3 assigned to definition 2 on the mpMRI CRF) and TRUS-guided biopsy (definition 1) in the subset of men found to have CS prostate cancer on TPM-biopsy according to definition 1. For the secondary analysis, all analyses that were carried out for definition 1 were repeated for definition 2 and, at the request of The Lancet reviewers, for any Gleason score of ≥ 7.

Exploratory analyses will be performed in the future to evaluate the impact on the sensitivity, specificity and predictive values of mpMRI when each of the 12 regions of interest are correlated between mpMRI and TPM-biopsy. Agreement between mpMRI and TPM-biopsy in identifying CS cancer in the same region will be based on a nearest neighbourhood approach. 91 Sensitivity to this approach will be tested by also presenting results according to complete match (the most stringent rule) and using a left/right rule (the less stringent rule). The sensitivity, specificity and predictive values of each of the individual MRI reporting sequence combinations, namely T2, T2 + DW and T2 + DW + DCE, will inform additional important tertiary analyses in subsequent publications.

Odds ratios (ORs) represent the odds of each test correctly detecting the presence or absence of disease. For specificity and NPV, the coding logic was reversed as the correct test result is a negative test result. Ratios are presented for TRUS-guided biopsy relative to mpMRI, so ratios of > 1 favour TRUS-guided biopsy and ratios of < 1 favour mpMRI.

Recruitment and screening

A total of 740 participants were registered to the trial from 11 NHS hospitals in England. Recruitment took place between May 2012 and November 2015. Details of the screening and participating centres are given in Appendix 1.

Flow of participants through the study

Figure 8 shows the flow of participants through the study. Of the 740 men registered, 164 subsequently withdrew from the study before completing all three tests. Reasons for withdrawal are presented in Table 6. Most withdrawals took place before the combined biopsy was carried out; the most common reason for withdrawal was the discovery of a large prostate volume (> 100 ml), which was a mandatory withdrawal criterion. A total of 576 men were included in the final analysis. For the analysis set, the median (range) time between mpMRI and the combined biopsy was 38 days (1–190 days). The median (range) time between registration and the end-of-study visit was 111 days (31–421 days).

FIGURE 8.

Participant flow through the study. Reproduced from Ahmed et al. 92 © 2015 The Authors. Published by Elsevier Inc. This is an Open Access article distributed in accordance with the terms of the Creative Commons Attribution (CC BY 4.0) license, which permits others to distribute, remix, adapt and build upon this work, for commercial use, provided the original work is properly cited. See: http://creativecommons.org/licenses/by/4.0/.

Reproduced from Ahmed et al. 92 © 2015 The Authors. Published by Elsevier Inc. This is an Open Access article distributed in accordance with the terms of the Creative Commons Attribution (CC BY 4.0) license, which permits others to distribute, remix, adapt and build upon this work, for commercial use, provided the original work is properly cited. See: http://creativecommons.org/licenses/by/4.0/

Baseline characteristics

Table 7 summarises the baseline characteristics for the 576 men who were included in the final analysis and the 164 men who withdrew after registration. The group of participants who withdrew had a similar age and PSA level to the included group, but were slightly less likely to report a family history of prostate cancer.

Cancer prevalence by template prostate mapping biopsy

The prevalence of any cancer as assessed by the reference test, TPM-biopsy, was 71% (95% CI 67% to 75%). The prevalence of CS cancer according to the primary definition (a Gleason score of ≥ 4 + 3 and/or a cancer core length of ≥ 6 mm) was 40% (95% CI 36% to 44%). The prevalence of CS cancer according to the secondary definition (a Gleason score of ≥ 3 + 4 and/or a cancer core length of ≥ 4 mm) was 57% (95% CI 53% to 62%). Using the exploratory definition of any Gleason pattern of ≥ 7, the prevalence of CS cancer was 53% (308/576).

Cancer prevalence according to each definition as detected by TPM-biopsy is summarised in Table 8, together with the number of cancers that were identified to have perineural or lymphovascular involvement. A summary of the cancer grades according to TPM-biopsy pathology results is given in Appendix 1.

Diagnostic accuracy of transrectal ultrasound-guided biopsy compared with template prostate mapping biopsy (primary outcome)

The diagnostic accuracy results for TRUS-guided biopsy compared with TPM-biopsy are shown in Table 9 and Figure 9. Cancer was detected in 286 out of 576 men (50%, 95% CI 45% to 54%), of whom 65 had perineural invasion and one had lymphovascular invasion. CS cancer according to the primary definition was detected in 124 out of 576 men (22%, 95% CI 18% to 25%), of whom 48 had perineural invasion and one had lymphovascular invasion.

FIGURE 9.

Diagnostic accuracy of the detection of CS cancer for TRUS-guided biopsy compared with TPM-biopsy.

A summary of the cancer grades according to TRUS-guided biopsy pathology results is given in Appendix 1. Appendix 1 also provides two case illustrations showing the correlation of TRUS-guided biopsy and mpMRI with TPM-biopsy, one for a man harbouring CS cancer and one for a man with no cancer.

Diagnostic accuracy of multiparametric magnetic resonance imaging compared with template prostate mapping biopsy (primary outcome)

Multiparametric magnetic resonance imaging results

In the group of 576 men included in the final analysis, the mean volume of the prostate was 48 ml [standard deviation (SD) 20 ml]. Six men had a prostate volume of > 100 ml as they had been entered into the trial before this threshold was adopted as an exclusion criterion mid-way through the study. The Trial Management Group decided that these six men should remain in the study as they had been fully sampled using TPM-biopsy.

The distribution of mpMRI scores is presented in Table 10. Agreement between mpMRI and TRUS-guided biopsy for the detection of CS cancer is shown in Appendix 1.

TABLE 10 - Distribution of MRI scores across all 576 men

Primary MRI score	Participants, n (%)
1: highly likely benign	23 (4)
2: likely benign	135 (23)
3: equivocal	163 (29)
4: likely malignant	120 (21)
5: highly likely malignant	135 (23)
Total	576 (100)

The diagnostic results for mpMRI compared with TPM-biopsy are shown in Table 11 and Figure 10. The sensitivity of mpMRI for CS cancer was 93% (95% CI 88% to 96%) and the NPV was 89% (95% CI 83% to 94%). The specificity of mpMRI was 41% (95% CI 36% to 46%), with PPV being 51% (95% CI 46% to 56%). The proportion of CS disease by mpMRI score is shown in Figure 11.

TABLE 11 - Diagnostic accuracy of the detection of CS cancer for mpMRI compared with TPM-biopsy

CNS, clinically non-significant; NC, no cancer.

Notes

Sensitivity = 93% (95% CI 88% to 96%); PPV = 51% (95% CI 46% to 56%).

Specificity = 41% (95% CI 36% to 46%); NPV = 89% (95% CI 83% to 94%).

Reproduced from Ahmed et al. 92 © 2015 The Authors. Published by Elsevier Inc. This is an Open Access article distributed in accordance with the terms of the Creative Commons Attribution (CC BY 4.0) license, which permits others to distribute, remix, adapt and build upon this work, for commercial use, provided the original work is properly cited. See: http://creativecommons.org/licenses/by/4.0/.

mpMRI	TPM-biopsy, n
	CS cancer (+)	NC or CNS cancer (–)	Total
CS cancer (+)	213	205	418
NC or CNS cancer (–)	17	141	158
Total	230	346	576

FIGURE 10.

Diagnostic accuracy of the detection of CS cancer for mpMRI compared with TPM-biopsy. Pie charts represent actual mpMRI scores from 1 to 5. Reproduced from Ahmed et al. 92 © The Authors. Published by Elsevier Inc. This is an Open Access article distributed in accordance with the terms of the Creative Commons Attribution (CC BY 4.0) license, which permits others to distribute, remix, adapt and build upon this work, for commercial use, provided the original work is properly cited. See: http://creativecommons.org/licenses/by/4.0/.

Reproduced from Ahmed et al. 92 © The Authors. Published by Elsevier Inc. This is an Open Access article distributed in accordance with the terms of the Creative Commons Attribution (CC BY 4.0) license, which permits others to distribute, remix, adapt and build upon this work, for commercial use, provided the original work is properly cited. See: http://creativecommons.org/licenses/by/4.0/

FIGURE 11.

Proportion of men with no cancer, insignificant cancer and significant cancer based on TPM-biopsy by mpMRI score. Reproduced from Ahmed et al. 92 © 2015 The Authors. Published by Elsevier Inc. This is an Open Access article distributed in accordance with the terms of the Creative Commons Attribution (CC BY 4.0) license, which permits others to distribute, remix, adapt and build upon this work, for commercial use, provided the original work is properly cited. See: http://creativecommons.org/licenses/by/4.0/.

Reproduced from Ahmed et al. 92 © 2015 The Authors. Published by Elsevier Inc. This is an Open Access article distributed in accordance with the terms of the Creative Commons Attribution (CC BY 4.0) license, which permits others to distribute, remix, adapt and build upon this work, for commercial use, provided the original work is properly cited. See: http://creativecommons.org/licenses/by/4.0/.

A negative mpMRI result was recorded for 158 men (27%). Of these, 17 were found to have CS cancer on TPM-biopsy (see Table 15 and Figure 10). All 17 had Gleason grade 3 + ≤ 4, and core lengths of between 6 mm and 12 mm (Table 12). Of the 119 significant cancers missed by TRUS-guided biopsy, 13 had a Gleason score of 4 + 3, 99 had a Gleason score of 3 + 4 and 7 had a Gleason score of 3 + 3 (see Table 12).

TABLE 12 - Histological characteristics on TPM-biopsy of CS cases missed by mpMRI and TRUS-guided biopsy

Note

Reproduced from Ahmed et al. 92 © 2015 The Authors. Published by Elsevier Inc. This is an Open Access article distributed in accordance with the terms of the Creative Commons Attribution (CC BY 4.0) license, which permits others to distribute, remix, adapt and build upon this work, for commercial use, provided the original work is properly cited. See: http://creativecommons.org/licenses/by/4.0/.

Gleason grade	Number of cases missed (range of maximum cancer core length, mm), by trial arm
	mpMRI (N = 17)	TRUS-guided biopsy (N = 119)
3 + 3	1 (8)	7 (6–11)
3 + 4	16 (6–12)	99 (6–14)
4 + 3	0	13 (3–16)

Head-to-head comparison of the diagnostic accuracy of multiparametric magnetic resonance imaging and transrectal ultrasound-guided biopsy compared with template prostate mapping biopsy (primary outcome)

Multiparametric magnetic resonance imaging was more accurate than TRUS-guided biopsy in terms of both sensitivity (93% compared with 48%, respectively; McNemar’s test ratio 0.52, 95% CI 0.45 to 0.60) and NPV (89% compared with 74%, respectively; GEE model estimate for OR 0.34, 95% CI 0.21 to 0.5; p < 0.0001). A total of 119 significant cancers were missed by TRUS-guided biopsy, of which 13 had a Gleason score of 4 + 3, 99 had a Gleason score of 3 + 4 and 7 had a Gleason score of 3 + 3 (see Table 12). However, compared with mpMRI, TRUS-guided biopsy showed better specificity (41% compared with 96%, respectively; McNemar’s test ratio 2.34, 95% CI 2.08 to 2.68; p < 0.0001) and a better PPV (51% compared with 90%; GEE model estimate for OR 8.2, 95% CI 4.7 to 14.3; p < 0.0001].

Diagnostic accuracy of transrectal ultrasound-guided biopsy and multiparametric magnetic resonance imaging for other definitions of clinically significant cancer

In total, 203 out of 576 cases (35%, 95% CI 31% to 39%) had CS cancer on TRUS-guided biopsy according to the secondary definition [University College London (UCL) definition 2: Gleason grade of ≥ 3 + 4 and/or any grade of cancer length of ≥ 4 mm]. Table 14 presents the definitions and prevalence according to TPM-biopsy. Of these 203 cases with UCL definition 2 disease, 63 had perineural invasion and one had lymphovascular invasion. For the exploratory definition (any Gleason pattern of ≥ 7), 151 out of 576 cases (26%, 95% CI 23% to 30%) had CS disease. Of these 151 cases with disease with a Gleason grade of ≥ 7, 54 had perineural invasion and one had lymphovascular invasion.

Diagnostic accuracy results for the secondary and exploratory definitions of CS cancer are shown in Table 14. Table 15 shows the histological characteristics of cancers missed by TRUS-guided biopsy and mpMRI for all histological definitions.

Interobserver and intraobserver variability in the reporting of multiparametric magnetic resonance imaging scores

Blinded, double reporting was available for the mpMRI scans of 132 men. For this group, agreement between radiologists in the detection of CS cancer according to the primary definition was 80% (Table 16). This corresponded to a Cohen’s kappa statistic of 0.5 [moderate agreement according to the Landis and Koch classification,95 in which agreement is graded as excellent (κ ≥ 0.80), good (κ = 0.60–0.79), moderate (κ = 0.40–0.59), poor (κ = 0.20–0.39) or very poor (κ < 0.20)]. Cohen’s kappa statistics indicate how much better the agreement is over the agreement that would have occurred by chance (the expected agreement). We propose to carry out further analyses on interobserver and intraobserver variability and all of the mpMRI scans have been archived for future training and quality assurance purposes.

Agreement between radiologists for prostate volume measurements by mpMRI is shown in Figure 12, using the Bland–Altman method, which plots the difference between measurements against the average of the measurements. The average prostate volume was 48 ml. Overall, the mean difference between volumes was +1.3 ml (95% CI –0.04 ml to +2.69 ml]. The plot indicates that the interobserver reproducibility between radiologists for prostate volume measurement was approximately ±15 ml. There was a tendency for poorer agreement with larger volumes and this association was significant according to Pitman’s test, as well as simple linear regression.

FIGURE 12.

Bland–Altman plot for volume measurements between two radiologists.

There was no difference in diagnostic accuracy between the lead radiology site (UCLH, responsible for training all other sites) and the other sites (see Appendix 1 for comparison).

Health-related quality of life

Participants filled in EuroQol-5 Dimensions, three-level version (EQ-5D-3L) questionnaires at enrolment, after undergoing mpMRI, after undergoing the CBP and at the end of the study, which was, on average, 42 days after the CBP. As part of the economic evaluation, the EQ-5D-3L profiles were converted into preference-based index scores using the UK tariff. 96 The index scores after each test were compared with the values at enrolment. There was no evidence of a change in index score between post mpMRI and baseline (change = 0.008, 95% CI –0.002 to 0.018). However, there was a large and statistically significant negative impact after the CBP, with a change of –0.176 (95% CI –0.203 to –0.149). The health-related quality of life (HRQoL) implications of the various diagnostic strategies are explored further in the economic evaluation (see Chapters 4–8).

Safety

This section summarises the safety information for the CBP, which involved TPM-biopsy followed by TRUS-guided biopsy under the same anaesthesia. The CBP was carried out for the purposes of the study only; it is important to note that the CBP would not be required in routine clinical practice under any protocols developed as a result of PROMIS.

Overview

The economic evaluation conducted for PROMIS is presented in four chapters. Chapter 4 states the aims and objectives of the economic evaluation and describes the approach used. Chapter 5 describes the evaluation of the short-term outcomes associated with using the different diagnostic tests to detect CS cancer. Chapter 6 describes the evaluation of the long-term outcomes associated with treatment compared with monitoring of men with prostate cancer. Chapter 7 combines the results of Chapters 5 and 6 and presents conclusions on the cost-effectiveness of the diagnostic strategies.

Introduction

Prostate cancer is the most common cancer in men, with approximately 47,000 new cases diagnosed in the UK in 2013. It is the second most common cause of death for men in the UK, with approximately 11,000 deaths per year. 1 As discussed in Chapter 1, prostate cancer is divided into risk groups based on a combination of PSA level, Gleason score (i.e. cancer grade) and stage (see Table 1). Low-risk cancer is considered to be CNS because it is unlikely to have a clinical impact during the man’s remaining lifetime. The term ‘clinically insignificant’ is also commonly used. Intermediate-risk cancer and high-risk cancer have a greater risk of progression and are considered to be CS.

Currently in the UK, all men considered to be at risk of having prostate cancer are invited to undergo a biopsy, usually of the TRUS-guided type, to identify and classify the cancer. The biopsy procedure has adverse health consequences, albeit mostly temporary, and misses a considerable proportion of CS cancers.

There are two possible ways of improving the current practice of TRUS-guided biopsy for all men considered to be at risk: (1) choose another test that is better at detecting CS cancer (better sensitivity) or (2) incorporate additional tests prior to biopsy that might inform the decision to proceed to biopsy (triage test) and/or might improve the sensitivity of biopsy. The mpMRI is in the latter category, because it may enable the selection of patients who should have a biopsy and also provide information to enable targeted biopsies of suspicious areas within the prostate. The mpMRI can be used in combination with TRUS-guided biopsy or other biopsy types to better detect CS cancer and avoid unnecessary biopsies.

The different ways of combining mpMRI and biopsy form a range of diagnostic strategies for prostate cancer. These diagnostic strategies have different costs and health outcomes, in both the short term and the long term. The short-term costs and health outcomes depend on the prices of the tests and their direct health effects and detection rates in men with and without CS cancer. The long-term costs and health outcomes relate to the downstream consequences of treatment decisions, which in turn depend on the detection rates of each test. The short- and long-term costs and health outcomes together determine the most cost-effective way in which to use mpMRI and biopsy (i.e. the cost-effective diagnostic strategy).

To inform decisions on how best to use the tests in combination, decision modelling is required to combine the data generated in PROMIS with external information. The clinical data generated in PROMIS provide information on how accurate mpMRI and biopsy are at detecting CS and CNS prostate cancer. However, the data do not provide information on the effect of using mpMRI and biopsy in combination (e.g. the proportion of CS cancers detected when using mpMRI to direct the TRUS-guided biopsy to suspicious lesions). External information is also required to predict the long-term costs and health outcomes of the alternative strategies, based on the proportions who are correctly diagnosed with CS prostate cancer, that is, the quality-adjusted survival and lifetime costs of men according to their cancer status and diagnostic classification. Decision modelling structures the available evidence using a mathematical model, combining the different pieces of information to produce an estimate of the costs and health outcomes of the different diagnostic strategies. Health outcomes are typically expressed as quality-adjusted life-years (QALYs), a measure that weights life expectancy with its HRQoL.

An evaluation of the effectiveness of the alternative strategies can rely on the health outcomes associated with each, the most effective strategy being the one conferring the most health to the potential population of users. To ascertain cost-effectiveness, however, the consequences of the additional costs that some of the strategies impose on the NHS also need to be considered. This requires information on the health opportunity cost of diverting resources to more costly interventions. The health opportunity cost is the health benefit forgone by other patients if their interventions are no longer funded in order to release resources for other more costly diagnostic strategies. The cost-effective diagnostic strategy is the strategy that achieves the greatest health gain net of the health opportunity cost.

Previous research98 has estimated that £13,000 in additional costs displaces 1 QALY. This is the health opportunity cost for the NHS of offering a more costly intervention and is often referred to as the cost-effectiveness threshold. Therefore, the most cost-effective diagnostic strategy is the one that achieves the most QALYs net of its costs, converted into QALYs forgone using the health opportunity cost of £13,000. NICE uses a higher value, of between £20,000 and £30,000 per QALY, to make recommendations to the NHS on the cost-effectiveness of new interventions. 99 This means that an additional £20,000–30,000 is assumed to impose on the NHS a loss of 1 QALY.

Aims and objectives of the economic evaluation

The aim of the economic evaluation was to identify the most cost-effective diagnostic strategy in men with suspected localised prostate cancer from the perspective of the UK NHS, using information on TRUS-guided biopsy, mpMRI and TPM-biopsy. The evaluation makes use of information generated in PROMIS, supplemented by external evidence as appropriate, and compares the various ways that these diagnostic tests can be used to diagnose prostate cancer.

The specific objectives of the economic evaluation were to:

• define potential diagnostic strategies involving combinations of TRUS-guided biopsy, mpMRI and TPM-biopsy that could reasonably be used in clinical practice

• estimate the proportions of men correctly identified with prostate cancer, both CNS and CS, and men with cancer who have been missed, for each diagnostic strategy

• estimate the direct HRQoL and NHS costs associated with each diagnostic strategy

• evaluate the expected quality-adjusted survival and NHS costs of men depending on their diagnostic classifications, treatment decisions and true disease status

• compare the quality-adjusted survival and NHS costs of all diagnostic strategies incrementally to identify the most cost-effective strategy for a given cost-effectiveness threshold range

• identify the main drivers of cost-effectiveness and explore how they affect the results

• characterise the uncertainty around the results and identify the key areas of uncertainty, which should be prioritised for future research.

Framework of analyses

The cost-effectiveness of the different diagnostic strategies involving mpMRI, TRUS-guided biopsy and TPM-biopsy depends on two aspects. First, it depends on the diagnostic performance, direct impact on HRQoL (mainly through AEs) and costs (mainly depending on the price and level of use of each diagnostic technology) of the combination of tests forming each diagnostic strategy in correctly classifying men with their true disease status. These form the short-term outcomes of testing. Second, it depends on the long-term health benefits and costs of managing the disease, which in turn depend on the men’s true disease status and clinical pathway. These form the long-term outcomes of testing. The short-term and long-term outcomes of testing, once combined, allow the cost-effectiveness of the diagnostic strategies to be compared. For this reason, the decision model is divided into two components: (1) to evaluate short-term outcomes and (2) to evaluate long-term outcomes. The decision models were developed in Microsoft Excel^® 2013 (Microsoft Corporation, Redmond, WA, USA), with some auxiliary calculations on diagnostic accuracy carried out in Stata version 13.0. The results of the short-term and long-term models are combined to evaluate the cost-effectiveness of the different diagnostic strategies over the patients’ lifetime.

Methods

A decision tree (see Figure 13) is used to represent the diagnostic process and to evaluate the probability that men are classified as having no cancer, CNS cancer and CS cancer, conditional on their true disease status, for every given diagnostic strategy. The diagnostic strategies consist of clinically feasible combinations of mpMRI, TRUS-guided biopsy and TPM-biopsy; each are evaluated using the different cancer definitions and cut-off points.

For each diagnostic strategy, the decision tree evaluates:

• the proportion of men classified as having no cancer, CNS cancer and CS cancer, conditional on their true disease status

• the proportion of men undergoing each diagnostic test

• the direct costs of diagnosis, which consist of the costs of testing and the cost of managing the adverse effects of the tests

• the direct health outcomes of diagnosis, in terms of the HRQoL impact of the tests

• diagnostic efficiency, that is, the trade-off between the accuracy of the diagnostic strategies in detecting CS cancer and their direct costs.

Population

The target population was men with suspected prostate cancer who had been referred to secondary care for further investigation. Their true disease status was defined according to the cancer risk categories recommended by NICE13 and used in the randomised controlled study comparing expectant management with active treatment. 6 The groups are:

• men with no cancer

• men with low-risk cancer – a PSA level of ≤ 10 ng/ml and a Gleason score of ≤ 6

• men with intermediate-risk cancer – a PSA level of > 10 ng/ml and ≤ 20 and a Gleason score of 7

• men with high-risk cancer – a Gleason score of ≥ 8.

In order to inform the model inputs, the true disease status of men was ascertained using the results of TPM-biopsy and complemented by TRUS-guided biopsy if cancer had not been detected on TPM-biopsy but had been detected on TRUS-guided biopsy.

Diagnostic tests

PROMIS classifies men with cancer as having either CS cancer or CNS cancer. The definitions of CS cancer are discussed in Chapter 2 and are restated in the following section. These definitions correspond to the classification by cancer risk recommended by NICE. 13 Low-risk cancer is assumed to be equivalent to CNS cancer. Intermediate-risk and high-risk cancer form the group with CS cancer.

Diagnostic strategies

Each of the three diagnostic tests (mpMRI, TRUS-guided biopsy and TPM-biopsy) can be used in a different sequence to diagnose prostate cancer. The sequence of tests is determined by whether or not, and when, to use mpMRI; the type of biopsy (TRUS-guided or TPM); and whether or not men can be rebiopsied. Although TPM-biopsy can correctly diagnose all men, it involves the use of more health-care resources and hence has greater costs and may affect HRQoL. Therefore, although it may not be cost-effective to use TPM-biopsy in all men, it may be worthwhile using it in men selected according to their mpMRI results, TRUS-guided biopsy results or a combination of the two.

Four key assumptions are made in the design of the diagnostic strategies under evaluation:

1. The only tests considered are mpMRI, TRUS-guided biopsy and TPM-biopsy. This follows from PROMIS, which compared mpMRI and TRUS-guided biopsy with TPM-biopsy.

2. There can be up to three tests in one diagnostic strategy. Diagnostic episodes may be repeated over time but this is not explicitly modelled in this analysis.

3. A diagnostic strategy can include up to two biopsies.

4. If included in the strategy, mpMRI can be used only once.

Tables 21–24 identify the 32 diagnostic strategies compared in this work. Table 21 shows the strategies starting with mpMRI prior to TRUS-guided biopsy (M1–M7) and Table 23 shows the strategies starting with TRUS-guided biopsy prior to TRUS-guided biopsy (T1–T9). Strategies M1–M7 and T1–T9 do not include TPM-biopsy. Tables 22 and 24 show the equivalent strategies but including TPM-biopsy as the last biopsy, replacing TRUS-guided biopsy. Therefore, strategies N1–N7 start with mpMRI and include TPM-biopsy as the last biopsy in the test sequence and strategies P1–P7 start with a biopsy and include TPM-biopsy as the last biopsy in the test sequence. This means that in strategy P1, which includes only one biopsy for all men, all men receive TPM-biopsy. In strategy P2, all men receive TRUS-guided biopsy and men who were classified as having no cancer receive TPM-biopsy. Men are classified as having ‘de facto’ no cancer in strategies in which men with suspected CNS cancer on mpMRI are not retested with TRUS-guided biopsy. In other words, men with suspected CNS cancer on mpMRI are not referred for active surveillance because they have not been diagnosed with CNS cancer. Instead, they are discharged to primary care where their PSA level is monitored annually. In contrast, men in whom TRUS-guided biopsy detected CNS cancer are referred for active surveillance.

Figure 13 shows how these strategies can be translated into decision trees, using strategies M1 and T1 as examples. The black circles represent chance nodes, in which the outcome depends on a probability. In strategy M1, all men receive mpMRI. Men identified as having suspected CS cancer on mpMRI receive TRUS-guided biopsy. The TRUS-guided biopsy can classify these men as not having cancer if no cancer is detected or as having CNS or CS cancer, depending on the characteristics of the lesions. In strategy T1, men receive only one single test, TRUS-guided biopsy. The biopsy can classify men as having no cancer, CNS cancer or CS cancer.

FIGURE 13.

Decision trees. (a) Strategy M1: mpMRI in all men and TRUS-guided biopsy in men with suspected CS cancer; and (b) strategy T1: all men have TRUS-guided biopsy.

This study evaluates each of the 32 diagnostic strategies using each of the two diagnostic definitions for mpMRI, TRUS-guided biopsy and TPM-biopsy (see Tables 18 and 19) and between two and five cut-off points on the mpMRI Likert scale for suspicion of cancer. In total, the 32 different diagnostic strategies form 383 substrategies:

• In strategies M1–M7, all men undergo mpMRI, which will then indicate whether one or two TRUS-guided biopsies are required. Strategies M1–M7 are tested over two mpMRI definitions, four cut-off points and two TRUS-guided biopsy definitions, forming 112 possible substrategies.

• In strategies N1–N7, all men undergo mpMRI to inform the need for a TRUS-guided biopsy or TPM-biopsy. In strategies N1 and N2, all men receive mpMRI and, depending on the result, some men receive TPM-biopsy. Therefore, in strategies N1 and N2, two mpMRI definitions can be used, each over four cut-off points. This forms 16 substrategies. In N3–N7, the two mpMRI definitions, four cut-off points and two TRUS-guided biopsy definitions apply, forming 80 substrategies. Therefore, in total, there are 96 substrategies.

• In strategies T1–T9, all men undergo a TRUS-guided biopsy, which can be subsequently followed by mpMRI and potentially a second TRUS-guided biopsy. Strategies T1–T4 include only TRUS-guided biopsy and, therefore, are tested over two definitions of CS cancer, forming eight possible combinations. Strategies T5–T9 are tested over two mpMRI definitions, four cut-off points and two TRUS-guided definitions, forming 80 possible combinations. Therefore strategies T1–T9 form 88 possible substrategies.

• In strategies P1–P9, all men undergo either a TRUS-guided biopsy or a TPM-biopsy and may undergo mpMRI to inform whether or not they should have a second biopsy (TRUS-guided or TPM). Strategy P1 consists of a TPM-biopsy for all men, which constitutes the reference test and uses the primary outcome as the definition for CS cancer against which all other tests are compared. Strategies P2–P4 are tested over the two TRUS-guided biopsy definitions. Strategies P5–P9 are tested over the two mpMRI definitions, four cut-off points and two TRUS-guided definitions. As a result, there are 87 possible substrategies.

The diagnostic strategies were labelled according to their test combination first (M1–M7, N1–N7, T1–T9 or P1–P9) and then according to their TRUS-guided biopsy definition (1 or 2), MP-MRI definition (1 or 2) and cut-off point (2–5). For example, substrategy M1 125 refers to test combination M1, in which all men were first assessed using MP-MRI definition 2 and cut-off point 5 and then those with suspected CS cancer were followed up with TRUS-guided biopsy definition 1.

Strategy P1 (testing all men with TPM-biopsy) represents the value of a perfect immediate test. TPM-biopsy is taken here as the reference test, which correctly identifies the true disease status of all men. TPM-biopsy is unlikely to be feasible in clinical practice, given its HRQoL and resource implications and the large number of men referred to secondary care with suspected prostate cancer. The inclusion of strategy P1 aims to show the difference in benefits between (1) a perfect immediate test that is costly (P1) and (2) less costly and less accurate strategies composed of sequences of tests, in which TPM-biopsy can be used in a small proportion of men.

Substrategy M7 123 corresponds to the PROMIS-proposed diagnostic pathway. This involves using mpMRI in all men followed by TRUS-guided biopsy in men whose mpMRI result indicates suspected CS cancer according to definition 2 and cut-off point 3. TRUS-guided biopsy is evaluated according to definition 1. Men in whom TRUS-guided biopsy does not detect CS cancer are rebiopsed and classified according to definition 1.

Substrategy T7 123 corresponds most closely to the strategy recommended in the 2014 NICE clinical guidelines. 13 It involves TRUS-guided biopsy for all men. Men in whom CS cancer according to definition 1 was not detected receive mpMRI. Men with suspected CS cancer on mpMRI using definition 2 and cut-off point 3 receive a second TRUS-guided biopsy, evaluated according to definition 1. However, in some hospitals, mpMRI is not available for the diagnosis of prostate cancer and the usual diagnostic strategy is T4 1––. In T4 1––, all men receive TRUS-guided biopsy assessed according to definition 1. Men in whom CS cancer was not detected are rebiopsied.

Model inputs

Analytical methods

The decision tree compares the diagnostic outcomes, short-term costs and health outcomes (measured as HRQoL) of the different diagnostic strategies. The results are probabilistic in that they represent the average over 1000 Monte Carlo simulations sampled from the distributions of the parameter inputs. 113 The 95% CIs are calculated as the 97.5 and 2.5 percentiles of the Monte Carlo simulations.

Results are presented for the 32 strategies over the two definitions of TRUS-guided biopsy for CS cancer, the two definitions of mpMRI for CS cancer and the four cut-off points. These form 16 substrategies, as detailed in Table 31. The substrategies are labelled as follows:

• The first number indicates the definition of CS cancer for TRUS-guided biopsy: the primary definition is definition 1 and the secondary definition is definition 2. Substrategies 112–5 and 122–5 use TRUS-guided biopsy definition 1, the primary CS definition.

• The second number indicates the definition of CS cancer for mpMRI: the primary definition is definition 2 and the secondary definition is definition 1. Substrategies 122–5 and 222–5 use definition 2, the primary CS cancer definition.

• The third number indicates the cut-off point: ≥ 2, ≥ 3, ≥ 4 and 5. For example, type 112 uses the TRUS-guided biopsy definition 1, the mpMRI definition 2 and the cut-off point of ≥ 2.

• There are substrategies for which one or more of the definitions or cut-off points are not applicable; for example, in T1, only the definition of TRUS-guided biopsy is applicable (1 or 2) because mpMRI is not performed. These substrategies have a dash (–) to represent where the definition or cut-off point is not applicable. Therefore, T1 under the TRUS-guided biopsy definition 1 is represented as T1 1––.

Results

The results are presented in terms of the men referred for testing:

• the proportion of men with CS cancer detected; this corresponds to the sensitivity of the diagnostic strategies in detecting CS cancer

• the number of TRUS-guided biopsies; this can vary from zero to two

• the proportion of men who receive TPM-biopsy

• the HRQoL loss attributable to the adverse effects of testing; this applies only to TPM-biopsy

• diagnostic efficiency as the proportion of CS cancers detected per pound spent in testing.

The diagnostic frontier is evaluated by plotting the costs and outcomes of the alternative diagnostic substrategies. The line formed by the strategies with the most CS cancers detected per pound spent is the efficiency frontier and identifies the strategies that represent an efficient use of resources for a certain level of funding. Strategies above the line will detect the same proportion of CS cancers but at a higher cost.

Detection of clinically significant cancer

Figures 14–17 show the probability of detecting CS cancer (i.e. the strategies’ true-positive rate or sensitivity) for each diagnostic substrategy. The diagnostic strategies are represented on the horizontal axis and the probability of detecting CS cancer is represented on the vertical axis. The different substrategies across the different definitions and mpMRI cut-off points are represented as points. The vertical bars represent the 95% CI for the probability of detecting CS cancer for each substrategy. The figures are organised by mpMRI cut-off point, from ≥ 2 to 5, to aid interpretation. Appendix 3 (see Tables 49 and 50) presents the same data in tabular form.

FIGURE 14.

Sensitivity of strategies with a mpMRI cut-off point of ≥ 2.

FIGURE 15.

Sensitivity of strategies with a mpMRI cut-off point of ≥ 3.

FIGURE 16.

Sensitivity of strategies with a mpMRI cut-off point of ≥ 4.

FIGURE 17.

Sensitivity of strategies with a mpMRI cut-off point of 5.

The probability of detecting CS cancer varies between strategies and within the same strategy, depending on the CS cancer definitions and cut-off points used. The primary TRUS-guided definition for CS cancer (definition 1) detects up to 0.29 fewer CS cancers than the secondary definition (definition 2). This reflects the stringent primary definition for CS cancers. The implication is that the substrategies using definition 2 always detect more CS cancers than those using definition 1, for a given strategy, mpMRI definition and mpMRI cut-off point.

The difference in the detection of CS cancer between the two mpMRI definitions for CS cancer is smaller than that between the TRUS-guided definitions and depends on the mpMRI cut-off point. On average, the primary CS cancer definition for mpMRI (definition 2: lesions with a volume of ≥ 0.2 ml and/or a Gleason score of ≥ 3 + 4; substrategies 122–5 and 222–5) detects 0.01–0.18 more CS cancers than the secondary definition (definition 1: lesions with a volume of ≥ 0.5 ml and/or a Gleason score of ≥ 4 + 3; substrategies 112–5 and 212–5). This is because definition 2 (primary) is less stringent and hence more sensitive in detecting CS cancers.

Lower mpMRI cut-off points detect a large proportion of CS cancers. For example, substrategies using a mpMRI cut-off point of ≥ 2 (substrategies 112, 122, 212 and 222; see Figure 14) detect nearly all CS cancers. In contrast, in substrategies using a high mpMRI cut-off point, such as those using a cut-off point of 5 (substrategies 115, 125, 215 and 225; see Figure 17), a smaller proportion of cancers is detected.

The most sensitive combination of definitions and mpMRI cut-off point to detect CS cancer includes the secondary TRUS-guided biopsy definition (definition 2), the primary mpMRI definition (definition 2) and a PI-RADS cut-off point of ≥ 2 (i.e. substrategy 222). This will be referred to hereafter as the ‘most sensitive definition’. The least sensitive combination of definitions and mpMRI cut-off point uses the primary TRUS-guided biopsy definition (definition 1), the secondary mpMRI definition (definition 1) and a mpMRI cut-off point of 5 (i.e. substrategy 115). This will be referred to hereafter as the ‘least sensitive definition’.

As evident in Figures 14–17, strategies including TPM-biopsy detect the most CS cancers. Strategies P1 (all men are tested with TPM-biopsy) and P4 (all men in whom TRUS-guided biopsy did not detect cancer or detected CNS cancer are retested with TPM-biopsy) detect all CS cancers.

Some strategies at specific cut-off points and definitions detect most CS cancers:

• P9 consists of TRUS-guided biopsy in all men, mpMRI in men in whom no CS cancer was detected and rebiopsy with TPM-biopsy in men whose mpMRI result was suspicious of CS cancer (in men for whom TRUS-guided biopsy detected CNS cancer) or suspicious of any cancer (in men for whom TRUS-guided biopsy did not detect cancer). P9 using the most sensitive definition (substrategy P9 222) detects all CS cancers. The proportion of CS cancers detected decreases to a minimum of 56% when using the least sensitive definition for mpMRI and TRUS-guided biopsy, definition 1 (substrategy P9 115).

• P7 consists of testing all men with TRUS-guided biopsy; men in whom no cancer or CNS cancer is detected are retested with mpMRI and men with suspected CS cancer are rebiopsied with TPM-biopsy. This detects 99% of CS cancers under the most sensitive definition for mpMRI and TRUS-guided biopsy, definition 2 (substrategy P7 222), and 52% under the least sensitive definition (substrategy P7 115).

• N1 consists of testing all men with mpMRI and conducting TPM-biopsy in all men with suspected CS cancer. It detects 98% of CS cancers under the most sensitive definition of mpMRI, definition 2 (substrategy N1 222), and 28% under the least sensitive definition (substrategy N1 115).

• N2 consists of testing all men with mpMRI and conducting TPM-biopsy in all men with suspected cancer. It detects 99% of CS cancers under the most sensitive definition of mpMRI, definition 2 (substrategy N2 222), and 44% under the least sensitive definition (substrategy N2 115).

• T7 and T9 detect 95% of all CS cancers under the most sensitive definitions. In T7 and T9, all men receive TRUS-guided biopsy. In T7, men in whom CS cancer was not detected undergo mpMRI and those with suspected CS cancer are rebiopsied. In T9, men in whom CS cancer was not detected undergo mpMRI. Men whose first biopsy indicated no cancer receive a second biopsy if the mpMRI result is suspicious for any cancer. Men whose first biopsy indicated CNS cancer receive a second biopsy if the mpMRI result is suspicious for CS cancer.

• M7 consists of testing all men with mpMRI, conducting TRUS-guided biopsy in men with suspected CS cancer and rebiopsying those in whom CS cancer was not detected. This strategy detects 95% of CS cancers under the most sensitive definition (substrategy M7 222). Under the least sensitive definition, M7 detects 26% of CS cancers (substrategy M7 115).

Given that the accuracy data for each diagnostic strategy are drawn from sampled data, there is some uncertainty regarding the sensitivity of each substrategy in detecting CS cancer. This is greater in strategies with smaller test sequences that do not include TPM-biopsy, such as T1 [which detects 38% of CS cancers (95% CI 33% to 43%)] and M1 [which detects 44% of CS cancers (95% CI 36% to 51%)].

The PROMIS-proposed substrategy, M7 123, detects 82% of CS cancers (95% CI 75% to 87%). M7 consists of testing all men with mpMRI, conducting TRUS-guided biopsy in men with suspected CS cancer and rebiopsying those in whom CS cancer was not detected. This compares favourably with T4 1––, the standard care substrategy, which does not include mpMRI and which detects 52% of CS cancers (95% CI 45% to 61%). T4 consists of testing all men with TRUS-guided biopsy and repeating this biopsy in those men in whom CS cancer was not detected. The model calculates that the substrategy most similar to the NICE recommendations,13 T7 123, detects 85% of CS cancers (95% CI 78% to 91%). In T7, men in whom CS cancer was not detected undergo mpMRI and those men with suspected CS cancer are rebiopsied.

Biopsies

Number of transrectal ultrasound-guided biopsies

One of the aims of introducing mpMRI in the diagnostic pathway for prostate cancer is to avoid unnecessary biopsies. Figure 18 shows the number of TRUS-guided biopsies carried out per man referred for diagnosis for each substrategy. Note that this number can vary only from zero to two because the diagnostic strategies were constrained to include up to two biopsies. Strategies P1, N1 and N2 do not include TRUS-guided biopsy and hence the average number of TRUS-guided biopsies per referred man is zero.

FIGURE 18.

Number of TRUS-guided biopsies carried out per man referred for testing for each substrategy.

The strategies starting with TRUS-guided biopsy involve at least one TRUS-guided biopsy per patient; however, some involve more. Strategy T4, under TRUS-guided biopsy definition 1, has the highest mean number of TRUS-guided biopsies per man referred, at 1.79. In this strategy, all men have TRUS-guided biopsy and all men in whom CS cancer is not detected undergo a second TRUS-guided biopsy. Because, under CS cancer definition 1, fewer CS cancers are detected, 79% of men receive a second TRUS-guided biopsy. Other strategies starting with TRUS-guided biopsy with a high number of TRUS-guided biopsies per referred man are strategies T7 and T9, with a maximum of 1.75 and 1.78, respectively. In strategy T7, all men receive TRUS-guided biopsy and men in whom CS cancer was not detected receive mpMRI. Men with suspected CS cancer on mpMRI receive a second TRUS-guided biopsy. Strategy T9 is similar, but also includes rebiopsy in men in whom the mpMRI result showed suspected CNS cancer after an initial TRUS-guided biopsy that did not detect cancer. Strategies with TRUS-guided biopsy definition 1 for CS cancer and low cut-off points for the mpMRI score have a greater proportion of men referred for a second biopsy. Therefore, the average number of TRUS-guided biopsies per referred man increases.

Strategies starting with mpMRI (M1–M7 and N1–N7) have a generally lower mean number of TRUS-guided biopsies per referred man. However, in two strategies, > 50% of men receive two TRUS-guided biopsies; these are strategies M6 and M7, under the TRUS-guided biopsy definition 1 and cut-off point of ≥ 2, in which the average number of TRUS-guided biopsies is 1.51 and 1.66, respectively. In strategy M6, men with suspected cancer on mpMRI receive a TRUS-guided biopsy and are rebiopsied if no cancer is detected. In strategy M7, men with suspected CS cancer on mpMRI receive a TRUS-guided biopsy and are rebiopsied if CS cancer is not detected.

Number of template prostate mapping biopsies

Figure 19 shows the proportion of men who undergo TPM-biopsy. As expected, given the structure of the strategies, strategies M1–M7 and T1–T9 do not involve TPM-biopsy. Strategies N1–N7 and P1–P9 include TPM-biopsy as the last biopsy in the series. Some strategies involve TPM-biopsy in all men referred for diagnosis: P1, which consists of offering TPM-biopsy to all men with no prior testing, and N2, in which men are first tested with mpMRI and men with suspected cancer receive TPM-biopsy, under the cut-off point of ≥ 2. Under this cut-off point, all men referred for diagnosis have suspected cancer and, therefore, receive TPM-biopsy.

FIGURE 19.

Proportion of men who receive a TPM-biopsy.

Cost of testing

Figure 20 shows the costs of testing (see Appendix 3, Tables 51 and 52, for the results in tabular form). The least costly test (at £244, 95% CI £232 to £257) is strategy M1 (mpMRI for all men and TRUS-guided biopsy for men in whom there is suspicion of CS cancer) using the least sensitive definition. This strategy is the least costly because it includes only one biopsy and, using this definition and cut-off point, mpMRI results in only 15% of men being referred for TRUS-guided biopsy. Using more sensitive definitions, a greater proportion of men are referred for TRUS-guided biopsy.

FIGURE 20.

Costs of testing.

The most costly substrategy is P9 112, at £1654 (95% CI £1604 to £1702) per man referred. This substrategy consists of TRUS-guided biopsy in all men, mpMRI in men in whom no CS cancer was detected and rebiopsy with TPM-biopsy in men whose mpMRI result was suspicious of CS cancer (in men in whom TRUS-guided biopsy detected CNS cancer) or suspicious of any cancer (in men in whom TRUS-guided biopsy did not detect cancer) using the TRUS-guided biopsy definition 1, the secondary mpMRI definition 2 and the cut-off point of ≥ 2.

The 95% CIs are relatively narrow, at up to 9% greater or lower than the average cost. Uncertainty regarding the cost of testing arises from uncertainty regarding the probability that each test correctly classifies men, which determines whether or not men receive a subsequent test, and the risk of AEs for strategies involving biopsy.

The PROMIS-proposed substrategy, M7 123, costs £687 (95% CI £657 to £719). This is less than the cost of T4 1––, the standard care substrategy without mpMRI, at £742 (95% CI £727 to £757), and the cost of the NICE-proposed substrategy T7 123, at £780 (95% CI £760 to £801).

Diagnostic efficiency

Figure 21 shows the relationship between the proportion of CS cancers detected and the expected cost of testing. The strategies forming the line (called a frontier) are efficient because they offer the greatest detection of CS cancers for different costs of testing. Strategies above the efficiency frontier are inefficient because they are expected to detect fewer CS cancers for the same cost. Table 32 details the strategies forming the efficiency frontier. None of these strategies, except P4, has HRQoL implications because they do not include TPM-biopsy. Neither the PROMIS-proposed substrategy (M7 123) nor any of the standard care substrategies (T4 1–– and T7 123) form the diagnostic efficiency frontier.

FIGURE 21.

Diagnostic efficiency.

TABLE 32 - Diagnostic strategies in the efficiency frontier

NC, no cancer.

Strategy	TRUS-guided biopsy definition	mpMRI definition	mpMRI cut-off point	Detection of CS cancer, mean (95% CI) (%)	Cost (£) of testing, mean (95% CI)
M1: mpMRI for all men and TRUS-guided biopsy in men with suspected CS cancer	2	1	5	22 (17 to 27)	244 (232 to 257)
M3: mpMRI for all men and TRUS-guided biopsy in men with suspected CS cancer; men with CNS cancer at the first biopsy receive a second TRUS-guided biopsy	2	1	5	24 (20 to 29)	250 (237 to 264)
	2	2	5	35 (30 to 41)	290 (273 to 306)
M4: mpMRI for all men and TRUS-guided biopsy in men with suspicion of any cancer; men with suspected CS cancer on mpMRI and in whom CNS cancer was detected at the first biopsy receive a second TRUS-guided biopsy	2	2	5	37 (32 to 43)	296 (280 to 312)
M7: mpMRI for all men and TRUS-guided biopsy in men with suspected CS cancer; rebiopsy with TRUS-guided biopsy for those in whom CS cancer was not detected	2	2	5	40 (35 to 46)	307 (287 to 328)
M3: mpMRI for all men and TRUS-guided biopsy in men with suspected CS cancer; men with CNS cancer at the first biopsy receive a second TRUS-guided biopsy	2	2	4	57 (51 to 64)	385 (366 to 404)
M4: mpMRI for all men and TRUS-guided biopsy in men with suspicion of any cancer; men with suspected CS cancer on mpMRI and in whom CNS cancer was detected at the first biopsy receive a second TRUS-guided biopsy	2	2	4	60 (54 to 67)	400 (382 to 420)
M7: mpMRI for all men and TRUS-guided biopsy in men with suspected CS cancer; rebiopsy with TRUS-guided biopsy for those in whom CS cancer was not detected	2	2	4	65 (60 to 71)	430 (404 to 458)
T6: TRUS-guided biopsy for all men; men classified with CNS cancer receive mpMRI and men with suspected CS cancer receive a second TRUS-guided biopsy	2	2	3	74 (69 to 78)	488 (473 to 503)
	2	2	2	75 (71 to 79)	500 (484 to 517)
M7: mpMRI for all men and TRUS-guided biopsy in men with suspected CS cancer; rebiopsy with TRUS-guided biopsy for those in whom CS cancer was not detected	2	2	3	85 (81 to 89)	628 (597 to 660)
T7: TRUS-guided biopsy for all men; men classified as having NC or CNS cancer receive mpMRI and men with suspected CS cancer receive a second TRUS-guided biopsy	2	2	3	91 (86 to 94)	709 (688 to 730)
M7: mpMRI for all men and TRUS-guided biopsy in men with suspected CS cancer; rebiopsy with TRUS-guided biopsy for those in whom CS cancer was not detected	2	2	2	95 (92 to 98)	807 (777 to 833)
P4: TRUS-guided biopsy in all men and TPM-biopsy for men in whom the biopsy did not detect CS cancer	2	Not applicable	Not applicable	100 (100 to 100)	1332 (1278 to 1385)

Summary of findings

• Multiparametric magnetic resonance imaging, TRUS-guided biopsy and TPM-biopsy can be used in 32 different combinations to diagnose prostate cancer. Multiparametric magnetic resonance imaging can be used as a first test for all men referred to secondary care or following a negative TRUS-guided biopsy result. Likewise, TRUS-guided biopsy can be used as a first test or as a targeted test following mpMRI. The different combinations have been evaluated for different diagnostic definitions and cut-off points, forming 383 diagnostic substrategies. Our analysis was informed by the evidence obtained in PROMIS, together with published studies on repeated TRUS-guided biopsy and TRUS-guided biopsy in combination with mpMRI. 103–105

• The short-term outcomes relate to the immediate consequences of using the different diagnostic strategies in men with suspected CS cancer. These are the proportion of CS cancers detected (sensitivity), the number of TRUS-guided biopsies, the proportion of men receiving TPM-biopsy and the cost of testing (which includes the cost of managing any adverse effects from testing). The proportion of CS cancers detected and the cost of testing form the cost-efficiency plane and frontier.

• The strategies considered represent a range of levels of use of TRUS-guided biopsy, from 0 to 1.79 biopsies per man. The strategies considered also represent a range of levels of use of TPM-biopsy, from strategies in which no men receive TPM-biopsy to those in which all men receive this test.

• The proportion of CS cancers detected depends on the diagnostic definition and mpMRI cut-off point. The most sensitive strategies use the TRUS-guided biopsy definition 2, mpMRI definition 2 and cut-off point 2. The least sensitive strategies use the TRUS-guided biopsy definition 1, mpMRI definition 1 and cut-off point 5.

• P1 and P4 detect all CS cancers. Some strategies under the most sensitive definitions and cut-off point detect almost all CS cancers: N1, N2, P7 and P9, involving TPM-biopsy; T7 and T9, involving TRUS-guided biopsy as the first test; and M7, involving mpMRI as the first test.

• The diagnostic strategies cost between £244 and £1654 per man to the UK NHS.

• A limited number of strategies form the cost-efficiency frontier. These strategies detect the greatest proportion of CS cancer by level of spending (on testing and adverse effects management). Of these, only four substrategies detect ≥ 80% of CS cancers: M7 (substrategies 223 and 222), T7 (substrategy 223) and P4 (substrategy 2––).

  • In M7, men whose mpMRI result is suspicious of CS cancer receive TRUS-guided biopsy and men in whom no CS cancer was detected receive a second biopsy.

  • In T7, all men receive TRUS-guided biopsy and men in whom TRUS-guided biopsy does not detect CS cancer undergo mpMRI. Men in whom the mpMRI result is suspicious of CS cancer receive a second biopsy.

  • In P4, all men receive TRUS-guided biopsy and men in whom TRUS-guided biopsy does not detect CS cancer undergo TPM-biopsy.

Evidence on long-term outcomes

Methods

Modelling approach

Step 1: life-years

Appendix 4 reports the modelling approach in detail. In brief, the survival curves reported in the supplementary material of Wilt et al. 8 for each subgroup allocated to observation were digitised using WebPlotDigitizer version 3.12 (Ankit Rohatgi, Austin, TX, USA), a free online tool. Alternative parametric distributions were assessed for goodness of fit with the available data. The Weibull distribution was selected across the three cancer risk subgroups. In the extrapolation period, the Weibull distribution predicted lower hazards of death than those of the general population. Therefore, at relevant times, the hazard of death of the general population was used to predict survival.

Step 2: transition probabilities

Step 1 obtains the life-years for each subgroup and each treatment option. However, the prediction of quality-adjusted survival (i.e. QALYs) and costs requires information about the health state in which these life-years are lived, namely whether with progression-free or metastatic disease. This requires information on the transition probabilities from progression-free disease to metastasis or death and from metastasis to death. This is obtained with a Markov model with three health states (progression-free disease, metastatic disease and death), evaluated by calibration (Figure 22). The model is calibrated to the life-years predicted in step 1, the proportion of patients with metastases in Wilt et al. 8 and a plausible range of values for the probability of dying after metastasis from the standard care arm of the STAMPEDE trial. 125

FIGURE 22.

Structure of the long-term calibration model.

Specifically, the calibration model aimed to find a plausible set of transition probabilities that fit the above criteria. It randomly draws numbers between 0 and 1 from the transitions from progression-free disease to metastatic disease and death. For the transition from metastatic disease to death, the model draws a random number from within the 95% CI of the cumulative incidence of all-cause death from the metastatic disease subgroup allocated to the standard care arm of the STAMPEDE trial. 125 A 10% leeway is given to allow for differences in the disease severity of patients and their management between the STAMPEDE trial and PIVOT. STAMPEDE is a RCT set in the UK that compares treatments for men with locally advanced and metastatic prostate cancer. James et al. 125 report the survival outcomes of men with newly diagnosed metastatic prostate cancer who were allocated to the current standard of care, androgen deprivation therapy.

The calibration model records the transition probabilities that meet three conditions of plausibility: (1) the life-years are within the 95% CI predicted in step 1, (2) the cumulative incidence of metastases at 12 years is within the 95% CI of PIVOT and (3) the transition probability from progression-free disease to death is smaller than the transition probability from metastatic disease to death. The calibration model is run until 1000 plausible sets of transition probabilities for each subgroup are found.

Step 3: lifetime quality-adjusted life-years and costs

Lifetime QALYs and costs are predicted by the same three-state Markov model and the plausible set of transition probabilities from step 2. A 3.5% discount rate is used to calculate the present value of future benefits (QALYs) and costs, in line with current NICE guidance. 99

Model inputs

The inputs for HRQoL and costs are reported in Table 35. The model assumes that the active treatment of choice is RP. Men diagnosed with CNS cancer receive active surveillance, as recommended by NICE. 13 HRQoL in men with no cancer is calculated as a function of the men’s age using the Ara and Brazier127 regression model. Men with localised cancer are assumed to have the same HRQoL as similarly aged men with no cancer. In other words, localised cancer is assumed to have no effect on men’s HRQoL. The treatment and management of localised cancer is also assumed to have no HRQoL implications. This is based on the findings of Korfage et al. ,129 who found no effect of treatment on HRQoL post RP. The decrement from metastatic disease is calculated from Torvinen et al. 126 Torvinen et al. 126 assessed HRQoL in different stages of prostate cancer using the EQ-5D, 15D and European Organization for Research and Treatment of Cancer Quality of Life Questionnaire-C30 (EORTC-30) in Finland between 2009 and 2010. Men with prostate cancer were divided into five mutually exclusive groups: (1) local disease in the first 6 months after diagnosis (n = 47), (2) local disease 0.5–1.5 years after diagnosis or recurrence (n = 158), (3) local disease > 1.5 years after diagnosis (n = 317), (4) metastatic disease (n = 89) and (5) palliative care (n = 19). The decrement for metastatic disease is estimated for the model as the difference between the weighted average of the local disease EQ-5D score and the weighted average of the metastatic disease EQ-5D score.

Radical treatment is costed as the cost of RP. The cost of RP includes the cost of surgery and follows the approach taken for the 2014 NICE clinical guideline. 13 It includes the procedure cost, taken as the average PbR tariff for ‘LB21Z Bladder neck open procedures – male’ and ‘LB22Z Laparoscopic bladder neck procedures – male’ (both combined day case/ordinary elective spell tariff), the urology follow-up appointment (‘WF01A: Follow-up attendance single-professional’), a first surgical consultation (‘WD01B: First attendance single-professional’) and a follow-up surgical consultation (‘WF01A: Follow-up attendance single-professional’). Active surveillance is costed as an annual cost corresponding to a once-yearly outpatient appointment (‘WF01A: Follow-up attendance single-professional’) and three PSA tests in primary care.

Every year, men incur an annual cost for the management of AEs. This cost is calculated as the weighted average of the incidence of AEs multiplied by the unit cost of managing those AEs. The incidence of AEs is obtained from PIVOT. Wilt122 reports patient-reported urinary, erectile and bowel dysfunction at 2 years. The cost of managing AEs is obtained from the PbR tariff for the management of erectile dysfunction (PbR tariff for LB43Z) and from the 2014 NICE clinical guideline13 for the management of urinary incontinence and bowel dysfunction (inflated to 2014–15 prices).

Men with no cancer are modelled to have the life expectancy and HRQoL of the UK general population127,130 and incur the costs of a once-yearly PSA test in primary care. 111

Results

Summary

• The 2014 NICE clinical guideline13 recommends that men with low-risk cancer, equivalent to CNS cancer, should receive active surveillance and be offered radical treatment if their cancer progresses. It recommends that men with intermediate- and high-risk cancer, equivalent to CS cancer, should receive radical treatment if there are reasonable chances of a cure. These recommendations are based on limited quantitative evidence on the cost-effectiveness of active surveillance and active treatment for these patients.

• Two good-quality RCTs were identified on the clinical effectiveness of RP compared with watchful waiting for the management of prostate cancer by risk subgroup: the SPCG-4 trial6,115,119,123 and PIVOT. 8,122 The patients taking part in PIVOT were found to be more comparable to the patients in PROMIS; therefore, PIVOT was selected to inform the prediction of long-term outcomes. No evidence was found on the clinical effectiveness of other treatments by risk subgroup. Consequently, RP was taken as the treatment of choice in patients referred for radical treatment.

• Men with low-risk cancer managed with active surveillance have a life expectancy of 16.26 years on average. Men with intermediate-risk cancer who receive active surveillance have a life expectancy of 13.38 years on average; this increases to 15.65 years if they receive radical treatment. Men with high-risk cancer who receive active surveillance have a life expectancy of 11.15 years on average; this increases to 13.23 years if they receive radical treatment.

• Radical treatment was found to be cost-effective compared with active surveillance in men with intermediate-risk and high-risk cancer, with an ICER of £3067 per QALY gained and £3602 per QALY gained, respectively. The probability that radical treatment is cost-effective ranges between 0.74 and 0.93.

Analytical methods

In order to consider uncertainty, analyses were run probabilistically. The 1000 simulations from the model predicting short-term outcomes were considered together with the 1000 simulations of the model predicting long-term outcomes to calculate the distribution of overall QALYs and costs associated with each diagnostic strategy. Results are presented in terms of their averages and 95% CIs, calculated using the percentile method. The expected overall costs and QALYs are shown in the cost-effectiveness plane and a cost-effectiveness frontier is drawn. The cost-effectiveness frontier is formed by the substrategies that offer the most health outcomes per unit of cost. Substrategies to the left of the frontier are not cost-effective, because those in the frontier achieve more health benefits for the same cost. The probability that a strategy is cost-effective is shown in the cost-effectiveness acceptability frontier. The probability that a strategy is cost-effective was calculated as the proportion of simulations in which this strategy offers the most health benefits net of its costs, for a given value of health opportunity cost. 113 The cost-effectiveness acceptability frontier indicates the probability that the cost-effective alternative, on average (i.e. the strategy with the highest average net health value), will be cost-effective. 131 Remaining sensitivity analyses were conducted deterministically unless significant changes were observed that warranted in-depth exploration with probabilistic sensitivity analysis.

Sensitivity analysis

Base-case results

Health outcomes

Figure 23 shows the health outcomes achieved by each diagnostic substrategy; 95% CIs are not shown because they greatly overlap (see Appendix 3, Tables 53 and 54, for full results). The health outcomes vary between 8.30 and 8.75 QALYs.

FIGURE 23.

Base-case results: health benefits of each diagnostic strategy.

The substrategy that achieves the least favourable health outcomes is M1 115, at 8.29 QALYs (95% CI 7.94 to 8.68 QALYs). In M1 115, all men receive mpMRI and men with suspected CS cancer according to the mpMRI definition 1 and cut-off point 5 receive TRUS-guided biopsy. Men who have CS cancer using the TRUS-guided biopsy definition 1 receive radical treatment; otherwise, men receive active surveillance. This is the strategy that detects the fewest CS cancers (sensitivity 0.15, 95% CI 0.12 to 0.19; see Appendix 3, Tables 49 and 50). Therefore, using this strategy, the majority of men with CS cancer are missed and do not receive immediate treatment.

The substrategy that achieves the most favourable health outcomes is P4 2––, at 8.74 QALYs (95% CI 8.41 to 9.06 QALYs). All men receive TRUS-guided biopsy at definition 2 of clinical significance, which is the most sensitive in detecting CS cancers. Men in whom CS cancer was not detected receive TPM-biopsy. P4 2–– detects all CS cancers and minimises the proportion of men who receive TPM-biopsy and who experience the associated reduction in HRQoL.

The PROMIS-proposed substrategy (M7 123) achieves 8.65 QALYs (95% CI 8.35 to 8.95 QALYs). The standard care substrategy without mpMRI (T4 1––) achieves 8.49 QALYs (95% CI 8.19 to 8.80 QALYs). The strategy deemed to be the most similar to the NICE-recommended strategy13 achieves 8.66 QALYs (95% CI 8.36 to 8.97 QALYs).

Costs

Figure 24 shows the lifetime costs of using each diagnostic substrategy; 95% CIs are not shown to improve readability (see Appendix 3, Tables 55 and 56, for full results). The costs vary between £3500 and nearly £6300 per man referred for testing. Similar to the results for health outcomes, the substrategies using the lowest mpMRI cut-off point (≥ 2, represented in black), mpMRI definition 2 and TRUS-guided definition 2 have the highest costs. These substrategies have the highest costs of testing (see Figure 20 and Appendix 3, Tables 51 and 52) and the highest treatment costs, because more men with CS cancer are referred for treatment.

FIGURE 24.

Base-case results: costs of each diagnostic strategy.

The substrategy with the lowest costs is M1 115, at £3497 (95% CI £3007 to £4115), which is also the substrategy that achieves the least favourable health outcomes, as discussed in Health outcomes. The smallest proportion of men are referred for TRUS-guided biopsy using the mpMRI definition 1 and cut-off point 5. The smallest proportion of men are referred for radical treatment, which is more costly than active surveillance, using the TRUS-guided biopsy definition 1.

The substrategy with the greatest costs is P9 122, at £6277 (95% CI £5858 to £6808). In P9 122, all men receive TRUS-guided biopsy and men in whom TRUS-guided biopsy at definition 1 did not detect CS cancer receive mpMRI. Men in whom mpMRI at definition 2 and cut-off point of ≥ 2 was suspicious of CS cancer are referred to TPM-biopsy. P9 122 detects almost all CS cancers (0.99, 95% CI 0.98 to 1.00) (see Appendix 3, Table 49) and these men then receive radical treatment, which is more costly than active surveillance. P9 122 is also more costly in the short term as a result of the large number of tests involved.

The PROMIS-proposed substrategy (M7 123) costs £5027 (95% CI £4609 to £5512). The standard care substrategy without mpMRI (T4 1––) costs £4603 (95% CI £4174 to £5044). The strategy deemed to be the most similar to the NICE-recommended strategy13 costs £5173 (95% CI £4755 to £5664).

Cost-effectiveness results

Figure 25 shows each diagnostic substrategy in terms of its average costs and QALYs and defines the strategies that achieve the most QALYs per pound spent (i.e. the cost-effectiveness frontier, represented as a black line). Substrategies above the cost-effectiveness frontier cannot be considered to be cost-effective. This figure does not include the uncertainty around the results to improve readability. Table 40 presents the same information, including uncertainty, in a tabular format.

FIGURE 25.

Cost-effectiveness plane.

TABLE 40 - Diagnostic strategies in the cost-effectiveness frontier

NC, no cancer.

Strategy	Definition		mpMRI cut-off point	QALYs, mean (95% CI)	Costs, mean (95% CI) (£)	ICER (£)	Net health at
	TRUS-guided biopsy	mpMRI					£13,000 per QALY	£20,000 per QALY	£30,000 per QALY
M1: mpMRI for all men and TRUS-guided biopsy in men suspicious for CS cancer	1	1	5	8.29 (7.94 to 8.68)	3497 (3007 to 4115)		8.024	8.118	8.176
	2	1	5	8.33 (7.99 to 8.69)	3608 (3137 to 4191)	3081	8.051	8.148	8.208
M3: mpMRI for all men and TRUS-guided biopsy in men with suspected CS cancer; men with CNS cancer on the first biopsy receive a second TRUS-guided biopsy	2	1	5	8.34 (8.01 to 8.70)	3648 (3180 to 4218)	3630	8.059	8.157	8.218
M4: mpMRI for all men and TRUS-guided biopsy in men with suspicion of any cancer; men with suspected CS cancer at mpMRI and in whom CNS cancer was detected on the first biopsy receive a second TRUS-guided biopsy	2	2	5	8.41 (8.11 to 8.74)	3909 (3477 to 4415)	3738	8.109	8.214	8.279
M7: mpMRI for all men and TRUS-guided biopsy in men with suspected CS cancer; rebiopsy with TRUS-guided biopsy for those in whom CS cancer was not detected	2	2	5	8.42 (8.13 to 8.74)	3965 (3539 to 4455)	3867	8.119	8.226	8.292
M3: mpMRI for all men and TRUS-guided biopsy in men with suspected CS cancer; men with CNS cancer on the first biopsy receive a second TRUS-guided biopsy	2	2	4	8.52 (8.23 to 8.82)	4325 (3915 to 4774)	3921	8.183	8.300	8.372
M4: mpMRI for all men and TRUS-guided biopsy in men with suspicion of any cancer; men with suspected CS cancer at mpMRI and in whom CNS cancer was detected on the first biopsy receive a second TRUS-guided biopsy	2	2	4	8.53 (8.24 to 8.83)	4390 (3979 to 4833)	4031	8.194	8.312	8.386
M7: mpMRI for all men and TRUS-guided biopsy in men with suspected CS cancer; rebiopsy with TRUS-guided biopsy for those in whom CS cancer was not detected	2	2	4	8.56 (8.27 to 8.86)	4502 (4094 to 4937)	4250	8.212	8.333	8.408
T6: TRUS-guided biopsy for all men; men classified as having CNS cancer receive mpMRI; men with suspected CS cancer receive a second TRUS-guided biopsy	2	2	3	8.60 (8.31 to 8.90)	4693 (4289 to 5127)	4393	8.241	8.367	8.445
	2	2	2	8.61 (8.32 to 8.91)	4732 (4327 to 5166)	4633	8.246	8.373	8.452
M7: mpMRI for all men and TRUS-guided biopsy in men with suspected CS cancer; rebiopsy with TRUS-guided biopsy for those in whom CS cancer was not detected	2	2	3	8.66 (8.37 to 8.97)	5021 (4612 to 5496)	5501	8.276	8.412	8.495
T7: TRUS-guided biopsy for all men; men classified as having NC or CNS cancer receive mpMRI; men with suspected CS cancer receive a second TRUS-guided biopsy	2	2	3	8.69 (8.38 to 9.00)	5194 (4780 to 5682)	5778	8.293	8.433	8.519
M7: mpMRI for all men and TRUS-guided biopsy in men with suspected CS cancer; rebiopsy with TRUS-guided biopsy for those in whom CS cancer was not detected	2	2	2	8.72 (8.40 to 9.04)	5367 (4947 to 5876)	7076	8.304	8.449	8.538
P4: TRUS-guided biopsy in all men and TPM-biopsy in men in whom CS cancer was not detected	2	Not applicable		8.74 (8.41 to 9.06)	5968 (5550 to 6490)	30,084	8.278	8.439	8.538

Across the three values of health opportunity cost used (£13,000, £20,000 and £30,000 per QALY gained), the most cost-effective substrategy is M7 222. P4 2–– is on the margin of cost-effectiveness using the health opportunity value of £30,000 per QALY; the difference in net health between P4 2–– and M7 222 at this value is only 0.0001 QALYs.

In the restricted comparison between the PROMIS-proposed substrategy of M7 123, the standard care no-mpMRI strategy of T4 1–– and the substrategy deemed to be the most similar to the NICE-recommended substrategy13 (T7 123), T7 123 is most cost-effective (ICER £8520 per QALY vs. M7 123). The ICER of M7 123 compared with T4 1–– is £2730 per QALY. In the complete comparison of the 383 strategies, these three strategies have zero probability of being cost-effective.

Figure 26 shows the cost-effectiveness acceptability frontier. The cost-effectiveness acceptability curve plots the options that are cost-effective in terms of their probability of cost-effectiveness. The most cost-effective option is the option that offers the most benefits net of its cost, for a given value of health opportunity cost. For example, at the health opportunity cost of £20,000 per QALY, the cost-effective option is M7 222. M7 222 is the option that offers the most benefits on average. In some of the simulations, other strategies are cost-effective. Therefore, the probability that M7 222 is cost-effective corresponds to the proportion of simulations in which M7 222 offers the most benefits, net of its costs. The cost-effectiveness acceptability curve is useful in evaluations with a large number of options, which is the case here.

FIGURE 26.

Cost-effectiveness acceptability frontier.

At health opportunity costs lower than £3800 per QALY, M1 115 is the cost-effective substrategy (the substrategy that detects the fewest CS cancers but is also the least costly). Between the health opportunity costs of £3000 and £7500 per QALY, a number of substrategies appear to be cost-effective (M1 215 to T6 222 in Figure 25), but these are associated with a low probability of being cost-effective. At health opportunity costs of > £7500 up to £30,084 per QALY, M7 222 is the most cost-effective substrategy; the probability of this strategy being cost-effective ranges from 0.31 to 0.48 depending on the value of the health opportunity cost. At the health opportunity cost of ≥ £30,084, P4 2–– is the substrategy that is most likely to be cost-effective. The probability ranges between 0.44 and 0.63 for values of health opportunity cost up to £50,000 per QALY.

Bivariate sensitivity analysis on the cost of diagnostic tests

Figure 27–38 show the results of the bivariate sensitivity analysis on the unit costs of TRUS-guided biopsy and mpMRI, by the cost of TPM-biopsy. The unit costs are varied by ± 50%; the unit cost of TRUS-guided biopsy is represented on the vertical axis, whereas the unit cost of mpMRI is represented on the horizontal axis. Each figure refers to a unit cost of TPM-biopsy (base case at £1370, +25% at £1713, –25% at £1028 and the cost estimated for PROMIS at £1872) and a health opportunity cost (£13,000, £20,000 and £30,000 per QALY). The square delineated in black represents the base-case unit cost of TRUS-guided biopsy and mpMRI. The figures show the cost-effective substrategy for a combination of unit costs, distinguished by colours.

FIGURE 27.

Bivariate sensitivity analysis: TPM-biopsy cost = £1370, health opportunity cost = £13,000 per QALY.

FIGURE 28.

Bivariate sensitivity analysis: TPM-biopsy cost = £1370, health opportunity cost = £20,000 per QALY.

FIGURE 29.

Bivariate sensitivity analysis: TPM-biopsy cost = £1370, health opportunity cost = £30,000 per QALY.

FIGURE 30.

Bivariate sensitivity analysis: TPM-biopsy cost = £1713, health opportunity cost = £13,000 per QALY.

FIGURE 31.

Bivariate sensitivity analysis: TPM-biopsy cost = £1713, health opportunity cost = £20,000 per QALY.

FIGURE 32.

Bivariate sensitivity analysis: TPM-biopsy cost = £1713, health opportunity cost = £30,000 per QALY.

FIGURE 33.

Bivariate sensitivity analysis: TPM-biopsy cost = £1028, health opportunity cost = £13,000 per QALY.

FIGURE 34.

Bivariate sensitivity analysis: TPM-biopsy cost = £1028, health opportunity cost = £20,000 per QALY.

FIGURE 35.

Bivariate sensitivity analysis: TPM-biopsy cost = £1028, health opportunity cost = £30,000 per QALY.

FIGURE 36.

Bivariate sensitivity analysis: TPM-biopsy cost = £1872, health opportunity cost = £13,000 per QALY.

FIGURE 37.

Bivariate sensitivity analysis: TPM-biopsy cost = £1872, health opportunity cost = £20,000 per QALY.

FIGURE 38.

Bivariate sensitivity analysis: TPM-biopsy cost = £1872, health opportunity cost = £30,000 per QALY.

Figure 27 shows the cost-effective substrategy using the base-case cost for TPM-biopsy at the health opportunity cost of £13,000. M7 222 is the cost-effective substrategy unless the cost of mpMRI is increased. The cost-effective substrategy is T9 222 if the unit cost of mpMRI is increased and the unit cost of TRUS-guided biopsy is reduced. If both the unit cost of mpMRI and the unit cost of TRUS-guided biopsy are increased, the cost-effective substrategy is T7 222. At the higher opportunity cost of £20,000 per QALY (see Figure 28), the cost-effective substrategy is robust as M7 222, unless very high unit costs of mpMRI are coupled with very low unit costs of TRUS-guided biopsy. At £30,000 per QALY (see Figure 29), the cost-effective substrategy is also M7 222, unless the costs of mpMRI and TRUS-guided biopsy increase. Figures 30–32 and Figures 36–38, using higher unit costs for TPM-biopsy, show a similar pattern. Figures 33–35 show the results for a lower cost of TPM-biopsy (£1028). The cost-effective strategy is more sensitive to changes in the unit costs of TRUS-guided biopsy and mpMRI using this unit cost. At the health opportunity cost of £13,000 per QALY, M7 222 is cost-effective. However, this can change to T9 222, T7 222 or P1 depending on the unit costs of mpMRI and TRUS-guided biopsy. At the health opportunity cost of £20,000 per QALY, M7 222 is cost-effective only at lower costs of mpMRI. The cost-effective substrategy for a wider range of unit costs is P4 2–– or P1.

There is some variation in the unit costs of the tests. These figures can be used to indicate the cost-effective substrategy for a given combination of unit costs. For example, the average unit cost of TRUS-guided biopsy using the code LB76Z is £403. However, 20,502 men received it as a day case procedure, at a cost of £533, whereas 17,846 men received it as an outpatient procedure, at a cost of £223. At the TPM-biopsy cost of £1397, similar to the base-case average NHS reference cost,110 M7 222 is cost-effective unless the cost of mpMRI increases to > £197 per scan.

The results suggest that the unit costs of the tests, particularly TPM-biopsy, are a driver of the cost-effectiveness results. Higher TPM-biopsy costs favour the cost-effectiveness of highly sensitive strategies, such as M7 222, which do not include TPM-biopsy. Very high mpMRI costs favour strategies that start with TRUS-guided biopsy rather than mpMRI. However, at lower TPM-biopsy costs, and depending on the health opportunity cost adopted, the cost-effective strategy may involve TPM-biopsy for a large number of men, such as in P1 and P4 2––.

Scenario: using the payment-by-results tariff for the unit costs of the tests

Figure 39 shows the cost-effectiveness plane for the scenario using the PbR tariff for the unit costs of the tests. The substrategies in the cost-effectiveness frontier are similar to those in in the cost-effectiveness frontier in the base case. Four strategies forming the frontier start with mpMRI (M1, M3, M4 and M7) and five strategies start with biopsy (T1, T6, T7, P4 and P1). The cost-effective strategy depends on the value of the health opportunity cost. At £13,000 per QALY gained, the cost-effective substrategy is M7 222. On average, M7 222 achieves the highest net health, at 8.307 QALYs. At £20,000 and £30,000 per QALY gained, the cost-effective substrategy is P4 2––. However, the difference between P4 2–– and P1, the next best substrategy, is small, at 0.001 QALYs.

FIGURE 39.

Cost-effectiveness plane for the scenario using the PbR tariff.

Figure 40 shows the cost-effectiveness acceptability frontier. As with the base-case results, the substrategy most likely to be cost-effective at values of health opportunity cost of < £3000 per QALY gained is M1 115. There is a region of uncertainty around £3000–6000 per QALY gained, where a number of substrategies appear in the frontier. M7 222 is most likely to be cost-effective up to a value of health opportunity cost of £17,500 per QALY. P4 2–– is the substrategy most likely to be cost-effective at values of > £17,500 per QALY gained.

FIGURE 40.

Cost-effectiveness acceptability frontier for the scenario using the PbR tariff.

The threshold sensitivity analyses described earlier were also conducted using the PbR tariff for the unit costs (see Appendix 6). In general, the results are similar to those of the base case. In addition to M7 222 and P4 2––, the substrategies that emerge as potentially cost-effective are typically T7 222, T9 222 and P1.

Generalisability and cancer prevalence

Despite the number of withdrawals following consent, there were no significant differences between those who withdrew after registration and those who completed all of the tests. We would expect this to be the case as the mpMRI findings were blinded to patients and physicians. Withdrawal after consent of the men with large prostates may have led to a reduction in the number of men with no cancer or CNS cancers, because larger prostates tend to lead to an elevated PSA level as a result of benign prostatic hyperplasia. Although a significant number of men were screened in order to recruit our cohort, the screening log data indicated that there were no differences in age or PSA level between patients who were screened as eligible for the study but who did not enter the study and those who did enter the study.

The prevalence of any cancer or CS cancer was much higher than we anticipated. This was a function of the combined biopsy strategy – a diagnostic approach that has rarely been tried before. The effect was the identification of a large, and almost certainly representative, sample of men with prostate cancer from the cohort referred for biopsy. This figure might have also been influenced by our withdrawal of men with large prostates (> 100 ml), which could not be biopsied using the reference test.

Nonetheless, our data are not outliers with respect to the literature. Previous TPM-biopsy studies have found a higher prevalence of CS prostate cancer than studies using TRUS-guided biopsy. 82,132 No single sampling system identifies all disease. The combination of TRUS-guided biopsy and TPM-biopsy probably represents the best detection strategy that is available.

The degree to which the prevalence of CS prostate cancer is contingent on the sampling method used does raise questions about where the threshold of clinical significance should be placed. This highlights the uncertainty around what constitutes clinical significance in prostate cancer. 133 PROMIS was conducted within a health-care system that does not have a formal prostate cancer screening programme in place. In jurisdictions where screening has been in place for some time, it is likely that the prevalence of CS cancer among men undergoing biopsy will be lower than in the UK, where a formal population-based PSA screening programme is not recommended.

Diagnostic accuracy: findings and implications

On TPM-biopsy, 408 men (71%) who were tested had cancer (CNS and CS cancer), of which 230 cases (40%) were CS according to the primary definition. The primary definition of clinical significance used in our study was a Gleason score of ≥ 4 + 3 of any length and/or a maximum cancer core length of ≥ 6 mm of any grade in any location on TPM-biopsy. Using our primary reporting schema for mpMRI (as described in Chapter 3), with a score of ≥ 3 considered a suspicious scan, mpMRI was more sensitive than TRUS-guided biopsy (93%, 95% CI 88% to 96% vs. 48%, 95% CI 42% to 55%; p < 0.0001) and less specific (41%, 95% CI 36% to 46% vs. 96%, 95% CI 94% to 98%; p < 0.0001). Furthermore, using these primary definitions and thresholds, the NPV of mpMRI was 89% and the PPV was 51%.

The relatively low specificity and PPV seen with mpMRI mean that 73% of patients would test positive on mpMRI and would still require a biopsy. Because of the low specificity of mpMRI, 50% of this 73% (or 36% of men in absolute terms) would undergo an unnecessary biopsy after mpMRI triage. In the current diagnostic pathway, 100% of men have biopsies and about 60% (346/576) of these biopsies might be considered unnecessary as they were carried out in men without CS cancer.

Our results show that, of the 27% of men who would not receive a biopsy on the basis of their mpMRI result, 10% (17 men in our sample) would have CS cancer missed.

Using mpMRI to triage men might allow 158 out of 576 (27%) men to avoid a primary biopsy, resulting in 28 out of 576 (5%) fewer CNS cancers being diagnosed. If subsequent TRUS-guided biopsy needles were directed by mpMRI findings, an estimate of up to 102 out of 576 more cases (18%) of CS cancer might be detected compared with the standard pathway of TRUS-guided biopsy for all.

PROMIS was unable to test the strategy of targeting biopsies to suspicious areas because of the blinded nature of the study design. In the above paragraph, we assumed that deployment of the needles should be guided by the mpMRI findings and might be equivalent to the high accuracy of a TPM-biopsy strategy. The PRECISION trial has recently tested this assumption by randomising men to either systematic TRUS-guided biopsy or a MRI-targeted approach. It found that fewer men were biopsied, fewer cancers of a Gleason score of 6 were found and greater numbers of men with cancer of a Gleason score of ≥ 7 were detected by using mpMRI prior to biopsy. 134

Context within the literature

Our results support the findings of recent systematic reviews assessing the diagnostic accuracy of mpMRI. 49,135 The reviews found sensitivities of 58–96%, NPVs of 63–98% and specificities of 23–87%. The broad ranges are attributable to the studies being conducted at single centres, each of which used different target conditions on different reference standards and different thresholds for designating a positive scan. Other limitations common to most of the studies were retrospective analysis, non-blinding of imaging findings (making them vulnerable to incorporation and reporting biases) and comparison of mpMRI with reference tests that were inaccurate (i.e. TRUS-guided biopsy) or inappropriate (i.e. radical treatment).

One other prospective study compared mpMRI with TPM-biopsy. 136,137 This study reported a sensitivity of 96%, specificity of 36%, NPV of 92% and PPV of 52% for the detection of CS cancer, which was defined as a Gleason score of 7–10 with > 5% at Gleason grade 4, ≥ 20% positive cores or ≥ 7 mm cancer. This was a single-centre, non-blinded study that permitted scanners with two magnetic field strengths (1.5 T or 3.0 T), used fewer sampling cores in the TPM-biopsy protocol and did not include TRUS-guided biopsy, which is the current standard test. 138

Service provision and health-care system issues

We have shown that high-quality mpMRI can be delivered in a multicentre setting in the UK NHS. However, we have not considered the impact that this could have on health-care settings from a capacity perspective in terms of access to high-quality mpMRI and have not formally evaluated aspects of radiology training to ensure high-quality reporting nor the need or otherwise of high-quality MRI-targeted biopsies.

In health-care systems where rates of prior PSA testing are higher, the proportion of men with CS prostate cancer according to our definition is likely to be lower. If mpMRI were used as a triage test in these settings, the proportion of men who could potentially avoid a primary biopsy might be larger. Furthermore, the recent ProtecT trial9 demonstrates no cancer-specific survival benefit over active monitoring when men are treated with RP or radical radiotherapy, although there was a beneficial effect in reducing time to metastases. 139 With the majority of patients having low-risk disease in ProtecT, weight is added to the need for strategies, such as mpMRI, that reduce the diagnosis of CNS cancer and are better able to identify higher-risk disease. We acknowledge that, in the longer term, the number and type of patients being referred to secondary care may change depending on the future validation of potential screening strategies. 140

We note that, although a positive MRI result is associated with a very high OR of harbouring CS prostate cancer and, as a result, has dominated prediction models that incorporate MRI and blood or urine biomarkers,141 there is still much work to do in this area to explore the incremental value of using an inexpensive and easily applied biomarker as a triage test prior to MRI or as a companion test, particularly in relation to an ‘indeterminate’ MRI scan. Many of the MRI studies being planned will incorporate urine and blood-based biomarkers as well as existing validated risk calculators. 140

Limitations of the diagnostic study

PROMIS had some limitations. First, TPM-biopsy is limited by the size of the template grid equipment and interference with needle insertion from the bony pubic arch. This meant that some men with very large prostates (> 100 ml) had to be excluded from the study, which may result in a slight decrease in the proportion of true negatives in the study population. However, although it is too invasive for routine use in the clinic, TPM-biopsy offered the precision required for a highly accurate reference test by virtue of its uniform 5-mm sampling density over the entire prostate. 142

Second, we acknowledge that PROMIS represents a selected group, although it is encouraging that men who were subsequently withdrawn from the study did not differ from those who completed it in terms of the baseline data we collected. We also acknowledge that these men may differ in terms of other characteristics that we did not measure.

Third, some systematic error may have been introduced into the standard test by the sequence in which TPM-biopsy and TRUS-guided biopsy were performed, and we were unable to control for this. Carrying out TPM-biopsy before TRUS-guided biopsy may have contributed to the poor accuracy of the standard test because of swelling, distortion and tissue disruption. The sequencing was based on patient safety considerations and on the need to preserve the integrity of the reference test. Infection is a major risk associated with TRUS-guided biopsy because of breaching of the rectal mucosa. Participants might have been exposed to excess risk if bacteria had been introduced into the prostate prior to multiple needle deployments of an inherently sterile TPM-biopsy.

Fourth, because of the need for blinding, the targeting of suspicious lesions identified in MRI was not possible, and we cannot accurately assess the clinical utility of a MRI-targeted biopsy approach. Fifth, although we included some measurements of interobserver variability, these were between two expert readers. Further work is required to measure the interobserver variability of expert and non-expert reporters. Last, we acknowledge that likelihood ratios and area under the receiver operating characteristic curves were not part of the prespecified analysis plan. These metrics provide an overall measure of test performance and clarify the relative strengths and weaknesses of each test, particularly as likelihood ratios are independent of disease prevalence.

The mpMRI scans were performed using a magnetic field strength of 1.5 T. This was chosen because it is widely available and because we assumed that any benefit shown was likely to be at least maintained at higher magnetic field strengths, where it is possible to achieve greater signal-to-noise ratios.

Finally, we acknowledge the benefits of communicating risk using the 1–5 ordinal scale conferred by the Gleason grade group. 143 Our end points cannot, unfortunately, be reclassified using Gleason grade group as they incorporated histopathology for our primary end point (Gleason grade 4 + 3 or Gleason grade group 3) and secondary end point (Gleason grade 3 + 4 or Gleason grade group 2), and the cancer core length thresholds, which are not part of the Gleason grade group definitions. One of the principal outcomes of PROMIS is that a normal MRI result was not associated with any Gleason grade group of ≥ 3.

Recommendations for further research

PROMIS is a unique data set that will continue to be an important resource for further research. Future work is likely to include the following.

• modelling of prognostic factors for CS cancer and development and validation of a risk score calculator for the presence of CS cancer

• radiological issues:

  • optimisation of sequencing (T2, DW, DCE)

  • reporting methodology

  • training and quality assurance for scanners and radiologists

• pathology issues:

  • characterisation of disease

  • correlation between TPM-biopsy and MRI lesion location

• biomarker/translational research analyses using the subset of blood and urine samples that have been collected.

Conclusions

Transrectal ultrasound-guided biopsy performs poorly as a diagnostic test for CS prostate cancer. Multiparametric magnetic resonance imaging identifies a large proportion of men with suspected CS cancer who will then have TRUS-guided biopsy for diagnosis, in which the biopsy strategy should be guided by the mpMRI findings. Using mpMRI as a triage test avoids biopsy in up to 27% of men with non-CS cancer or no cancer who were referred for testing. However, mpMRI as a triage test missed 17 men with CS cancer (3% of the whole PROMIS cohort) as a result of a biopsy not being performed.

Main findings

Multiparametric magnetic resonance imaging, TRUS-guided biopsy and TPM-biopsy can be used in a range of combinations and with different definitions and cut-off points to detect CS cancer in men referred to secondary care for further investigation. The combinations, definitions and cut-off points form 383 diagnostic strategies, which detect 15–100% of CS cancers, depending on the way that the tests re used in combination, their definitions for CS cancer and the cut-off point used. Strategies using the TRUS-guided biopsy definition 2, mpMRI definition 2 and cut-off point 2 detect the largest proportion of CS cancers compared with the same strategies using other definitions and cut-off points. The HRQoL consequences of testing are small and relate only to TPM-biopsy. The diagnostic strategies cost between £244 and £1653 for the UK NHS.

The quality-adjusted life expectancy and health-care costs of men diagnosed with prostate cancer at an average age of 67 years depend on the severity of their cancer and on their treatment. Men with low-risk cancer who are managed with active surveillance have an average life expectancy of 16.26 years. Men with intermediate-risk cancer who receive active surveillance have an average life expectancy of 13.38 years; this increases to 15.65 years if they are treated with radical treatment. Men with high-risk cancer who receive active surveillance have an average life expectancy of 11.15 years; this increases to 13.23 years if they are treated with radical treatment. These estimates translate to discounted lifetime health benefits and costs, which informed the outcomes of men undergoing each diagnostic strategy. In order to assess the cost-effective strategy, the results of the short-term model on the costs and outcomes of testing were combined with the results of the long-term model on the costs and outcomes of radical treatment and observation.

In the base case, using NHS reference costs,110 the cost-effective strategy at the health opportunity costs of £13,000, £20,000 and £30,000 per QALY gained involves testing all men with mpMRI at definition 2, cut-off point 2 for CS cancer, using MRI-targeted TRUS-guided biopsy at definition 2 to detect CS cancer and rebiopsying those men in whom CS cancer was not detected (substrategy M7 222). M7 222 achieves 8.72 discounted QALYs (95% CI 8.40 to 9.04 discounted QALYs) and costs £5367 (95% CI £4947 to £5876). M7 222 detects 95% of CS cancers (95% CI 92% to 98%).

A main driver of cost-effectiveness is the unit cost of each test, particularly TPM-biopsy. At a higher TPM-biopsy unit cost (e.g. £1872 per test), the cost-effective substrategy is M7 222 unless the unit cost of mpMRI is almost doubled and the unit cost of TRUS-guided biopsy is maintained or reduced. At a lower TPM-biopsy unit cost (e.g. £931), M7 222 is cost-effective if the unit cost of mpMRI is reduced, unless the £13,000 per QALY health opportunity cost is used. A reduced unit cost of TPM-biopsy favours the substrategies that include TPM-biopsy and that detect more CS cancers, such as P4 2––.

Under base-case assumptions, the results are generally robust to uncertain assumptions. Apart from M7 222 and P4 2––, the substrategies that emerge as potentially cost-effective in the sensitivity analysis are generally T7 222, T9 222 and P1. With T7 222 and T9 222, all men receive TRUS-guided biopsy as the first test and men classified as having no cancer or CNS cancer receive mpMRI. In T7, men with suspected CS cancer receive a second TRUS-guided biopsy (MRI guided). In T9, men with suspected CS cancer, as well as men with suspected CNS cancer on mpMRI but a prior no cancer result on the first biopsy, receive a second TRUS-guided biopsy (MRI guided). These strategies appear to be potentially cost-effective according to the TRUS-guided biopsy definition 2, mpMRI definition 2 and cut-off point of 2. In P1–––, all men receive TPM-biopsy and no further tests.

The value-of-information analysis indicates the magnitude of health losses because of uncertainty in the parameter estimates. It suggests that up to 0.015 QALYs per man referred for testing could be gained if better information was available. Given the large number of men referred for testing each year in the UK, this relatively small health loss translates into a large loss at the population level. Future research that resolves this uncertainty is worth up to £9.3M. More research is required to be conducted on which specific parameters most contribute to the health losses attributable to uncertainty, in order to better target investments in future research.

Comparison with other studies

The cost-effectiveness of mpMRI in combination with TRUS-guided biopsy has been subject to increasing research. A recent systematic review on the cost-effectiveness of mpMRI for the diagnosis of prostate cancer144 found five studies, although only three compared mpMRI as a first screening test in combination with TRUS-guided biopsy in an incremental analysis. 13,145,146 In addition, another study has been published more recently. 147 These studies compared up to two ways of using mpMRI, either as a first screening test to determine which men should receive MRI-targeted TRUS-guided biopsy13,146,147 or as MRI-targeted TRUS-guided biopsy for all men13 or for men with a previous negative biopsy result. 145 None of the studies had access to individual patient data; therefore, all inputs were obtained from the literature or expert opinion. As in the current evaluation, mpMRI was found to be cost-effective as a first test before TRUS-guided biopsy.

Two of these studies were conducted in a UK setting. 13,145 Mowatt et al. 145 compared targeted TRUS-guided biopsy with systematic extended core TRUS-guided biopsy. The population was men who had suspected prostate cancer with a prior negative or inconclusive biopsy and with an indication for repeat biopsy. This study found T2-weighted MRI-targeted TRUS-guided biopsy to be cost-effective, although the health gains and additional costs compared with the least costly strategy (systematic extended-core TRUS-guided biopsy) were very small.

The economic evaluation conducted for the 2014 NICE clinical guideline13 compared a strategy of mpMRI followed by MRI-targeted TRUS-guided biopsy in men with a MRI scan suspicious of cancer with TRUS-guided biopsy. The population was newly referred men with symptoms of prostate cancer. The mpMRI + TRUS-guided biopsy strategy achieved lower health benefits, but was also less costly, than the TRUS-guided biopsy alone. Although the ICER for MP-MRI+TRUS-guided biopsy compared with TRUS-guided biopsy alone was approximately £16,000 per QALY, which is generally considered to be cost-effective, the cost savings were found to be insufficient to justify the lower health benefits.

Strengths

To our knowledge, this is the first study to examine the cost-effectiveness of all of the possible ways of using mpMRI, TRUS-guided biopsy and TPM-biopsy to diagnose prostate cancer and distinguish between CNS and CS cancer. Having access to the individual patient data from the PROMIS clinical study facilitated the comparison of not only the different combinations of tests, but also different definitions and diagnostic cut-off points to define the disease. This makes this study the most comprehensive cost-effectiveness analysis of alternative diagnostic strategies for prostate cancer to date.

This evaluation was informed primarily by the diagnostic information collected in PROMIS. PROMIS is the first multicentre study to evaluate the diagnostic performance of TRUS-guided biopsy and mpMRI in a blinded fashion compared with a reference test of TPM-biopsy. Therefore, it forms the best available evidence on TRUS-guided biopsy and mpMRI. Additional information on the diagnostic performance of repeat tests and tests performed in combination was sought from the literature, and recent systematic reviews were used whenever possible. 23,105

This study explicitly distinguishes between the short-term outcomes of the diagnostic strategies and the long-term outcomes from the subsequent management decisions informed by diagnosis. Furthermore, it overcomes the limitations in the available evidence on long-term outcomes with a novel calibration method to obtain the quality-adjusted survival and lifetime costs. Extensive sensitivity analysis explores the sensitivity of the results to parameter uncertainty and to assumptions regarding the management of men in clinical practice. Probabilistic sensitivity analysis indicates the impact that parameter uncertainty had on the results. Value-of-information analysis shows the magnitude of health losses attributable to parameter uncertainty and the potential for investment in future research.

Limitations

This study has some limitations, mostly related to the fact that evidence on the long-term outcomes of men with prostate cancer and on the diagnostic performance of tests in repeated testing episodes is itself limited. Data on the long-term outcomes of men with prostate cancer, by risk subgroup and management, were obtained from a recently published RCT, PIVOT. 8 Given that only summary data were available, we developed a method to predict quality-adjusted survival and costs from information on overall survival, cumulative incidence of metastases and mortality following metastases. It assumes that the probabilities of disease progression and death are constant over time, which is unrealistic. However, this is unlikely to have a major impact on the results because the aim is to obtain long-term outcomes on average for radical treatment or active surveillance, rather than selecting between treatments.

Importantly, the long-term outcomes of men with prostate cancer who are incorrectly classified as having no cancer were assumed to be equivalent to those of men allocated to active surveillance. These men would not be recruited for studies on prostate cancer; therefore, the consequences of misdiagnosis are, by definition, unknown. The cost-effectiveness of less-sensitive diagnostic strategies is favoured if the long-term outcomes of men who are incorrectly classified as having no cancer were overestimated. The implication is that the base case may be biased towards M7 222 and against P4 2–– and P1–––, which detect all cancers. For this reason, the sensitivity analysis establishes the minimum reduction in health outcomes that misclassified men would need to experience to change the cost-effective strategy. A reduction as small as 0.01 QALYs, equivalent to 3.7 days in full health, changes the cost-effective strategy at the health opportunity cost of £30,000 per QALY. Reductions of ≥ 0.1 QALYs, equivalent to 37 days in full health, change the strategies at £13,000 and £20,000 per QALY. As a benchmark, 1–2% increases in the hazard of all-cause death (HR 1.01–1.02) reduce life expectancy by 0.07–0.13 years in men with intermediate-risk cancer. The implication is that small changes in survival may have a large impact on the results. Therefore, the follow-up of men with suspected prostate cancer but in whom cancer was not detected requires further investigation and may have an impact on the most cost-effective diagnostic strategy.

It was not possible to account for the impact of repeated tests over time on the outcomes and costs of men who are managed with active surveillance. As discussed earlier, these men are likely to be different in their cancer characteristics from the men first referred for diagnosis. Furthermore, the cancer may progress in size and differentiation at an unknown rate. To some extent, the impact of repeated testing is included in the long-term outcomes, having been based on the active surveillance group of PIVOT,8 which included repeated testing and the possibility of treatment over time. However, if the sensitivity of the watchful waiting protocol in PIVOT is lower than that of the active surveillance protocol recommended by NICE,13 the long-term outcomes of men misclassified as having low-risk cancer may have been underestimated. This was the rationale of the threshold sensitivity analysis to find the smallest proportion of reclassified men that would change the cost-effective strategy. However, it is difficult to relate this minimum proportion to the sensitivity of the active surveillance protocol recommended by NICE. Consequently, more research is required on the sensitivity of repeated testing and its implications upstream to the cost-effectiveness of diagnostic strategies.

Remaining areas of uncertainty and areas for further research

The cost-effective diagnostic strategy depends on the cost-effectiveness of active treatment; however, the long-term outcomes of men with prostate cancer remain uncertain and the evidence base, perhaps, remains insufficient. The 2014 NICE clinical guideline13 recommends that men with low-risk cancer should be offered active surveillance, whereas men with intermediate- and high-risk cancer should be offered radical treatment. These recommendations are mostly based on expert consensus rather than formal quantitative evaluations to identify for whom treatment should be recommended, how to monitor men with lower-risk cancer and which treatments should be prioritised. The lack of evidence on management decisions has upstream implications for the cost-effectiveness of diagnostic strategies because their costs and benefits are partly determined by the costs and benefits of active surveillance and treatment. If treatment is more beneficial than suggested by the evidence from PIVOT,8 the benefits of more-sensitive strategies have been underestimated. Conversely, if treatment is less beneficial, the benefits of less-sensitive strategies have been underestimated. This is explored in the threshold sensitivity analysis, which showed that the less-sensitive and less-costly diagnostic strategies become cost-effective for a 15–20% reduction in the clinical effectiveness of radical treatment.

The ProtecT study9, which compared RP, radiotherapy and active monitoring in men with prostate cancer, may help resolve this uncertainty. At the time of conducting this analysis, the 10-year follow-up results are available, but only for the entire trial population. Active treatment showed no effect on mortality, although few deaths occurred, but a significant reduction in cancer progression (HR 0.39, 95% CI 0.27 to 0.54). This is similar to the results of PIVOT8 (relative risk for bone metastasis 0.44, 95% CI 0.25 to 0.76). However, this refers to the entire trial population, almost 80% of whom had low-risk cancer and, therefore, would not be offered active treatment according to NICE’s recommendations. 13 Evidence is required on the effect of active treatment specifically in men with intermediate- and high-risk (CS) cancer, for whom the NICE guidelines13 recommend treatment.

Moreover, the predicted long-term outcomes are affected by the misclassification of men by TRUS-guided biopsy in PIVOT. 8 In PIVOT, patients were diagnosed, enrolled and classified by risk on the basis of a biopsy result and not on the basis of their true disease status. The biopsy is an imperfect test in that it produces a considerable proportion of false-negative results, with evidence from PROMIS suggesting that only 34% of intermediate-risk patients are identified as having CS cancer. The implications are twofold. First, the men enrolled in PIVOT may not be representative of the men with cancer in the general population. This is only an issue if the patients who were missed by the biopsy have a different prognosis from the prognoses of the patients enrolled in PIVOT. Second, the PIVOT risk groups may have been misclassified and this may introduce bias in the prediction of long-term outcomes and the relative clinical effectiveness of radical treatment, and, therefore, may introduce bias to the benefits of correctly detecting CS cancer. Quality-adjusted survival may have been underestimated in the high-risk subgroup if the sensitivity of TRUS-guided biopsy is higher for higher cancer severity; hence, the high-risk subgroup may consist of men with the most severe cancer, who have worse prognoses. Because only 18 out of 576 men in PROMIS had high-risk cancer, any bias arising is unlikely to affect the results.

Quality-adjusted survival may also have been underestimated in the intermediate- and low-risk subgroups in PIVOT if these were contaminated by more severe patients. This may be an issue particularly for long-term outcomes post radical treatment if its effects are more pronounced in the intermediate-risk subgroup. If the quality-adjusted survival of men treated with radical treatment was more underestimated than that of men managed with observation, the benefits of more-sensitive strategies in detecting CS cancer may have been underestimated. This issue can be resolved only with better-quality evidence on the outcomes of men with prostate cancer, based on a perfect test such as TPM-biopsy for their diagnosis and classification.

There is some uncertainty regarding the costs of the tests, particularly TPM-biopsy, which affects the results. The NHS reference cost for TPM-biopsy is approximately 25% higher than the PbR tariff, which makes strategies with TPM-biopsy less likely to be cost-effective.

In principle, the unit costs of resource use should reflect the cost to the budget holder under whose perspective the analysis is conducted (i.e. the UK NHS). 148 This would suggest that appropriate unit costs should reflect what the NHS actually pays external individuals or organisations – in this case, what hospitals pay staff and the suppliers of equipment, services and consumables. Hence, unit costs would be based on the mean reference costs reported by relevant hospital trusts. However, the UK NHS is devolved into a range of organisations: NHS England, clinical commissioning groups, hospital trusts, etc. When a patient is managed in secondary care, the hospital is generally reimbursed by the relevant clinical commissioning group that has referred them, according to the PbR tariff. The PbR tariff is informed by NHS reference costs, adjusted for inflation and changes in efficiency. If the specific perspective of the analysis is that of clinical commissioning groups, for example, the PbR tariff may be the more relevant basis for estimating unit costs. The impact of changes in the unit costs is provided in the multivariate sensitivity analysis (see Chapter 7, Threshold sensitivity analysis).

The value-of-information analysis explored the expected costs to population health associated with parameter uncertainty and the maximum expected benefit associated with investment in further research. It shows that, although the expected health loss attributable to parameter uncertainty is small at the per-patient level, it sums to a considerable amount at the population level over 5 years, given the large number of men referred for investigation. In order to inform which research to commission in the future, the value-of-information analysis should be developed further. This could take a number of avenues. The expected value of perfect parameter information can indicate which parameters are driving the uncertainty and have greater consequences, and hence should be prioritised for investment. The expected value of sample information can be explored in order to help inform the most cost-effective study design and sample size to obtain information on a specific parameter. There is also an important question regarding the structural uncertainties in the model. Specifically, the model assumes that men receiving expectant management have the outcomes of men in the observation arm of PIVOT. However, in practice, these men may receive active treatment over time. This was explored in the scenario analysis. However, future research could extend the model to account for this assumption as a parameter and explore its impact in the value-of-information analysis. There is little guidance on how to develop this work in practice and future methodological research may address this gap.

The results suggest that the most cost-effective strategy is using mpMRI as the first test, rather than TRUS-guided biopsy. It may be difficult to implement this change, because it implies a shift in resources from histopathology to imaging departments. Although financial resources are more straightforward to transfer, there may be issues around capacity, namely related to skilled staff and capital investment in equipment. Consequently, more consideration and research may be required on how best to implement this policy and manage change.

This study suggests that mpMRI is cost-effective as the first test for the diagnosis of prostate cancer when followed by MRI-targeted TRUS-guided biopsy in men in whom mpMRI suggests suspicion for CS cancer and a second biopsy if no CS cancer is found. The cost-effective strategy uses the most sensitive definitions and cut-off point for CS cancer: definition 2 for TRUS-guided biopsy (any Gleason pattern of ≥ 4 and/or cancer core length of ≥ 4 mm) and definition 2 for mpMRI (lesion volume of ≥ 0.2 ml and/or a Gleason score of ≥ 3 + 4) at the cut-off point of ≥ 2 on the Likert scale for suspicion of CS cancer.

This study makes it clear that the value of the diagnostic test depends on the value of the treatment options that follow. Here, the most cost-effective strategy is one that detects almost all men with CS cancer because active treatment was found to be highly cost-effective in this population. The findings are sensitive to the costs of the tests, hence the motivation for the bivariate sensitivity analysis indicating the most cost-effective strategy for a wide range of unit costs of each test. This can inform decisions in contexts facing different sets of prices. The findings are also sensitive to the sensitivity of MRI-targeted TRUS-guided biopsy compared with TRUS-guided biopsy alone, which was tested in the threshold sensitivity analysis. As with any research, the findings are grounded on the evidence used. The cost-effectiveness of diagnostic tests depends on the cost-effectiveness of subsequent treatment decisions. If radical treatment is less beneficial than PIVOT suggests, the scope for investment in diagnosis is smaller, and less-sensitive and less-costly strategies are favoured. This study makes clear how the value of treatment can have upstream implications for the value of the diagnostic tests.

Trial Management Group

• Professor Mark Emberton (Chief Investigator; Urologist), UCLH.

• Mr Hashim Ahmed (Co-chief Investigator; Urologist), UCLH.

• Dr Ahmed El-Shater Bosaily (Clinical Research Fellow), UCLH.

• Dr Alex Kirkham (Radiologist), UCLH.

• Dr Alex Freeman (Pathologist), UCLH.

• Dr Charles Jameson (Pathologist), UCLH.

• Mr Richard Graham Hindley (Urologist), Basingstoke and North Hampshire Hospital.

• Dr Christopher Parker (Translational Research), Royal Marsden Hospital.

• Professor Colin Cooper (Translational Research), Royal Marsden Hospital.

• Robert Oldroyd (Patient Representative).

• Professor Richard Kaplan (Programme Lead; Oncologist), Medical Research Council Clinical Trials Unit (MRC CTU at UCL).

• Dr Louise Brown (Project Lead; Statistician), MRC CTU at UCL.

• Dr Rhian Gabe (Statistician), University of York.

• Dr Yolanda Collaco-Moraes (Clinical Operations Manager), MRC CTU at UCL.

• Cybil Adusei and Katie Ward (Trial Managers), MRC CTU at UCL.

• Sophie Stewart, Katie Thompson, Claire Mulrenan, Hannah Gardner and Alex Maytum (Data Managers), MRC CTU at UCL.

• Carlos Diaz-Montana (Data Programmer), MRC CTU at UCL.

• Dr Chris Coyle (Clinical Fellow), MRC CTU at UCL.

• Professor Mark Sculpher (Health Economics), University of York.

• Ms Rita Faria (Health Economics), University of York.

• Dr Eldon Spackman (Health Economics), University of York.

Trial Steering Committee (also acted as Data Monitoring Committee)

• Dr David Guthrie (Chairperson; Oncologist), Derbyshire Royal Infirmary.

• Professor John Chester (Oncologist), University of Cardiff.

• Professor Richard Cowan (Oncologist), Christie Hospital Manchester.

• Professor Michael Jewitt (Urologist), University of Toronto.

Participating centres (number of men registered and team members)

• University College London Hospital (304): H Ahmed [Principal Investigator (PI)], A El-Shater Bosaily, M Emberton, A Kirkham, S Punwani, A Freeman, C Jameson, M Hung, J Coe, V Abu, R Scott and M Ahmed.

• Basingstoke and North Hampshire Hospital (130): R Hindley (PI), P Aslet, A Emara, A Chetwood, D Peppercorn, A Thrower, H El-Mahallawi, A Mustajab, H Alkhazaraji, A Edwards and J Smith.

• Southampton General Hospital (55): T Dudderidge (PI), J Dyer, J Smart, K Tung, H Markham, B Birch, R Lockyer, A Lodge, L Dando and J Gwilt.

• Southmead Hospital, Bristol (44): R Persad (PI), M Elmahdy, S Pandian, D Gillatt, J Ash-Miles, M Sohail, C Shiridzinomwa and A Treasure.

• Charing Cross Hospital, Imperial College London (41): M Winkler (PI), T Barwick, V Stewart, L Honeyfield, N Qazi, B Statton, N Ngo, K Ansu, S Edwards and E Temple.

• Wrexham Maelor Hospital (39): I Shergill (PI), S Agarwal, K Pradeep, M Atkinson and S Ackerley.

• Royal Hallamshire Hospital, Sheffield (36): D Rosario (PI), J Catto, F Salim, S Morgan and J Howson.

• Musgrove Park Hospital, Taunton (36): N Burns-Cox (PI), K Gordon, A Birring, A Maccormick, P Burn, D Paterson and H Routley.

• Frimley Park Hospital (25): S Bott (PI), H Evans, G Kousparos, AM Silvanto, A Mann, J Amero, A Pilcher and N Pereira.

• The Whittington Hospital (18): M Ghei (PI), TT Shah, J Kumaradevan, A Trinidade, R Katz, D Arul, L Harbin, V Conteh and A Verjee.

• Maidstone Hospital (12): A Henderson (PI), S Ghosh, P McMillan, J James, G Russell, E Bernsten, R Casey, T Nolan and D Day.

Contributions of authors

Louise Clare Brown contributed to the diagnostic study design, the statistical analysis, the first draft of the manuscript (diagnostic study section) and the interpretation and critical review of the manuscript (diagnostic study section).

Hashim U Ahmed contributed to the diagnostic study design, the statistical analysis, the first draft of the manuscript (diagnostic study section) and the interpretation and critical review of the manuscript (diagnostic study section).

Rita Faria developed the economic evaluation design, undertook the economic modelling and analysis, wrote the first and subsequent drafts of the economic chapters, and critically revised the economic chapters.

Ahmed El-Shater Bosaily contributed to the acquisition of data, diagnostic test reporting and the first draft of the manuscript (diagnostic study section).

Rhian Gabe contributed to the diagnostic study design and statistical analysis.

Richard S Kaplan contributed to the diagnostic study design, the statistical analysis and the interpretation and critical review of the manuscript (diagnostic study section).

Mahesh Parmar contributed to the interpretation and critical review of the manuscript (diagnostic study section).

Yolanda Collaco-Moraes contributed to the acquisition of data and diagnostic test reporting.

Katie Ward contributed to the acquisition of data and diagnostic test reporting.

Richard Graham Hindley contributed to the acquisition of data (as a PI in a participating site).

Alex Freeman contributed to the acquisition of data and diagnostic test reporting.

Alexander Kirkham contributed to the acquisition of data and diagnostic test reporting.

Robert Oldroyd contributed to the diagnostic study concept and initial design and the interpretation and critical review of the manuscript (diagnostic study section).

Chris Parker contributed to the diagnostic study concept and initial design and the interpretation and critical review of the manuscript (diagnostic study section).

Simon Bott contributed to the acquisition of data (as a PI in a participating site).

Nick Burns-Cox contributed to the acquisition of data (as a PI in a participating site).

Tim Dudderidge contributed to the acquisition of data (as a PI in a participating site).

Maneesh Ghei contributed to the acquisition of data (as a PI in a participating site).

Alastair Henderson contributed to the acquisition of data (as a PI in a participating site).

Rajendra Persad contributed to the acquisition of data (as a PI in a participating site).

Derek J Rosario contributed to the acquisition of data (as a PI in a participating site).

Iqbal Shergill contributed to the acquisition of data (as a PI in a participating site).

Mathias Winkler contributed to the acquisition of data (as a PI in a participating site).

Marta Soares supervised the economic evaluation design, the economic modelling and analysis, and critically reviewed the economic chapters.

Eldon Spackman commented on the economic evaluation design, modelling and analysis, conducted the analysis of the impact of the tests on health-related quality of life, and critically reviewed the economic chapters.

Mark Sculpher oversaw the economic evaluation design, modelling and analysis, and critically reviewed the economic chapters.

Mark Emberton contributed to the diagnostic study concept and initial design, the first draft of the manuscript (diagnostic study section) and the interpretation and critical review of the manuscript (diagnostic study section).

All authors reviewed and approved the final manuscript.

Publications

El-Shater Bosaily A, Parker C, Brown LC, Gabe R, Hindley RG, Kaplan R, et al. PROMIS – prostate MR imaging study: a paired validating cohort study evaluating the role of multi-parametric MRI in men with clinical suspicion of prostate cancer. Contemp Clin Trials 2015;42:26–40.

Ahmed HU, El-Shater Bosaily A, Brown LC, Gabe R, Kaplan R, Parmar MK, et al. Diagnostic accuracy of multi-parametric MRI and TRUS biopsy in prostate cancer (PROMIS): a paired validating confirmatory study. Lancet 2017;389:815–22.

Faria R, Soares MFO, Spackman DE, Hashim A, Brown L, Kaplan R, et al. Optimising the diagnosis of prostate cancer in the era of multi-parametric magnetic resonance imaging: a cost-effectiveness analysis based on the Prostate MR Imaging study (PROMIS). Eur Urol 2018;73:23–30.

Wilt TJ, Ahmed HU. Prostate cancer screening and the management of clinically localized disease. BMJ 2013;346:f325.

El-Shater Bosaily A, Valerio M, Hu Y, Freeman A, Jameson C, Brown L, et al. The concordance between the volume hotspot and the grade hotspot: a 3-D reconstructive model using the pathology outputs from the PROMIS trial. Prostate Cancer Prostatic Dis 2016;19:258–63.

Data sharing statement

University College London encourages data sharing and we propose to make PROMIS an openly accessible research study for future investigators. All available data can be obtained by contacting the corresponding author.

Patient data

This work uses data provided by patients and collected by the NHS as part of their care and support. Using patient data is vital to improve health and care for everyone. There is huge potential to make better use of information from people’s patient records, to understand more about disease, develop new treatments, monitor safety, and plan NHS services. Patient data should be kept safe and secure, to protect everyone’s privacy, and it’s important that there are safeguards to make sure that it is stored and used responsibly. Everyone should be able to find out about how patient data are used. #datasaveslives You can find out more about the background to this citation here: https://understandingpatientdata.org.uk/data-citation.

Disclaimers

This report presents independent research funded by the National Institute for Health 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, NETSCC, 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, NETSCC, the HTA programme or the Department of Health and Social Care.

Centres

Participants were recruited from the following NHS hospitals in England: UCLH (lead centre), Basingstoke and North Hampshire Hospital, Imperial College/Charing Cross Hospital, Musgrove Park Hospital, Maidstone Hospital, Southmead Hospital, Whittington Hospital, Wrexham Maelor Hospital, Royal Hallamshire Hospital, Frimley Park Hospital and Southampton General Hospital.

Screening

Screening log data were of adequate quality in 6 of the 11 centres. These centres provided 608 of the 740 registrations. In summary, 608 out of 1618 men (38%) who were screened at these centres were registered into the study. However, about half of the 1010 non-registered men were ineligible according to the entry criteria (n = 450) and in 68 men there was a valid clinical reason for non-entry. For 378 men (37%), the reason for not registering was patient or clinician refusal; in addition, 11 men (1%) experienced delays but would have been eligible. According to the baseline characteristics we were able to collect on these non-registered men, there were no strong differences in age or PSA level between the 389 men who were either delayed or refused entry and those who consented to be registered into PROMIS.

Results of diagnostic tests (supplementary)

Case illustrations

Figure 41 illustrates a case showing correlation of TRUS-guided biopsy and mpMRI with TPM-biopsy for a man harbouring CS cancer. Axial mpMRI scans are shown in a man with a PSA level of 12.4 ng/ml and normal DRE. The mpMRI shows a 39-ml prostate volume and a right-sided lesion (Likert score of 5/5) on (clockwise from top left) T2-weighted, diffusion-weighted high b-value and ADC map and dynamic contrast enhanced images (all other areas had a Likert score of 1–2/5). TRUS-guided biopsy detected no cancer and TPM-biopsy found a Gleason score of 4 + 3 and a cancer core length of 4 mm in a total of 8 out of 70 scores taken (primary definition of CS prostate cancer) that was concordant with the mpMRI. On the TPM-biopsy diagram, each white or coloured box represents an area biopsied with one core. TRUS-guided biopsy showed no cancer.

FIGURE 41.

A case showing correlation of TRUS-guided biopsy and mpMRI with TPM-biopsy for a man harbouring CS cancer. (a) mpMRI; (b) TPM-biopsy; and (c) TRUS-guided biopsy. ASAP, atypical small acinar proliferation; G3, Gleason 3; HGPIN, high-grade prostate intraepithelial neoplasia; Inflamm, inflammation; ISUP, International Society of Urological Pathology; Lymphovas, Lymphovascular.

Figure 42 illustrates a case showing correlation of TRUS-guided biopsy and mpMRI with TPM-biopsy for a man harbouring no cancer. It shows a series of images demonstrating mpMRI in a man with a PSA level of 8.9 ng/ml and a normal DRE, showing a 45-ml prostate volume and no suspicious area lesion on mpMRI (T2-weighted image only shown). This was reported as a Likert score of 2. TPM-biopsy and TRUS-guided biopsy detected no cancer.

FIGURE 42.

A case showing correlation of TRUS-guided biopsy and mpMRI with TPM-biopsy for a man harbouring no cancer. (a) mpMRI; (b) TPM-biopsy; and (c) TRUS-guided biopsy. ASAP, atypical small acinar proliferation; G3, Gleason 3; HGPIN, high-grade prostate intraepithelial neoplasia; Inflamm, inflammation; ISUP, International Society of Urological Pathology; Lymphovas, Lymphovascular.

Clinical implications of introducing multiparametric magnetic resonance imaging

Two scenarios were explored. In the best-case scenario (Figure 43), the results of mpMRI would be used to direct TRUS-guided biopsy. In the worst-case scenario (Figure 44), TRUS-guided biopsy after mpMRI would be conducted without direction from mpMRI.

FIGURE 43.

Flow chart showing best-case scenario (TRUS-guided biopsy directed by mpMRI).

FIGURE 44.

Flow chart showing worst-case scenario (TRUS-guided biopsy not directed by mpMRI).

Agreement between sites for multiparametric magnetic resonance imaging measurements

In order to determine whether or not the lead radiology site (UCLH, responsible for training all other sites) demonstrated different diagnostic accuracy from the non-UCLH sites, the results were compared for the primary outcome analysis between mpMRI definition 2, cut-off point 3 and TPM-biopsy definition 1 variables for the detection of CS cancer. The results were almost identical (Tables 47 and 48 and Figures 45 and 46). For both groups, the proportion of CS cases that were missed was about 3% and the proportion of men who may be able to avoid a biopsy was about 27%.

FIGURE 45.

Results for detection of CS cancer for mpMRI (definition 2, cut-off point 3) relative to TPM-biopsy (definition 1) for the UCLH site alone.

FIGURE 46.

Results for detection of CS cancer for mpMRI (definition 2, cut-off point 3) relative to TPM-biopsy (definition 1) for non-UCLH sites.