The INTRABEAM® Photon Radiotherapy System for the adjuvant treatment of early breast cancer: a systematic review and economic evaluation

Breast cancer aetiology

Breast cancer, in common with all other cancers, is caused by deoxyribonucleic acid (DNA) mutations that disrupt the normal maintenance of cellular identity, growth and differentiation. 3 The majority of breast and other cancers develop from somatic mutations3,4 resulting from errors in processes such as DNA replication, DNA modification or DNA repair,4,5 which in turn may be influenced by environmental and/or dietary factors. 6 A small proportion of cancer types arise from inheritable single-gene disorders,3 for example BRCA1 (breast cancer 1) and BRCA2 (breast cancer 2) are genes associated with inheritable breast cancer. 4,7–9

There are two main forms of breast cancer: non-invasive, in which the cancer cells have not spread; and invasive, in which the breast cancer cells can potentially spread to the surrounding breast tissue or beyond. Approximately 10% of newly diagnosed breast cancer cases are non-invasive, the majority (approximately 90%) being ductal carcinoma in situ (DCIS). 10 In DCIS, cancer cells have developed inside milk ducts but have not yet developed the ability to spread beyond the ducts. DCIS is usually identified by mammography as it rarely presents as a lump. The remaining 90% of newly diagnosed breast cancer cases are various types of invasive breast cancer.

When breast cancer is diagnosed, information is gathered to describe and classify it according to a variety of characteristics. Much of the information required can be obtained only from samples taken during surgical removal of the primary tumour. Key aspects include:11

• histological type (e.g. invasive ductal carcinoma, invasive lobular carcinoma)

• histological grade, ranging from low (generally slow growing) to high (generally fast growing)

• stage, based on the tumour node metastasis (TNM) classification (Tables 1 and 2)

• oestrogen receptor (ER) alpha status

• human epidermal growth factor receptor-2 (HER-2) status

• DNA profile.

This information is essential for deciding what local and systemic treatments may be required and provides information about prognosis. The focus of this assessment is the treatment of early breast cancer; however, it should be noted that there is no internationally agreed single definition of early breast cancer (e.g. in terms of TNM stage). Typically, however, early breast cancer would be classified as TNM stage I or II (either IIa or IIb), with potentially some stage III tumours (those for which treatment could be curative).

The aim of treatment for early breast cancer is to provide a cure. As already stated, there are two major categories of early breast cancer: non-invasive (in situ) disease (predominantly in the form of DCIS) and invasive cancer. 11 For invasive cancer to be categorised as early breast cancer, the tumour should not have spread beyond the breast or the lymph nodes (which remain mobile) in the armpit ipsilateral to (on the same side as) the affected breast. 13 Once an invasive cancer has spread to distant sites (which may occur after initial treatment with curative intent), it is no longer curable, but can be treated to control symptoms.

Breast cancer epidemiology

In England, in 2011, the age-standardised rates of breast cancer incidence per 100,000 of the population were 124.8 for women and 0.9 for men. 1 For the period 2008–10 the age-standardised rate for women in England was 125.7 [95% confidence interval (CI) 125.0 to 126.4]. 14 The strongest risk factor for breast cancer is increasing age and, consequently, over 80% of new diagnoses of breast cancer in England are in women aged > 50 years1 and in men aged > 60 years. 1 Other important risk factors include obesity, alcohol consumption and lack of physical activity, which are estimated to be linked to about 18.5% of UK female breast cancer cases. 15

There were 9702 deaths of women and 64 deaths of men from breast cancer in England in 2011. 16 The UK age-standardised mortality rate from breast cancer per 100,000 women in 2008–10 was 25.3 (95% CI 25.0 to 25.6 per 100,000 women). 14 For women diagnosed with breast cancer during 2004–6 and followed up to 2011, the age-standardised 1-year survival rate for all breast cancers was 94.7% and the 5-year survival was 83.3%. 17 Between 2002 and 2006, a statistically significant annual increase in 1-year survival of 0.3% and in 5-year survival of 0.9% was observed. 17 The rise in survival estimates has been due to earlier detection and improved treatment of breast cancer in women. 2 An analysis of survival by stage at diagnosis for women in the UK diagnosed with invasive breast cancer (DCIS was excluded) during 2000–718 reported 1-year and 3-year net survival as shown in Table 3.

Breast cancer diagnosis

In England, the main routes to diagnosis for the majority of breast cancer cases are via the NHS Breast Cancer Screening Programme or urgent (2-week wait) referrals from a general practitioner (GP) due to a suspicion of cancer. The Breast Cancer Screening Programme targets women aged 50–69 years (with extension from 47 years to 73 years ongoing, and expected to be completed after 2016). In 2006–8, just over 50% of breast cancer cases in the 50–69 years age group were diagnosed through screening, whereas, in other age groups (< 50 years and ≥ 70 years), over 50% of cases were diagnosed through the urgent GP referral route. 19 Breast cancer screening aims to detect cancers at an early stage when they are too small to cause changes to the breast that can be observed or felt. In England in 2011–12, 40.7% (6403) of all the breast cancers detected by screening were invasive but small (< 15 mm in diameter). 20 In the case of breast cancers detected via routes other than screening, there are no regularly published data on stage of breast cancer at diagnosis;21 however, evidence suggests that the majority (at least 80%) of women are diagnosed with early disease (stage I or stage II) whatever their route to diagnosis. 22

The 2009 National Institute for Health and Care Excellence (NICE) guideline Early and Locally Advanced Breast Cancer: Diagnosis and Treatment11 provides recommendations for breast cancer diagnosis. Diagnosis is made after triple assessment consisting of a clinical assessment, mammography and/or ultrasound imaging, and core biopsy and/or fine-needle aspiration cytology. 11 A multidisciplinary team should review and discuss the test results and, if a cancer diagnosis is pathologically confirmed, suggest a treatment plan.

Breast cancer natural history and prognosis

The natural history of breast cancer is variable and incompletely understood. 23 If left untreated, a typical invasive breast cancer might progress in the following manner. Initially, the breast cancer cells multiply, thereby increasing the size of the tumour;24 as the tumour proliferates, the risk that metastatic cells will be generated increases. 25 A key route for metastatic spread of breast cancer cells is via the lymphatic system. If a breast cancer spreads, the first place it spreads to is often the first lymph node (or nodes) receiving direct lymphatic drainage from the tumour;24,25 this lymph node is called the sentinel lymph node. 26 The tumour can also spread to more distant lymph nodes and to systemic sites via the bloodstream (e.g. bone, lung, liver, brain). It is also possible for tumour cells to metastasise via the vascular system directly to systemic sites;25 however, not all breast cancers metastasise. Evidence from screening studies suggests that some screen-detected breast cancers may regress spontaneously27 and natural history may vary according to a variety of factors, for example genotype,28 hormone receptor status29 and race. 30

The heterogeneous nature of breast cancer natural history has an impact when trying to provide a prognosis and tools have been developed which aim to predict invasive breast cancer outcome. For example, the Nottingham Prognostic Index (NPI)31 (Table 4) is a tool that combines information on the size of the tumour, the number of lymph nodes involved and the histological grade to produce an overall score, with a higher score indicating a worse prognosis. Other models have been developed which aim to more accurately predict outcome by including alternative indicators and/or more explanatory factors, for example Predict33 and the Galway Index of Survival. 34 The program Adjuvant! enables prognostic estimates of outcome either with or without therapy to be produced based on estimates of individual patient prognosis and data on the efficacy of a range of adjuvant therapy options and is available online (www.adjuvantonline.com/index.jsp). 35

Impact of breast cancer

Psychological distress, chiefly in the form of anxiety, may be experienced by women from the initial diagnostic procedures for a suspected breast cancer36 through all stages of treatment and beyond. 37,38 In addition to psychological aspects, women may experience a range of physical problems, for example arm and breast symptoms and/or lymphoedema39,40 and fatigue. 40

An analysis of patients’ free-text comments from the Cancer Patient-Reported Outcome Measures (PROMs) Survey in England41 identified a range of issues that may affect patients diagnosed with breast cancer. These included poor body image following breast surgery, ongoing problems following surgery such as pain and lymphoedema and problems associated with other non-surgical treatments, for example hot flushes related to hormone treatments, burns following radiotherapy and neuropathy during and following chemotherapy. In addition, some patients found that existing comorbidities such as arthritis and osteoporosis were exacerbated by their treatment. Some survey respondents highlighted that, during and/or following treatment, a lack of energy affected their everyday life, and some found that they had cognitive problems and memory loss. Both during and after treatment some patients suffered from feelings of depression, loneliness and isolation. A continuing fear of recurrence was also an issue for some. Other problems highlighted by the survey were social and financial issues, for example relating to employment and obtaining insurance.

The impact of breast cancer for the NHS is likely to increase across all facets of the breast cancer care pathway in the future. This is because the population of England is growing in both size and age, which will lead to increasing rates of breast cancer given that the strongest risk factor for breast cancer is age.

Study design

• For the systematic review of clinical effectiveness, RCTs were eligible for inclusion. If the data from available RCTs were incomplete (e.g. absence of data on outcomes of interest), evidence from good-quality CCTs was eligible for consideration.

• For the systematic review of cost-effectiveness, full economic evaluations (cost-effectiveness, cost–utility or cost–benefit analyses) reporting on measures of both costs and consequences were eligible for inclusion.

• For the systematic review of HRQoL, primary research studies based in the UK, Europe, North America and Australasia were eligible for inclusion.

• Abstracts or conference presentations of studies were eligible for inclusion only if sufficient details were presented to allow an appraisal of the methodology and the assessment of results to be undertaken.

• Case series, case studies, narrative reviews, editorials and opinions were excluded, as were non-English-language studies. Systematic reviews and clinical guidelines were used only as a source of references.

Intervention(s)

• INTRABEAM Photon Radiotherapy System with or without post-operative WB-EBRT.

Comparator(s)

• External beam radiotherapy delivered by a linear accelerator.

Population

• For the systematic review of clinical effectiveness, people with early operable breast cancer (as defined by the trials).

• For the systematic review of HRQoL, people with breast cancer (not limited to early-stage breast cancer).

• People with a local recurrence were excluded.

Outcomes

Studies were included if they reported on one or more of the following outcomes:

• overall survival

• disease-free survival

• ipsilateral local recurrence

• adverse effects of treatment

• HRQoL

• cost-effectiveness [expressed in natural units such as life-years gained (cost-effectiveness analysis), quality-adjusted life-years (QALYs) (cost–utility analysis), or in monetary units (cost–benefit analysis)].

Quantity and quality of research available

Titles and, where available, abstracts of a total of 655 citations were screened and full copies of 44 references were obtained. Of these, 38 were excluded after inspection of the full article (see Appendix 2). The most common primary reason for exclusion was that the reference was an abstract containing insufficient details to allow appraisal of methodology and/or results (n = 25); a further eight records were excluded chiefly because the outcome was not relevant to the review, three records were excluded chiefly because of an incorrect intervention, one record was excluded on the basis of study design and one record was excluded because it related to an ongoing study (see Chapter 4, Ongoing studies). One RCT, the TARGeted Intraoperative radioTherapy Alone (TARGIT-A) trial, met the inclusion criteria for the review (Figure 1). The primary and secondary outcomes for the whole trial population were described by two full papers and three linked abstracts. Five substudies of the TARGIT-A trial which report outcome data from participants at just one or two centres were identified. Four of these substudies were excluded from this systematic review on the grounds of outcome (see Appendix 2). One substudy has been included which reports data on HRQoL from patients at one TARGIT-A trial centre. 63 Table 6 provides a summary description of the TARGIT-A study publications included in the clinical effectiveness systematic review.

FIGURE 1.

Flow chart for the identification of studies. a, See Chapter 4, Ongoing studies.

Overview of the TARGIT-A trial

The key characteristics of the TARGIT-A trial64,65 are shown in Table 7 with further details in the data extraction form (see Appendix 3). The TARGIT-A trial is the pivotal trial evaluating the concept of delivering a single dose of targeted IORT at the time of surgery using the mobile INTRABEAM Photon Radiotherapy System.

Design

The TARGIT-A trial is an international, multicentre, non-inferiority RCT that recruited participants in 33 centres in 11 countries including the UK (six centres), Europe (17 centres in six countries), the USA (seven centres), Canada (one centre) and Australia (two centres). The trial evaluated IORT using the INTRABEAM device compared with conventional WB-EBRT. The planned follow-up for trial participants was at least 10 years. 69 Median follow-up achieved for the most recent 2014 publication65 is 2 years 5 months.

As a non-inferiority trial, the RCT sought to determine whether or not INTRABEAM treatment was no worse than WB-EBRT. The pre-stated non-inferiority margin was an absolute difference of 2.5% in the primary end point (local recurrence) between groups. The 2.5% non-inferiority margin was chosen at the trial outset because it seemed clinically acceptable to both clinicians and patients. 64 However, it should be noted that, when the non-inferiority margin was chosen, the estimated local recurrence rate (LRR) (based on the literature available in 1999)70,71 was 6%, and since then recurrence rates have fallen. Two patient preference studies72,73 suggest that patients would be willing to accept an increase in the risk of local recurrence for the convenience of INTRABEAM treatment but it should be noted that these studies were conducted in countries in which WB-EBRT is typically delivered over 5–6 weeks and it is not known whether or not patient preference would be similar in England where WB-EBRT is typically delivered over 3 weeks.

The trial randomised participants in three strata: pre pathology, post pathology and contralateral breast cancer. In the initial 2010 publication,64 pre-pathology entry accounted for two-thirds of patients, post pathology approximately 30% and contralateral breast cancer patients < 4%. It is not clear if these proportions were maintained in the additional patient numbers reported in the updated 2014 publication. 65 The baseline stratification data show differences between centres in the number of patients entering the trial according to the three timings of delivery strata, particularly pre and post pathology (see Appendix 3 for further details). Patients who entered the trial in the pre-pathology stratum were randomised to either INTRABEAM or WB-EBRT prior to WLE of the primary tumour (Figure 2a). The trial was pragmatic in that if participants randomised to INTRABEAM were subsequently found to have unfavourable pathological features (unexpected lobular carcinoma, extensive intraductal component, positive margins at first excision), and hence were at high risk of recurrence elsewhere in the breast, they received WB-EBRT in addition (i.e. INTRABEAM + WB-EBRT, approximately 15% of INTRABEAM patients). The protocol also allowed for post-pathology entry of patients whereby patients underwent initial surgery and then, providing no unfavourable pathological features were identified, were randomised in a second stratum to receive INTRABEAM delivered as a second procedure or WB-EBRT (Figure 2b). Post-pathology entrants to the trial were randomised within 30 days after lumpectomy and the median time between initial lumpectomy and post-pathology INTRABEAM treatment was 37 days. The timing of INTRABEAM delivery was not specified in the intervention description within the NICE scope and, therefore, the post-pathology participants are included in this systematic review. Additionally, patients with a history of previous contralateral breast cancer were also included and randomised in a third stratum. Treatment for breast cancer in the contralateral breast is not an exclusion criterion for this review and, therefore, these participants are also judged to meet the criteria for inclusion.

FIGURE 2.

Flow diagram for the two main trial strata. (a) Pre pathology; and (b) post pathology.

Quality assessment of TARGIT-A trial

Overall, the methodological quality of the TARGIT-A trial was judged to be good with a low risk of bias. Table 8 shows the judgements of risk of bias in the various domains. For the HRQoL substudy, the assessment of selection bias and reporting bias for the main trial was judged to apply. For the remaining criteria it was judged that the HRQoL substudy could potentially differ from the main trial and, therefore, separate assessments were conducted (see Table 8). Overall, the substudy was judged to be at a high risk of bias owing to the lack of blinding and it is not clear how representative the results are for the total trial population because the substudy represents only about 2.5% of the overall trial population. Therefore, the substudy results should be interpreted with caution.

Randomisation schedules that were generated by computer and held securely in two centres, with requests for randomisation made by telephone or fax, meant that the risk of selection bias was low.

Owing to the nature of the interventions, it was not feasible to blind the patients or investigators in the trial, which could potentially introduce performance bias. However, given that the main trial outcomes (recurrence and survival) were objective measures, it was deemed unlikely that patients or investigators were influenced by the lack of blinding and thus performance bias was judged to be low. Similarly, for the main trial, although not all outcome assessors were blinded, the risk of detection bias was judged to be low because the main trial outcomes (recurrence and survival) were objective measures. For the substudy,63 the lack of patient and investigator blinding led to a judgement of a high risk of performance bias, and detection bias was judged as unclear owing to a lack of information.

The risk of attrition bias (differences between groups in withdrawals from the study) was deemed to be low in the TARGIT-A trial. There was a low proportion of withdrawals, and the rate appeared similar between treatment groups (0.5% INTRABEAM, 1.6% WB-EBRT). 65 Similar numbers of patients in the two treatment groups received their allocated treatment (91% INTRABEAM, 92% WB-EBRT)65 and all randomised patients were included in an ITT analysis for most outcomes. However, as noted above (see Overview of the TARGIT-A trial, Outcomes), an additional analysis by treatment received would have been desirable. The substudy63 was deemed to be at low risk of attrition bias because only one patient was reported as lost to follow-up.

The risk of bias due to selective reporting was deemed low as all outcomes specified in the trial protocol69 were reported in either the original or updated publication. 64,65 No other sources of bias in the total trial population were identified. The substudy63 used a retrospective questionnaire without reporting baseline measurements and was therefore deemed to be at unclear risk of other sources of bias.

Assessment of clinical effectiveness

The majority of the results presented in the following section are the most recent data for the TARGIT-A trial reported in the updated publication by Vaidya et al. 65 Results are presented for ipsilateral local recurrence, overall survival, and morbidity and toxicity. The main trial outcome data are supplemented with some morbidity data from the initial trial publication (see Vaidya et al. 64). The TARGIT-A trial presented outcomes of recurrence and survival for the whole trial population, and separately for the pre- and post-pathology strata. The separate analysis of these two strata was pre-specified. No data were presented from the third stratum (participants with a history of previous contralateral breast cancer) and no data on HRQoL have been published for the whole trial population. However, limited data on the secondary outcome of QoL are provided by a substudy at one trial centre. 63

Critical appraisal of the economic evaluations

The included cost-effectiveness studies were assessed against a critical appraisal checklist (Table 16), which appraised the quality of the studies and their generalisability to the UK. Any concerns identified by the assessment group (AG) are described below.

Both studies clearly defined the decision problem and used the relevant intervention and comparator for the purpose of this review, although the number of fractions used in the comparator arm of WB-EBRT was not relevant to UK practice (a standard of 33 fractions was used by Alvarado et al. ,106 whereas standard UK practice is 15 fractions over 3 weeks; the number of fractions was not reported by Shah et al. 108). The patient groups of interest as well as the perspective of the studies (societal) were stated; however, as the studies were based in the USA, they are not generalisable to the UK NHS setting. It is to be noted that the TARGIT-A trial, on which both the economic evaluations were based, included pre- and post-pathology patients. The study type and modelling methodology adopted by Alvarado et al. 106 are appropriate for the decision problem in this review. Shah et al. ,108 on the other hand, do not describe the methodology but do state that the methodologies are described elsewhere.

The study by Alvarado et al. 106,107 was transparent with respect to the information on model inputs and the assumptions used. Health state-specific costs109–118 and utilities119 were populated from published literature, although it was unclear if systematic reviews were conducted to inform these parameters. Both direct and indirect costs were reported. 106,107 The utilities associated with the health states in the base-case model were obtained via standard gamble technique in the source study119 and health outcomes were reported in terms of QALYs and life-years gained. A 10-year time horizon was used; this is considered inappropriate as risk of local recurrence continues over a lifetime. A series of one-way and two-way sensitivity analyses were conducted to assess uncertainty. In addition, scenario analysis of the 3-week accelerated WB-EBRT schedule of 16 fractions was performed. Although the results of the one-way sensitivity analyses favoured INTRABEAM over WB-EBRT in the treatment of patients with early-stage breast cancer, the robustness of the results still remains questionable as a probabilistic sensitivity analysis (PSA) was not conducted. The external validity of the economic model was assessed by comparing the findings with the published results of TARGIT-A, as well as against an online tool for adjuvant therapy and published cost-effectiveness evidence in the disease area using WB-EBRT as one of the comparator arms. The results of the base-case model were comparable with these sources.

Shah et al. 108 reported that all assumptions and methodology adopted in the analyses were based on, and consistent with, previously published articles, references of which were obtained and examined by the AG. The methodologies adopted to estimate reimbursement costs as well as the assumptions used in cost estimations were adequately described in the references provided. The study reported health outcomes in terms of QALYs. The time horizon for the analysis was not clearly stated but based on the estimation of mean utility by reimbursement technique it was assumed to be 10 years. No sensitivity or validation checks were reported, thus raising questions about the robustness of the results presented.

Description and results of the published economic evaluations

Summary of cost-effectiveness studies

• Two cost-effectiveness studies, reported in three publications,106–108 were identified.

• Both studies were based on the findings of the TARGIT-A trial.

• Cost–utility analyses were performed to evaluate QALYs, costs and ICERs of INTRABEAM compared with WB-EBRT.

• The analyses were conducted for a time horizon of 10 years in one study;106,107 for the other study108 it is assumed that a similar time horizon was adopted, although this was not clearly stated.

• The quality of utility data used in both the studies is questionable. The source study by Hayman et al. 119 was an old publication and more recent data would have been appropriate, such as those identified in Southampton Health Technology Assessments Centre’s systematic review of health-related quality-of-life studies. It was also not clear whether or not a systematic approach was adopted to identify this source.

• The perspectives, settings and comparators of both studies were not generalisable to the UK setting.

• Alvarado et al. 106 found INTRABEAM to be a more valuable strategy with less cost and greater QALYs than WB-EBRT. Shah et al. 108 concluded that although INTRABEAM represented a potential cost-saving alternative compared with WB-EBRT, the latter represented a cost-effective modality compared with INTRABEAM when additional medical and non-medical costs were factored in.

Critical appraisal of the included studies

A summary of the critical appraisal of the included studies is presented in Appendix 9.

The designs of the included studies varied: three were RCTs,126,128,133 two were single-cohort studies,127,129 one was a longitudinal study,130 one was a retrospective descriptive study131 and two were cross-sectional studies. 132,134

All nine included studies defined the study question and explained the treatment strategies. Across the studies, the study designs as well as the methods of recruiting participants were clearly outlined. The studies were transparent with regard to the information provided for the methodologies applied. One study did not include patients aged < 65 years,128 another excluded those aged > 65 years130 and three studies did not state clearly if any individuals relevant to this review were excluded. 129,131,134 One study131 did not describe participant characteristics. With respect to the sample size, only two studies126,129 provided an appropriate justification for the study sample size. The response rates to EQ-5D were not reported in two studies128,133,134 thereby raising questions on the validity of the reported results as a lower response rate could possibly result in biased outcomes. Similarly, loss to follow-up was not reported by four studies127,129,131,134 and loss to follow-up would also impact on the validity of the results.

The included studies were assessed on the basis of their relevance to the NICE reference case. Of the nine included studies, only three128,126,132 met all of the criteria (see Appendix 8). Five studies did not meet one of the criteria, as valuations of HRQoL were not undertaken from the general UK population. 127,129,131,133,134 The population characteristics in the remaining study did not match those described in the decision problem as they included women with a poor prognosis (stage II/III). 130

Of the included studies, only one study reported utility value for disease free after WLE. 126 This study was UK based and included patients aged ≥ 18 years. Three studies reported utility values for the WLE + WB-EBRT health state, of which one was based in the UK128 and two were US based. 127,129 Patients in the study by Freedman et al. 127 were > 18 years of age and those in the study by Serra et al. 129 ranged from 28 years to 77 years of age. The UK-based study by Prescott et al. 128 excluded women aged < 65 years and the mean age of the baseline cohort was 72 years. It was observed that the baseline patient characteristics with respect to age differed across the three studies. Freedman et al. 127 included women with early-stage breast cancer for their analysis, which was similar to the population of interest for the independent model. In addition, they reported outcomes at a longer follow-up of up to 15 years.

The utility values for the health state of mastectomy and immediate reconstruction were reported by two studies. 130,131 Robertson et al. 131 conducted a retrospective study based on Swedish breast cancer patients who had undergone immediate breast reconstruction with implants. Conner-Spady et al. ,130 on the other hand, conducted a longitudinal study in Canadian women with stage II or III breast cancer and at high risk of relapse. The study by Robertson et al. 131 had advantages over Conner-Spady et al. 130 with respect to larger sample size, recent publication date and longer follow-up period. Furthermore, women aged > 65 years were not included in the Canadian study. 130

Three studies reported utility associated with distant metastases,132,133,134 one of which also reported utility associated with disease-free status after local recurrence. 132 Sample size ranged from 345132 to 497. 134 In two of these studies, the median age of population was reported and was 57 years132 and 59 years;134 no information related to age was provided in the other study. 133 Lidgren et al. 132 included women with a previous diagnosis of breast cancer, while Sherrill et al. 133 focused on those with advanced or metastatic HER-2-positive (HER-2+) breast cancer who had progressive disease. Hildebrandt et al. 134 included both male and female patients affected by breast, cervical, endometrium, ovarian and other gynaecological cancer, and reported data separately for each disease.

Results

The utility values for the potentially relevant health states extracted from the nine included studies are tabulated in Table 19.

Summary and conclusions of the health-related quality-of-life review

The key findings of this systematic review are summarised below.

• Nine studies met the inclusion criteria of the HRQoL systematic review.

• Two studies were UK based and the remaining studies were based in Europe and North America.

• The included studies were diverse with respect to their aims, population of interest, geographical locations, interventions, comparators, study designs and methodologies adopted.

• The review identified utilities that could be used to inform the independent cost-effectiveness model for five out of seven potentially relevant health states: disease free after WLE; WLE + WB-EBRT; disease free after local recurrence; mastectomy and immediate reconstruction; and distant recurrence.

• The review did not identify any relevant study to populate the utilities for two potentially relevant health states: WLE + INTRABEAM or WLE + INTRABEAM + WB-EBRT.

Modelling approach

A multistate Markov model, developed in Microsoft Excel, was used in the submission. The model used a cohort of breast cancer patients aged ≥ 55 years who were disease free after WLE. The economic model was based on the results of the pre-pathology stratum of the TARGIT-A trial65 with 2298 patients. This was because results were less favourable in post-pathology stratum (see Chapter 4, Assessment of effectiveness) and the submission recommended that INTRABEAM be used in pre-pathology patients only (MS, pp. 3–4).

It was not reported whether the model was constructed de novo or adapted from another previously existing model. The model consisted of four health states:

• disease free

• local recurrence treated by mastectomy/lumpectomy

• non-breast cancer death

• breast cancer death.

Patients in the disease-free state could remain in that state or transition to either local recurrence or non-breast cancer death. Those in the local recurrence state could remain in that state or die from either non-breast cancer or breast cancer-related deaths. The two death states were the absorbing states. The analysis was conducted for a time period of 20 years with an annual cycle length.

Assumptions

The manufacturer’s model made the following assumptions:

• After local recurrence, INTRABEAM patients would undergo salvage lumpectomy.

• After local recurrence, WB-EBRT patients would undergo salvage mastectomy. There is also an undocumented assumption that all patients undergoing mastectomy have reconstruction, which is reflected in the high cost of mastectomy.

• The death rate in disease-free patients was equal to that in the general population.

• An average of 23 fractions of WB-EBRT per patient were delivered, based on 15–30 fractions in the clinical practice.

• All patients were given INTRABEAM concurrent with initial lumpectomy (pre-pathology stratum of TARGIT-A trial).

A few of the model assumptions are not relevant to UK practice. The model assumed that INTRABEAM patients would undergo salvage lumpectomy after local recurrence; however, clinical experts advised that in the UK most patients would undergo mastectomy after local recurrence instead. Furthermore, the undocumented assumption that all mastectomy patients would undergo reconstruction is not in line with UK practice, as only around 31% of the patients undergoing mastectomy will have reconstructions, as shown in the independent model discussed in Methods for economic analysis. In addition, the assumption of using an average of 23 fractions of WB-EBRT per patient was not appropriate as the current standard UK practice is 15 fractions.

Critical appraisal of model

The manufacturer’s economic evaluation was appraised for methodological quality and generalisability to the UK NHS using a checklist adapted from the NICE reference case requirements and the Philips et al. 58 checklist (Table 20). The evaluation met half of the requirements for methodological quality and generalisability, and the remaining criteria were either not met or unclear; therefore, the evaluation did not fully meet the NICE reference case. A brief description is presented below.

The manufacturer’s evaluation provided a clear statement of the decision problem to be addressed, which appeared to follow the scope for the appraisal issued by NICE. Although the comparator included WB-EBRT, which is routinely used within the NHS, its appropriateness is questionable as the number of WB-EBRT fractions used in the UK practice is 15 compared with 23 fractions used in the model.

Six out of 33 centres in the TARGIT-A trial were based in the UK and centres were allowed to follow local policy for WB-EBRT delivery. The MS reported 23 fractions as the average of the range between 15 and 30 fractions being used in all the countries in the trial, but it was not clear if this was a weighted average of fractions used in the trial or a midpoint. The perspective adopted in the model was appropriate and, although the MS reported that the analysis was UK based, limited details were provided on the baseline characteristics of the patient population. A Markov modelling methodology was used, which seemed appropriate given the clinical nature of breast cancer; however, the AG considered that the reported model was a simplified structure with only four health states and that an additional health state for progressed disease would have been appropriate. Another limitation was that a lifetime horizon was not adopted.

Patients entering the model were aged 55 years (on average) and were followed for 20 years. This time span might not reflect the entire follow-up period of the disease. Patients transitioned through the health states in annual cycles, which is an appropriate assumption. The model structure was presented diagrammatically but no justification of the key assumptions and description of the data inputs used was provided. Measures of clinical effectiveness were obtained from a single study;65 however, no other relevant trials were identified by the SHTAC’s systematic review. Benefits for the model were measured in QALYs using standard gamble for measuring utility, although the source study was dated 1997. 135 It was not clear if a systematic review was conducted to identify the study. The model extrapolated local recurrence and survival data beyond 5 years by tacitly assuming an exponential fit to time to local recurrence; however, the AG considers that a log-normal distribution would be the best fit based on comparison with external data (see Data sources). All benefits and costs were discounted at 3.5% as outlined in NICE guidance. Uncertainty was assessed through PSA and no one-way or scenario analyses were conducted. Finally, no details around model validation were provided.

Estimation of effectiveness

Data on effectiveness for both the intervention (INTRABEAM) and the comparator (WB-EBRT) were derived from a single RCT (TARGIT-A) by Vaidya et al. 65 and 5-year cumulative risks reported in the source study were converted to annual probabilities and populated in the model. It was not reported whether or not a systematic review was conducted to identify the source study; however, no other relevant trials were identified by the SHTAC’s systematic review (see Chapter 4, Quantity and quality of research available). No adverse events were included in the analysis, which was considered appropriate by the AG.

Estimation of quality-adjusted life-years

Health-related quality-of-life utility values were assigned to patients in the disease-free state, those undergoing salvage lumpectomy and those undergoing salvage mastectomy. A utility value of 0.92 was assigned to patients in the disease-free state, a value of 0.87 to patients undergoing salvage lumpectomy and a value of 0.82 to those undergoing salvage mastectomy. The MS obtained these values from a single study by Hayman et al. published in 1997. 135 No details were provided of the method of deriving these values or the rationale used. The source study135 used a standard gamble approach to estimate utility values, which were not obtained from the general population. This is a limitation as it was shown in the systematic review of HRQoL (see Southampton Health Technology Assessments Centre’s systematic review of health-related quality-of-life studies) that there were several other more recent and relevant HRQoL studies that used the EQ-5D measure.

Estimation of costs

Treatment unit costs were obtained from the following sources: expert opinion, reference costs 2012–13,136 payments by results tariff 2013–14,137 and the study by Wolowacz et al. 138 As with clinical effectiveness and utilities, the methods of deriving the costs were not adequately described. The costs associated with travel/parking/accommodation were appropriately not included within the WB-EBRT arm (it was stated that these expenses might range from £50 to £100 per patient per fraction delivered).

The validity of the costs estimates is questionable. The cost of INTRABEAM per patient was obtained from expert opinion and although the manufacturer provided the cost compositions of INTRABEAM, it was not transparent in explaining the assumed cost per patient. In addition, cost of WB-EBRT was obtained from an inappropriate Healthcare Resource Group (HRG) code, the code used in the model for WB-EBRT was for ‘other radiotherapy treatment’, whereas the AG considers that the HRG code description required for the purpose of this analysis is ‘deliver a fraction of radiotherapy on a megavoltage machine’, which includes WB-EBRT delivered by linear accelerator, as per the NICE scope. The AG considers that HRG codes SC22Z and SC23Z are required for treatment delivery, and SC45Z, SC46Z, SC47Z and SC48Z are required for WB-EBRT (see Data sources and Table 31). Costs were only varied by ± 10% in PSA. There were also inconsistencies in the sources used to populate the reported costs; for instance, the costs of treating post-INTRABEAM local recurrence (salvage lumpectomy) and that of treating post-WB-EBRT local recurrence (salvage mastectomy) were obtained from payments by results tariff 2013–14, whereas the cost of WB-EBRT was obtained from the reference costs 2012–13. 137 The use of reference costs is preferable and would be considered standard practice.

Cost-effectiveness results

The base-case results from the submission are shown in Table 21 and indicate that INTRABEAM is associated with higher QALYs and lower costs. The submission states that the incremental analysis shows dominance of INTRABEAM over WB-EBRT.

One-way sensitivity analyses and scenario analyses were not conducted. A PSA was undertaken using Monte Carlo simulation with 1000 iterations. The cost parameters in the model were assigned to beta-project evaluation and review technique (PERT) distributions and beta distributions were assigned to utilities. For the cost parameters, the AG considers that gamma distribution would have been a more standard choice. It is not usual practice to assign beta-PERT distribution; however, it is expected that this would have little impact. For the PSA, at the £20,000 and £30,000 willingness-to-pay (WTP) thresholds, INTRABEAM has the highest probability of being cost-effective, at 100% for both thresholds.

Critique of the manufacturer’s submission

• The MS provides very limited information on model structure, baseline characteristics of the patient population and setting.

• Limited information is provided with respect to input parameters such as costs and utilities. The MS is not transparent in providing the methodology adopted to inform the input parameters.

• The method to derive costs of INTRABEAM is not clear.

• No rationale is provided for using the specific distributions assigned to the parameters.

• The method of extrapolation of local recurrence and survival data is not justified.

• The number of fractions for the WB-EBRT arm used in the model (23 fractions) is higher than UK practice; this will lead to an overestimation of WB-EBRT costs.

• The manufacturer’s model assesses health benefit in terms of QALYs, which is a valid measure of health in the UK NHS setting. The source study135 used standard gamble from a 1997 publication to estimate utilities. No details were provided as to whether or not a systematic search was conducted to identify utilities for the model.

• Model validation was not conducted; hence, the generalisability of model results remains questionable.

• Probabilistic sensitivity analysis was conducted for only 1000 simulations and no one-way or scenario analyses were conducted. Limited sensitivity analyses conducted around the base-case model results raise questions on the robustness of the model predictions.

• In summary, results of the MS model should be viewed with caution owing to the methodological and reporting limitations outlined above.

Overview

We developed a new model to estimate the costs, benefits and cost-effectiveness of the INTRABEAM Photon Radiotherapy System compared with WB-EBRT for early operable breast cancer.

The effects of the intervention on the clinical course of the disease are obtained from the TARGIT-A trial included in the systematic review of clinical effectiveness (see Chapter 4). The patient population included in the economic model reflects the patient population in the pre-pathology stratum of this trial. This is because the TARGIT-A study recommends INTRABEAM concurrent with lumpectomy as an alternative to post-operative WB-EBRT65 but does not recommend the use of post-operative INTRABEAM as an alternative to WB-EBRT (as non-inferiority was not established in this stratum). Furthermore, use of the pre-pathology stratum provides consistency with the manufacturer’s economic model, which is also based on the results of the pre-pathology stratum.

The analysis takes the perspective of the NHS and PSS in the UK. The model adopts a lifetime (40-year) horizon to estimate costs and benefits from each treatment. Future costs and benefits are discounted at 3.5% per annum as recommended by the UK Treasury. 139 The outcome of the economic evaluation is reported as the cost saved per QALY lost.

Methods for economic analysis

The model uses transition probabilities obtained from the clinical literature to simulate the progression of early operable breast cancer in a cohort of patients and to estimate the cost-effectiveness of the radiotherapy treatments under consideration. The model was constructed using the TreeAge Pro 2014 software (TreeAge Software, Inc., Williamstown, MA, USA). The model structure was informed by a review of other published models of early breast cancer106,109,120,140–142 and the evidence available to inform disease progression, which is drawn from the only existing RCT, the TARGIT-A trial65 (see Chapter 4).

The model structure follows the disease pathway for early-stage breast cancer. It is slightly modified from an economic model structure used in a previous HTA report to NICE140 in order to reflect the clinical evidence available. The structure is also similar to the model structure adopted by Alvarado et al. 106 in their cost-effectiveness analysis of IORT. The SHTAC’s model uses six distinct health states: recurrence free, local recurrence, disease free after local recurrence, any other recurrence, death from breast cancer, and death from other causes (Figure 5). The local recurrence, disease free after local recurrence and any other recurrence health states were chosen pragmatically in order to match the definitions and data supplied by the TARGIT-A trial publication. 65

FIGURE 5.

Influence diagram for the SHTAC’s breast cancer cost-effectiveness model.

Local recurrence is defined in the TARGIT-A trial as recurrence in the conserved breast while any other recurrence incorporates regional recurrence (axilla plus supraclavicular), contralateral breast recurrence and distant recurrence. 65 The AG notes that regional recurrence, contralateral recurrence and distant recurrence have very different prognoses and costs but they are not modelled separately as no data were available to inform possible transitions to or from these health states.

Non-death health states are associated with a HRQoL utility and a cost estimate.

All patients start the model in the recurrence-free state and may then either stay in the recurrence-free state, have a local recurrence and move to the local recurrence state, have another type of recurrence and move to the any other recurrence state, or die from non-breast cancer causes. From the local recurrence state, a patient may move to the disease free after local recurrence state, suffer any other recurrence or die from other causes. A patient in the disease free after local recurrence state may remain either in this state, suffer any other recurrence or die from other causes. From the any other recurrence state, it is possible to die from breast cancer, die from other causes or stay in the state. The local recurrence state is temporary and it is only possible to remain here for one cycle.

Model cycle length is 1 year and a lifetime horizon of 40 years was adopted in the base case, which is sufficiently long to capture all clinically and economically important events. A half-cycle correction was applied.

The baseline disease progression parameters used in the model were obtained from the TARGIT-A trial (see Chapter 4). 65 These inform the annual probabilities of local recurrence, any other recurrence while recurrence free, and death from breast cancer. Data from de Bock et al. 143 were used to inform the probability of any other recurrence given local recurrence at the suggestion of the advisory group. Data from the Office for National Statistics (ONS) were used to inform the probability of all-cause mortality by age. 144 Parametric curves were fitted to Kaplan–Meier data in order to provide the probabilities of local recurrence in both treatment arms.

The costs included in the model are those for initial radiation treatment and repeat lumpectomy and mastectomy and reconstruction, with or without radiation treatment, at local recurrence. Full details of the costs used in the model are given in Data sources.

The model includes the following assumptions:

• All patients enter the model in the recurrence-free state after initial BCS and radiation therapy.

• It is not possible to die from breast cancer while in the local recurrence state or the disease free after local recurrence state. It is only possible to die from breast cancer while in the any other recurrence state.

• Only one local recurrence is allowed; repeat local recurrence is not modelled.

• Death rates for non-breast cancer causes are based on mortality statistics for England and are applied across all health states.

• The survival of patients with recurrence of any sort is assumed to be independent of the time of recurrence.

A further simplification is that, owing to data limitations, the costs of post-progression therapies are not included in the base case.

In each cycle, the total costs and QALYs are calculated by multiplying the individual costs and HRQoL of each model state by the proportion of the model cohort in that state, for each of the radiotherapy types. The total discounted lifetime costs and QALYs are calculated by aggregating the costs and QALYs for all cycles. The ICER is calculated as:

where convention therapy A is the current standard of care and therapy B is a new therapy. The associated incremental net monetary benefit (NMB) of a specific treatment compared with a comparator may be calculated as:

when the incremental QALYs and incremental costs are simply the denominator and numerator, respectively, of equation (1) and WTP is the maximum amount a decision-maker is prepared to pay per QALY gained. 57 As long as the incremental NMB is more than zero, then a treatment is cost-effective and larger NMBs represent greater cost-effectiveness than smaller NMBs.

Data sources

Recurrence-free state: probability of local recurrence

The baseline risk of local recurrence in the economic model is taken from the pre-pathology subgroup of the TARGIT-A trial. 65 The TARGIT-A trial was the only trial included in the review of clinical effectiveness (see Chapter 4) and as such is the main source of evidence of the clinical efficacy of INTRABEAM.

Local recurrence probabilities in the pre-pathology substratum for INTRABEAM and WB-EBRT were extracted from a Kaplan–Meier plot in the trial publication65 using the digitising software PlotDigitizer (© 2000–14 Joseph A Huwaldt) and the method of survival curve reconstruction described in Guyot et al. 145 Parametric survival models were then fitted to the observed data using Stata software version 11.0 (StataCorp LP, College Station, TX, USA) in order to extrapolate local recurrence beyond the 5 years reported. 65 In line with the recommendation of Latimer146 all of the ‘standard’ parametric models were considered (exponential, Weibull, Gompertz, log-logistic, log-normal).

Akaike information criterion (AIC) values obtained for each distribution are given in Table 22, which shows that the log-normal, log-logistic and Weibull distributions provide the best fit to the data based on this criterion. The Gompertz and exponential distributions fit the data less well. The log-normal and Weibull fits are compared graphically with the 5 years of trial data in Figure 6. The log-logistic fit is similar to the log-normal and is not considered further. Figure 6 demonstrates that the log-normal and Weibull fits are similar over this time period. Figure 7 shows the behaviour of the log-normal and Weibull fits over the model time horizon of 40 years and it can be seen that local recurrences continue to occur throughout the time horizon with both models, but that the proportion with local recurrence after 40 years is much higher under the Weibull model than under the log-normal model. Previous economic evaluations to NICE have assumed that patients who have experienced an episode of early-stage breast cancer but are in remission after 15 years will have the same risk of progression as the general population. 140 However, clinical advice to the AG is that the risk of local recurrence continues throughout life and is relatively linear over time. Data on local recurrence at 9 years from the ELIOT trial,147 and the study of Kreike et al. ,148 which follows up BCS + radiotherapy patients for 15 years, also suggest that risk of local recurrence does not decrease over time.

TABLE 22 - Values of AIC obtained for parametric survival models fitted to reconstructed local recurrence data from TARGIT-A triala,65

Lower values of AIC indicate a better fit to the data.

Model	AIC
Log-normal	213.0
Log-logistic	214.2
Weibull	214.2
Gompertz	217.6
Exponential	219.2

FIGURE 6.

Kaplan–Meier plot of local recurrence in the pre-pathology subgroup of the TARGIT-A trial65 compared with fitted log-normal and Weibull local recurrence curves.

FIGURE 7.

Kaplan–Meier plot of local recurrence in the pre-pathology subgroup of the TARGIT-A trial65 compared with fitted log-normal and Weibull local recurrence curves over 40 year time horizon.

The model adopts the log-normal curve in the base case. Not only is this a better fit by the AIC criterion, but the rate of local recurrence does not increase as steeply over time as in Weibull model (see Figure 7; see also Figure 6 for more detail of the first 5 years). This behaviour means that median survival is longer under this model and, thus, it provides a better fit to other published data on survival after breast cancer (see Model validation). Coefficients of the fitted log-normal regression model are given in Table 23.

Model validation

The overall survival predictions from the model base case are compared with the trial-observed Kaplan–Meier data for the pre-pathology subgroup in Figure 8. The model overall survival predictions in Figure 8 were obtained using TARGIT-A trial data to model non-breast cancer death for the first five model cycles and provide a good fit to the observed data. Data from the TARGIT-A trial show that overall survival in the INTRABEAM treatment arm is somewhat better than overall survival in the WB-EBRT arm at 5 years, and this is reflected in the model predictions (see Figure 8). The model thus appears to be performing satisfactorily in this respect.

FIGURE 8.

Kaplan–Meier plot of overall survival in the pre-pathology subgroup of the TARGIT-A trial65 compared with overall survival predicted by the SHTAC’s economic model using TARGIT-A trial data to model non-breast cancer death for first five cycles.

The model base case does not use trial-observed data for non-breast cancer death, for reasons given in Data sources. Figure 9 gives the model predictions for overall survival in each of the treatment arms in the pre-pathology subgroup when only ONS mortality data are used to model non-breast cancer death. Figure 9 shows that, when using these data, predicted overall survival in the INTRABEAM treatment arm is worse than observed in the trial, although the overall survival prediction for the WB-EBRT arm is still a good fit. This is to be expected because ONS age-specific all-cause mortality rates are higher than the non-breast cancer mortality rates seen on the INTRABEAM arm in the TARGIT-A trial. The model predictions change in reflection of these differences (compare Figure 8 with Figure 9) and so, again, the model appears to be working satisfactorily.

FIGURE 9.

Kaplan–Meier plot of overall survival in the pre-pathology subgroup of the TARGIT-A trial65 compared with overall survival predicted by the SHTAC’s economic model using ONS mortality data to model non-breast cancer death in all cycles.

It may be seen from Figures 8 and 9 that median overall survival predicted by the model base case for early operable breast cancer patients is approximately 21.5 years and that overall survival is approximately 56% at 20 years. Relative survival at 20 years is 82% and at 25 years is 77%. Relative survival compares the survival of people with the cancer to that of people without cancer in order to help correct for deaths from things other than breast cancer. Exact comparison with other data sources is difficult; however, the SEER (Surveillance, Epidemiology, and End Results) database of the US National Cancer Institute has 20-year relative survival of 64.7% in breast cancer patients aged ≥ 50 years diagnosed between 1985 and 1989. 157 Figures from Cancer Research UK for England and Wales indicate that relative survival from breast cancer at 20 years is 64.5%. 158 Thus, the relative survival of 82% at 20 years given by the model is somewhat higher than these estimates, but this is to be expected as treatment has improved in the 25 or so years since the patients on whom these estimates are based were diagnosed.

Relative survival compares the survival of people with the cancer to that of people without cancer in order to help correct for deaths from things other than breast cancer. Thus, it is reasonable that the overall survival of 56% in the model is lower than these published estimates of relative survival because it does reflect deaths from other causes.

Results of independent economic analysis

This section reports the cost-effectiveness of INTRABEAM compared with WB-EBRT in a cohort of early operable breast cancer patients. Base-case discounted cost-effectiveness summary results are given in Table 34 and are broken down by health state in Table 35. Results with no discounting of costs and outcomes are given in Table 36. INTRABEAM is less expensive but also less effective than WB-EBRT as it has lower total costs but also fewer total QALYs. Therefore, the ICERs given in Tables 34 and 36 represent the money saved per QALY lost that is associated with replacing WB-EBRT by INTRABEAM.

In situations in which a new intervention (INTRABEAM) is both less costly and less effective than the current standard of care (WB-EBRT), the ICER for INTRABEAM to replace WB-EBRT must lie above the usual NICE cost-effectiveness thresholds of £20,000 and £30,000 per QALY if INTRABEAM is to be considered a cost-effective alternative to WB-EBRT. However, the ICER value of £1596 saved per QALY lost, shown in Table 34, indicates that WB-EBRT is the cost-effective treatment option within the WTP threshold of £20,000 per QALY. Over the 40-year time horizon of the model, it is associated with more QALYs at broadly similar overall cost. WB-EBRT is also cost-effective in the undiscounted analysis in which incremental QALYs are nearly twice those seen in the discounted results and the ICER (£ saved/QALY lost) is smaller (see Table 36).

Sensitivity analysis

Deterministic and probabilistic sensitivity analyses were conducted in order to investigate the effect of uncertainty in model parameter values on the cost-effectiveness results. DSA was used to highlight the most influential parameters while the effect of uncertainty and interaction in multiple parameters was examined using PSA. Scenario analysis was used to investigate the effect of uncertainty in model assumptions and structure.

Each parameter was assumed to follow a probability distribution and these are given, with the distribution parameters, in Table 37. For beta distributions, the distribution parameters were fitted using either the method of moments or information on the sample size and number of events when available. Distribution parameters were fitted to the gamma distributions using the method of moments. In cases for which a standard error (SE) or standard deviation (SD) was not supplied in the source literature, the SE was calculated using an arbitrary ± 20% from the base-case value. Correlation between the parameters of the log-normal distribution used to inform time to local recurrence was incorporated by sampling from a multivariate normal distribution with covariance matrix as specified in Table 37.

TABLE 37 - Parameters, distributions and associated upper and lower values used in probabilistic and DSA

QC, quality control.

Distribution calculated after arbitrary ± 20% variation applied to mean to obtain SE.

On log scale.

Parameter	Distribution	Distribution parameters	Mean/base case	2.5th percentile for mean	97.5th percentile for mean
Costs
INTRABEAM commissioninga	Gamma	α = 96.04; β = 26.51	£2546	£2062	£3080
One WB-EBRT delivery	Gamma	α = 18.36; β = 6.45	£118	£71	£178
WB-EBRT planning	Gamma	α = 4.10; β = 78.97	£324	£90	£704
INTRABEAM setup costsa	Gamma	α = 96.04; β = 62.31	£5984	£4847	£7239
Mastectomy and reconstruction	Gamma	α = 99.63; β = 78.51	£7822	£6362	£9431
Mastectomy	Gamma	α = 147.71; β = 16.99	£2510	£2122	£2931
One hour in operating theatrea	Gamma	α = 96.04; β = 5.92	£569	£461	£688
Pre-treatment QC INTRABEAMa	Gamma	α = 96.04; β = 0.26	£25	£20	£31
Staff time per hour in theatre during INTRABEAM deliverya	Gamma	α = 96.04; β = 1.57	£150	£122	£182
Staff time per hour in theatre during INTRABEAM planninga	Gamma	α = 96.04; β = 2.61	£250	£203	£303
Annual staff training in radiation protectiona	Gamma	α = 96.04; β = 9.58	£920	£745	£1113
Staff time in support of INTRABEAM deliverya	Gamma	α = 96.04; β = 0.79	£76	£61	£92
Repeat lumpectomy	Gamma	α = 95.55; β = 16.13	£1542	£1248	£1866
Survival curve parameters
Time to local recurrence	Multivariate normalb
β(treatment arm)	Covariance matrix0.081–0.0770.531–0.0080.1310.035		Covariance matrix			0.081			–0.077	0.531		–0.008	0.131	0.035	–0.256	–0.815	0.307
Covariance matrix
0.081
–0.077	0.531
–0.008	0.131	0.035
Constant	4.97	3.553	6.383
Sigma	0.436	0.072	0.797
Probabilities
Other recurrence INTRABEAM from recurrence free (5 years)	Beta	α = 19.1; β = 378	0.048	0.029	0.071
Other recurrence WB-EBRT from recurrence free (5 years)	Beta	α = 16.7; β = 337.9	0.047	0.028	0.071
Other recurrence after local recurrence (10.2 years)	Beta	α = 129; β = 181	0.416	0.362	0.471
INTRABEAM patient receives WB-EBRT	Beta	α = 239; β = 1332	0.152	0.135	0.170
Mastectomy patient has reconstruction	Beta	α = 5120; β = 11365	0.311	0.304	0.318
INTRABEAM patient has mastectomy at local recurrencea	Beta	α = 18.4; β = 4.6	0.800	0.618	0.933
INTRABEAM patient dies from breast cancer (5 years)	Beta	α = 10.6; β = 310.8	0.033	0.016	0.055
WB-EBRT patient dies from breast cancer (5 years)	Beta	α = 11.3; β = 407.8	0.027	0.014	0.045
Incident breast cancer patients suitable for INTRABEAMa	Beta	α = 294; β = 1528	0.161	0.145	0.179
Resource use
INTRABEAM delivery timea	Normal	Mean = 33; SE = 3.37	33	26.40	39.60
INTRABEAM planning timea	Normal	Mean = 6; SE = 0.61	6	4.80	7.20
Utilities
Recurrence free after the first year	Beta	α = 2400; β = 558.5	0.811	0.8	0.83
Recurrence free in the first year	Beta	α = 2161; β = 635.3	0.773	0.76	0.79
Other recurrence	Beta	α = 171; β = 78.7	0.685	0.63	0.74
Other
Catchment population served by one INTRABEAM devicea	Normal	Mean = 1,000,000; SE = 102,041	1,000,000	800,004	1,199,996

The model parameters were varied in DSA between the 2.5th and 97.5th percentiles of the assumed parameter distribution of the mean value and these are given in Table 37. Table 38 gives upper and lower bounds for parameters examined in DSA when these are different from the upper and lower bounds examined in PSA.

TABLE 38 - Lower and upper parameter values examined in DSA (when different from 2.5th and 97.5th percentiles given in Table 37)

Parameter	Base case	Lower value	Upper value
Proportion of incident breast cancer patients suitable for INTRABEAM	0.16	0.1	0.5
Fractions of WB-EBRT required to complete a course of treatment	15	5	23
Lifetime of INTRABEAM device (years)	10	5	10
Proportion of INTRABEAM patients requiring radiation shield	1	0.25	1
Age of cohort entering model (years)	62	55	72
Discount rate for costs (%)	3.5	0.0	6.0
Discount rate for health (%)	3.5	0.0	6.0

Deterministic sensitivity analysis

Table 39 shows the results of the deterministic sensitivity analyses for INTRABEAM compared with WB-EBRT for the most influential parameters. A tornado diagram depicting the range in incremental NMB given in this table is given in Figure 10. A complete set of DSA results is given in Appendix 10.

TABLE 39 - Key DSA results for INTRABEAM compared with WB-EBRT. WTP set to £20,000 per QALY

Variable description	Low value	High value	Low value incremental NMB (£)	High value incremental NMB (£)	Range (£)
5-year probability of any other recurrence INTRABEAM	0.029	0.071	5781	–9171	14,952
5-year probability of any other recurrence WB-EBRT	0.028	0.071	–8760	5977	14,737
Beta coefficient for INTRABEAM arm time to local recurrence	–0.815	0.307	–4512	118	4630
5-year probability of death from breast cancer WB-EBRT	0.014	0.045	–4150	–346	3804
5-year probability of death from breast cancer INTRABEAM	0.016	0.055	1051	–2518	3569
Constant (time to local recurrence)	3.553	6.383	–3367	–836	2531
Discount rate for utilities (%)	0	6	–3192	–1042	2150
Number of WB-EBRT deliveries required in course of treatment	5	23	–2604	–832	1772
Starting age of model cohort (years)	55	72	–2273	–757	1516
Cost of delivering one fraction WB-EBRT (£)	71	178	–2211	–877	1334
Proportion of incident cases which are suitable for INTRABEAM	0.1	0.5	–2064	–1128	936
Sigma (time to local recurrence)	0.072	0.797	–1110	–2018	908

FIGURE 10.

Tornado diagram showing key results of DSA for INTRABEAM vs. WB-EBRT. Bars indicate spread in incremental NMB between upper and lower parameter bounds (£). WTP set to £20,000 per QALY.

The incremental NMB rather than the ICER is used in Table 39 and Figure 10 as the ICER for INTRABEAM compared with WB-EBRT is sometimes negative (Figure 11) and incremental NMB has a more straightforward interpretation. A WTP of £20,000 and equation (2) (see Methods for economic analysis) were used to calculate the incremental NMB.

FIGURE 11.

Scatterplot of the costs and health benefits from PSA: INTRABEAM compared with WB-EBRT.

Table 39 and Figure 10 compare INTRABEAM incrementally with WB-EBRT in order to be consistent with the base case (see Table 34). Thus, a negative incremental NMB indicates that INTRABEAM is not cost-effective compared with WB-EBRT (or, conversely, that WB-EBRT is cost-effective compared with INTRABEAM). A positive incremental NMB indicates that INTRABEAM is cost-effective compared with WB-EBRT (or, conversely, that WB-EBRT is not cost-effective compared with INTRABEAM).

The results show that the incremental NMB is, above all, very sensitive to the probability of any other recurrence, which is assumed for both WB-EBRT and INTRABEAM as there is a very wide difference in the incremental NMB between the low and high values of these parameters. The differences lead to a switch in which treatment is considered cost-effective at a WTP threshold of £20,000 per QALY. At a low probability of any other recurrence on the INTRABEAM arm, INTRABEAM is cost-effective compared with WB-EBRT at a WTP of £20,000 (shown by positive incremental NMB in Table 39). At high probability of any other recurrence on the INTRABEAM arm, WB-EBRT is a cost-effective treatment option at the £20,000 per QALY WTP threshold (shown by negative incremental NMB in Table 39). With low probability of any other recurrence on the WB-EBRT arm, WB-EBRT is a cost-effective treatment option compared with INTRABEAM at the £20,000 per QALY WTP threshold, but this is reversed with high probability of any other recurrence on the WB-EBRT arm, that is, INTRABEAM becomes cost-effective at the £20,000 per QALY WTP threshold (see Table 39).

The model is also somewhat sensitive to the probability of death from breast cancer on the INTRABEAM arm and, again, this difference leads to a switch in which treatment is considered cost-effective at a WTP threshold of £20,000 per QALY. At low values for probability of death from breast cancer on the INTRABEAM arm, INTRABEAM is cost-effective at a WTP of £20,000 per QALY, but it is not cost-effective compared with WB-EBRT at high values for probability of death from breast cancer on the INTRABEAM arm (see Table 39).

Change in which treatment is considered cost-effective at a WTP threshold of £20,000 per QALY also occurs between the low and high parameter values considered for the beta coefficient for the INTRABEAM arm in the log-normal model of time to local recurrence (see Table 39). At low values of this coefficient, WB-EBRT is cost-effective compared with INTRABEAM but at the highest values considered, INTRABEAM becomes slightly more cost-effective than WB-EBRT.

In summary, the results of the DSA indicate that there is a degree of uncertainty surrounding the base-case results. In the case of four parameters, the difference between upper and lower values results in a switch in the treatment option, which is considered cost-effective at a WTP of £20,000 per QALY.

Probabilistic sensitivity analysis

A total of 10,000 PSA simulations were run, and the mean results for these simulations are presented in Table 40 and are similar to results for the base case given in Table 34. The scatterplot for cost and health outcomes is shown in Figure 11 and, similar to the DSA findings, indicates considerable uncertainty in the results. There are many points in the north-west quadrant of Figure 11, which demonstrates that in a large number of the PSA simulations INTRABEAM is less effective than WB-EBRT, as well as being more costly. Conversely, in many of the PSA simulations WB-EBRT is more effective and cheaper than INTRABEAM, shown by the large number of points in the south-east quadrant of Figure 11.

TABLE 40 - Baseline PSA cost-effectiveness results

LYG, life-years gained.

INTRABEAM is both cheaper and less effective than WB-EBRT; therefore, the ICER represents the £ saved per QALY lost associated with replacing WB-EBRT with INTRABEAM.

Intervention	Total costs (£)	Total LYG	Total QALYs	Incremental cost (£)	Incremental QALYs	ICER (£ saved/QALY lost)
WB-EBRT	2398	20.73	11.327	–	–	–
INTRABEAM	2272	20.52	11.240	–126	–0.087	1447a

The CEAC calculated from the PSA simulations is given in Figure 12 and indicates that at the £20,000 WTP threshold WB-EBRT has the highest probability (61.3%) of being cost-effective. WB-EBRT also has the highest probability of being cost-effective (61.4%) at a WTP of £30,000 per QALY. INTRABEAM has a higher probability of being cost-effective than WB-EBRT at WTP thresholds of around £5000 per QALY or less (see Figure 12).

FIGURE 12.

Cost-effectiveness acceptability curve from the PSA.

Scenario analysis

In addition to the sensitivity analyses, five scenarios were examined to investigate the uncertainty surrounding the structural assumptions made by the model.

Trial-observed non-breast cancer mortality data

The model base case uses ONS all-cause mortality tables to give the probability of non-breast cancer death. As an alternative to using ONS data in all model cycles, the use of non-breast cancer mortality data from the TARGIT-A trial was examined. A Weibull fit to TARGIT-A Kaplan–Meier data65 was used to obtain trial-observed non-breast cancer mortality probabilities for the first five model cycles. ONS mortality data were used thereafter. INTRABEAM dominates WB-EBRT in this scenario, as it is associated with lower total costs and greater total QALYs (Table 41).

TABLE 41 - Cost-effectiveness results using trial-observed non-breast cancer mortality data for first five model cycles

LYG, life-years gained.

Intervention	Total costs (£)	Total LYG	Total QALYs	Incremental cost (£)	Incremental QALYs	ICER (£/QALY)
WB-EBRT	2366	20.58	11.259	–	–	–
INTRABEAM	2234	20.83	11.425	–132	0.166	Dominating

Population served by one device

The manufacturer’s model assumes that 100 patients are treated with INTRABEAM each year in a district general hospital (P Pinilla-Dominguez, NICE, 2014, personal communication). To replicate this assumption in the SHTAC’s model requires a corresponding assumption about the typical catchment population of a hospital offering INTRABEAM. In the base case, the SHTAC’s model assumes that the catchment population is one million, which implies 126 INTRABEAM procedures a year (see Table 33). A catchment population of 795,000 is required to give 100 INTRABEAM procedures a year. Results using this catchment population are given in Table 42. The table shows that INTRABEAM is now dominated by WB-EBRT as it is associated with slightly higher total cost, but fewer QALYs.

TABLE 42 - Cost-effectiveness results using a population served by one INTRABEAM device of 795,000

LYG, life-years gained.

Intervention	Total costs (£)	Total LYG	Total QALYs	Incremental cost (£)	Incremental QALYs	ICER (£/QALY)
WB-EBRT	2368	20.72	11.329	–	–	–
INTRABEAM	2414	20.51	11.241	47	–0.088	Dominated

Mastectomy disutility

The manufacturer’s model uses a utility of 0.87 for lumpectomy at local recurrence and a utility of 0.82 for mastectomy. These figures imply a disutility for mastectomy of 0.05. The AG considers that it is unclear from the literature if mastectomy is associated with significant disutility to HRQoL as measured with EQ-5D. 160,161 A scenario analysis was conducted to examine the effect of a mastectomy disutility of 0.05 on model outcomes. In the SHTAC’s model it is assumed that this disutility is a weighted average of the disutilities associated with mastectomy and mastectomy and reconstruction.

The results are given in Tables 43 and 44. Table 43 shows results obtained when it is assumed that the mastectomy utility decrement applies to both the local recurrence and disease free after local recurrence health states; Table 44 shows the results obtained when it is assumed that the mastectomy utility decrement applies to the local recurrence health state only. Applying the utility decrement to both the local recurrence and disease free after local recurrence health states has more impact on final ICER than applying the decrement to the local recurrence state alone, but in neither case does the utility decrement make an appreciable difference to model outcome. The ICER decreases by less than £50 per QALY compared with the base case (see Table 34).

TABLE 43 - Cost-effectiveness results using a utility decrement of 0.05 for mastectomy (applied to local recurrence and disease free after local recurrence health states)

LYG, life-years gained.

INTRABEAM is both cheaper and less effective than WB-EBRT; therefore, the ICER represents the £saved per QALY lost associated with replacing WB-EBRT with INTRABEAM.

Intervention	Total costs (£)	Total LYG	Total QALYs	Incremental cost (£)	Incremental QALYs	ICER (£ saved/QALY lost)
WB-EBRT	2368	20.72	11.304	–	–	–
INTRABEAM	2227	20.51	11.214	–140	–0.090	1563a

TABLE 44 - Cost-effectiveness results using a utility decrement of 0.05 for mastectomy (applied to local recurrence health state only)

LYG, life-year gained.

INTRABEAM is both cheaper and less effective than WB-EBRT; therefore, the ICER represents the £ saved per QALY lost associated with replacing WB-EBRT with INTRABEAM.

Intervention	Total costs (£)	Total LYG	Total QALYs	Incremental cost (£)	Incremental QALYs	ICER (£ saved/QALY lost)
WB-EBRT	2368	20.72	11.326	–	–	–
INTRABEAM	2227	20.51	11.238	–47	–0.088	1592a

The decrease in ICER compared with the base case indicates that WB-EBRT becomes more cost-effective than INTRABEAM in this scenario. Although in the base case a smaller proportion of INTRABEAM patients undergo mastectomy for local recurrence (80% compared with 100% for WB-EBRT), more INTRABEAM patients experience a local recurrence. The net effect is that the total mastectomy utility decrement is greater on the INTRABEAM arm and, consequently, the incremental QALYs associated with WB-EBRT are slightly higher than in the base case.

Alternative set of health state utilities

The health state utilities used in the model base case are the same in the local recurrence health state and the recurrence-free health state after the first year (see Table 24). Although these utilities are based on the studies of Lidgren et al. 132 and Turnbull et al. ,126 it is arguably not appropriate that these two health states should have the same utility. Their identical values may arise because EQ-5D is not a particularly sensitive instrument to use when examining QoL in early breast cancer patients as found, for example, by Hildebrandt et al. 134 An alternative set of health state utility values used in a previous HTA report to NICE was examined. 140 These were valued by either patients or clinical experts using the TTO and are given in Table 45.

TABLE 45 - Alternative health state utility values examined in scenario analysis

Health state	Utility value	Source
Recurrence free	0.78	Hind et al.140
Local recurrence	0.61
Disease free after local recurrence	0.71
Any other recurrence	0.42

The results for the scenario are given in Table 46. These show that, although total QALYs decline in both treatment arms with use of the alternative utility set, the incremental QALYs do not change appreciably from the base case. Thus, the overall ICER is very similar to the base case: £1517 saved per QALY lost, compared with £1596 in the base case (see Table 34).

TABLE 46 - Cost-effectiveness results using alternative set of health state utilities from Hind et al.140

LYG, life-years gained.

INTRABEAM is both cheaper and less effective than WB-EBRT; therefore, the ICER represents the £ saved per QALY lost associated with replacing WB-EBRT with INTRABEAM.

Intervention	Total costs (£)	Total LYG	Total QALYs	Incremental cost (£)	Incremental QALYs	ICER (£ saved/QALY lost)
WB-EBRT	2368	20.72	10.812	–	–	–
INTRABEAM	2227	20.51	10.719	–140	–0.093	1517a

Costs post progression

The base case does not include costs of treatment after any other recurrence because of lack of information on the types of recurrence within this category. The trial publication reports the proportions with regional recurrence (1.1% INTRABEAM compared with 0.9% WB-EBRT) and distant recurrence (3.9% INTRABEAM compared with 3.2% WB-EBRT) for all patients, but does not give these data for the pre-pathology stratum. 65 However, the costs of treating these types of recurrence are quite different. 140 Using costs given in the HTA report of Hind et al. ,140 inflated to 2013 using the Hospital and Community Health Services prices index,153 the AG calculated the annual cost of metastatic disease (active treatment and supportive care) as £12,122 and the cost of end-of-life care for a breast cancer patient as £3669. In contrast, the costs of contralateral disease are more similar to those incurred at local recurrence. 140

For illustrative purposes, the AG has considered a scenario in which 60% of recurrences in the ‘any other recurrence’ health state are assumed to be distant recurrences and where mortality following any other recurrence is the same in both treatment arms (using the probability for WB-EBRT in the base case; see Table 23). This assumption is necessary because trial data show that mortality following any other recurrence is higher for INTRABEAM and, consequently, including costs for this state without such adjustment would simply result in additional incremental cost for WB-EBRT (as WB-EBRT patients live longer in this state). A figure of 60% with distant recurrence was estimated based on data given in the TARGIT-A publication for all patients and data in the literature. 140 The costs of distant recurrence are the major costs in the any other recurrence health state and as a simplification costs for the types of recurrence in this category were not considered. Using the costs given above for distant recurrence and end-of-life care, the results shown in Table 47 were obtained. Health state costs for this scenario are given in Table 48.

TABLE 47 - Illustrative cost-effectiveness results using post-progression costs

LYG, life-years gained.

INTRABEAM is both cheaper and less effective than WB-EBRT; therefore, the ICER represents the £ saved per QALY lost associated with replacing WB-EBRT with INTRABEAM.

Intervention	Total costs (£)	Total LYG	Total QALYs	Incremental cost (£)	Incremental QALYs	ICER (£ saved/QALY lost)
WB-EBRT	4652	20.72	11.329	–	–	–
INTRABEAM	4662	20.51	11.268	–10	–0.061	157a

TABLE 48 - Costs by health state including post-progression costs

Health state	WB-EBRT total costs (£)	INTRABEAM total costs (£)
Recurrence free	2100	1882
Local recurrence	268	345
Disease free after local recurrence	0	0
Any other recurrence	1795	1897
Dead background mortality	0	0
Dead breast cancer	499	527

Table 47 shows that the base-case conclusion does not change when post-progression costs for distant disease and end-of-life care are considered, that is, INTRABEAM is not cost-effective compared with WB-EBRT at a WTP threshold of £20,000 per QALY. However, the cost saving associated with replacing WB-EBRT with INTRABEAM is much smaller as the ICER is reduced from £1596 saved per QALY lost in the base case to £157 saved per QALY lost in the scenario. INTRABEAM is only £10 less expensive than WB-EBRT per patient over the 40-year time horizon considered in the model.

Discussion

INTRABEAM is less expensive but also less effective than WB-EBRT as it is associated with lower total costs but fewer total QALYs. The base case ICER for replacing WB-EBRT with INTRABEAM is £1596 saved per QALY lost (this represents the money saved per QALY lost that is associated with replacing WB-EBRT by INTRABEAM). INTRABEAM is therefore not cost-effective compared with WB-EBRT at the WTP threshold of £20,000 per QALY as the cost saved per QALY lost is less than £20,000.

The CEAC calculated from PSA indicates that at the £20,000 WTP threshold WB-EBRT has a greater probability than INTRABEAM of being cost-effective, at 61.3%. WB-EBRT also has the highest probability of being cost-effective (61.4%) at a WTP of £30,000 per QALY.

The base-case result is subject to a degree of uncertainty as the disease progression parameters included in the model are largely drawn from the TARGIT-A trial. 65 As discussed in Chapter 4, Assessment of effectiveness, and Chapter 7, Statement of principal findings and Strengths and limitations of the assessment, this trial has relatively short follow-up. The numbers experiencing local recurrence in the pre-pathology stratum, which is used to inform the economic model, are also quite small. Results of DSA show that the base-case finding that INTRABEAM is not cost-effective at a WTP of £20,000 per QALY compared with WB-EBRT would be reversed if the probability of experiencing any other recurrence on the INTRABEAM arm was at the low end of its likely range, or if the probability of death from breast cancer on the INTRABEAM arm was at the low end of its likely range.

A strength of the economic model is that it is based on data identified from systematic searches for clinical effectiveness, cost-effectiveness and QoL evidence. Other strengths are that QoL/health state utility weights are taken from studies using the EQ-5D and valued using the UK general population tariff, and that a transparent approach was taken to costing the use of INTRABEAM per procedure by considering all elements of the cost base.

Possible weaknesses of the model are that the systematic review of QoL did not find EQ-5D values to populate all of the model health states and that the clinical effectiveness data used to inform disease progression in the model are drawn largely from one study which has a relatively short follow-up time. 64,65 This study also has a small number of events for the primary outcome in the pre-pathology stratum and the base-case results are therefore subject to some uncertainty. Owing to data limitations, the model does not include costs for the any other recurrence health state in the base case.

Comparison of the economic models

A key structural difference between the Carl Zeiss economic model and the SHTAC’s model is that the Zeiss model has four health states while the SHTAC’s model has six. The SHTAC’s model includes an additional (temporary) health state at local recurrence and also an ‘any other recurrence’ health state which includes metastatic disease. A further structural difference is that the Zeiss model uses an exponential assumption to extrapolate trial local recurrence data over the time horizon of the model, while the SHTAC’s model assumes a log-normal fit to these data. The Zeiss model is run over a 10-year time horizon rather than the 40-year horizon used in the SHTAC’s model.

Different cost and utility data were also used. The Zeiss model uses expert opinion to inform the cost of each INTRABEAM procedure while the SHTAC’s model uses a microcosting approach. The Zeiss model assumes that at local recurrence all INTRABEAM patients have salvage lumpectomy and that all WB-EBRT patients have salvage mastectomy. The cost of salvage mastectomy in the Zeiss model appears to include the cost of breast reconstruction for all patients. In contrast, the SHTAC’s model considers that most INTRABEAM patients will have mastectomy at local recurrence and that of patients having mastectomy, not all of them will have reconstruction.

Utilities used in the Zeiss model were obtained via standard gamble and were not obtained from the general population. Utilities used in the SHTAC’s model were obtained using the EQ-5D and valued with the UK tariff.

Cost-effectiveness

• The systematic review identified two relevant economic evaluations,106,108 both of which were based on the TARGIT-A trial. Both studies were associated with a number of limitations.

• Alvarado et al. 106 developed a Markov decision-analytic model with six health states. Costs and benefits were discounted at 3%, costs were expressed in US$ and the price year was 2011. INTRABEAM was found to be associated with less cost and greater QALYs than WB-EBRT.

• Shah et al. 108 analysed cost-effectiveness through reimbursement models and conducted a cost-minimisation analysis. Methods and assumptions were based on previously published articles. The authors concluded that although INTRABEAM represented a potential cost-saving alternative, WB-EBRT represented a cost-effective modality compared with INTRABEAM based on cost per QALY analyses when additional medical costs and non-medical costs associated with INTRABEAM were factored in.

• Both studies were based in the USA and adopted a societal perspective and are, therefore, not generalisable to the UK NHS.

• The time horizon was 10 years in one study106 and not clearly stated in the other study108 (but assumed to be 10 years based on the estimation of mean utility), which is inappropriate as the risk of local recurrence continues over a lifetime.

• Alvarado et al. 106 used a standard 33 fractions of WB-EBRT in their model, which is more than the current standard UK practice of 15 fractions and will lead to an overestimation of WB-EBRT costs. The number of fractions of WB-EBRT was not reported by Shah et al. 108

• The quality of utility data used in both the studies is questionable. The source study119 was a publication dated 1998, and more recent data would have been appropriate, such as those identified in Chapter 5, Southampton Health Technology Assessments Centre’s systematic review of health-related quality-of-life studies.

Quality of life

• The systematic review on HRQoL studies was conducted with an aim to identify utility data for the SHTAC’s independent model. Nine studies were identified, which were diverse with respect to their aims, interventions, comparators, study designs and methodologies. When assessing the studies on the basis of their relevance to the NICE reference case, only three met all of the criteria (details in Appendix 8). 126,128,132

• The studies provide a source of EQ-5D data for five of the seven health states identified a priori as being potentially relevant for the SHTAC’s independent model. EQ-5D data were not identified for the health states WLE + INTRABEAM or WLE + INTABEAM + WB-EBRT.

Manufacturer’s submission

• The MS evaluated the cost-effectiveness of INTRABEAM in early breast cancer patients when compared with radiotherapy usually given in the UK over 3–6 weeks as WB-EBRT. The total costs, QALYs gained and cost-effectiveness associated with the intervention and comparator under consideration in the appraisal were reported. A multistate Markov model consisting of four health states was constructed. The analysis was conducted for a time period of 20 years with an annual cycle length. The perspective was that of the NHS and benefits and costs were discounted at 3.5%.

• The base-case results indicate that INTABEAM is associated with greater QALYs and lower costs than WB-EBRT. One-way sensitivity analyses and scenario analyses were not conducted. PSA found that at the £20,000 and £30,000 WTP thresholds, INTRABEAM has the highest probability of being cost-effective, at 100% for both thresholds.

• Limited information on the model structure and input parameters is provided in the MS and the AG has raised a number of concerns regarding the methods used; as a consequence the results of the MS model should be viewed with caution.

Southampton Health Technology Assessments Centre’s model

• INTRABEAM is less expensive but less effective than WB-EBRT. The base-case ICER for replacing WB-EBRT with INTRABEAM is £1596 saved per QALY lost. INTRABEAM is, therefore, not cost-effective compared with WB-EBRT at the WTP threshold of £20,000 per QALY.

• At the £20,000 WTP threshold WB-EBRT has a greater probability than INTRABEAM of being cost-effective, at 61.3%. WB-EBRT also has the highest probability of being cost-effective (61.4%) at a WTP of £30,000 per QALY.

• The base-case result is subject to a degree of uncertainty. For four model parameters, the difference in their upper and lower values causes a switch in the treatment option, which is considered cost-effective at a WTP of £20,000 per QALY. Model outcomes are particularly sensitive to the probability of any other recurrence.

• Alternative model health state utility values examined in scenario analysis do not substantively change the base-case findings. Other scenario analyses show that INTRABEAM is dominated by WB-EBRT if it is assumed to serve a smaller catchment population than the base case, and that INTRABEAM dominates WB-EBRT if trial-observed mortality data are used for the first five model cycles.

Clinical effectiveness

1. exp Breast Neoplasms/

2. Carcinoma, Intraductal, Noninfiltrating/

3. (“ductal carcinoma* in situ” or DCIS).tw.

4. (breast* adj5 (neoplasm* or cancer* or tumo?r* or carcinoma* or adenocarcinoma* or sarcoma* or dcis or ductal* or infiltrat* or intraductal* or lobular or medullary or malignan*.tw.

5. (mammar* adj5 (neoplasm* or cancer* or tumo?r* or carcinoma* or adenocarcinoma* or sarcoma* or dcis or ductal* or infiltrat* or intraductal* or lobular or medullary or malignan*)).tw.

6. exp “Neoplasms, Ductal, Lobular, and Medullary”/

7. (breast or mammar*).tw.

8. 6 and 7

9. or/1-5,8

10. intrabeam*.af.

11. Radiosurgery/ or radiosurg*.tw.

12. Radiotherapy, Adjuvant/

13. (radiother* or irradiat* or radiat* or xray or “x-ray”).tw.

14. or/12-13

15. “during surg*”.tw.

16. “radio* guided surg*”.tw.

17. (intraoperativ* or “intra operativ”).tw.

18. (“single dose” or “single fraction*”).tw.

19. or/15-18

20. 14 and 19

21. IORT.tw.

22. (intraoperativ* adj5 radiotherap*).tw.

23. TARGIT*.tw.

24. “tumo?r bed”.tw.

25. (boost* or target*).tw.

26. 13 and 24 and 25

27. 9 and (10 or 11 or 20 or 21 or 22 or 23 or 26)

28. Randomized Controlled Trials as Topic/

29. randomized controlled trial.pt.

30. controlled clinical trial.pt.

31. Controlled Clinical Trial/

32. placebos/

33. random allocation/

34. Double-Blind Method/

35. Single-Blind Method/

36. (random* adj2 allocat*).tw.

37. placebo*.tw.

38. ((singl* or doubl* or trebl* or tripl*) adj (blind* or mask*)).tw.

39. crossover studies/

40. (crossover* or (cross adj over*)).tw.

41. Research Design/

42. ((random* or control*) adj5 (trial* or stud*)).tw.

43. Clinical Trials as Topic/

44. random*.ab.

45. or/28-44

46. 27 and 45

Cost-effectiveness

1. exp Breast Neoplasms/

2. Carcinoma, Intraductal, Noninfiltrating/

3. (“ductal carcinoma* in situ” or DCIS).tw.

4. (breast* adj5 (neoplasm* or cancer* or tumo?r* or carcinoma* or adenocarcinoma* or sarcoma* or dcis or ductal* or infiltrat* or intraductal* or lobular or medullary or malignan*)).tw.

5. (mammar* adj5 (neoplasm* or cancer* or tumo?r* or carcinoma* or adenocarcinoma* or sarcoma* or dcis or ductal* or infiltrat* or intraductal* or lobular or medullary or malignan*)).tw.

6. exp “Neoplasms, Ductal, Lobular, and Medullary”/

7. (breast or mammar*).tw.

8. 6 and 7

9. or/1-5,8

10. intrabeam*.af.

11. Radiosurgery/ or radiosurg*.tw.

12. Radiotherapy, Adjuvant/

13. (radiother* or irradiat* or radiat* or xray or “x-ray”).tw.

14. or/12-13

15. “during surg*”.tw.

16. “radio* guided surg*”.tw.

17. (intraoperativ* or “intra operativ”).tw.

18. (“single dose” or “single fraction*”).tw.

19. or/15-18

20. 14 and 19

21. IORT.tw.

22. (intraoperativ* adj5 radiotherap*).tw.

23. TARGIT*.tw.

24. “tumo?r bed”.tw.

25. (boost* or target*).tw.

26. 13 and 24 and 25

27. 9 and (10 or 11 or 20 or 21 or 22 or 23 or 26)

28. exp economics/

29. exp economics hospital/

30. exp economics pharmaceutical/

31. exp economics nursing/

32. exp economics medical/

33. exp “Costs and Cost Analysis”/

34. Cost Benefit Analysis/

35. exp models economic/

36. exp fees/ and charges/

37. exp budgets/

38. (economic* or cost or costs or costly or costing or price or prices or pricing or pharmacoeconomic*).tw.

39. (value adj1 money).tw.

40. budget$.tw.

41. or/28-40

42. ((energy or oxygen) adj cost).tw.

43. (metabolic adj cost).tw.

44. ((energy or oxygen) adj expenditure).tw.

45. or/42-44

46. 41 not 45

47. (letter or editorial or comment or historical article).pt.

48. 46 not 47

49. 27 and 48

    Lines 50–54 added to strategy on 25/09/2013. Nothing extra found as a consequence.

50. accelerated partial breast irradiation.mp. 430

51. APBI.tw. 266

52. 50 or 51

53. 48 and 52

54. 53 not 49

Health-related quality of life

1. exp Breast Neoplasms/

2. (breast* adj3 (neoplasm* or cancer* or tumo?r* or carcinoma* or adenocarcinoma* or sarcoma* or dcis or ductal* or infiltrat* or intraductal* or lobular or medullary or malignan*)).tw.

3. (mammar* adj3 (neoplasm* or cancer* or tumo?r* or carcinoma* or adenocarcinoma* or sarcoma* or dcis or ductal* or infiltrat* or intraductal* or lobular or medullary or malignan*)).tw.

4. or/1-3

5. (sf36 or sf 36 or short form 36 or shortform 36 or sf thirtysix or sf thirty six or shortform thirstysix or shortform thirty six or short form thirty six or short form thirtysix or short form thirty six).ti,ab.

6. (euroqol or euro qol or eq5d or eq 5d).ti,ab.

7. (hui or hui1 or hui2 or hui3).ti,ab.

8. (health adj3 utilit$ ind$).mp.

9. “EORTC QLQ-BR23”.tw.

10. “FACT-B”.tw.

11. “Functional Assessment of Cancer Therapy Breast”.tw.

12. “BCQ”.tw.

13. “breast cancer chemotherapy questionnaire”.tw.

14. or/5-13

15. 4 and 14

General comments

• Generalisability: women with early breast cancer (although definition of ‘early’ is vague); international study with 6 out of 33 centres in the UK. Unsure whether or not population is typical of those with early breast cancer; also unclear how similar the WB-EBRT treatment is to standard WB-EBRT in the UK.

• Outcome measures: outcomes reported are appropriate. Outcomes reported in linked publications, but are from only one or two participating centres, not for the whole trial population.

• Intercentre variability: teams at each centre were trained and audited by a member of the trial ISC. 64 Observation of the baseline stratification data64 shows differences between centres in the number of patients entering the trial according to the three timings of delivery strata, particularly pre pathology and post pathology. Seven centres had patients in all three strata, 10 centres had patients in two strata (pre-pathology and post pathology, n = 3; pre pathology and contralateral, n = 6; post pathology and contralateral, n = 1), and 11 centres had patients in one stratum only (pre pathology, n = 8; post pathology, n = 3). 64 Centres were allowed to restrict the inclusion criteria beyond the core protocol (e.g. age, tumour size, grade, node) and to stipulate local policy for the delivery of WB-EBRT. Results are not presented by treatment centre nor any comment made in the text so intercentre variability in outcomes is unknown.

• Conflict of interests: appear the same for both the 201064 and 201465 papers. Lead author received a research grant from Photoelectron Corp and Carl Zeiss and also honoraria; one author receives monthly consultancy fees from Carl Zeiss; one author has received a research grant and two authors have received honoraria from Carl Zeiss; Carl Zeiss sponsors most of the travel and accommodation costs for meetings/conferences relating to TARGIT. Only three authors’ travel/accommodation had not been sponsored by Carl Zeiss.

• Other: pivotal trial for TARGIT (INTRABEAM). Registered with ClinicalTrials.gov number NCT00983684.

General comments

• Generalisability: this substudy reports on only 46 IORT and 42 WB-EBRT group participants from the TARGIT-A trial representing only about 2.5% of the total trial population of 3451 randomised participants (1721 TARGIT, 1730 WB-EBRT). 65 It is not clear how generalisable the results are to the remainder of the TARGIT-A trial population or to UK breast cancer patients.

• Outcome measures: questionnaire response rate was 96–99%. The five functioning and symptom scales of the QLQ-C30 and QLQ-BR23 questionnaires were preselected during the design of the study based on a pilot study and relevance for radiation-related QoL in breast cancer. Other subscales and items of the questionnaires were not presented. In addition, states that four other QoL scales were used: the Hospital Anxiety and Depression Scale, the Functional Assessment of Cancer Therapy – Fatigue subscale, the Rosenberg Self-Esteem Scale and the Body Image Scale to control for differences that may inherently exist between treatment groups. Scores for each questionnaire were summed for each scale. However, the paper only narratively comments on differences between groups for these scales (no data).

• On observation of the data, ITT and as-treated QoL outcomes seem similar for the WB-EBRT group, but are difficult to judge for the IORT group because of the way data are presented; for ITT results, IORT and IORT + WB-EBRT are presented as a single group whereas for as-treated results, IORT alone and IORT + WB-EBRT are reported separately.