Benefit, Harm, and Cost-effectiveness Associated With Magnetic Resonance Imaging Before Biopsy in Age-based and Risk-stratified Screening for Prostate Cancer

This decision analytical model uses a hypothetical cohort of men in England to evaluate whether magnetic resonance imaging before biopsy for prostate cancer screening is associated with improvements in benefit-harm and cost-effectiveness profiles compared with biopsy-first screening.

eAppendix. eTable 1. Age-specific cumulative proportion eligible for screening by 10-year absolute risk threshold eTable 2. Outcomes of biopsy-first no screening and biopsy-first screening for prostate cancer eTable 3. Outcomes of biopsy-first no screening and biopsy-first screening for prostate cancer per 10,000 men eTable 4. Outcomes of MRI-first no screening and MRI-first screening for prostate cancer per 10,000 men eFigure 1. Population age distribution eFigure 2. Cost-effectiveness acceptability curves showing MRI-first screening strategies eFigure 3. Cost-effectiveness acceptability frontier showing MRI-first screening strategies eFigure 4. Net monetary benefit of MRI-first and biopsy-first strategies eTable 5. Outcomes of no screening and MRI-first screening for prostate cancer per 10,000 men under assumptions from the PRECISION trial eFigure 5. Benefit/harm profile per 10,000 men with MRI-first screening using parameter estimates from the PRECISION trial eFigure 6. Net monetary benefit per 10,000 men of MRI-first screening and no screening using parameter estimates from the PRECISION trial eFigure 7. Net monetary benefit per 10,000 men of MRI-first screening and no screening at different costs of MRI eFigure 8. Net monetary benefit per 10,000 men of MRI-first screening and no screening at different costs of polygenic risk stratification eFigure 9. Net monetary benefit per 10,000 men of MRI-first screening and no screening at 75% uptake of PSA screening and polygenic risk stratification eFigure 10. Overdiagnosed cancers and prostate cancer deaths prevented per 10,000 men of MRIfirst risk-stratified compared to MRI-first age-based screening when overdiagnosis varies by polygenic risk eFigure 11. Observed vs predicted incidence rate eFigure 12. Observed vs predicted mortality rate eReferences This supplementary material has been provided by the authors to give readers additional information about their work.

Polygenic risk
In England, the 10-year absolute risk of developing prostate cancer rises from 1.3% at age 50 to a peak of 7.2% at the age of 71. 1 Common genetic susceptibility variants explain between 52% and 63% of the population variation in prostate cancer risk, making it the most heritable form of cancer. 2,3 Common susceptibility loci are assumed to interact log-additively, leading to a lognormal distribution of polygenic risk in the population on a relative risk scale. 4 Known prostate cancer susceptibility variants define a log relative risk distribution of 0.68. 5 From this we derived the age-specific proportion of men above each 10-year absolute risk threshold, and the proportion of all cancers that would be diagnosed in these men. 1 These proportions were used to calculate the age-specific relative risk of developing prostate cancer in those men above and below the 10-year absolute risk thresholds.
Men in the highest 10% of this risk distribution have a risk of developing prostate cancer of ~2.5 relative to the mean whilst those in the lowest 10% of the distribution have a relative risk of approximately 0.2.
The cumulative age-specific proportions receiving at least one screen at 10-year absolute risk thresholds varying from 2% to 10% are shown in eTable 1.

eTable 1: Age-specific cumulative proportion eligible for screening by 10-year absolute risk threshold
The cumulative proportion of the population eligible for a screen by age and 10-year absolute risk threshold. For example, at a risk threshold of 2%, at the age of 55 46.4% of the population are eligible for screening. At 56, 53.4% of individuals are eligible for screening. There are 341,434 men aged 55 and 330,382 aged 56. Therefore, the cumulative proportion screened by age 56 is (341,434*0.464 + 330,382*0.534) / (341,434+330,382) = 0.5. The age-specific population and the proportion eligible for screening are available with the source code for this analysis https://github.com/callta/PRIMARI.

Costing descriptions
A detailed description of the costs used, and their derivation, has been published elsewhere. 1 We have adjusted the cost of cancer staging and assessment, and that of active surveillance to reflect the updated 2019 National Institute for Health and Care Excellence (NICE) guidelines. 6 Prostate biopsy costs were based on NICE estimates of the cost of a TRUS and perineal biopsy with relevant histopathology (£311.79 and £652.40, respectively). 7 The 2018 National Prostate Cancer Audit showed that 88% of men with prostate cancer in England had a TRUS biopsy and 12% a perineal biopsy. An assumed 1.4% will be admitted to hospital post-biopsy with complications, 8  Following the ProtecT trial, 11 we assumed that 45% of those eligible will have active surveillance for a period of 10-years, with the remaining 55% having either radical prostatectomy or radiotherapy.

Resource use Magnetic Resonance Imaging
In the screened cohorts, we estimated the number of multi-parametric Magnetic Resonance Imaging (MRI) scans by multiplying the number of individuals screened by the age-specific proportion estimated to have a PSA of ≥3ng/ml, derived from the ProtecT trial. 12

Biopsies
In the non-screened cohort, the number of biopsies was estimated from PROMIS, in which there were 1.88 biopsies for every cancer detected in a clinically-suspected cohort diagnosed using MRI prior to biopsy. 13 This was modelled in probabilistic analyses using a normal distribution and a standard deviation of 0.1.
Two recent meta-analyses have shown that MRI prior to biopsy leads to a third of biopsies being avoided amongst cohorts with clinically-suspected prostate cancer. 14, 15 We assume that under real-world scenarios 20% of those with an MRI score of 1-2 will go on to have biopsy, so the final reduction in the number of biopsies was estimated to be 26.4% (33% x 0.8). This real-world scenario assumption is taken from the NICE 2019 update to the prostate cancer diagnosis guidelines, which used expert opinion to derive the 20% figure. 16

Incidence
By comparison with a biopsy-first approach, MRI followed by biopsy leads to a reduction in the number of clinically insignificant cancers (Gleason 6) detected, and an increase in the number of clinically significant cancers (Gleason ≥7) detected. 14,15 There have been > 25 studies reporting relevant results. In Drost and colleagues Cochrane systematic review and meta-analysis, the pooled absolute difference in the proportion of men detected with clinically significant and clinically insignificant cancers by MRI and systematic biopsy were: 14 -8.2% (95% CI: 6% to 10.3%) reduction in clinically insignificant cancers -2% (95% CI: 1.1% to 4.6%) increase in clinically significant cancers We applied these values to the: baseline incidence of prostate cancer, taking into account the age-specific proportion of cancers detected by stage as seen in the ECRIC database of all clinically detected cases in East Anglia, England. 17 relative increase in incidence of prostate cancer with screening, adjusting for the age-specific proportion of cancers detected by stage in the ProtecT trial. 17 When the tumour grade of a prostate cancer is incorrectly assigned, this is known as misclassification. Cancers misclassified as insignificant in a screening programme are likely to go on to become interval cancers, some of which will be clinically detected. Using the results of the Trio study comparing MRI-targeted, systematic, and both MRI-targeted and systematic biopsy, we have estimated the proportion of cancers incorrectly classified as insignificant (Gleason 6) as 2.76%, 95% CI: 2.06% to 3.46%. 18

Mortality
By detecting more clinically significant disease that might otherwise have been detected at a more advanced stage, alongside evidence for the potential of MRI to enhance pre-therapeutic risk assessment to improve outcomes, 19 the use of MRI prior to biopsy is expected to translate into a reduction in mortality. We have adjusted baseline population prostate-cancer-specific mortality, taking into account the age-specific distribution of cancers at diagnosis, to reflect the impact of MRI.
As there are no data to support a change in relative risk of mortality from prostate cancer with screening in the context of MRI, we have not adjusted this relative risk. Our results are therefore likely to be conservative. However, this reduces the number of assumptions where empirical evidence is not available.

Overdiagnosis
Age-specific overdiagnosis was estimated by multiplying the number of cases by (-0.62 + age x 0.014), derived from Pashayan and colleagues estimates of overdiagnosis with PSA screening. 20 To account for the impact of MRI, we then adjusted the proportion of cases estimated to be overdiagnosed with screening by multiplying this figure by the decrease in clinically insignificant cancers detected.
We calculated the ratio of overdiagnosed cancers to prostate cancer deaths prevented by dividing overdiagnosed cancers by the change in total number of prostate cancer deaths between no screening and the relevant screening strategy. In all scenarios shown in eTable 3, both clinically-detected and screen-detected cancers were assumed to have an MRI prior to biopsy. Net monetary benefit at a willingness-to-pay threshold of £20,000 ($26,000) per QALY gained. Based on 10,000 simulations. To convert to US $, multiply by approximately 1.3. Table 3 in the main manuscript is equivalent to eTable 4, but showing results for the full 4.48 million men. Abbreviations: QALYs, quality-adjusted life-years.

eFigure 2: Cost-effectiveness acceptability curves showing MRI-first screening strategies
Cost-effectiveness acceptability curves (CEAC) of selected MRI-first screening strategies shown. Each CEAC shows the probability at a specific willingness-to-pay threshold of that strategy having a higher net monetary benefit than no screening. To convert to US $, multiply by approximately 1.3. Abbreviation: QALY, quality-adjusted life-year.

eFigure 3: Cost-effectiveness acceptability frontier showing MRI-first screening strategies
Cost-effectiveness acceptability frontier showing the screening strategy with the highest net monetary benefit at each willingness-to-pay (WTP) threshold, and the probability at that given WTP threshold that the strategy has a higher net monetary benefit than no screening. The strategy is indicated in writing towards the top of the graph (risk-stratified screening at absolute risk thresholds from 10% to 5%

Sensitivity Analyses
We performed sensitivity analyses to explore the impact of alternative assumptions regarding change in clinically significant and insignificant cancers detected with MRI, as well as different costs of MRI and polygenic risk assessment.
We based our alternative assumptions on the impact of MRI on clinically significant and insignificant cancers on the PRECISION trial. 21 This showed a decrease in the number of clinically insignificant cancers detected of -13% (95% CI: -7% to -19%), and an increase in clinically significant cancers detected of 12% (95% CI: 4% to 20%). These values were chosen as their confidence intervals include those seen in PROMIS. 13 We re-ran our probabilistic model using these estimates.