Polygenic risk-tailored screening for prostate cancer: A benefit-harm and cost-effectiveness modelling study.

BACKGROUND: The United States Preventive Services Task Force supports individualised decision-making for prostate-specific antigen (PSA)-based screening in men aged 55-69. Knowing how the potential benefits and harms of screening vary by an individual's risk of developing prostate cancer could inform decision-making about screening at both an individual and population level. This modelling study examined the benefit-harm tradeoffs and the cost-effectiveness of a risk-tailored screening programme compared to age-based and no screening. METHODS AND
FINDINGS: A life-table model, projecting age-specific prostate cancer incidence and mortality, was developed of a hypothetical cohort of 4.48 million men in England aged 55 to 69 years with follow-up to age 90. Risk thresholds were based on age and polygenic profile. We compared no screening, age-based screening (quadrennial PSA testing from 55 to 69), and risk-tailored screening (men aged 55 to 69 years with a 10-year absolute risk greater than a threshold receive quadrennial PSA testing from the age they reach the risk threshold). The analysis was undertaken from the health service perspective, including direct costs borne by the health system for risk assessment, screening, diagnosis, and treatment. We used probabilistic sensitivity analyses to account for parameter uncertainty and discounted future costs and benefits at 3.5% per year. Our analysis should be considered cautiously in light of limitations related to our model's cohort-based structure and the uncertainty of input parameters in mathematical models. Compared to no screening over 35 years follow-up, age-based screening prevented the most deaths from prostate cancer (39,272, 95% uncertainty interval [UI]: 16,792-59,685) at the expense of 94,831 (95% UI: 84,827-105,630) overdiagnosed cancers. Age-based screening was the least cost-effective strategy studied. The greatest number of quality-adjusted life-years (QALYs) was generated by risk-based screening at a 10-year absolute risk threshold of 4%. At this threshold, risk-based screening led to one-third fewer overdiagnosed cancers (64,384, 95% UI: 57,382-72,050) but averted 6.3% fewer (9,695, 95% UI: 2,853-15,851) deaths from prostate cancer by comparison with age-based screening. Relative to no screening, risk-based screening at a 4% 10-year absolute risk threshold was cost-effective in 48.4% and 57.4% of the simulations at willingness-to-pay thresholds of GBP£20,000 (US$26,000) and £30,000 ($39,386) per QALY, respectively. The cost-effectiveness of risk-tailored screening improved as the threshold rose.
CONCLUSIONS: Based on the results of this modelling study, offering screening to men at higher risk could potentially reduce overdiagnosis and improve the benefit-harm tradeoff and the cost-effectiveness of a prostate cancer screening program. The optimal threshold will depend on societal judgements of the appropriate balance of benefits-harms and cost-effectiveness.

Introduction

Screening with a prostate-specific antigen (PSA) test could reduce death from prostate cancer in some men but at the cost of substantial numbers overdiagnosed, as well as false positive results [1]. Overdiagnosed cancers are the screen-detected cancers that would not otherwise have come to light during an individual’s lifetime [2], whose diagnosis and related treatment can lead to avoidable physical and psychological harms whilst also incurring an economic cost [3]. The updated 2018 guidelines of the US Preventive Services Taskforce recommend consideration of screening for certain at risk men between the ages of 55 and 69 [1]. Understanding the variation in the potential benefits and harms of screening by an individual’s risk of developing prostate cancer could inform decision-making about screening at both the individual and population level.

Genome-wide association studies have identified common susceptibility loci that together explain approximately 37% of the familial relative risk of prostate cancer [4]. Individually, these loci have little clinical significance [5]. However, together they define a risk distribution with a variance of 0.68 and area under the curve (AUC) of 0.72 that can be used to stratify individuals into groups at higher and lower risk of developing prostate cancer [2,6], enabling tailored screening by risk group. Individuals in the first and 99th percentiles of the polygenic risk distribution have relative risks of developing prostate cancer of 0.09 and 5.52, respectively, compared to the population mean. Almost 49% of prostate cancers occur in those men in the highest 20% of the polygenic risk distribution, and only 7% occur amongst those in the lowest 20% of the risk distribution (Fig C in S1 Appendix). Avoiding screening of those at lower risk may consequently reduce the harms of screening without a commensurate loss of benefit. The proportion of prostate cancers overdiagnosed varies inversely with polygenic risk, with almost 50% lower overdiagnosis in men in the highest quartile compared with the lowest quartile of polygenic risk [2]. This approach to screening would require genotyping of all men for the purpose of risk assessment. However, the additional costs of population-wide genotyping may be offset by lower levels of overtreatment.

The aims of this study were to assess the balance of benefit and harms, as well as the cost-effectiveness, of the introduction of a polygenic risk-tailored screening programme for prostate cancer.

Methods

Model structure

We developed a life-table model (Fig A in S1 Appendix) to simulate cohorts of men under 3 scenarios: no screening, age-based screening with PSA, and risk-tailored screening (henceforth, precision screening). Our life table is a cohort-based Markov model that estimates the age-specific incidence of prostate cancer, deaths from prostate cancer, and deaths from other causes. Each hypothetical cohort consisted of 4.48 million men aged 55 to 69, representing the mean population of men of this age in England in 2013–2016 [7]. All cohorts were assumed to be prostate-cancer–free on entering the model and were followed up to the age of 90.

Age-based screening involved PSA testing every 4 years between the ages of 55 and 69, reflecting the screening interval used in the core analyses of the European Randomized Study of Screening for Prostate Cancer [8]. In the precision screening cohort, we estimated the 10-year absolute risk of developing prostate cancer based on age and polygenic profile. From the age of 55 to 69, this cohort of men started quadrennial PSA screening at the age at which they reached a specified risk threshold, which was varied between a 2% and 10% 10-year absolute risk of developing prostate cancer. We set a PSA cutoff of 3.0 ng/mL for suspected prostate cancer requiring further assessment for both the age-based and precision screening cohorts.

Precision screening

The distribution of polygenic risk on a relative risk scale in the population is log-normal, assuming a log-additive interaction between loci [6]. The variance of the polygenic risk distribution was calculated as 0.68, based on known prostate cancer susceptibility variants (see S1 Appendix for further details) [6,9].

We calculated the log relative risk of developing prostate cancer for each risk threshold respective to the background 10-year absolute risk of developing prostate cancer in the absence of screening. By applying the log relative risk of developing prostate cancer to the polygenic risk distribution, we then determined the proportion above the risk threshold and the proportion of all cases of prostate cancer accounted for amongst those above the threshold. From this, we derived the relative risk of developing prostate cancer in those who were screened or not screened.

Model parameters

Model parameters are shown in Table 1. We used Office for National Statistics data and DevCan software to calculate the age-specific incidence of prostate cancer, mortality from prostate cancer, and mortality from other causes, accounting for competing risks (Table B in S1 Appendix) [10]. The incidence of prostate cancer in the nonscreened cohorts was adjusted by 10% to reflect the estimated proportion of cancers that are diagnosed as a result of opportunistic screening in England (see S1 Appendix for further details). Costs were estimated from the perspective of the National Health Service (NHS) in 2016 prices using Unit Costs of Health and Social Care and NHS Reference Costs, which have been previously validated for use in costing prostate cancer care (Table 1 and Table A in S1 Appendix) [11-13]. We determined the proportion receiving different treatments by stage at diagnosis from the National Cancer Registration and Analysis Service, the National Prostate Cancer Audit, and National Institute for Health and Care Excellence (NICE) guidelines [14-16]. The cost of polygenic risk stratification was based on an empirical estimate.

Table 1

Model parameters.

Parameter	Value (95% CI)a	Distribution Used in Probabilistic Analyses (α, β)b	Description	Source
Life table
RR of prostate-cancer–specific mortality with screening	0.79 (0.69 to 0.91)	Log-normal [SE: 0.06]	The relative reduction in mortality seen with screening with PSA in the ERSPC.	[8]
RR of incidence of prostate cancer with screening	1.23 (1.03 to 1.48)	Log-normal [SE: 0.18]	Relative increase in the incidence of prostate cancer in the presence of screening with PSA, derived from a meta-analysis of randomised controlled trials of PSA screening.	[18]
Proportion overdiagnosed	−0.62 + age × 0.014	Beta [SE: 0.001]	Derived from linear regression of estimates for the risk of overdiagnosis in 5-year age groups.	[2]
RR of advanced cancer at diagnosis if screened	0.85 (0.72 to 0.99)	Log-normal [SE: 0.07]	Relative decrease in the proportion of cancers that are considered advanced (stages III or IV) if screen-detected, derived from a meta-analysis of randomised controlled trials of PSA screening.	[18]
Utility values
General population utility	0.8639 (0.852 to 0.875)	0.83 + Gamma (4, 0.06) × 0.167	A yearly utility decline of 0.0048 (0.004 to 0.006) was estimated from linear regression of the mean health state values in 5-year intervals from 45 to 90 against age.	[19]
Relative reduction in utility for those with prostate cancer	0.93, range: [0.88 to 1.0]	0.88 + Gamma (5, 0.05) × 0.2	Average over 10 years. Sampled from a right-skewed distribution in probabilistic analysis.	[20]
Costs (GBP£ in 2016 prices)c,d
PSA testing	11 (7 to 15)	Gamma (33.9, 0.3)		[11,12]
Polygenic risk stratification	25 (17 to 33)	Gamma (33.9, 0.7)	Empirical estimate calculated from the laboratory costs of genotyping a similar number of SNPs.
Biopsy	388 (260 to 516)	Gamma (33.9, 11.5)		[11,21,22]
Declined biopsy, but had a PSA ≥ 3.0 ng/ml	105 (70 to 140)	Gamma (33.9, 3.1)	Individuals who declined biopsy but had a PSA ≥ 3.0 ng/ml were assumed to have one urological appointment.e	[11]
Staging of diagnosed cancer	770 (516 to 1,024)	Gamma (33.9, 22.7)	Cost of staging with MRI and an isotope bone scan.	[11,21,22]
Active surveillance	4,341 (2,908 to 5,774)	Gamma (33.9, 128.1)	Average over 10 years, assuming that 55% will go on to have radical therapy during this time period.	[11,23,24]
Radical prostatectomy	8,173 (5,476 to 10,870)	Gamma (33.9, 241.2)	Incorporating the costs of complications and follow-up over 5 years.	[11,21,22]
Radical radiotherapy	5,385 (3,608 to 7,162)	Gamma (33.9, 158.9)	Incorporating the costs of complications and follow-up over 5 years.	[11,21,22]
Brachytherapy	1,527 (1,023 to 2,031)	Gamma (33.9, 45.1)		[11,21,22]
Chemotherapy	7,426 (4,975 to 9,877)	Gamma (33.9, 219.2)		[11,21,25]
Androgen deprivation therapy	559 (375 to 744)	Gamma (33.9,16.5)	Derived from the NICE costing statement of its prostate cancer, inflated to 2015–2016 prices, with the addition of 1 urological appointment as follow-up.	[16]
Palliation and death from prostate cancer	6,837 (535 to 20,257f)	Gamma (1.8, 3854.9)	Inflated from 2013–2014 estimated costs to the healthcare system in the last 12 months of life.	[26]

a95% CI unless otherwise stated.

bα and β parameters shown unless otherwise stated.

cIn sensitivity analyses, it was assumed that 95% of the costs are likely to vary no more than approximately one-third from the calculated baseline value [27].

dAll costs are in 2016 GBP£.

eExcept in sensitivity analysis, all men eligible for biopsy were assumed to have one.

f95% credible interval—see Table A in S1 Appendix for further details. Abbreviations: CI, confidence interval; ERSPC, European Randomized Study of Screening for Prostate Cancer; NICE, National Institute for Health and Care Excellence; PSA, prostate-specific antigen; RR, relative risk; SE, standard error; SNP, single nucleotide polymorphism.

We developed utility weights for those with prostate cancer based on treatment modality (see S1 Appendix for further details). We estimated the number of overdiagnoses by multiplying the number of screen-detected cases by the age-specific proportion estimated to be at risk of overdiagnosis [17].

Model outputs

We calculated costs, the number of quality-adjusted life-years (QALYs), life-years, prostate cancer cases, overdiagnoses, deaths from prostate cancer, and the number of overdiagnoses per prostate cancer death averted. Both costs and benefits were discounted at 3.5% per annum, as per NICE guidelines [28].

We calculated incremental cost-effectiveness ratios (ICERs) by dividing the difference in mean costs between the compared scenarios by the difference in mean QALYs, derived from 10,000 probabilistic simulations [29]. The net monetary benefit (NMB) was calculated by subtracting the costs from the product of the QALYs and the willingness-to-pay threshold. We ranked the cost-effectiveness of each screening strategy using different risk thresholds by NMB, using willingness-to-pay thresholds of £20,000 ($26,000) and £30,000 ($39,000) per QALY gained because these reflect the range of thresholds used by NICE in the consideration of the cost-effectiveness of an intervention [28].

Probabilistic and deterministic sensitivity analyses

To account for uncertainty in the parameters, we ran 10,000 simulations for each screening scenario; the distributions used for each input varied are shown in Table 1. The results presented are probabilistic unless otherwise stated. In sensitivity analyses, we evaluated precision screening starting at the age of 45, 50, and 60; the impact of different levels of screening uptake and adherence; overdiagnosis when adjusted for polygenic risk; and the cost of polygenic testing. We used Python v3.6.5 for all analyses; the code is available at https://github.com/callta/precision_screening_pca. The CHEERS checklist [30] was used in the preparation of this manuscript.

Results

Risk of developing prostate cancer

Our analyses show that the background 10-year absolute risk of developing prostate cancer in the absence of screening rose from 2.6% to 7.1% between the ages of 55 and 69 (Fig B in S1 Appendix). Just under half of men aged 55 (49.1%) had a 10-year absolute risk of developing prostate cancer of ≥2% based on age and polygenic risk profile, increasing to 90.6% of men aged 69. The proportion of men by age at or above each 10-year absolute risk threshold is shown in Fig E in S1 Appendix.

Benefits and harms of screening

In a hypothetical cohort of 4.48 million men, age-based screening led to 39,272 fewer deaths from prostate cancer and to 94,831 overdiagnoses, representing 2.4 overdiagnoses per prostate cancer death averted, as well as 764,446 additional biopsies over 35 years follow-up, by comparison with no screening. The tradeoff between the benefits and harms of precision screening varied by risk threshold (Fig 1, Table 2, Table C in S1 Appendix). In the precision screening cohort, at a 2% 10-year absolute risk threshold, 36,534 deaths from prostate cancer were prevented at the expense of 84,681 overdiagnoses by comparison with no screening. As the risk threshold was raised to 10%, 14,507 deaths were prevented and 26,791 overdiagnosed by comparison with no screening. This represented a drop in the ratio of overdiagnosed cases to prostate cancer deaths prevented from 2.3 to 1.8 at risk thresholds of 2% and 10%, respectively (Fig F in S1 Appendix), and there was a reduction in the number of additional biopsies performed from 652,177 to 150,635 (Table C in S1 Appendix).

Fig 1

Overdiagnosed cases and prostate cancer deaths prevented with precision screening as compared to age-based screening.

Results based on 10,000 simulations.

Table 2

Outcomes of age-based and precision screening compared with no screening.

Screening Strategy	Prostate Cancer Cases (n)	Difference with No Screening	Overdiagnosed Cases (n)	Deaths from Prostate Cancer (n)	Difference with No Screening	QALYs (n)	Difference with No Screening	Costs (£ Millions)	Difference with No Screening (£ Millions)	ICER (£/QALY Gained)	Cumulative Percentage Screeneda (%)
No screening	537,870		N/A	192,623		46,682,945		2,975			0
Age-based screening	644,047	106,177	94,831	153,351	−39,272	46,699,360	16,416	3,549	574	34,952	100
Precision screening (10-year AR)
2.0%	622,733	84,863	84,681	156,089	−36,534	46,702,653	19,709	3,572	597	30,297	75.4
2.5%	614,230	76,360	79,620	157,723	−34,900	46,703,346	20,401	3,537	562	27,542	66.7
3.0%	606,014	68,144	74,419	159,482	−33,141	46,703,788	20,844	3,503	527	25,290	58.9
3.5%	598,318	60,448	69,298	161,275	−31,348	46,704,012	21,067	3,469	494	23,446	51.9
4.0%	591,244	53,375	64,384	163,045	−29,578	46,704,054	21,109	3,438	463	21,924	45.8
4.5%	584,818	46,949	59,743	164,759	−27,864	46,703,950	21,006	3,409	434	20,659	40.5
5.0%	579,026	41,156	55,406	166,397	−26,226	46,703,733	20,788	3,383	407	19,598	35.9
5.5%	573,830	35,960	51,379	167,947	−24,676	46,703,427	20,482	3,358	383	18,704	31.9
6.0%	569,186	31,316	47,656	169,407	−23,216	46,703,054	20,109	3,336	361	17,947	28.4
6.5%	565,045	27,175	44,224	170,775	−21,848	46,702,631	19,686	3,316	341	17,303	25.4
7.0%	561,358	23,488	41,065	172,055	−20,568	46,702,172	19,227	3,298	322	16,755	22.7
7.5%	558,079	20,209	38,160	173,250	−19,373	46,701,687	18,743	3,281	305	16,289	20.4
8.0%	555,165	17,295	35,490	174,364	−18,258	46,701,187	18,242	3,265	290	15,894	18.4
8.5%	552,577	14,707	33,036	175,403	−17,220	46,700,677	17,732	3,251	276	15,560	16.6
9.0%	550,279	12,410	30,779	176,371	−16,251	46,700,163	17,218	3,238	263	15,281	15.0
9.5%	548,241	10,371	28,703	177,274	−15,349	46,699,649	16,704	3,227	251	15,050	13.6
10.0%	546,432	8,562	26,791	178,116	−14,507	46,699,140	16,195	3,216	241	14,862	12.3

Outcomes shown in hypothetical cohorts of 4.8 million men aged 55 to 69 in England, followed up to age 90. Abbreviations: AR, absolute risk; ICER, incremental cost-effectiveness ratio; N/A, not applicable; QALY, quality-adjusted life-year.

aTotal cumulative proportion eligible for screening by the aged 69. See Fig E in S1 Appendix for an indication of the proportions eligible for precision screening as they age by different risk thresholds. Each ICER represents that specific screening strategy compared with no screening.

Cost-effectiveness

Age-based screening led to an additional 16,416 QALYs by comparison with no screening, at a cost of £574 million ($746 million) over 35 years, such that the ICER was £34,952 ($45,437) per QALY gained. In 40.7% of simulations, the ICER of age-based compared with no screening was ≤£20,000 ($26,000) per QALY gained.

In the precision screening model, below a 4.5% 10-year absolute risk threshold, there was a plateau in the QALYs gained by comparison with no screening that subsequently began to fall as the risk threshold was raised, whilst the costs of precision screening continued to drop as the risk threshold increased (Fig 2). At all 10-year absolute risk thresholds below 10%, precision screening led to a greater number of incremental QALYs gained than age-based screening whilst incurring fewer additional costs at all risk thresholds above 2% (Fig 2).

Fig 2

Incremental cost and QALYs of precision and age-based screening compared with no screening.

Results based on 10,000 simulations. The solid lines describe the incremental costs incurred and QALYs gained of precision screening versus no screening, whilst the dashed lines represent the incremental costs and QALYs of age-based versus no screening. QALY, quality-adjusted life-year.

At a 2% 10-year absolute risk threshold, the ICER was £30,297 ($39,386) per QALY gained. This declined as the risk threshold increased, reaching a plateau at a threshold of approximately 7%, at which point the ICER was £16,755 ($21,781) per QALY gained. Precision screening was cost-effective at a willingness-to-pay threshold of £20,000 ($26,000) per QALY gained compared to no screening at all 10-year absolute risk thresholds above 4.5%. At a risk threshold of 5%, precision screening had a 51.5% probability of being cost-effective and an ICER of £19,598 ($25,478), rising to a 62.5% probability of being cost-effective and an ICER of £14,862 ($19,320) at a threshold of 10%.

NMB of screening strategies

Comparing all precision and age-based strategies with no screening, the highest NMB at willingness-to-pay thresholds of £20,000 ($26,000) and £30,000 ($39,000) per QALY was seen with precision screening at a 10% and 8% 10-year absolute risk threshold, respectively. Screening strategies by NMB are presented in Fig H in S1 Appendix, whilst the cost-effectiveness acceptability planes, curves, and frontier are shown in Figs I–J in S1 Appendix. Age-based screening had a lower NMB than all precision screening strategies.

Sensitivity analyses

The results of sensitivity analyses are available in S1 Appendix.

Discussion

This modelling analysis has shown that precision screening based on age and polygenic risk could reduce overdiagnosis whilst preserving most of the mortality benefits of age-based screening for prostate cancer. At all risk thresholds studied, precision screening had a higher NMB, lower ICER, and fewer overdiagnosed prostate cancers than age-based screening. The cost-effectiveness of screening increases as the risk threshold rises, plateauing at a risk threshold of approximately 7%. However, the greatest QALY gains are at a 10-year absolute risk threshold of 4%. A precision screening programme using a 4% risk threshold would reduce overdiagnosis by one-third, yield more QALYs, and cost less whilst maintaining the benefits of screening. The ideal strategy will depend on both a society’s willingness to pay for each QALY gained as well as the tradeoff between benefits and harms considered acceptable both at an individual and population level [31].

A plateau is also seen in the cost-effectiveness as the risk threshold for precision screening rises (Table 2, Fig 2). This reflects the fact that fewer deaths are being prevented relative to the increased number of prostate cancer cases, and therefore the greater number of years lived with prostate cancer, as men at higher risk are screened. The incremental QALYs gained with precision screening begin to drop at a 10-year risk threshold above 4%. However, the ICER of precision screening does not begin to plateau until the 10-year absolute risk threshold is raised to 7%. Together, this suggests that a strategy of precision screening at a 10-year absolute risk threshold of between 4% and 7% may provide the most appropriate balance of harms and benefits, considering prostate cancer deaths prevented, cases overdiagnosed, and QALYs gained for the additional costs of screening.

Screening men at a higher risk of prostate cancer lowers the proportion of overdiagnosed cases, the number of additional biopsies performed, and the ratio of overdiagnosed cases to prostate cancer deaths averted. As the risk threshold rose, a smaller proportion of men became eligible for screening. Overdiagnosis dropped as the risk threshold increased. With fewer men screened, there were fewer prostate cancer deaths averted compared to age-based screening. However, the extent of the drop in overdiagnosis was greater than the extent of prostate cancer deaths not prevented, leading to an improvement in the benefit–harm profile as the risk threshold rose (Fig G in S1 Appendix).

In the UK, NICE considers interventions with an ICER of ≤£20,000 per QALY gained as cost-effective, a threshold that was reached with all precision screening strategies above a 5% 10-year absolute risk threshold [28]. A precision screening strategy using a 5% 10-year absolute risk reflects the average risk of developing prostate cancer in men aged 61 in England. A programme employing this strategy would screen 1 in 10 men (11.4%) at the age of 55, rising to just over half (50.5%) by the age of 69. This strategy would reduce overdiagnosis by a 41.6%, yield more QALYs, and cost less, at the expense of 8.5% fewer prostate cancer deaths averted by comparison with age-based screening.

Precision screening for prostate cancer would involve an evolution in screening services, with logistical and ethical implications. Risk tailoring implies that different individuals are invited to screening at different ages, with potential knock-on effects on screening delivery [32]. In addition, although the disclosure of genetic material to insurance companies is covered under a moratorium in England [33], the broader impact of risk-tailored screening using genetic material on individuals and society requires further research. Finally, the introduction of screening programmes could risk widening health inequalities between both socioeconomic classes and ethnic groups, which can occur as a result of varied uptake amongst different socioeconomic strata and ethnicities of screening [34,35]. Research to mitigate this occurrence should be considered alongside prospective studies of polygenic risk-based screening.

In a precision screening programme, in addition to altering the screening start age, the frequency of screening could be varied according to risk. For example, men at higher risk may receive more frequent screening and men at lower risk receive less frequent or no screening. Intensified screening could improve the benefit–harm tradeoffs if the sojourn time—the time it takes to progress from preclinical screen-detectable cancer to clinically detectable cancer—varies with risk level [36]. In the absence of these data in the context of polygenic risk, the impact of varying screening intervals was not estimated.

There have been no comparable studies estimating the impact of polygenic risk-tailored screening in prostate cancer. However, the conclusions reached from our precision and age-based screening models compare favourably with a microsimulation model from the US [37], attesting to the underlying robustness of our model in spite of the differences in model design and assumptions. Using a microsimulation model and a selective treatment strategy involving initial conservative management for those with localised cancer (Gleason score <7 and stage T2a), Roth and colleagues state that quadrennial age-based screening between the ages of 55 and 69 with a PSA cutoff of 3 ng/ml would lead to 30 additional QALYs per 10,000 men screened (37 in this analysis) [37].

Limitations

Follow-up of individuals ended at the age of 90, reflecting the increasing uncertainty in estimates regarding the incidence and mortality from prostate cancer beyond this age. A life-table approach using aggregate data has been used because of uncertainty in how the natural history of prostate cancer varies by risk. Recent data suggest that there is limited variation in health-related quality of life between stages of prostate cancer, with diminishing utility at higher stages of disease [38]. Because screening leads to a greater proportion of cases detected at early stages, the QALYs recorded in screened cohorts in our model may be underestimated. However, our approach produces similar results to microsimulation models taking into account natural history for age-based screening [37]. Simplifying the model structure in this way minimises the number of underlying assumptions, whilst parameter uncertainty is accounted for with probabilistic sensitivity analyses. The estimates of resource use are limited by the absence of data regarding whether the stage at diagnosis of screen-detected cancers and whether response to treatment varies by polygenic risk.

Because there are no data on how overdiagnosis varies for each percentile of the risk distribution, we have assumed the proportion overdiagnosed is equivalent to that seen with PSA screening alone. In sensitivity analyses in which we assume overdiagnosis to vary with polygenic risk, the balance of benefits and harms of precision screening is substantially improved, suggesting that our model underestimates the potential benefits of precision screening (Fig M in S1 Appendix). We have also assumed that precision screening does not lead to a greater relative risk reduction in mortality than age-based screening. Polygenic hazard scores have been shown to be predictive of aggressive cancer, which potentially could disproportionately benefit from screening [39]. This may lead to a more conservative estimate of the benefit/harm ratio attributed to precision screening in this model.

Conclusion

Our analyses show that precision screening based on age and polygenic risk profile could improve the benefit/harm tradeoff and cost-effectiveness of a screening programme for prostate cancer. Offering screening to men at a 10-year absolute risk threshold between 4% and 7% could lead to greater QALYs, lower costs, and a 32.1% to 56.7% reduction in overdiagnosis when compared to age-based screening. These findings require verification by a prospective randomised evaluation.

Completed CHEERS checklist of items to report in a health economic evaluation.

CHEERS, Consolidate Health Economic Evaluation Reporting Standards.

(DOCX)

Click here for additional data file.

Detailed methods and supplementary results.

(DOCX)

Click here for additional data file.

17 Sep 2019

Dear Dr. Callender,

Thank you very much for submitting your manuscript "Polygenic Risk-Tailored Screening for Prostate Cancer: A Benefit-Harm and Cost-Effectiveness Analysis" (PMEDICINE-D-19-02276) for consideration at PLOS Medicine.

Your paper was evaluated by a senior editor and discussed among all the editors here. It was also discussed with an academic editor with relevant expertise, and sent to independent reviewers, including a statistical reviewer. The reviews are appended at the bottom of this email and any accompanying reviewer attachments can be seen via the link below:

[LINK]

In light of these reviews, I am afraid that we will not be able to accept the manuscript for publication in the journal in its current form, but we would like to consider a revised version that addresses the reviewers' and editors' comments. Obviously we cannot make any decision about publication until we have seen the revised manuscript and your response, and we plan to seek re-review by one or more of the reviewers.

In revising the manuscript for further consideration, your revisions should address the specific points made by each reviewer and the editors. Please also check the guidelines for revised papers at http://journals.plos.org/plosmedicine/s/revising-your-manuscript for any that apply to your paper. In your rebuttal letter you should indicate your response to the reviewers' and editors' comments, the changes you have made in the manuscript, and include either an excerpt of the revised text or the location (eg: page and line number) where each change can be found. Please submit a clean version of the paper as the main article file; a version with changes marked should be uploaded as a marked up manuscript.

In addition, we request that you upload any figures associated with your paper as individual TIF or EPS files with 300dpi resolution at resubmission; please read our figure guidelines for more information on our requirements: http://journals.plos.org/plosmedicine/s/figures. While revising your submission, please upload your figure files to the PACE digital diagnostic tool, https://pacev2.apexcovantage.com/. PACE helps ensure that figures meet PLOS requirements. To use PACE, you must first register as a user. Then, login and navigate to the UPLOAD tab, where you will find detailed instructions on how to use the tool. If you encounter any issues or have any questions when using PACE, please email us at PLOSMedicine@plos.org.

We expect to receive your revised manuscript by Oct 02 2019 11:59PM. Please email us (plosmedicine@plos.org) if you have any questions or concerns.

***Please note while forming your response, if your article is accepted, you may have the opportunity to make the peer review history publicly available. The record will include editor decision letters (with reviews) and your responses to reviewer comments. If eligible, we will contact you to opt in or out.***

We ask every co-author listed on the manuscript to fill in a contributing author statement, making sure to declare all competing interests. If any of the co-authors have not filled in the statement, we will remind them to do so when the paper is revised. If all statements are not completed in a timely fashion this could hold up the re-review process. If new competing interests are declared later in the revision process, this may also hold up the submission. Should there be a problem getting one of your co-authors to fill in a statement we will be in contact. YOU MUST NOT ADD OR REMOVE AUTHORS UNLESS YOU HAVE ALERTED THE EDITOR HANDLING THE MANUSCRIPT TO THE CHANGE AND THEY SPECIFICALLY HAVE AGREED TO IT. You can see our competing interests policy here: http://journals.plos.org/plosmedicine/s/competing-interests.

Please use the following link to submit the revised manuscript:

https://www.editorialmanager.com/pmedicine/

Your article can be found in the "Submissions Needing Revision" folder.

To enhance the reproducibility of your results, we recommend that you deposit your laboratory protocols in protocols.io, where a protocol can be assigned its own identifier (DOI) such that it can be cited independently in the future. For instructions see http://journals.plos.org/plosmedicine/s/submission-guidelines#loc-methods.

Please ensure that the paper adheres to the PLOS Data Availability Policy (see http://journals.plos.org/plosmedicine/s/data-availability), which requires that all data underlying the study's findings be provided in a repository or as Supporting Information. For data residing with a third party, authors are required to provide instructions with contact information for obtaining the data. PLOS journals do not allow statements supported by "data not shown" or "unpublished results." For such statements, authors must provide supporting data or cite public sources that include it.

We look forward to receiving your revised manuscript.

Sincerely,

Adya Misra, PhD

Senior Editor

PLOS Medicine

plosmedicine.org

-----------------------------------------------------------

Requests from the editors:

Abstract-please explain the life table model briefly

Abstract-please clarify what the analysis from the health service perspective would entail. If this is intention to pay or cost burden, please include this information.

Abstract-in the methods/findings section please include a sentence about the limitations of your methodology

Abstract-please quantify all results using p-values and 95% confidence intervals

Abstract-in the conclusions section please clarify that these are based on a modelling analysis

Abstract-please tone down the language in the abstract to reflect that there is no clinical application of polygenic risk scores in prostate cancer screening

Author summary- At this stage, we ask that you include a short, non-technical Author Summary of your research to make findings accessible to a wide audience that includes both scientists and non-scientists. The Author Summary should immediately follow the Abstract in your revised manuscript. This text is subject to editorial change and should be distinct from the scientific abstract. Please see our author guidelines for more information: https://journals.plos.org/plosmedicine/s/revising-your-manuscript#loc-author-summary

References-please use square brackets and remove italics from citations. Please use the "Vancouver" style for reference formatting, and see our website for other reference guidelines https://journals.plos.org/plosmedicine/s/submission-guidelines#loc-references

Introduction-Please include a sentence about PSA testing, why there is a substantial risk of false positive diagnosis. The second sentence requires clarification, please revise or consider removing.

Introduction-please introduce polygenic risk at first view and explain how this may enable tailored screening

Methods-Please ensure that the correct link to the Python code has been provided as the current link is not working

Methods-Please provide paragraph or line numbers in the CHEERS checklist as page numbers are likely to change

Methods-Please briefly explain the life table model

Discussion- please include something like “according to our modelling analysis...” at the start of the section

Last sentence of the conclusion- please revise to clarify that and RCT would be essential, instead of valuable

Overall please use careful language to indicate that the results are based on a model

Comments from the reviewers:

Reviewer #1: In this article, the authors use a life-table based model to estimate the benefits, harms and cost-effectiveness of a program that screens men in the U.K. for prostate cancer based on their risk (based on age and polygenic profile) versus age-based screening and no screening.

The authors found that risk-based screening using a risk threshold for screening commencement of 5% or over was cost-effective (using a willingness-to-pay threshold of £20,000 per QALY) when compared with age-based screening.

Overall this is an interesting study that is addressing the important question of how to optimize prostate cancer screening and limit the harms associated with the high proportion of overdiagnosed cases. However I believe that the modelling approach used in this paper is not the most appropriate and restricts the possible characteristics of a risk-based screening program that can be explored and evaluated.

For example, risk-based screening programs using different screening intervals and/or PSA thresholds cannot be evaluated because the relative risk of prostate cancer mortality in the presence of screening is taken directly from the ERSPC trial and therefore the model in this manuscript is constrained to do what was done in the trial.

Well calibrated and validated microsimulation models that use a more detailed natural history of prostate cancer are a better alternative to perform a comprehensive evaluation of a risk-based screening program for prostate cancer. Such models have been developed for some countries (e.g. the United States and the Netherlands - see https://cisnet.cancer.gov/prostate/) and have been successfully used in several evaluations, some of which are cited in the manuscript. This alternative approach would for example

* Allow different screening intervals, PSA thresholds, starting/stopping ages for screening to be evaluated

* Evaluate emerging triage technologies (i.e. not all men with PSA > 3ng/ml would be sent directly to biopsy - this could potentially further reduce overdiagnosis);

* Directly assert from the simulation whether a prostate cancer case is an overdiagnosed case as opposed to using a separate model for this classification;

* Better model the introduction of screening and the transitory effects that could occur.

A microsimulation model of prostate cancer that accounted for individual risk profile would be an interesting and worthwhile addition to the current status of prostate cancer modelling research.

.

The authors rightly state in the appendix that a life table based model is simpler and limits the number of unknown parameters when compared with microsimulation models. However there are various methods that are regularly used to parametrize microsimulation models together with uncertainty analysis methods that look at how sensitive the outcomes are to the fitted parameter sets.

Lastly, I tried to access the Python code in https://github.com/callta/precision_screening_pca but the repository was empty (tried to access this on 2019-08-30).

Reviewer #2: Callender et al provide a timely and interesting comparative effectiveness analysis of prostate cancer screening. They show convincingly that a precision approach offers substantial benefits in terms of reduced costs per QALY and reduced over-diagnosis, at given thresholds of lives lost to prostate cancer. They argue that risk-based screening based instead of on age, on combined genetics and age-adjusted 10-year risk of 4% to 7% provides the optimal balance that is likely to meet willingness-to-pay recommendations for the UK.

There is however one important element of the study that is missing, and that is any evaluation of the performance of polygenic risk assessment. The main reference [7] given for justifying the parameter they include in their modeling is a 2002 paper on breast cancer, and the only other reference to a PRS is [28] in the last sentence of the manuscript. In addition to that study, a quick search identified two more (PMID: 30366021, PMID: 29892016) with polygenic scores incorporating from 7 to 147 SNPs, and it should now be possible based on the data in the latter study to generate a truly genome-wide score based on millions of variants. I think it is essential to consider in this paper the impact of the likely improvement in diagnostic utility of PRS as they explain more of the (genetic) variance of disease. Relative risks in the top percentiles are now approaching 3X greater than the median; and similarly at the low end of the scale it should be possible to identify individuals whose negative predictive value is such that it provides even more utility than positive predictive value. Presumably a composite score combining age, polygenic risk, and other prostate cancer risk factors (eg family history, prior PSA tests), and ancestry will do even better.

Several recent reviews have addressed the use of PRS in clinical contexts. This paper is a really nice example of combining such assessments with real-world prediction of utility. It would be even stronger with commentary on the impact of the discriminatory power of the PRS.

Reviewer #3: This is a promising paper but it needs a few major modifications before finer points can even be assessed.

The first major issue is that it is unclear that the model has necessarily converged. Only 10,000 model runs were used and not all of the line-by-line differences in Table 2 were in the expected directions. As a result, I suspect that this hasn't converged, and that results found may be slightly different were a large number of have been used and convergence achieved. (If I have misunderstood and this is the authors' expected directions, greater clarity over expectations would be helpful.)

The abstract of the paper is very poor in comparison to the elements of the analysis. Given that the paper's results would find that age-based screening is not cost-effective (ICER at 34k per QALY) vs no-screening, the comparison of age-based and precision screening is largely irrelevant. The comparisons against "no screening" in Table 2 are better but this should be incremental.

This lack of a truly incremental analysis is the biggest (and most telling) flaw in the paper. The paper does not identify a clear enough decision problem, start with a true decision problem, or present results that answer a clear decision problem. Table 2 provides really helpful information for decision making but does not do so in a very systematic (or indeed a truly incremental) fashion. If each of these lines were taken to be a separate decision option (as they are) then this would be much stronger --- here, with back of the envelope calculations (and removing the options that appear to be dominated or extended dominated), it looks to me that the most cost-effective option is to provide screening a risk threshold of 10% with a CE threshold of 20k per QALY and at 8% for a CE threshold of 30k per QALY. I can identify these figures by using Table 2 and calculating ICERs ... this is what the paper should present (it does indirectly where NMB is presented but only as an aside in the discussion, rather than as the results).

QALYs Incr QALYs Costs Incr Costs ICER

NS 46,682,945 2,975,391,145

10% 46,699,140 16,195 3,216,074,093 240,682,948 14,862

9.50% 46,699,649 509 3,226,784,314 10,710,221 21,042

9.00% 46,700,163 514 3,238,490,220 11,705,906 22,774

8.50% 46,700,677 514 3,251,296,757 12,806,537 24,915

8.00% 46,701,187 510 3,265,319,785 14,023,028 27,496

7.50% 46,701,687 500 3,280,686,485 15,366,700 30,733

7.00% 46,702,172 485 3,297,535,256 16,848,771 34,740

6.50% 46,702,631 459 3,316,014,724 18,479,468 40,260

6.00% 46,703,054 423 3,336,281,231 20,266,507 47,911

5.50% 46,703,427 373 3,358,493,767 22,212,536 59,551

5.00% 46,703,733 306 3,382,804,647 24,310,880 79,447

4.50% 46,703,950 217 3,409,343,088 26,538,441 122,297

4.00% 46,704,054 104 3,438,186,910 28,843,822 277,344

Note that on the results provided, age-based screening is dominated and doesn't appear in a sensible table of overall results (but would appear in a CEAC).

In an ideal world, the paper should be reframed around this type of view, since it's the standard approach for economic evaluations to consider all relevant options together and to do so incrementally. The paper should also consider some additional points between 9.5% and 10%, and between 7.5% and 8%, in order to optimise this targetted figure a little more closely.

Once this is done then this should really be presented on a CEAC and with a cost-effectiveness frontier; both analyses need to appear as they're standard practice - but neither do at present.

Given all this, the authors suggestion at the beginning of the discussion that a threshold of 4% to 7% is optimal is utterly indefensible on economic grounds when this purports to talk about a willingness to pay of 20k or 30k per QALY. At 4%, their own results suggest that this is cost-effective only if society is willing to pay in excess of 250k per QALY. The authors have, it appears, done a nice piece or work here but don't appear to have correctly used the methods that they state --- this is, however, very easily fixed!

Reviewer #4: I confine my remarks to statistical aspects of this paper.

These were generally fine, but I do have a couple of suggestions.

p 5 - Instead of four yearly put "four times a year" or "every four years".

I'm a little confused as to the data. If it was simulated (top of p. 5) then how could people be invited to do something (middle of p. 5)?

Table 1 - I can't comment on the values chosen, but the distributions seem reasonable.

Table 2 - maybe put QALY in millions and costs in thousands?

Figure 1 a) I am not sure how the bottom two rows mesh with the relative risk rows. It seems like there would be age based screening and no screening rates for each 10 year risk. b) I would make this into a line graph with risk on the x axis, and frequency on the y axis, with lines as needed.

Figure 2. Dual axis graphs are not good (see the work of William Cleveland). Either make this into two graphs or show e.g. the ratio of QALY to cost.

Peter Flom

Any attachments provided with reviews can be seen via the following link:

[LINK]

8 Oct 2019

Submitted filename: Response to reviewers letter.docx

Click here for additional data file.

1 Nov 2019

Dear Dr. Callender,

Thank you very much for re-submitting your manuscript "Polygenic Risk-Tailored Screening for Prostate Cancer: A Benefit-Harm and Cost-Effectiveness Analysis" (PMEDICINE-D-19-02276R1) for review by PLOS Medicine.

I have discussed the paper with my colleagues and the academic editor and it was also seen again by xxx reviewers. I am pleased to say that provided the remaining editorial and production issues are dealt with we are planning to accept the paper for publication in the journal.

The remaining issues that need to be addressed are listed at the end of this email. Any accompanying reviewer attachments can be seen via the link below. Please take these into account before resubmitting your manuscript:

[LINK]

Our publications team (plosmedicine@plos.org) will be in touch shortly about the production requirements for your paper, and the link and deadline for resubmission. DO NOT RESUBMIT BEFORE YOU'VE RECEIVED THE PRODUCTION REQUIREMENTS.

***Please note while forming your response, if your article is accepted, you may have the opportunity to make the peer review history publicly available. The record will include editor decision letters (with reviews) and your responses to reviewer comments. If eligible, we will contact you to opt in or out.***

In revising the manuscript for further consideration here, please ensure you address the specific points made by each reviewer and the editors. In your rebuttal letter you should indicate your response to the reviewers' and editors' comments and the changes you have made in the manuscript. Please submit a clean version of the paper as the main article file. A version with changes marked must also be uploaded as a marked up manuscript file.

Please also check the guidelines for revised papers at http://journals.plos.org/plosmedicine/s/revising-your-manuscript for any that apply to your paper. If you haven't already, we ask that you provide a short, non-technical Author Summary of your research to make findings accessible to a wide audience that includes both scientists and non-scientists. The Author Summary should immediately follow the Abstract in your revised manuscript. This text is subject to editorial change and should be distinct from the scientific abstract.

We expect to receive your revised manuscript within 1 week. Please email us (plosmedicine@plos.org) if you have any questions or concerns.

We ask every co-author listed on the manuscript to fill in a contributing author statement. If any of the co-authors have not filled in the statement, we will remind them to do so when the paper is revised. If all statements are not completed in a timely fashion this could hold up the re-review process. Should there be a problem getting one of your co-authors to fill in a statement we will be in contact. YOU MUST NOT ADD OR REMOVE AUTHORS UNLESS YOU HAVE ALERTED THE EDITOR HANDLING THE MANUSCRIPT TO THE CHANGE AND THEY SPECIFICALLY HAVE AGREED TO IT.

Please ensure that the paper adheres to the PLOS Data Availability Policy (see http://journals.plos.org/plosmedicine/s/data-availability), which requires that all data underlying the study's findings be provided in a repository or as Supporting Information. For data residing with a third party, authors are required to provide instructions with contact information for obtaining the data. PLOS journals do not allow statements supported by "data not shown" or "unpublished results." For such statements, authors must provide supporting data or cite public sources that include it.

If you have any questions in the meantime, please contact me or the journal staff on plosmedicine@plos.org.

We look forward to receiving the revised manuscript by Nov 08 2019 11:59PM.

Sincerely,

Adya Misra, PhD

Senior Editor

PLOS Medicine

plosmedicine.org

------------------------------------------------------------

Requests from Editors:

Title- Suggest adding a study descriptor. “A benefit-harm and cost effectiveness analysis of polygenic risk-tailored screening for Prostate Cancer: a modelling study” or similar

Please add p values to the 95%CIs throughout

Please add a sentence of limitations in the final sentence if the ‘methods and findings’ section of the abstract The R2R mentions this has been added but we are unable to see it.

I think the author summary overreaches what can be concluded from the results…”Genome wide association studies have identified more than 160 common genetic variants that, when combined together as a polygenic risk score, can be used to develop a tailored screening programme for prostate cancer.”

And "Based on this model, we show that a polygenic risk-tailored screening programme would reduce overdiagnosis, maintain the mortality benefits of age-based screening, and improve the cost-effectiveness of a screening programme for prostate cancer"

Please tone down the conclusions throughout, clarifying the results are based on a model

The square brackets are in the wrong place (after full stop)

The data statement, contributor and so on can be removed from the main text as they get pulled in automatically from EM (page 21 etc)

Abstract background and Introduction-last sentence suggests this is a clinical trial, please revise to clarify a modelling approach was used

Results first sentence should begin with “according to our model” or “our analyses show” or similar

Comments from Reviewers:

Reviewer #1: I would like to thank the authors for their replies. However, these did not change my believe that the life table modelling approach used in this paper is not the most appropriate to explore the effectiveness and cost-effectiveness of risk-based screening programs for prostate cancer.

Reviewer #2: I agree that current genome-wide polygenic risk scores are only marginally better than those based on GWAS-significant SNPs. However, my point was that these will continue to improve as sample sizes increase and the variance explained grows. I feel that there is still more that you could comment on the impact of the PRS performance on your modeling now and in the future. But it just a suggestion.

Reviewer #4: The authors have addressed my concerns and I now recommend publication.

Peter Flom

Any attachments provided with reviews can be seen via the following link:

[LINK]

19 Nov 2019

Dear Dr. Callender,

On behalf of my colleagues and the academic editor, Dr. Steven D. Shapiro, I am delighted to inform you that your manuscript entitled "Polygenic risk-tailored screening for prostate cancer: A benefit-harm and cost-effectiveness modelling study" (PMEDICINE-D-19-02276R2) has been accepted for publication in PLOS Medicine.

PRODUCTION PROCESS

Before publication you will see the copyedited word document (in around 1-2 weeks from now) and a PDF galley proof shortly after that. The copyeditor will be in touch shortly before sending you the copyedited Word document. We will make some revisions at the copyediting stage to conform to our general style, and for clarification. When you receive this version you should check and revise it very carefully, including figures, tables, references, and supporting information, because corrections at the next stage (proofs) will be strictly limited to (1) errors in author names or affiliations, (2) errors of scientific fact that would cause misunderstandings to readers, and (3) printer's (introduced) errors.

If you are likely to be away when either this document or the proof is sent, please ensure we have contact information of a second person, as we will need you to respond quickly at each point.

PRESS

A selection of our articles each week are press released by the journal. You will be contacted nearer the time if we are press releasing your article in order to approve the content and check the contact information for journalists is correct. If your institution or institutions have a press office, please notify them about your upcoming paper at this point, to enable them to help maximize its impact.

PROFILE INFORMATION

Now that your manuscript has been accepted, please log into EM and update your profile. Go to https://www.editorialmanager.com/pmedicine, log in, and click on the "Update My Information" link at the top of the page. Please update your user information to ensure an efficient production and billing process.

Thank you again for submitting the manuscript to PLOS Medicine. We look forward to publishing it.

Best wishes,

Adya Misra, PhD

Senior Editor

PLOS Medicine

plosmedicine.org