Estimating the Harms and Benefits of Prostate Cancer Screening As Used in Common Practice Versus Recommended Good Practice: A Microsimulation Screening Analysis

Abstract

BACKGROUND

Prostate-specific antigen (PSA) screening and concomitant treatment can be implemented in several ways. We investigated how the net benefit of PSA screening varies between common practice versus “good practice.”

METHODS

We used MIcrosimulation SCreening ANalysis (MISCAN) to evaluate the effect on quality-adjusted life-years (QALYs) if 4 recommendations were followed: limited screening in older men; selective biopsy in men with elevated PSA; active surveillance for low-risk tumors; and treatment preferentially delivered at high-volume centers. Outcomes were compared to a base model with annual screening starting at ages 55–69, simulated using the European Randomized Study of Screening for Prostate Cancer (ERSPC) data.

RESULTS

In terms of QALYs gained compared to no screening, per 1000 screened men followed over their lifetime, recommended good practice led to 73 life-years (LYs) and 74 QALYs gained compared to 73 LYs and 56 QALYs for the base model. In contrast, common practice led to 78 LYs gained but only 19 QALYs gained; more than a 75% relative reduction in QALYs gained from unadjusted LYs gained. The poor outcomes for common practice were influenced predominantly by use of aggressive treatment for low-risk disease, with PSA testing in older men also strongly reducing potential QALY gains.

CONCLUSIONS

Commonly-used PSA screening and treatment practices are associated with little net benefit. Following a few straightforward clinical recommendations, particularly greater use of active surveillance for low-risk disease and reducing screening in older men, would lead to an almost 4-fold increase in the net benefit of prostate cancer screening.

Graphical Abstract

Precis for use in TOC: To compare the benefit and harm of PSA-based prostate cancer screening, the authors used MIcrosimulation SCreening ANalysis and compared common practice to recommended “good practice.” They found that common screening and treatment practices are associated with little net benefit, whereas following a few straightforward clinical recommendations, particularly greater use of active surveillance for low-risk disease and reducing screening in older men, would lead to an almost 4-fold increase in the net benefit of PSA screening.

INTRODUCTION

The European Randomized Study of Screening for Prostate Cancer (ERSPC) demonstrated that regular prostate-specific antigen (PSA) screening every 2-4 years leads to a relative reduction in prostate cancer (PC)-specific mortality of 21% at 13 years of follow-up.¹ However, this benefit is offset by harms, in terms of over-diagnosis and consequent side-effects from treatment, hence the clear recommendation against PSA screening from the United States Preventive Services Task Force in 2012.² Using MIcrosimulation SCreening ANalysis (MISCAN), we have previously shown that over a lifetime, screening leads to a 28% relative reduction in PC-specific mortality and 8.4 life-years gained per averted death.³ However, this benefit is mitigated by a loss in quality-adjusted life-years (QALYs)—a 23% reduction from life-years gained—primarily because of side-effects of treatment such as urinary and erectile dysfunction.³

There have been considerable advances in our understanding of PC and PSA since the ERSPC was initiated in the early 1990s. Empirical data suggest that the ratio of benefit-to-harm could be improved by restricting screening to appropriate age ranges, restricting biopsy and treatment to men at highest risk, and shifting treatment to higher-volume centers.^4-6 These relatively uncontroversial findings have been incorporated in many guidelines. In contrast, research into common clinical practice has found frequent PSA testing among older men with limited life expectancy,^7-8 aggressive use of curative treatment for low-risk tumors,⁹ and surgical treatment largely performed by low-volume providers.¹⁰

We hypothesize that the benefit-to-harm ratio from PSA screening and subsequent treatment would be improved by following a straightforward set of simple good practice guidelines. We sought to quantify the effects of implementing these recommendations upon the outcomes of PC screening using MISCAN. We compared a “recommended good practice” model versus a model reflecting common screening and treatment practices, with a base model using ERSPC data.

METHODS

MISCAN

The MISCAN model, described in detail elsewhere,³ simulates individual life histories with and without PSA screening, and with and without development of PC. The “tumor growth model” simulates PC natural history, which progresses from no disease, to preclinical screen-detectable PC, to clinical cancer at various stages. Thereafter, the tumor is screen-detected, clinically diagnosed, or progresses to another stage. The model is calibrated using raw data from the core age group (55–69 years) of the Rotterdam and Göteborg sections of the ERSPC. This includes follow-up data through 2008 (median 11 years) and a stage-dependent cure rate estimated for the observed PC-specific mortality reduction of 29% among attendees to screening in ERSPC.³ The model was subsequently validated using data from all centra in the ERSPC, for both the screening and the control arms (thus accounting for a low contamination rate), as described earlier.³

The effectiveness of radical prostatectomy (RP) compared to watchful waiting was assigned a relative risk of PC-specific mortality of 0.65 based on Scandinavian Prostate Cancer Group-4 data¹¹; a similar effect was assumed for radiotherapy (RT). Survival was modeled using the Gleason score–dependent Albertsen data¹² as well as Surveillance, Epidemiology, and End Results (SEER) data.³

QALYs were calculated by multiplying utility estimates for various health states, where 0 is death (or worst imaginable health) and 1 is full health, by the duration and number of men in the state. Utility estimates were obtained from the Cost-Effectiveness Analysis Registry¹³ or the literature, or were based on assumptions. For active surveillance (AS), we assumed an estimated utility of 0.97 for the base case. A complete justification and references to the assumptions used in the base model were reported previously.³

Model building

MISCAN relies on certain parameter inputs, which can be changed. We simulated lifetime outcomes for those who underwent PSA screening versus controls who did not undergo screening, for a male population aged 0 to 100 years, with an age distribution based on the European Standard Population.³ We changed some of MISCAN's inputs to investigate the effects of the different models upon QALYs.

The base model uses annual PSA screening, as often practiced in the U.S. It follows a population of men aged 0–100 over their lifetimes and screens them, with 80% participation rate, between ages 55 and 69; matching the ERSPC core age group where a significant effect on PC-specific mortality was demonstrated in favor of screening.^{1, 14-15} The base model also uses: positive predictive value (PPV) of biopsy of 22.7% as in the ERSPC; primary treatment distribution (RP, RT, or AS with deferred treatment) based on age, T-stage and Gleason score as in the ERSPC; and complication rates after curative treatment as seen in U.S. population-based series.³

We created 2 additional models: “recommended good practice,” which amended the base model by incorporating 4 simple recommendations on screening and treatment found in many guidelines, and “common practice,” in which we incorporated data from empirical studies of contemporary U.S. practice patterns. Table 1 lists the assumptions changed from the base model.

Table 1.

Parameters Assigned to 3 MISCAN-based Models of PSA Screening and Treatment

	Base Model (Heijnsdijk et al3)		Recommended Good Practice Model		Common Practice Model
	Parameter	Source	Parameter	Source	Parameter	Source
1. Ages of men screened	55–69 years	Schröder et al14	55–69 years	Schröder et al14	55–90 years	Drazer et al7
2. PPV of biopsy	22.7%	Schröder et al14	40%	Vickers et al18	40%	Vickers et al18
3. Use of AS	AS rates depending on age, T stage and Gleason score as in ERSPC for both low-risk and non–low-risk PC; about 30% for low-risk	ERSPC data	AS rates for non–low-risk PC same as base model	ERSPC data	AS rates for non–low-risk PC same as base model	ERSPC data
			90% AS for men with low-risk tumors	Assumption	9.2% AS for men with low-risk tumors	Cooperberg et al9
4. Rate of side-effects	Population-based rates: 6% urinary leakage problems, 30% overall sexuality problems	Sanda et al27	Rates as seen in high-volume centers: 5% urinary leakage problems, 19% overall sexuality problems	Vickers et al26	Population-based rates: 6% urinary leakage leaking problems, 30% overall sexuality problems	Sanda et al27

Abbreviations: AS, active surveillance; ERSPC, European Randomized Study of Screening for Prostate Cancer; MISCAN, Microsimulation Screening Analysis; PC, prostate cancer; PPV, positive predictive value; PSA, prostate-specific antigen.

Age for screening

The ERSPC found no evidence of benefit for men who start PSA screening at age ≥70, with the lower bound of the 95% CI excluding the central estimate for risk reduction for men aged <70.¹⁵ Similarly, the American Urological Association does not recommend routine PSA screening in men aged ≥70 years.¹⁶ For the common practice model, where some men were assumed to continue screening after age 70, we used age-dependent screening rates from an empirical study of health behaviors in the U.S.: ages 70–74: 47%; 75–79: 44%; 80–84: 43%; 85+: 26%.⁷ As that study included all ages over 84 into a single category, we assumed the 26% rate of screening for this category applied to ages 85–90, with no screening above age 90.

Biopsy criteria

The ERSPC study protocol stated that men with a positive screening test (PSA ≥ 3.0 ng/mL) should be recommended for biopsy. The proportion of test-positive men who had evidence of cancer on biopsy was only 22.7%.¹⁴ In common urologic clinical practice, patients with elevated PSA are evaluated for benign disease and subject to repeat PSA testing before the decision to biopsy.¹⁷ We investigated how screening outcomes would change if men with elevated PSA were biopsied more selectively, based on clinical work-up. Instead of a PPV of 22.7% for biopsy after a positive PSA test, we applied a PPV of 40%, in line with U.S. clinical cohorts,¹⁸ for both the “recommended good practice” model and the “common practice” model.

Active surveillance (AS)

Recent data clearly indicate that not all men with PC need immediate treatment, and low-risk tumors can be safely managed by the approach known as active surveillance, with repeat biopsy and routine monitoring of the disease.¹⁹ Several guideline groups, such as the National Comprehensive Cancer Network, now recommend AS for low-risk PC.¹⁷ We investigated how QALYs were affected if men with low-risk disease (clinical stage T1, Gleason score 6) were enrolled in AS. In the base model, AS usage depended on age, and averaged 30% across all men with low-risk tumors. For cumulative proportions of men leaving AS each year, we used data from Klotz's series: year 1: 8%, year 2: 16%, year 3: 20%, year 4: 24%, year 5: 28%, year 6: 29%, and year 7: 30%.¹⁹ For the recommended good practice model, we applied a 90% rather than a 100% AS rate to men with low-risk disease, given that there may be clinical reasons to treat some low-risk men. For the common practice model, we applied an AS rate of 9.2% for men with low-risk disease, obtaining this estimate from the Cancer of the Prostate Strategic Urologic Research Endeavor (CaPSURE) registry 1990–2008, and also reflecting what has been practice for many years.⁹

High-volume centers

There is a considerable literature on the volume-outcome relationship, suggesting decreased complications and side-effects and improved outcomes for patients treated by high-volume providers.^20-23 Shifting treatment trends so that more patients are treated by high-volume surgeons could, therefore, possibly improve cancer control and decrease complications. There have been widespread calls for “regionalization”²⁴; that is, increasing the proportion of patients treated at high-volume centers.²⁵ We investigated how QALYs were affected if impotence and incontinence rates after RP were in line with rates seen at high-volume centers.²⁶ The MISCAN model used a representative, multiregional, U.S. cohort as the source for estimates of overall sexual problems and urinary leakage problems at 24 months post-RP, taking baseline functioning into account.²⁷ The base model assumed 30% overall sexual bother, 6% urinary bother, and 0% bowel bother post-RP.³ Although different rates were used for RT (20% sexual, 5% urinary, and 8% bowel bother), when multiplied with utilities, total utility ended up being similar for the two treatment modalities. These estimates may seem lower than many reported in the literature because they are marginal—that is, they take into account that some men would have dysfunction without surgery/RT. Also, these estimates reflect bother not function, and not all men experiencing dysfunction report lowered utility.

Estimates for functional outcomes after RP for surgeons at a high-volume center were derived from empirical data using case-mix-adjusted outcomes,²⁶, giving rates of sexual and leakage problems of 19% and 5%, respectively.

Sensitivity analysis

A sensitivity analysis was performed, comparing QALYs gained between the 3 models. In an attempt to reflect the effect of the different strategies on a population level, rather than an individual level, we varied the utility estimates (more vs less extreme) by about half those previously published.³ We compared 5 different scenarios per model using different combinations of utilities for screening procedures versus treatment and terminal illness, ie, reflecting varying population-level trade-offs for tolerability of screening procedures versus down-stream consequences.

Since the use of AS for men with low-risk disease has increased over the past years, another sensitivity analysis was performed, with a 34% AS rate, as reported in a recent update from the CaPSURE registry for the time period 2008–2013.²⁸

RESULTS

Effect of modeling on QALYs

Table 2 shows quality-adjusted effects of the 3 screening models, compared to no screening, given various health states. The recommended good practice model displayed favorable effects at the biopsy stage. Compared to the base model, the good practice model had more QALYs lost in the AS health state due to its increased AS rate (3.2 vs. 9.7 QALYs per 1000 men); however, this was balanced by fewer QALYs lost from side-effects after RT and RP. The opposite was true for the common practice model, with few QALYs lost for AS, but substantial losses in QALYs due to the higher rate of treatment with RT and RP.

Table 2.

QALYs Gained By the 3 Screening and Treatment Models At Various Health States

	Utility Estimates			Quality of Life Adjustmenta No. of life-years
Health Stateb	Base case	More extreme	Less extreme	Base modelc	Recommended good practice	Common practice
Screening	0.99	0.99	1	–1.6	–1.6	–1.4
Biopsy	0.90	0.885	0.94	–1.7	–0.5	–1.6
Cancer diagnosis	0.80	0.775	0.85	–0.7	–0.7	–2.1
Radiotherapy
At 2 months after procedure	0.73	0.72	0.82	–0.2	–0.0	–2.7
At >2–12 months	0.78	0.695	0.83	–0.9	–0.2	–11.0
Radical prostatectomy
At 2 months after procedure	0.67	0.615	0.785	–2.0	–0.6	–3.9
At >2-12 months	0.77	0.735	0.84	–6.9	–2.1	–13.7
Active surveillance	0.97	0.91	0.985	–3.2	–9.7	–1.4
Post-recovery periodd (1–10 years after treatment)
Overdiagnosise	0.95d,f	0.94	0.975	–10.8	–5.6	–24.8
No overdiagnosis	0.95d,f	0.94	0.975	–5.5	5.5	–19.8
Palliative therapy	0.60	0.73	0.42	14.1	14.2	18.4
Terminal illness	0.40	0.40	0.32	2.6	2.6	3.2
Total number of life-years gained	Full model	Full model	Full model	73	73	78
Total number of QALYs gained	Full model	Full model	Full model	56	74	19

Abbreviation: QALYs, quality-adjusted life-years.

^aNumbers are over the lifetimes of 1000 men aged 0–100. Minus sign indicates number of years to be subtracted from the life-years gained in order to get the QALYs gained.

^bFor a complete list of sources of the utility values and the duration of temporary health states, see Heijnsdijk et al.³ The more and less extreme utilities used for the sensitivity analysis are assumed to be half those previously reported, to reflect the effects of a policy on a population level, rather than the effects on the individual level.

^cThe difference in life-years for each health state has been multiplied by the utility loss to calculate the adjustment for quality of life.

^dThe following utilities translate into an aggregated utility of 0.95: urinary leakage, 0.83; bowel problems, 0.71; and sexuality problems, 0.89.

^eOverdiagnosis implies diagnosis of prostate cancer, which in a situation without screening would not have been clinically diagnosed within the lifespan of a typical man.

^f0.96 for the recommended good practice model.

The predicted effects of the screening approaches are shown in Table 3. Compared to the base model, recommended good practice led to an improvement in QALYs gained, from 56 to 74, largely related to increased use of AS. This approach also substantially reduced the number of biopsies performed, from 605 to 407 per 1000 men. In contrast, common practice with screening up to age 90 years and with a 9.2% AS rate, led to 78 life-years gained but only 19 QALYs gained. This is more than a 75% relative reduction in QALYs gained from unadjusted life-years gained. Of the QALYs lost by following common practice compared to recommended good practice, about 24 were related to overtreatment of low-risk disease, 34 due to screening older men, and 3 due to treatment at low-volume centers (Table 4). Note that these figures do not add up to the 55 QALY difference between common practice and recommended good practice because of interaction effects, such as the impact of overtreatment in older men.

Table 3.

Predicted Effects of the 3 Screening and Treatment Models, Compared to No Screening^a

	No Screening	Base Model	Recommended Good Practice	Common Practice
Biopsies performed	313	605	407	595b
Negative biopsies	201	448	250	359
Cancers diagnosed	112	157	157	236
Relative reduction in prostate cancer–specific mortality	-	37%	37%	41%
Life-years gained	-	73	73	78
QALYs gained	-	56	74	19
Relative reduction in life-years gained after adjustment for quality of life	-	23%	–1%	76%

Abbreviations: QALYs, quality-adjusted life-years.

^aNumbers are over the lifetime of 1000 men aged 0–100.

^bSome men undergo more than one biopsy.

Table 4.

QALYs Gained/Lost By Different Aspects of Practice

	Recommended Good Practice		Common Practice
Parameter	Aspect of practice	QALYsa	Aspect of practice	QALYsa
1. Age for screening	Limit screening in older men	Same as base model	Widespread screening of older men	34.2 (–21.8)
2. Biopsy criteria	Restrictive biopsy criteria	57.2 (+1.2)	Restrictive biopsy criteria	57.2 (+1.2)
3. AS	AS for most low-risk cancers	73.2 (+17.2)	Little use of AS	49.1 (–6.9)
4. Regionalization	Most treatment at high-volume centers	59.3 (+3.3)	Much treatment at low-volume centers	Same as base model
Total	All four of the above factors	74.0 (+18.0)	All four of the above factors	19.0 (–37.0)

Abbreviations: AS, active surveillance; QALYs, quality-adjusted life-years.

^aNumber in parentheses indicates incremental/decremental effect on QALYs as compared to base model's 56.0 QALYs gained.

In a sensitivity analysis varying the more and less extreme utility estimates in an attempt to reflect the effect on QALYs of the different strategies at a population level, did not show recommended good practice leading to worse outcomes than the base or common practice models (Table 5).

Table 5.

Sensitivity Analysis: Effects of Various Modeling Assumptions on Total QALYs Gained

	Base model		Recommended Good Practice		Common Practice
	Scenario 5e		Scenario 5e		Scenario 5e
	Terminal illness 0.4	Terminal illness 0.32	Terminal illness 0.4	Terminal illness 0.32	Terminal illness 0.4	Terminal illness 0.32
Scenario 1a	39.8	40.1	49.7	50.0	–7.8	–7.4
Scenario 2b	77.6	77.9	88.5	88.8	61.1	61.5
Scenario 3c	60.5	60.9	55.7	56.1	42.9	43.4
Scenario 4d	55.2	55.6	80.7	81.1	5.2	5.6

In the base model, screening men between the ages of 55–69 years yields 56 QALYs gained over a lifetime. This is based on assumptions of the utilities for the modeled health states; screening attendance, biopsy, diagnosis, treatment, post-recovery period, palliative treatment, and terminal illness. These utilities can be varied from less extreme to more extreme values (Table 2).

^aTreatment and procedures less tolerable (ie, low utilities for everything but terminal illness)

^bTreatment and procedures more tolerable (ie, high utilities for everything but terminal illness)

^cCancer worry less tolerable, treatment side effects more tolerable (ie, low utilities for active surveillance, biopsy, diagnosis, and screening; high utilities for treatment and recovery)

^dVice versa of Scenario 3 (ie, high utilities for active surveillance, biopsy, diagnosis and screening; low utilities for treatment and recovery)

^eScenarios 1–4, with different utility value for terminal illness

Increasing the use of AS to 34%, to reflect more contemporaneous rates, yielded an overall 30 QALYs gained for current practice compared to 74 QALYs for recommended good practice.

DISCUSSION

This study examined the effect upon QALYs of widespread implementation of 4 widely-accepted screening and treatment recommendations, compared to common clinical practice. Microsimulation modeling showed that care following the good practice recommendations – restricting screening in elderly men, selective biopsy, AS for low risk tumors and preferential referral to high-volume centers – led to a large improvement in QALYs gained per 1000 men, up to 74 from 56 for the base model. In contrast, common screening and treatment practice was estimated to lead to only 19 QALYs gained, translating into a more than 75% relative loss in potential QALYs gained.

Naturally, any modeling study is only as good as the model used. The MISCAN model has been shown to adequately predict PC incidence and PC-specific mortality in the Netherlands.³ When applied to the U.S. population and compared to other models, differences are relatively minor (eg, lead time of 7.9 vs. 6.9 years). In comparison with 2 other models, MISCAN may be conservative, that is, may overestimate some screening harms.²⁹ We have also previously argued that the European data may underestimate the benefits of screening due to sub-optimal treatment efficacy in the ERSPC, where both radiation doses and surgeon volumes were much lower than would be optimal.³⁰ Note that we did not include higher cure rates associated with referral to high-volume centers in our “recommended good practice” model, perhaps underestimating the benefits for more regionalized care. Furthermore, the differences in urinary and sexual problems between standard care and care at high-volume centers were relatively modest in our model: 1% and 11%, respectively, in absolute terms. Again, this may lead to some underestimation of the effects of regionalized treatment.

There has been considerable recent interest in the use of risk-stratified methods of evaluating men with elevated PSA-levels before biopsy, such as reflex blood tests or multiparametric magnetic resonance imaging. The PPV associated with these tests is likely even higher than the 40% figure used in our models. The QALYs gained with recommended good practice may, therefore, be a slight underestimation. However, we do not expect this to make a large difference to our findings as the near 20-point increase in PPV used in the main analysis led to only a minor improvement in QALYs gained (+1.2 QALYs).

There is some evidence that current practice is changing. Across community-based urology practices in Michigan, half of men with low-risk PC now receive initial AS.³¹ We expect there will be more pronounced changes throughout the U.S. in the near future. Changing the use of AS to 34%, as reported in the most recent update from the CaPSURE registry,²⁸ did increase overall QALYs gained from 19 to 30. These are promising signs that changes in urologic practice will make a large difference to quality-of-life outcomes of screening.

There is also evidence that screening practices in older men have been changing for the better. For instance, incidence data from SEER have indicated that the age-, race- and ethnicity-adjusted rate of early-stage PC among men ≥75 fell from 443 to 330 per 100,000 (−25.4%; p<0.001) between 2007 and 2009.³² While encouraging, these changes go only a small way toward the major shift in screening and treatment practices needed for U.S. practice to be compliant with good practice recommendations.

Critics of PSA screening claim that it has little benefit and causes significant harm. This may be the case as PSA screening is currently implemented in the US, but does not take into account the potential benefit of screening that follows good practice recommendations. Addressing the problems of screening in older men and aggressive treatment of low-risk disease might be expected to strongly increase the benefit of PSA screening.

A limitation of the present study is that results based on the MISCAN model are relevant for Caucasian men and may not apply to men of other ethnicities.

CONCLUSIONS

Common practices for PSA screening and subsequent PC treatment are associated with considerable harm and moderate benefit. Changing practices to conform to established recommendations would lead to an estimated 4-fold increase in the net benefit of screening.