Prostate‐specific antigen‐based screening for prostate cancer: A systematic review to inform an American Cancer Society guideline

Abstract

The purpose of this systematic review was to identify and synthesize the evidence on the benefits and harms of prostate‐specific antigen (PSA)‐based screening for prostate cancer to inform the American Cancer Society’s updated guideline. PubMed and the Cochrane Central Register of Controlled Trials were searched for published and gray literature through December 4, 2023, and updated through October 29, 2025, which allowed for approximately 3 months of overlap with searches conducted for the previous guideline. Eligible studies included randomized and nonrandomized comparisons of PSA‐based screening strategies to no screening or other screening strategies. Low– and medium–risk of bias systematic reviews were included, and their findings were supplemented with studies published after the systematic review. One investigator abstracted study data, and a second confirmed. Two investigators independently assessed the risk of bias and strength of evidence, and discrepancies were resolved by consensus. Twenty‐nine publications were included: two publications from a 2018 systematic review of PLCO, CAP, and ERSPC; 23 subsequent publications from six trials (PLCO, CAP, ERSPC, Göteborg‐1, Rotterdam Pilot 1, and ProScreen); and four observational studies. PSA‐based screening was associated with increased detection of low‐risk prostate cancer, reduced incidence of advanced or metastatic disease at diagnosis, and reduced cumulative incidence of metastatic disease. Screening was associated with a reduction in prostate cancer–specific mortality in some trials, although findings were inconsistent and likely depended on screening protocols, PSA thresholds, participant characteristics, and other factors. Harms of screening included overdiagnosis, false‐positive test results, biopsy‐related complications, and downstream treatment harms.

Short abstract

This systematic review found that prostate‐specific antigen–based screening was associated with increased detection of low‐risk prostate cancer, reduced incidence of advanced or metastatic disease at diagnosis, reduced cumulative incidence of metastatic disease, and reduced prostate cancer–specific mortality in some trials. Harms of screening included overdiagnosis, false‐positive test results, biopsy‐related complications, and downstream treatment harms.

INTRODUCTION

Prostate cancer is the most commonly diagnosed cancer, and the second‐leading cause of cancer death among men in the United States, accounting for an estimated 333,830 new cases and 36,320 deaths in 2026, representing approximately 11% of cancer deaths in men. ¹ African American and Black men are disproportionately affected, with approximately two to three times the likelihood of being diagnosed and two to four times the likelihood of dying from prostate cancer compared with any other racial or ethnic group, as well as the highest probability of developing prostate cancer at all ages. ¹ , ²

Cancer incidence rates reflect not only the relative frequency of a cancer’s diagnosis but also the contemporary clinical landscape for detection. Prostate‐specific antigen (PSA)‐based screening has been widely used to detect prostate cancer since the late 1980s, after it was introduced as a test to monitor patients with prostate cancer for disease progression in 1986, and was approved by the US Food and Drug Administration for the detection of prostate cancer in 1994. ³ The early 1990s saw a rapid rise in the incidence of prostate cancer, which reflected the increased use of PSA‐based screening. ⁴ , ⁵ The incidence of early‐stage prostate cancer declined after the 2008 US Preventive Services Task Force (USPSTF) recommended against screening in men aged older than 75 years, and the 2012 USPSTF recommended against screening for all men. ⁶ , ⁷ Prostate cancer incidence decreased by 6.4% annually from 2007 to 2014 but has increased approximately 3% annually since 2014, with steeper increases in regional‐stage (4.6% per year in 2013–2021) and distant‐stage disease (4.8% per year in 2016–2021). ² In 2023, the prevalence of prostate cancer screening in men aged 65–69 years was higher in White men (41%) than in Black (33%), Asian (24%), or Hispanic men (27.5%). ⁸ The increasing incidence despite low rates of screening, along with diagnoses at later stages when treatment options are more limited and the disproportionate impact on Black men, highlights the need to identify better screening strategies. PSA testing is the first step in a screening continuum that includes secondary testing after an elevated PSA, followed by treatment decisions for cancer detected during screening. Studies of screening strategies incorporating risk stratification, starting screening earlier, or using age‐relevant PSA thresholds for additional testing are ongoing. ⁹ , ¹⁰ Although PSA‐based screening may lead to earlier detection, it also increases the likelihood of “overdiagnosing” tumors that may never become clinically significant. ¹¹ , ¹² Because many screen‐detected tumors may never cause symptoms in a person’s lifetime, conservative management with active surveillance is recommended for treatment of low‐risk, localized prostate cancer. ¹³

The 2010 American Cancer Society (ACS) guideline recommends that asymptomatic men with at least a 10‐year life expectancy should have an opportunity to make an informed decision about whether to undergo PSA‐based screening for prostate cancer. ¹⁴ An informed decision‐making process should address the uncertainties, risks, and potential benefits associated with screening at age 50 years if at average risk. The process should begin earlier for those at higher risk: age 45 years for Black men or those with a first‐degree relative diagnosed before age 65 years, or as early as age 40 years for those with multiple family members diagnosed with prostate cancer before age 65 years. ¹⁴ The American Urological Association, European Association of Urology, National Comprehensive Cancer Network, and others recommend PSA‐based screening, with some differences in age to start screening, screening interval, and age to stop screening (see Table S1A). ¹⁴ , ¹⁵ , ¹⁶ , ¹⁷ , ¹⁸ , ¹⁹ , ²⁰ , ²¹ , ²² Although most organizations recommend magnetic resonance imaging (MRI), risk calculators, or biomarkers to guide biopsy decisions after an abnormal PSA, there are no current recommendations that identify preferred biomarker assays. ²³ The USPSTF, unlike many other organizations, recommends prostate screening as an individual decision for men aged 55–69 years, and recommends against screening for men aged 70 years and older. The purpose of this systematic review was to identify and synthesize the evidence on the benefits and harms of PSA‐based screening for prostate cancer to inform the ACS’s 2026 updated guideline. ²⁴

MATERIALS AND METHODS

Topic development

In preparation for updating the prostate cancer screening guideline, the ACS Guideline Development Group (GDG) requested that we, the ACS Cancer‐related Evidence Synthesis Team (CrEST), perform this systematic review. The scope of the review, including research questions and analytic framework (see Appendix S2), was developed in consultation with the GDG and the ACS Guideline Development Team (GDT). This review is limited to the questions requested by the GDG, and represents one piece of a larger program of research used to inform the guideline update. Key questions (KQs) were reviewed by subject matter experts selected by the GDG as well as by content and methodological experts selected by the CrEST (see Appendix S3). KQs were revised on the basis of the input received, and final KQs were approved by the GDG before registering the protocol. The current publication includes questions about the benefits and harms of initial PSA‐based screening. The review was conducted as part of a larger systematic review protocol that included topics related to health equity, the screening continuum (i.e., biomarkers, imaging, and biopsy), and active surveillance for low‐risk, localized prostate cancer. Several topics were subsequently moved to separate protocols, and the protocol was revised.

The ACS CrEST was solely responsible for conducting all aspects of the systematic review; ACS GDG members and ACS GDT staff were not involved in decisions regarding study selection, data analysis, or strength of evidence (SOE). The protocol, which follows the Preferred Reporting Items for Systematic Reviews and Meta‐Analyses guideline, ²⁵ was registered with the PROSPERO systematic review registry before study initiation (CRD42023467439).

Data sources and searches

Search strategies were developed by a librarian with expertise in conducting searches for systematic reviews, and were peer reviewed by a second librarian with the Peer Review of Electronic Search Strategies assessment form. ²⁶ We searched PubMed and the Cochrane Central Register of Controlled Trials for published literature from April 2009 to December 4, 2023, to allow for approximately 3 months of overlap with searches conducted for the ACS Guideline for the Early Detection of Prostate Cancer: Update 2010 (see Table S4A). ¹⁴ We supplemented searches by reviewing conference programs, ClinicalTrials.gov, articles suggested by GDG members or content area experts, and reference lists of included and excluded articles. Searches were updated on October 29, 2025. Searches were conducted for an overarching early detection of prostate cancer evidence synthesis project that included topics beyond the scope of this publication (e.g., screening continuum, health equity, screening utilization, and treatment of low‐risk, localized prostate cancer). Final search strategies are listed in Appendix S4.

Study selection

We developed inclusion criteria based on the KQs and populations, interventions, comparators, outcomes, timing, and study designs and settings (PICOTSS) of interest (see Table S5A). We included randomized and nonrandomized trials or comparative observational studies (e.g., cohort and cross‐sectional) that compare PSA‐based prostate cancer screening strategies to no screening or other screening strategies. We included low– or medium–risk of bias (ROB) systematic reviews with similar PICOTSS in the current systematic review, and supplemented findings with studies published after the systematic review. Nonincluded systematic reviews were used to identify individual studies. Studies reporting results of simulated modeling or without a comparison group were excluded.

Search results were imported into EndNote for citation management and initial removal of duplicates and nonrelevant citations (e.g., animal studies). ²⁷ DistillerSR was used to manage the evaluation of abstracts and full texts for inclusion. ²⁸ Inclusion and exclusion criteria for selecting individual studies were based on the KQs and PICOTSS. Two research team members independently reviewed abstracts; full‐text articles were reviewed if either team member considered the abstract to be potentially includable. Full‐text articles were reviewed independently by two team members, and discrepancies were resolved via consensus discussions. Studies identified by reviewing reference lists or those suggested by experts or GDG members were evaluated with the established eligibility criteria. Detailed results of the search and study selection are available in Appendix S6. A list of studies excluded after full‐text review is available in Appendix S7.

Data abstraction and ROB assessment

One team member abstracted information from eligible studies via DistillerSR, and a second confirmed accuracy and completeness. Data for abstraction included study characteristics (e.g., study design, dates, setting, eligibility criteria, sample size, and funder), participant characteristics (e.g., age, race and ethnicity, family history, and risk factors), screening characteristics (e.g., test, interval, definition of a positive result, and contamination/crossover exposure), and study results relevant to the KQs and PICOTSS (e.g., mortality, stage at diagnosis, and quality of life [QOL]).

Two team members independently used predefined criteria to assess the ROB of each included study, and discrepancies were resolved by discussion and consensus. The ROB for randomized controlled trials (RCTs) and observational studies was assessed via methods developed by Cochrane and the Agency for Healthcare Research and Quality. ²⁹ , ³⁰ , ³¹ Included systematic reviews were assessed with elements of AMSTAR 2. ³² Detailed ROB assessments for included studies are available in Tables S8A–C. Each included study or systematic review was given an overall summary assessment of low, medium, or high ROB.

Data synthesis and analysis

We qualitatively synthesized the evidence for each KQ and compiled evidence tables of study characteristics and findings. We used a hierarchy‐of‐evidence approach, where the best evidence is the focus of data synthesis. ³³ Qualitative data were summarized with ranges and descriptive synthesis. We constructed 2 × 2 tables and calculated relative risks (RRs) for cancer incidence by Gleason score or disease stage when data were available. Quantitative synthesis was not feasible given the heterogeneity of the populations, screening techniques, and reported outcomes. We stratified analyses by sociodemographic characteristics or clinical factors whenever possible.

Rating the SOE and assessing applicability

We classified the overall SOE via a method that considers the ROB of individual studies, consistency of findings across studies, directness of the comparisons, precision, and reporting bias. ³⁴ The SOE was initially rated by one team member and verified via team discussion. Priority outcomes were rated as high, moderate, low, or insufficient (see Appendix S9 for SOE definitions and Table S10A for detailed SOE assessments).

We evaluated the applicability of the findings to US clinical practice, with special attention paid to the impact of study eligibility criteria, sociodemographic characteristics of the included sample in comparison to the target population, current clinical practice, and feasibility of screening tools or strategies in a clinical practice or setting. Concerns regarding applicability indicate potential limitations in the ability to generalize study results to target populations and settings.

RESULTS

Searches for the larger early detection of prostate cancer project identified 18,835 abstracts, of which 912 were selected for full‐text review. We included one medium‐ROB systematic review (two publications), ³⁵ , ³⁶ six RCTs (23 publications), ⁹ , ³⁷ , ³⁸ , ³⁹ , ⁴⁰ , ⁴¹ , ⁴² , ⁴³ , ⁴⁴ , ⁴⁵ , ⁴⁶ , ⁴⁷ , ⁴⁸ , ⁴⁹ , ⁵⁰ , ⁵¹ , ⁵² , ⁵³ , ⁵⁴ , ⁵⁵ , ⁵⁶ , ⁵⁷ , ⁵⁸ and four observational studies. ⁵⁹ , ⁶⁰ , ⁶¹ , ⁶² The systematic review, published in 2018, included three trials of PSA‐based screening (N = 647,906): the Prostate, Lung, Colorectal, and Ovarian Cancer Screening Trial (PLCO), Cluster Randomized Trial of PSA Testing for Prostate Cancer (CAP), and European Randomized Study of Screening for Prostate Cancer (ERSPC). ³⁵ , ³⁶ The remaining 27 publications included reports from the PLCO, CAP, ERSPC combined and individual sites, Göteborg‐1 trial, Rotterdam Pilot 1 trial, the ongoing ProScreen trial, and four observational studies not included in the systematic review.

Description of included studies

The included RCTs differed substantially in design, screening intensity, PSA threshold, adherence, and contamination (Table 1). Between 1993 and 2001, the PLCO randomized 76,683 US men aged 55–74 years to annual PSA testing plus digital rectal examination (DRE) for 5 years or usual care; the PSA threshold for referral to a urologist was 4.0 ng/mL. ⁵² , ⁵³ Participants were predominantly non‐Hispanic White (88%), and 7.5% reported a family history of prostate cancer. Follow‐up testing and subsequent treatment were determined by the patient’s provider. The PLCO was rated as high ROB because PSA testing was common in the usual care arm, with 74% of usual care participants receiving at least one PSA test during the screening phase, and PSA screening rates were nearly equal through the follow‐up period.

TABLE 1.

Characteristics of included PSA‐based screening trials.

Study Enrollment years ROB	Follow‐up, years	Screening versus control, No.	Age, years	PSA threshold for additional testing, ng/mL	Screening interval, years	Adherence a	Contamination
PLCO 1993–2001 High	Incidence, 15 Mortality, 16	38,340 versus 38,343	55–74	4.0	1	NR	74%
CAP 2001–2009 Medium	15.4	195,912 c versus 219,445	50–69	3.0	Single invitation	36%	10%–15%
ERSPC combined‐site analysis 1993–2003 High	23	72,890 versus 89,351 d	55–69 d	2.5–4.0 e	2–7	83% b	NR
ERSPC‐Finland 1996–1999 Medium	16	31,875 versus 48,301	55–67	4.0; 3.0 if free PSA < 0.16	4	91%	47.7% at 8 years 62.7% at 12 years
ERSPC‐France f 2001–2005 High	9.5	38,474 versus 38,161	55–69	3.0	4–6	31%	32%
ERSPC‐Netherlands 1993–2000 Medium	21	21,169 versus 21,136	55–74	3.0	4	91%	19.4%
ERSPC‐Spain 1996–1999 High	21	2415 versus 1861	45–70	3.0	4	75%	43.2% at 10 years g
Göteborg‐1 1995 Medium	22	9945 versus 9949 h	50–64	2.5–3.4 i	2	77%	72% j
Rotterdam Pilot 1 1991–1992 High	19	553 versus 581	55–74	3.0	4	NR	4.5%
ProScreen 2018–2020 Medium	3.2	15,201 versus 45,544	50–63	3.0	2–6 k	51% first round	NR

Abbreviations: CAP, Cluster Randomized Trial of PSA Testing for Prostate Cancer; ERSPC, European Randomized Study of Screening for Prostate Cancer; ng/mL, nanograms per milliliter; NR, not reported; PLCO, Prostate, Lung, Colorectal, and Ovarian Cancer Screening Trial; PSA, prostate‐specific antigen; ROB, risk of bias.

^a Adherence in the screening arm; contamination in the control arm.

^b At least one PSA test during the screening phase.

^c Cluster randomized 271 practices to screening invitation and 302 practices to control/no screening invitation.

^d Core age group (55–69 years).

^e PSA thresholds varied by site.

^f The French site was excluded from ERSPC combined‐site analysis because of incomplete follow‐up.

^g Contamination is reported in reference 76.

^h Portion of the Goteborg‐1 subset constitutes ERSPC‐Sweden: n = 5901 versus 5951.

ⁱ The PSA threshold was 3.4 (1995–1998), 2.9 (1999–2004), and 2.5 (after 2004).

^j Contamination is reported in reference 77.

^k The interval was determined by baseline PSA: 2 years if ≥3.0 ng/mL, 4 years if 1.5–2.99 ng/mL, and 6 years if <1.5 ng/mL.

The CAP randomized 573 primary care practices in the United Kingdom to a single invitation to PSA testing or usual care (N = 415,357; median age, 59 years; 2001–2009); biopsy was indicated for a PSA of 3.0 ng/mL or higher. ⁴⁵ , ⁴⁶ Most participants were White (98%), and family history of prostate cancer was not reported. Only 36% of men in the screening arm had PSA testing after the one‐time invitation, and cumulative contamination in the control arm was estimated at 10%–15% over 10 years. Participants in the screening arm with an abnormal PSA underwent ultrasound‐guided transrectal biopsy (TRUS), and those who were diagnosed with prostate cancer were invited to participate in the randomized Prostate Testing for Cancer and Treatment trial (ProtecT). The CAP was rated as medium ROB because of participant recruitment after randomization, low screening participation rates, and questions around methods for estimating contamination.

The ERSPC randomized 182,160 men across eight European countries (Belgium, Finland, France, Italy, the Netherlands, Spain, Sweden, and Switzerland) to PSA‐based screening or usual care from 1993 to 2005. ³⁹ , ⁴⁰ , ⁴¹ , ⁴² , ⁴³ , ⁴⁴ , ⁴⁸ , ⁵⁰ , ⁵¹ , ⁵⁴ , ⁵⁵ , ⁵⁶ , ⁵⁷ Enrollment began in the Netherlands and Belgium in 1993, Finland, Sweden, Italy, Spain, and Switzerland began enrolling between 1994 and 1996, and France began enrolling in 2000. Finland, Sweden, Italy, and France used effectiveness designs where participants were assigned to an invitation to screening or a control arm before consenting to participation, and the Netherlands, Spain, Belgium, and Switzerland used an efficacy design with randomization occurring after the participant consented to participate. All sites included men aged 55–69 years at baseline (core age group); however, Sweden enrolled men as young as 50 years old, Spain enrolled men as young as 45 years old, and others (the Netherlands, Belgium, Italy, and Spain) enrolled men up to age 74 years (full cohort). Screening intervals ranged from 2 to 7 years, PSA thresholds varied from 2.5 to 4.0 ng/mL at most sites, adherence (attending at least one screening during the screening phase) ranged from 74% to more than 90%, and only 5.8% of men in the screening arm never participated in screening. Men with abnormal PSA rates (thresholds differed by site) were recommended DRE, TRUS, and systematic prostate biopsy. Race and family history were not reported. We included published reports from the ERSPC of combined data from all sites (although most reports excluded French data because of low participation and short follow‐up times) ⁴³ , ⁵⁵ or reported findings from individual sites. ³⁹ , ⁴¹ , ⁴³ , ⁴⁴ , ⁴⁸ , ⁵⁰ , ⁵¹ , ⁵⁴ , ⁵⁶ , ⁵⁷ We rated the combined analyses of all ERSPC sites as high ROB because of concerns about heterogeneity across sites that included different screening protocols and populations, and concerns about randomization, allocation concealment, and contamination in the no‐screening arm that varied across sites. We assessed the ROB of individual ERSPC site analyses published independent of the combined analyses. Sites were rated as medium (Finland and the Netherlands) or high ROB (France and Spain) (see Appendix S8).

The Göteborg‐1 Randomized Population‐Based Prostate Cancer Screening Trial was established in 1995, and randomized 20,000 men aged 50–64 years to either invitation to PSA‐based screening every 2 years or a control group who were not invited for screening. ⁴⁰ , ⁴² Beginning in 1996, a subset of this trial population constituted the ERSPC Swedish site (n = 11,852). PSA levels at or above the threshold (3.4 ng/mL from 1995 to 1998, 2.9 ng/mL from 1999 to 2004, and 2.5 ng/mL after 2004) were followed up with DRE, TRUS, and sextant or 10‐core biopsy. In order to differentiate reports of this overlapping population, we refer to individual‐site analyses of the full population as the Göteborg‐1 trial, and refer to ERSPC‐Sweden when discussing the contributions in the context of the ERSPC combined‐site analyses. The Göteborg‐1 trial was rated as medium ROB because of the lack of allocation concealment and the high rates of PSA testing in the control group.

Rotterdam Pilot 1 was a small pilot study for the ERSPC; participants were not included in the ERSPC‐Netherlands cohort. ⁴⁹ The trial randomized men to screening (PSA, DRE, and TRUS) versus usual care from 1991 to 1992 (N = 1134; median age, 64 years). Participants with a PSA of ≥3.0 ng/mL underwent TRUS‐guided biopsy. The study was rated as high ROB because of unclear randomization methods and postrandomization exclusions of those with a high PSA, and is likely underpowered for mortality outcomes.

ProScreen is an ongoing Finnish trial of a single invitation to risk‐stratified screening via PSA, 4‐kallikrein testing for a PSA of ≥3.0 ng/mL, MRI for risk scores of more than 7.5%, and targeted biopsy for Prostate Imaging Reporting and Data Systems ratings of 3–5 or if MRI was negative and PSA density was elevated versus those not invited to screening (N = 60,745; mean age, 57.3 years; 2018–2020). The trial was rated as medium ROB because of low participation rates (first‐round participation, 51%), and mortality outcomes were not yet available. ⁹

We identified four observational studies that evaluated PSA‐based screening. One retrospective study used US Veterans Health Administration (VHA) administrative data (N = 45,834 non‐Hispanic Black and White men; study years, 2004–2017) to evaluate the association between PSA screening intensity in the 5 years before diagnosis and prostate cancer–specific mortality. ⁶² We rated this study as medium ROB because of the relatively short follow‐up period (median, 77 months) and concerns that findings might change with longer follow‐up (90% of patients remained alive at the time of analysis). A population‐based cohort study of the Stockholm PSA and Prostate Biopsy Registry (N = 174,636; age, 50–74 years; PSA tests were conducted in 2003–2015) examined the association between PSA testing intervals and tumor grade at diagnosis, as well as the likelihood of a false‐positive PSA result leading to biopsy recommendations. It was rated as low ROB given the near‐complete registry data for PSA testing and cancer outcomes. ⁶⁰ A small population‐based study in Stockholm evaluated long‐term prostate cancer–specific mortality 20 years after a single invitation to PSA‐based screening (DRE, TRUS, and PSA testing) in men aged 55–70 years (n = 1779 participated; n = 2400 were invited; 27,464 background population; recruitment, 1988–1989). The study was rated as high ROB because of the limited adjustment for confounders and potential baseline differences. ⁵⁹ A retrospective analysis of incidence and stage distribution in a nationwide PSA‐based screening program in Lithuania included 655,487 men aged 50–74 years (or 45–49 years with a family history of prostate cancer; PSA testing was conducted in 2006–2015). The study was rated as high ROB because of inadequate control for confounding, lack of discussion of missing data, and no examination of baseline differences between groups. ⁶¹

Prostate cancer incidence: Disease stage and tumor burden

There is moderate SOE that PSA‐based screening is associated with a higher incidence of a Gleason score of ≤6 and T1/T2 tumors and a lower incidence of Gleason ≥8. The 2018 systematic review did not report incidence by Gleason score or tumor size at diagnosis. There was a higher incidence of Gleason ≤6 in the screening arms and Gleason ≥ 8 in the usual care arms in the PLCO (median follow‐up, 15 years; Table 2) ⁵² , ⁵³ and CAP (median follow‐up, 15 years), ⁴⁵ , ⁴⁶ as well as in ERSPC‐Finland (median follow‐up, 16 years) ⁴⁸ and ERSPC‐Netherlands (median follow‐up, 21 years; N = 42,305). ⁴¹ There was no significant difference in incidence of Gleason ≤ 6 in the Rotterdam Pilot 1 trial (N = 1134). ⁴⁹ Localized T1/T2 cancer was more common with screening in the CAP after 15 years (3.6% vs. 3.1%; RR, 1.18 [95% CI, 1.14–1.22]) ⁴⁶ and ERSPC‐Netherlands after 21 years (13.6% vs. 7.4%; RR, 1.84 [95% CI, 1.72–1.96]); T3/T4 cancer was more likely in the control arm than the screening arm of the ERSPC‐Netherlands (1.8% vs. 2.3%; RR, 0.79 [95% CI, 0.68–0.92]). ⁴¹ The Stockholm PSA and Prostate Biopsy Registry study (N = 174,636) evaluated the impact of PSA testing intervals on Gleason scores at diagnosis in patients with a baseline PSA of <10 ng/mL; 67% of men were screened annually or biennially. ⁶⁰ Annual screening was more common for older men and those with higher baseline PSA. For men with a baseline PSA of >1.0 ng/mL, the risk of being diagnosed with Gleason ≥7 prostate cancer increased with screening interval (RR range vs. 1‐year interval: 2 years, 1.3–1.4; 3 years, 1.2–1.7; 4 years, 1.7–1.8; 5–8 years, 1.6–2.5). Risk of Gleason 6 cancer was not associated with testing interval. The effects of testing interval did not vary by age or family history.

TABLE 2.

Gleason score and disease stage at diagnosis in PSA‐based screening trials.

Study	Gleason ≤ 6	Gleason 7	Gleason ≥ 8	Localized, T1/T2	Advanced, T3/T4 a	Metastatic disease
	Screening versus control, No. (%) RR (95% CI)	Screening versus control, No. (%) RR (95% CI)	Screening versus control, No. (%) RR (95% CI)	Screening versus control, No. (%) RR (95% CI)	Screening versus control, No. (%) RR (95% CI)	Screening versus control, No. (%) RR (95% CI)
PLCO
Screening, n = 38,340	3095 (8.1) versus 2648 (6.9)	1510 (3.9) versus 1511 (3.9)	630 (1.6) versus 708 (1.8)	NR	NR	134 (0.3) versus 158 (0.4)
Control, n = 38,343	1.17 (1.11–1.23)	1.00 (0.93–1.07)	0.89 (0.80–0.99)			0.85 (0.67–1.06)
CAP
Screening, n = 189,386	4111 (2.2) versus 3482 (1.6)	NR	2247 (1.2) versus 2937 (1.3)	6874 (3.6) versus 6746 (3.1)	1705 (0.9) versus 2328 (1.1)	NR
Control, n = 219,439	1.37 (1.31–1.43) b		0.89 (0.84–0.94) b	1.18 (1.14–1.22) b	0.85 (0.80–0.90) b
ERSPC‐Finland
Screening, n = 31,866	2071 (6.5) versus 2034 (4.2)	877 (2.8) versus 1478 (3.1)	464 (1.5) versus 836 (1.7)	NR	NR	NR
Control, n = 48,278	1.54 (1.45–1.64) b	0.90 (0.83–0.98) b	0.84 (0.75–0.94) b
ERSPC‐Netherlands c
Screening, n = 17,442	1794 (10.3) versus 746 (4.3)	562 (3.2) versus 524 (3.0)	235 (1.3) versus 379 (2.2)	2364 (13.6) versus 1282 (7.4)	313 (1.8) versus 394 (2.3)	118 (0.7) versus 278 (1.6)
Control, n = 17,389	2.40 (2.21–2.60) b	1.07 (0.95–1.20) b	0.62 (0.53–0.73) b	1.84 (1.72–1.96) b	0.79 (0.68–0.92) b	0.42 (0.34–0.52) b , d
Rotterdam Pilot 1
Screening, n = 553	39 (7.1) versus 26 (4.5)	NR	NR	56 (10.1) versus 44 (7.6)	15 (2.7) versus 13 (2.2)	3 (0.5) versus 7 (1.2)
Control, n = 581	1.58 (0.97–2.55) b			1.34 (0.92–1.95) b	1.21 (0.58–2.52) b	0.45 (0.12–1.73) b
ERSPC‐Spain
Screening, n = 2415	NR	NR	NR	NR	NR	7 (0.3) versus 5 (0.3)
Control, n = 1861						1.08 (0.34–3.39) b

Abbreviations: CAP, Cluster Randomized Trial of PSA Testing for Prostate Cancer; CI, confidence interval; ERSPC, European Randomized Study of Screening for Prostate Cancer; NR, not reported; PLCO, Prostate, Lung, Colorectal, and Ovarian Cancer Screening Trial; PSA, prostate‐specific antigen; RR, relative risk; T1, T2, T3, and T4, tumor size categories in the TNM tumor staging system.

^a The CAP trial reported T4/N1/M1 disease combined (no T3); ERSPC‐Netherlands and Rotterdam Pilot 1 are T3/T4 combined. Rotterdam Pilot 1 T4 only is 1 versus 6.

^b RR was calculated by systematic review investigators.

^c Core age group (55–69 years).

^d Full age group (55–74 years): 161 of 20,984 versus 342 of 20,916 (RR, 0.47 [95% CI, 0.39–0.57]).

There is low SOE that PSA‐based screening is associated with a lower incidence of metastatic prostate cancer at diagnosis. In the systematic review, the ERSPC combined sites showed fewer metastatic cancers at diagnosis in the screening arm than in the control arm (3% vs. 10%; RR, 0.60 [95% CI, 0.52–0.70]), whereas the PLCO did not show a statistically significant difference in metastatic prostate cancer at diagnosis after nearly equivalent participation in PSA testing in the screening and usual care arms (2.3% vs. 2.9%; RR, 0.78 [95% CI, 0.59–1.02]). ³⁵ , ³⁶ In findings published after the systematic review, metastatic disease at diagnosis (Table 2) remained less frequent in the ERSPC‐Netherlands screening arm after 21 years in both the core age group (0.7% vs. 1.6%; RR, 0.42 [95% CI, 0.34–0.52]) and full age group (0.8% vs. 1.6%; RR, 0.47 [95% CI, 0.39–0.57]). ³⁹ Differences were not statistically significant in the PLCO after 15 years (0.3% vs. 0.4%; RR, 0.85 [95% CI, 0.67–1.06]), ⁵² , ⁵³ ERSPC‐Spain after 21 years (0.3% vs. 0.3%; p = .086), ⁴⁴ or Rotterdam Pilot 1 after 19 years (0.5% vs. 1.2%; RR, 0.45 [95% CI, 0.12–1.73]). ⁴⁹

There is moderate SOE that PSA‐based screening is associated with a lower cumulative incidence of metastatic prostate cancer. The systematic review reported lower cumulative metastatic disease in ERSPC combined sites (RR, 0.70 [95% CI, 0.60–0.82]). ³⁵ , ³⁶ In results from ERSPC individual sites published after the systematic review, cumulative metastatic disease at the Netherlands site after 21 years was lower in the screening arm than in the control arm in the core age group (1.7% vs. 2.6%; RR, 0.67 [95% CI, 0.58–0.78]) and full age group (1.9% vs. 2.6%; RR, 0.74 [95% CI, 0.66–0.85]). ³⁹ Cumulative diagnosis of metastatic disease in the Göteborg‐1 trial was 2.3% for all those invited to screening (1.2% for those who attended screening) versus 1.8% in the control arm (median follow‐up, 18 years). ⁴² In the CAP, invitation to screening was associated with a lower risk of advanced prostate cancer, defined as T4, N1, or M1 disease (0.9% vs. 1.1%; risk difference, −0.16%) after a median of 15 years. ⁴⁶ In Rotterdam Pilot 1, 1.3% of men in the screening arm and 2.8% in the control arm progressed to metastatic disease (RR, 0.46 [95% CI, 0.19–1.11]). ⁴⁹

Prostate cancer–specific mortality

There is low SOE that PSA‐based screening is associated with lower prostate cancer–specific mortality. The 2018 systematic review found no significant reduction in prostate cancer–specific mortality in the screening arms of the PLCO after 14.8 years of follow‐up (RR, 1.04 [95% CI, 0.87–1.24]) or CAP after 10 years (RR, 0.96 [95% CI, 0.85–1.08]) but found lower prostate cancer–specific mortality in the ERSPC combined‐site analysis after 13 years (RR, 0.79 [95% CI, 0.69–0.91]), with reductions limited to the Netherlands and Sweden. ³⁵ , ³⁶

Longer term follow‐up continued to show no prostate cancer–specific mortality reduction in the PLCO after nearly 17 years, which is reflective of the nearly equivalent rates of PSA testing in the screening and usual care arms (RR, 0.93 [95% CI, 0.81–1.08]; Table 3). Prostate cancer deaths in the screening versus usual care arms did not differ by Gleason score, D’Amico risk category, or cancer stage. ⁵² , ⁵³ In the CAP, extended follow‐up showed a small reduction in prostate cancer–specific mortality after a median of 15.4 years (0.6% vs. 0.7%; RR, 0.92 [95% CI, 0.85–0.99]). ⁴⁵ , ⁴⁶ A combined analysis of ERSPC sites found lower prostate cancer–specific mortality after 23 years in the core age group (55–69 years; 1.4% vs. 1.6%; rate ratio, 0.87 [95% CI, 0.80–0.95]) ⁵⁵ and after 15.5 years for all ages (rate ratio, 0.82 [95% CI, 0.74–0.91]). ⁴³ Similar to earlier results included in the systematic review, reductions were limited to the Netherlands and Sweden, although each trial site was not powered to assess mortality individually. Separate individual‐site analyses of the ERSPC’s core age group (55–69 years) found lower prostate cancer–specific mortality in the screening arm of ERSPC‐Netherlands (rate ratio, 0.73 [95% CI, 0.61–0.88]; median follow‐up, 21 years) ³⁹ and, after adjusting for age, prostate cancer mortality was lower in Finland’s screening arm (adjusted hazard ratio, 0.79 [95% CI, 0.74–0.84]). ⁴⁸ For all ages (50–74 years), ERSPC combined‐site analysis found lower prostate cancer–specific mortality in the screening arm (rate ratio, 0.82 [95% CI, 0.74–0.91]), ⁴³ as did individual‐site analyses from the Netherlands (rate ratio, 0.83 [95% CI, 0.71–0.97]; median follow‐up, 21 years) ³⁹ and the Göteborg‐1 trial (50–64 years at randomization; rate ratio, 0.71 [95% CI, 0.55–0.91]; median follow‐up, 21 years). ⁴⁰ There was no significant mortality difference between the arms of Rotterdam Pilot 1 after 19 years (RR, 0.48 [95% CI, 0.17–1.36]). ⁴⁹ The high‐ROB Swedish cohort study found similar prostate cancer mortality among men undergoing a single screening examination and the background population (3.3% vs. 3.4%; incidence rate ratio, 0.97 [95% CI, 0.71–1.23]). ⁵⁹

TABLE 3.

Prostate cancer–specific mortality in PSA‐based screening trials.

Study ROB	Enrollment years	Follow‐up, years	Prostate cancer–specific mortality
			Screening versus control, No. (%)	Rate ratio (95% CI)
PLCO High	1993–2001	17	333 (0.9) versus 352 (0.9)	0.93 (0.81–1.08)
CAP Medium	2001–2009	15.4	1199 (0.6) versus 1451 (0.7)	0.92 (0.85–0.99)
ERSPC combined‐site analysis a High	1993–2003	23	969 (1.4) versus 1385 (1.6)	0.87 (0.80–0.95)
ERSPC‐Finland a Medium	1996–1999	16	278 (0.8) versus 456 (0.9)	0.79 (0.74–0.84) c
ERSPC‐France a High	2001–2005	9.5	78 (0.2) versus 75 (0.2)	1.03 (0.75–1.42)
ERSPC‐Netherlands a Medium	1993–2000	21	198 (1.2) versus 268 (1.6)	0.73 (0.61–0.88)
ERSPC‐Spain b High	1996–1999	21	12 (4.3) versus 8 (3.3)	NR
Göteborg‐1 b Medium	1995	22	112 (1.1) versus 158 (1.6)	0.71 (0.55–0.91)
Rotterdam Pilot 1 High	1991–1992	19	5 (0.9) versus 11 (1.9)	0.48 (0.17–1.36)

Abbreviations: CAP, Cluster Randomized Trial of PSA Testing for Prostate Cancer; CI, confidence interval; ERSPC, European Randomized Study of Screening for Prostate Cancer; PLCO, Prostate, Lung, Colorectal, and Ovarian Cancer Screening Trial; PSA, prostate‐specific antigen; ROB, risk of bias.

^a Core age group (55–69 years) at randomization.

^b Full age group, not restricted to 55–69 years at randomization.

^c Age‐adjusted hazard ratio.

Comparisons by sociodemographic factors

Evidence was insufficient to assess the relationship between age, race, or family history and prostate cancer screening for prostate cancer mortality.

Age

The 2018 systematic review reported prostate cancer–specific mortality rates by age at randomization for two trials (PLCO and ERSPC). ³⁵ , ³⁶ There was no age and trial arm interaction for prostate cancer–specific mortality in the PLCO (p = .81). In the combined‐site analysis included in the systematic review, the ERSPC found lower prostate cancer–specific mortality rates in the screening arm for those aged 65–69 years (RR, 0.69 [95% CI, 0.55–0.87]; test for heterogeneity across age groups, p = .18) but no other significant differences by age. ³⁵ , ³⁶

In analyses published after the systematic review, the CAP reported no significant differences in prostate cancer–specific mortality between the screening and control arms for age groups 50–54, 55–59, 60–64, or 65–69 years (interaction, p = .75; median follow‐up, 15 years). ⁴⁶ In age‐stratified analyses published after the systematic review, the ERSPC combined sites found significant reductions in prostate cancer–specific mortality after a median of 15.5 years of follow–up for age 55–59 years (rate ratio, 0.76 [95% CI, 0.62–0.92]) and 64–69 years (rate ratio, 0.74 [95% CI, 0.62–0.90]) but no significant differences for other age groups. ³⁹ , ⁴² , ⁴³ In ERSPC‐Finland, older age at randomization was associated with a higher risk of prostate cancer–specific mortality after 16 years of follow‐up but there was no interaction between age at randomization and trial arm (interaction, p = .98). ⁴⁸ The Göteborg‐1 trial found significant reductions in prostate cancer–specific mortality in Sweden after 18 years of follow‐up for age 55–59 years at randomization (rate ratio, 0.47 [95% CI, 0.29–0.78]) but not for age 50–54 years (rate ratio, 0.50 [95% CI, 0.22–1.11]) or 60–65 years (rate ratio, 0.85 [95% CI, 0.58–1.26]). ³⁹ , ⁴² Age‐specific analyses of the CAP and ERSPC were likely underpowered to detect significant differences because of relatively few prostate cancer deaths in each 5‐year age group.

Race

The PLCO found that compared to men who were White, Black men in both the screening and usual care arms had higher PSA and a higher percentage of Gleason 8–10 cancers, and were significantly more likely to be diagnosed with metastatic prostate cancer. ⁴⁷ No other randomized trial reported screening effects by race. The retrospective cohort study using VHA administrative data of 45,834 non‐Hispanic Black and non‐Hispanic White men with intermediate‐, high‐, or very high risk prostate cancer found that higher PSA screening rates in the 5 years before diagnosis were associated with lower prostate cancer–specific mortality after a median of 6.4 years of follow‐up among Black men (subdistribution hazard ratio [sHR], 0.56 [95% CI, 0.41–0.76]) and White men (sHR, 0.58 [95% CI, 0.46–0.75]). Annual versus some screening (less than annual screening but more than none) was associated with lower mortality among Black men (sHR, 0.65 [95% CI, 0.46–0.92]) but not White men (sHR, 0.91 [95% CI, 0.74–1.11]) after 120 months of follow‐up. ⁶²

Family history

The systematic review reported no significant difference in prostate cancer–specific mortality between the screening and usual care arms in PLCO White participants with a family history of prostate cancer (n = 4833; hazard ratio [HR], 0.49 [95% CI, 0.22–1.10]). ³⁵ , ³⁶ In a report of the PLCO published after the systematic review, family history of prostate cancer in a first‐degree relative was predictive of prostate cancer–specific mortality in the usual care arm (HR, 1.89 [95% CI, 1.15–3.11]) but not in the screening arm (HR, 0.829 [95% CI, 0.42–1.63]). ³⁷

Harms of screening

Overdiagnosis

Overdiagnosis is screen‐detected prostate cancer that would not have been clinically diagnosed during a patient’s lifetime because of limited life expectancy and/or slow‐growing tumors or prostate cancer that would never have caused symptoms in a person’s lifetime. ¹¹ , ¹² There is no benefit to treating overdiagnosed cancer, and considerable harm is associated with exposure to, and the side effects of, unnecessary treatment with surgery, radiation, or androgen deprivation. Estimates of overdiagnosis are often based on either simulation models of “lead time”—the time from screening detection to clinical detection of a cancer and expected survival time—or applying estimates of the “excess incidence” of cancers diagnosed in the screening group over the number of cancers diagnosed in the control group with adequate long‐term follow‐up. ¹² , ⁶³ This review focuses on estimates of excess incidence from the included trials.

PSA‐based screening is associated with overdiagnosis, with estimates of the magnitude of overdiagnosis varying by trial, duration of follow‐up, and analytic approach (low SOE). In the 2018 systematic review, overdiagnosis as a percentage of all prostate cancers diagnosed was estimated at 16.4% in the PLCO (13‐year follow‐up), 40.7% in the CAP (10‐year follow‐up), and 33.2% in ERSPC combined sites (13‐year follow‐up). ³⁵ , ³⁶ Estimates for individual ERSPC sites were 23.6% in Finland (12‐year follow‐up), 36.0% in Spain (15.2‐year follow‐up), 46.8% in Sweden (14‐year follow‐up), and 47.9% in the Netherlands (12.8‐year follow‐up). As a percentage of screen‐detected cancers, overdiagnosis was estimated at 20.7% in the PLCO and 50.4% in ERSPC combined sites, with individual estimates of 25.6% in Finland and 58.9% in the Netherlands. ³⁵ , ³⁶ Excess incidence estimates in the 2018 systematic review likely represent inflated estimates of overdiagnosis because of short follow‐up durations. More recent estimates using the excess incidence approach with longer follow‐up from ERSPC‐Finland and a recent reanalysis of CAP data by authors not affiliated with the trial report lower rates of overdiagnosis. ERSPC‐Finland estimated a relative overdiagnosis over 18.6 years of follow‐up, with ranges of 1.2%–3.0%, 13.4%–15.7%, 10.2%–13.2%, and 10.3%–14.0% for the 1929–1932, 1933–1936, 1937–1940, and 1941–1944 birth cohorts, respectively. ⁵⁸ The reanalysis of CAP data estimated an overdiagnosis of 11.7% after a median 15‐year follow‐up (95% CI, 0%–26.7%) after a single screen. ³⁸ On the basis of their estimates of overdiagnosis, UK age‐based mortality rates, and a 15‐year sojourn time, they projected overdiagnosis for men diagnosed after a single PSA screening at ages 59, 64, 70, and 80 years to be 20%, 24%, 32%, and 58%, respectively.

False positives

There is moderate SOE that PSA‐based screening is associated with false‐positive results, with the extent and impact of false‐positive findings dependent on the screening approach. The 2018 systematic review reported that 10.4% of the 32,576 PLCO participants who completed at least one PSA screen in the first four rounds had at least one false‐positive PSA result. Across five ERSPC sites (n = 61,604), 17.8% of those attending at least one screening round had at least one false‐positive result. Biopsies after abnormal screening resulted in prostate cancer diagnosis in 32.3% of PLCO participants, 39.4% of CAP participants, 24.2% of ERSPC participants, and 37.2% of participants in a study of men aged older than 65 years with national VHA and Medicare data. ³⁵ , ³⁶ In a low‐ROB Swedish cohort study not included in the systematic review, the cumulative probability of at least one false‐positive PSA result followed by a biopsy recommendation increased with higher baseline PSA levels and shorter testing intervals (N = 174,636). ⁶⁰ False‐positive rates were slightly lower for 60‐ than 50‐year‐old men at all baseline PSA levels and screening intervals. For a 50‐year‐old man with a family history of prostate cancer and baseline PSA of 5–10 ng/mL, the probability of at least one false‐positive biopsy recommendation was 69.7% with annual screening, 58% with a 2‐year interval, 49.2% with a 3‐year interval, and 42.5% with a 4‐year interval; probabilities with a baseline PSA of 0–1 ng/mL were 15.7%, 6.3%, 3.5%, and 2.3%, respectively.

Biopsy harms

There is moderate SOE that biopsy harms include infection, bleeding, and urinary harms. The 2018 systematic review reported that biopsy‐related complications occurred after approximately 2% of 4861 biopsies in the PLCO, including 1.3% noninfectious and 0.7% infectious complications. In a VHA cohort study, 5.6% of 8313 men had infectious or urinary complications within 7 days of biopsy, and 1.6% were hospitalized. At the ERSPC‐Netherlands site, 22.6% of patients reported hematuria lasting more than 3 days, 7.5% reported significant postbiopsy pain, and 3.5% reported fever. In the UK‐based ProBE cohort study (N = 1147), 89.4% reported one or more infectious or hemorrhagic symptom within 7 days of biopsy. Pain (7.3%), fever (5.5%), and hematospermia (26.6%) were considered moderate or serious problems, and 1.3% of participants were admitted to the hospital for biopsy‐related complications within 35 days of biopsy. ³⁵ , ³⁶

Health‐related QOL or psychological harms

There is low SOE that PSA‐based screening may increase prostate cancer worry but does not affect overall health‐related QOL, anxiety, or depression. The systematic review reported that health and functional status remained stable after PSA screening, an abnormal PSA result, DRE, and TRUS biopsy (n = 215–386) in participants at the ERSPC‐Finland site. Overall SF‐36 (36‐Item Short Form Health Survey) scores were similar between men who received a normal and an abnormal PSA screening with benign biopsy results in one cross‐sectional study (N = 210). In one cohort (n = 400) and one cross‐sectional study (N = 210), men with an abnormal PSA followed by a benign biopsy reported more worry about prostate cancer than those with a normal PSA at 6–8 weeks and at 1‐year follow‐up. One cohort study (N = 569) found no increases from baseline in anxiety, depression, or the mental health components of the SF‐12 after an abnormal PSA screening, and one cross‐sectional study (N = 210) found no increase in trait anxiety 2 months after an abnormal PSA and benign biopsy. ³⁵ , ³⁶ A report from the ERSPC‐Finland’s 10‐year follow‐up not included in the systematic review found there were no clinically meaningful differences in health‐related QOL between men with cancer in the screening and control arms, and slightly less urinary bother for those in the screening arm (n = 1035). ⁵⁶

DISCUSSION

Benefits and harms of PSA‐based screening

In this systematic review comparing PSA‐based screening to no screening, we identified a 2018 systematic review of PLCO, CAP, ERSPC, 23 publications of evidence from six trials (PLCO [high ROB], CAP [medium ROB], ERSPC [high ROB combined; individual sites: two medium and two high], Göteborg‐1 [medium ROB], Rotterdam Pilot 1 [high ROB], and ProScreen [medium ROB]), and four observational studies (one low ROB, one medium ROB, and two high ROB) published after the review. We found that PSA‐based screening was associated with a higher incidence of low‐risk prostate cancer (Gleason ≤ 6 and T1/T2 tumors; moderate SOE), lower incidence of advanced (T3/T4 tumors; moderate SOE) or metastatic disease at diagnosis (low SOE), and lower cumulative incidence of metastatic disease (moderate SOE). We determined that prostate cancer screening is likely associated with lower prostate cancer–specific mortality (low SOE). However, there was inconsistency across trials and ERSPC sites, with PSA‐based screening associated with lower prostate cancer–specific mortality in the CAP and ERSPC sites in the Netherlands (Finland was significant only after adjusting for age). The Göteborg‐1 trial also reported lower prostate cancer–specific mortality in the screening arm, as reported in individual‐site results and in the ERSPC combined analyses. PSA‐based screening was not associated with prostate cancer–specific mortality in the PLCO. There was insufficient evidence to determine whether the effect of PSA‐based screening on prostate cancer–specific mortality varied by age, race, or family history.

Quantitative synthesis was not feasible because of heterogeneity, and interpretation of these findings is hampered by important methodological differences both across and within trials. Although the PLCO was the only trial conducted in the United States and thus has the most direct relevance to current US practice, its findings are compromised by the lack of a true control (no screening) arm. Participants were enrolled between 1993 and 2001 during a period of rapid increase of PSA use in clinical practice, and there was substantial contamination in the usual care arm. Nearly half of participants in the control arm received PSA testing during the screening phase, and the rates of PSA testing in the screening and control arms were nearly equal through the follow‐up period, which limited the ability of the trial to detect differences between the two study arms. Most of the ERSPC sites used a 4‐year screening interval, longer than the 1–2 years common in US clinical practice. Sweden, however, was an exception, and used an interval of 2 years. The PSA threshold for additional testing at ERSPC sites was primarily ≥3.0 ng/mL (range, 2.5–4.0 ng/mL), lower than the 4.0 ng/mL in the US‐based PLCO. Current clinical practice in the United States reflects the understanding that PSA increases with age, and PSA thresholds may be lower for younger men. In the CAP, screening uptake was relatively low after a single invitation to screening (36%), which resulted in less power to detect screening effects.

These sources of heterogeneity limit comparisons across RCTs and ERSPC sites, a limitation that may be mitigated via secondary analyses or modeling studies informed by trial data. One study explored the differences between the ERSPC Swedish and Finnish sites, and found that 122 cancers (and four resulting deaths) would have been missed in the Swedish cohort if the Finnish protocol (PSA threshold, ≥4.0 ng/mL, or 3.0 ng/mL if free PSA < 0.16; screening interval, 4 years) had been applied. ⁶⁴ There would have been 127 missed cancers (and 19 resulting deaths) had the Swedish protocol (PSA threshold, ≥2.5–3.4 ng/mL; screening interval, 2 years) been applied to the Finnish cohort. These results suggest that screening interval and PSA threshold may not fully explain the differences in prostate cancer–specific mortality between the two sites. In an effort to reconcile the differences between the PLCO and ERSPC, a study using microsimulation models that accounted for differences in implementation and setting found similar reductions in prostate cancer–specific mortality in both trials. ⁶⁵ These findings suggest that despite no significant prostate cancer–specific mortality benefit reported in the PLCO, PSA‐based screening reduces prostate cancer–specific mortality.

A major gap in the evidence examining PSA‐based screening is a lack of data available across sociodemographic and clinical subpopulations. RCTs were largely conducted in Europe, and enrolled predominantly White populations; in the PLCO, only 5% of participants were Black, and 1.8% were Hispanic. In addition, incidence and mortality results were rarely stratified by race or family history. Only the PLCO and one cohort study reported results stratified by family history. The PLCO found that a family history of prostate cancer in a first‐degree relative was predictive of prostate cancer–specific mortality in the usual care arm but not in the screening arm. However, the cohort study reported that the relationship between screening interval and disease stage did not differ for those with or without a family history of prostate cancer. Stratification by race was also limited to a report from the PLCO and one cohort study. The PLCO reported that Black men in both the screening and usual care arms had a higher PSA and higher percentage of Gleason 8–10 cancers, and were significantly more likely to be diagnosed with metastatic prostate cancer than White men. The cohort study of Black and White veterans receiving care in VHA settings found lower prostate cancer–specific mortality for those who had annual screening than for those who had some but not annual screening in the 5 years before diagnosis for non‐Hispanic Black men but not for non‐Hispanic White men. These limitations are particularly relevant given the disproportionate burden of prostate cancer among Black men. The current systematic review did not investigate the causes of the racial disparities, and future trials of screening for prostate cancer should include diverse populations. Furthermore, research is needed to better understand any biological differences in disease risk and progression alongside contributions from structural, social, and health system factors. ⁶⁶

Increased detection of earlier stage cancer and mortality reductions from screening must be considered alongside potential harms of screening, including overdiagnosis and overtreatment, false positives, and harms of biopsy. Overdiagnosis estimates based on excess incidence reported in the 2018 systematic review varied from 16% to more than 50% in the screening trials, which was likely a result of different screening protocols. Excess incidence estimations, however, may provide inflated estimates of overdiagnosis, particularly when sojourn time is long or unclear. More recent estimates with longer term follow‐up with excess incidence ranged from 2% to 15% in the ERSPC‐Finland site (low SOE for magnitude of overdiagnosis). Although modeling studies were not included in this systematic review, they should be considered alongside excess incidence estimates. Estimates of overdiagnosis for the CAP with the excess incidence methods ranged from 11% to upward of 40%, depending on age and follow‐up time. In a post hoc analysis with microsimulation models to estimate sojourn time, overdiagnosis was estimated to be 9.2% for 50‐ to 54‐year‐old men and 20.8% for 65‐ to 69‐year‐old men. ⁴⁶ However, overdiagnosis in this trial was limited to a single invitation to PSA‐based screening, and results have limited applicability to US clinical practice.

Another harm of PSA‐based screening is false‐positive PSA results, which can lead to exposure to additional diagnostic tests, including biopsy. False‐positive rates varied across trials, with higher rates of false positives in trials with lower PSA thresholds (moderate SOE). The US‐based PLCO reported that 10.4% of participants received a false‐positive PSA (≥4.0 ng/mL) in the first four screening rounds. Thresholds for referral to biopsy were lower in the ERSPC, where 17.8% received a false positive.

Harms of biopsy included infectious and noninfectious complications, including fever, bleeding, and pain (moderate SOE). Biopsy complications were reported for 2% of participants in the PLCO, and 5.6% reported infectious or urinary complications within 7 days in a US‐based cohort study. These results are likely applicable to the US population undergoing biopsy after an abnormal PSA. Rates were higher for the ERSPC and a UK cohort; however, those rates reflect older technology and biopsy techniques (e.g., no MRI and fewer biopsy cores), and may not reflect the use of modern approaches in current clinical practice in the United States. Although outside the scope of this review of PSA‐based screening, three recent trials found no significant differences in infection rates between MRI‐directed transperineal versus transrectal biopsy, ⁶⁷ , ⁶⁸ , ⁶⁹ and current recommendations from the American Urological Association endorse using either transperineal or transrectal biopsies for diagnosing prostate cancer. ¹⁹ Studies reporting psychological harms or health‐related QOL impacts were small but evidence suggests minimal harm from screening (low SOE).

Limitations

This systematic review, requested by the ACS GDG, includes studies comparing PSA‐based prostate cancer screening to no screening, and was developed as part of a larger program of research designed to support the development of the 2026 ACS prostate cancer screening guideline. ²⁴ There are several important limitations. First, we limited our search to English‐language publications, and may have missed studies published in other languages. We did not perform quantitative synthesis, given that there was significant heterogeneity across study populations, screening protocols, and reported outcomes.

To compare the benefits and harms of PSA‐based screening, we included only trials with no‐screening controls. The included trials were initiated in the 1990s or very early 2000s, and because of changes in contemporary clinical practice in recent years, the applicability of findings from the included screening trials is limited. Prebiopsy MRI, biomarkers, and other risk stratification approaches are increasingly incorporated into the prostate cancer screening continuum, and MRI‐guided transperineal and transrectal biopsies are associated with few harms and serve as alternatives to systematic or template biopsies. Only one included trial, the ongoing Finnish ProScreen trial, ⁹ compared men randomized to an invitation to risk‐stratified screening versus no invitation to screening; however, mortality outcomes are not yet available. Newer ongoing studies, such as the European‐based PROBASE, TRANSFORM, and PROFILE trials, are investigating risk‐adapted screening strategies incorporating MRI, biomarkers, genetic mutations, or family history of prostate cancer. ¹⁰ , ⁷⁰ , ⁷¹ Whether these approaches meaningfully reduce overdiagnosis or improve mortality outcomes remains uncertain. To provide the GDG with an overview of the evidence examining biomarkers, as a companion to this review, we also produced high‐level evidence maps to illustrate the body of evidence examining whether the use and/or sequence of post‐PSA biomarkers (with and without MRI) to guide biopsy decisions improves clinical utility. ²³

In addition to advances in the modern‐day screening continuum, disease management has also fundamentally changed since the bulk of the included PSA‐based screening trials were conducted. At the time, radical prostatectomy or radiation therapy were standard treatments. Today, low‐risk, localized prostate cancer is managed with active surveillance, ⁷² an approach associated with a more favorable balance between preserving function and maintaining low absolute risks of prostate cancer death. ³⁵ , ⁷³ , ⁷⁴ , ⁷⁵

CONCLUSIONS

PSA‐based screening reduces the incidence of advanced disease at diagnosis and can lower prostate cancer–specific mortality, although the magnitude of benefit varies by screening protocol and participant characteristics. Harms associated with PSA‐based screening include overdiagnosis, false‐positive test results, biopsy‐related complications, and downstream effects of treatment. Decisions regarding PSA‐based screening should weigh the benefits of earlier detection of clinically significant disease and reduced mortality alongside established harms, and consider strategies to mitigate the impact of overdiagnosis and reduce the likelihood of overtreatment.

AUTHOR CONTRIBUTIONS

Jessica C. Griffin: Conceptualization; methodology; data curation; investigation; validation; formal analysis; supervision; project administration; writing—original draft; writing—review and editing. Rita L. Shiau: Conceptualization; investigation; validation; formal analysis; writing—original draft; writing—review and editing. Clarissa A. B. Troutman: Data curation; investigation; validation; project administration; writing—review and editing; writing—original draft. Alanna Welsh: Investigation; validation; writing—review and editing. Emily Edwards: Investigation; validation; writing—review and editing. Ibironke A. Ajayi: Investigation; validation; writing—review and editing. Rose Relevo: Conceptualization; data curation; methodology; writing—review and editing. Karli K. Kondo: Conceptualization; validation; writing—original draft; writing—review and editing.

CONFLICT OF INTEREST STATEMENT

The views expressed here are those of the authors, and do not represent the American Cancer Society, the American Cancer Society Guideline Development Group or Guideline Development Program staff, or the American Cancer Society Cancer Action Network. Ibironke A. Ajayi reports consulting for Truveta. The other authors declare no conflicts of interest.

Supporting information

Supplementary Material