Initial Longitudinal Outcomes of Risk-Stratified Men in Their Forties Screened for Prostate Cancer Following Implementation of a Baseline Prostate-Specific Antigen

Prostate-specific antigen (PSA) screening provides an opportunity for the early detection of men harboring organ-confined prostate cancer (PCa) in whom cure is possible. Nevertheless, screening remains controversial because PSA elevations are difficult to interpret when confounded by benign processes [1]; PSA also cannot accurately differentiate men more likely to develop lethal PCa from those prone to lower-risk tumors [2, 3, 4, 5]. Ultimately, the routine use of PSA testing could result in unnecessary anxiety, potentially avoidable biopsies, and overtreatment of indolent cancers [6].

Professional societies continue to alter guidance on PSA screening by incorporating risk-adapted strategies to identify men likely to benefit most from early detection [7, 8, 9]. While shared decision-making for PCa testing is currently advocated for healthy men between the ages of 55 to 69 years [10], recent literature has suggested the potential utility of obtaining an early baseline PSA to establish lifetime cancer risk [2, 11, 12]. For men <50 years, a baseline PSA ≥1.5 ng/mL has been shown to confer increased lifetime PCa-associated risk and mortality [4, 13, 14]. These at-risk men may stand to benefit from early and more frequent urologic evaluation [11].

Important to the system-wide adoption of these practices is buy-in from primary care providers (PCPs) performing the majority of initial testing and systematic integration of new guidelines into practice. Recognizing this, the Duke Cancer Institute (DCI) and Duke Primary Care (DPC) collaborated to develop a previously described system-wide, low-cost, and risk-adapted PCa screening algorithm as a clinical decision support tool integrated into the electronic health record (EHR) on 2/2/2017 for men ages 40 to 75 years who were evaluated by a PCP [3]. The algorithm was implemented in the EHR, taking factors such as age at time of PCP examination, race, and family history (FH) into consideration to provide personalized urology referral recommendations for appropriate patients. Herein, we present the initial longitudinal outcomes of men in their forties who underwent early baseline PSA testing and were found to have an elevated PSA following initiation of this novel system-wide screening program.

Full patient breakdown is depicted in (Fig. 1). Our study cohort consisted of 564 men (4.7%) with a PSA ≥1.5 ng/mL between 2/2/2017–2/21/2018 (1 y after algorithm implementation) that triggered recommendation for urology referral. A total of 330 men (58.5%) were ultimately referred to Duke urologists, of whom 101 (30.6%) were AA and 72 (21.8%) had a FH of PCa (Table 1). Deferred referral was identified in 27 patients (4.8%) with a median 25-month delay (range 4–43 mo). The remainder of the cohort consisted of 234 men (41.5%) who were either not referred (n=206) or were referred for reasons unrelated to PCa suspicion (n=28). Further clinical parameters are presented in (Table 1). Of the 564 patients, 217 (38.5%) had at least one previous PSA value before algorithm implementation (118 from the referred group and 99 from the non-referred group). After an elevated PSA post-algorithm incorporation, 467 patients (82.8%) underwent confirmatory PSA testing, with 342 (60.6%) found to have a persistently abnormal value. Patients at time of referral for elevated PSA were found to have significantly higher mean PSA values than those not referred (2.97 vs. 1.98, p=0.001).

Table 1
Patient demographics and clinical outcomes

• Click for larger image
• Click for full table
• Download as Excel file

In our study cohort, 51 men (9.0%) underwent additional PCa risk assessment, including 48 receiving Prostate Health Index (PHI) testing (Beckman Coulter, Brea, CA, USA), one undergoing a 4Kscore^® test (Opko Health, Elmwood Park, NJ, USA), and two receiving urinary Exosome-Dx (ExosomeDx-Bio-Techne, Waltham, MA, USA). Mean PHI for patients undergoing such testing was 32.4±11.5. Implementing established score stratification per the manufacturer, 15 men had a 9.8% probability of cancer (PHI 0–26.9), 17 had a 16.8% probability (27.0–35.9), 14 had a 33.3% probability (36.0–54.9), and two had a 50.1% probability (55.0+). Within each category, rates of cancer detection were 0.0%, 0.0%, 28.6%, and 50.0%, respectively.

A total of 49 patients (8.7%) underwent prostate biopsies, of which 20 cases (40.8%) revealed PCa, resulting in a cohort detection rate of 3.5%. Average PSA at baseline and immediately prior to any PCa diagnosis at biopsy was 6.04 ng/mL (standard deviation [SD]=6.11 ng/mL) and 5.64 ng/mL (SD=3.32 ng/mL), respectively. Excluding PSA values taken prior to algorithm implementation, time from PSA draw post-algorithm implementation in 2017 to overall cancer diagnosis at biopsy was 16 months on average (SD=10.66 mo). csPCa was revealed in 11 (55.0%) of the men diagnosed with cancer, with three having ≥GGG4. Of men with csPCa, six were AA, and three had a FH of PCa. Mean PSA for these 11 men at baseline and immediately prior to csPCa diagnosis at biopsy was 7.76 ng/mL (SD=7.79 ng/mL) and 7.33 ng/mL (SD=3.63 ng/mL), respectively.

Management strategies varied for the 20 patients diagnosed with PCa, including five electing active surveillance (AS), 14 choosing radical prostatectomy (10 had csPCa), and one with csPCa opting for radiotherapy. Patients referred to urology based off of elevated PSA during the study period (2/2/2017–2/21/2018) but who did not undergo biopsy (n= 282) continued periodic PSA monitoring every 11 months on average (SD=9.0 mo), with a per-patient mean of 5.23 (SD=1.6) recorded PSA values since the algorithm was initiated. Men not referred to urology for elevated PSA (n=233) underwent PSA monitoring at mean interval of 14 months (SD=9.0 mo), although they underwent a statistically lower number of PSAs with a mean of 2.4 (SD=2.5, p<0.001) recorded values since algorithm implementation.

Multivariable analysis revealed that baseline PSA was associated with PCa of any grade on subsequent biopsy (odds ratio [OR], 1.07; p<0.05). While immediate referral to urology was associated with finding any PCa on biopsy compared to delayed referral (OR, 4.58; p<0.05), the presence of risk factors was not (p>0.05). Ultimately, of the men ages 40 to 49 years meeting screening criteria (n=6,397), the number of men needed to screen and detect all PCa and csPCA was 319.9 and 581.5, respectively (Table 2).

Table 2
Multivariable logistic regression for possible factors associated with any prostate cancer at subsequent prostate biopsy

• Click for larger image
• Click for full table
• Download as Excel file

We assessed outcomes of men aged 40 to 49 years who underwent PSA assessment utilizing the Duke PCa screening algorithm to identify patients who currently have or are at risk for subsequently developing potentially lethal PCa [2]. Initial longitudinal analysis led to overall cancer detection rate and csPCa detection rate of 3.5% and 2.0%, respectively; notably, 55.0% of PCa diagnoses were csPCa. While this small proportion of men benefited from early diagnosis within short follow-up, our overarching goal is to use a baseline PSA to risk stratify men remaining at increased lifetime risk even when the diagnosis of PCa is not imminent [12]. PSA itself on multivariable analysis, while significant, featured a lower OR than immediate referral for association with subsequently finding PCa, highlighting the importance of the algorithm compared to individual factors alone. Revealing these at-risk patients early provides the opportunity to intervene at a point in life when longevity is expected and the natural history of significant PCa can be impactfully changed. We believe our results confirm the impact of the algorithm as a tool to pre-emptively identify and risk-stratify men for early intervention.

While the current approach to PCa screening employs a combination of factors, PSA has been shown to be a stronger predictive factor for cancer diagnosis than either race or FH [5, 17]. Indeed, PCa-specific mortality rates have decreased by 40% during the past three decades due to widespread utilization of PSA assessment [2]. Nonetheless, screening has been controversial with concerns that excessive testing leads to unnecessary interventions and overdiagnoses of lower-risk PCa. These concerns were evidenced through the USPSTF initially recommending against routine PSA testing in 2012 [8]. Houston et al [18] highlighted the repercussions from reduction in screening, showing an increased incidence of advanced disease after that recommendation. Extended follow-up at 13 years of the European Randomized Study of Screening for Prostate Cancer (ERSPC) trial further demonstrated the importance of screening, as one cancer death was avoided per 781 men tested; noncompliance with screening protocols also contributed to a significant proportion of PCa deaths [19, 20]. Notably, the USPSTF since revised their recommendation in 2018, endorsing an individualized shared decision to screen.

While several professional societies currently recommend PSA screening initiation at age 55 years, the value of obtaining a baseline PSA in younger patients to determine lifetime risk and subsequent screening strategy has recently garnered interest. There is a perceived advantage of using baseline PSA to risk-stratify men in their forties, as there are fewer PSA-confounding factors aside from prostatitis and early benign prostatic hyperplasia [21]. Because PCa typically progresses slowly, younger men with higher baseline PSA may be more likely to develop csPCa over their lifetime [3]. Using a population-based sample of 21,277 previously unscreened patients, Lilja et al [11] demonstrated significant association between PSA at or before age 50 years and subsequent PCa risk (p<0.0005, area under the curve [AUC], 0.719) and advanced disease (AUC, 0.751) up to 30 years later. Indeed, men aged 44 to 50 years with PSA values above the median (0.63 ng/mL) accounted for 81% of diagnosed advanced cancer cases [11]. Vickers et al [12] further commented on the same cohort in their 2013 study, showing that 44% of observed deaths subsequently occurred in men with a PSA in the highest 10th of the distribution when they were age 45 to 49 years. Taken with the small but significant risk of metastasis at 15 years (0.09%) in their cohort, the authors suggested the use of a single PSA measurement in a man’s forties to predict long term csPCa risk and establish frequency of monitoring [12]. Our preliminary results support the findings of these studies; identification of those with an age-specific elevated baseline PSA and the subsequent 2.0% csPCa detection rate in our cohort identifies a key percentage of patients who benefited from early testing even within the short-term. Even referred patients who did not undergo biopsy had the opportunity to continue serial PSA testing. With continued follow-up to at least eight years in our study population, we expect early screening to impact cancer-related mortality in line with results previously shown by Vickers et al [12], further validating baseline PSA testing for men ages 40 to 49 years old.

The determination of an appropriate age-specific upper limit “normal” PSA value for early screening is also a source of debate, as there is no value below which PSA risk is absent [7]. Prior research shows a PSA cutoff of 4 ng/mL includes up to 14.9% of high-grade disease within accepted normal limits [7]. Previously published studies showed that age-specific PSA medians (0.6 ng/mL in AA and white men <50 y) should be used as a cutoff for risk stratification to detect PCa [16]. Tang et al [4], who identified a similar median PSA of 0.7 ng/mL in their cohort of men younger than 50 years old, found that an initial PSA ≥1.5 ng/mL is a more optimal risk-stratification cutoff than the age-specific median to determine men in their forties who are at higher risk for PCa and may benefit from further diagnostics or more frequent testing. The authors further showed this risk to be significantly higher and increasing over time for these patients with a PSA ≥1.5 ng/mL, with increased relative risk by 9.3-fold in AA men and 6.7-fold in Caucasian men [4]. Like this and other studies, our screening algorithm similarly utilizes a ≥1.5 ng/mL threshold to trigger a urologic evaluation for men in their forties [13, 14]. The utility of this cut point was further supported by a 2013 study finding that men with a value <1.0 ng/mL had only a 0.6% risk of developing lethal PCa; the authors suggested up to 75% of those men may be able to avoid further screening until 55 years of age [22].

The Duke PSA screening algorithm highlights the critical role of PCPs in PCa evaluation as the first-line for screening. Collaboration between urologists and PCPs is important to optimally coordinate care and specialty referrals, especially when contradictory recommendations can create confusion and inconsistency [2, 23, 24]. At Duke, an online survey was conducted the year following algorithm integration to assess PCP reaction [2]. PCPs were confident in their screening knowledge and use of shared-decision making to discuss patient options, citing the algorithm as the second most utilized guidance program after the USPSTF [2]. This finding highlights the need for continued quality improvement, as the USPSTF is largely PCP-based without representation from PCa specialists, and PCPs often lack access to specialty societies [2]. Indeed, immediate urologic referral after algorithm notification was associated with significantly higher detection of PCa. Yet, despite inconsistent urologic referral by PCPs in our study, the presence of independent PCP decision-making was evidenced through observed higher mean PSAs of those ultimately referred for specialist evaluation. While referred patients that continued to monitor without biopsy underwent a significantly higher number of PSA tests, PCPs still followed patients that were not referred with an average of 2.36 PSA tests at an interval of roughly 14 months. Abiding by a multidisciplinary approach when developing protocols and communication are important to optimize collaboration and outcomes.

Several limitations of this study merit mentioning. First, the study is inherently limited by its retrospective nature despite prospective data collection. Lack of complete adherence to algorithm-prompted urologic referral by PCPs highlights the continued need for improvement to maximize algorithm usage, better understand the reasoning behind the discrepancy, and foster collaboration between specialties. The scope of this study did not allow for the comparison of oncologic outcomes between those with a PSA above and below the baseline cut-off of 1.5 ng/mL. Longer follow-up is necessary, as the coronavirus pandemic may have inadvertently affected outcomes. In addition, the pandemic may have influenced the timing of urology referral and their general care for certain patients. Obtained PSAs were not always true “baseline” values, and the decision to refer or not refer a patient could have been confounded by previously obtained PSAs, FH, or prior urologic history. While all PSA testing was performed within our institutional system, the algorithm does not control for possible variability associated with different iterations of labs, machines, etc. In addition, biopsy type was not uniform, and could affect detection rates. Nearly 25% of the non-referred group and 15% of the referred group did not receive repeat confirmatory PSA testing, and only 9% in the entire cohort underwent auxiliary testing. The data is further limited by patients lost to follow-up or those who continue care outside of Duke; thus, it is likely that the prevalence of csPCa was underestimated. To further optimize patient outcomes, Duke has further expanded the comprehensive multiparametric magnetic resonance imaging center, image-targeted biopsy program, e-consult platform, and the AS protocol to avoid indolent cancer overtreatment. Furthermore, more longitudinal data will possibly provide insight on the most efficacious frequency of subsequent screening, as well as optimal timing of necessary intervention.

The deployment of a system-wide PCa screening algorithm identified an important percentage of patients ages 40 to 49 years with csPCa at preliminary longitudinal follow-up. Immediate referral was a significant predictor of identifying subsequent PCa, with high OR compared to other implicated factors, confirming the impact of the algorithm. Overall, this study adds to the body of literature providing validation for a baseline PSA measurement in a man’s forties as a means of risk assessment to identify those patients harboring potentially lethal disease who could benefit from early, potentially lifesaving intervention. Continued follow-up may potentially help determine optimal timing and frequency of subsequent monitoring protocols after initial testing.