Beyond Baseline: Serial PSA Measurements and Risk Stratification for Prostate Cancer

1. Introduction

Prostate‐specific antigen (PSA) testing remains a key method for prostate cancer screening, but a single PSA measurement lacks specificity and may result in false positives. The U.S. Preventive Services Task Force (USPSTF) has emphasized the need for long‐term follow‐up data to refine PSA screening strategies. Kovac et al. proposed that a baseline PSA level below 1.00 or 2.00 ng/mL could be sufficient to reduce or discontinue further PSA testing [1]. However, baseline PSA levels can be influenced by non‐cancer conditions, including obesity, inflammation, comorbidities, and medication use [2, 3, 4, 5]. The core principle of utilizing multiple PSA measurements is to accurately identify cases requiring immediate treatment, thus preventing overdiagnosis and unnecessary treatments. To better inform screening practices, we evaluated PSA trajectories over a 16‐year follow‐up before prostate cancer diagnosis by race and cancer status.

2. Methods

We analyzed data from 34,756 men (1713 Black and 33,043 White) in the Prostate, Lung, Colorectal, and Ovarian Cancer Screening Trial (PLCO). PSA measurements were evaluated over a 16‐year follow‐up before prostate cancer diagnosis. Participants aged 55–74 from the U.S. general population were randomized to the intervention arm or the control arm at 10 clinical centers between November 1993 and July 2001 [6]. Serial PSA values (T0–T5) were obtained at baseline and follow‐up visits. We calculated the time in years from each PSA test to either diagnosis of prostate cancer or trial exit. The PSA trajectories were analyzed using generalized estimating equation (GEE) with gamma distribution and log‐link function, accounting for within‐subject clustering and right‐skewness via PROC GEE in SAS 9.4. We also used a Cox proportional hazards model to evaluate baseline PSA as a covariate and longitudinal PSA measurements (16‐year follow‐up before prostate cancer diagnosis) as a time‐varying covariate.

3. Results

GEE models revealed that the three‐way (time × prostate cancer status × race) interaction was not significant (rate ratio (RR) = 1.054, 95% CI 0.975–1.140). However, significant two‐way interaction effects on PSA levels were found: (time × race) and (time × prostate cancer status). Compared to White men, Black men were expected to increase the mean PSA by 1.3% (RR = 1.013, 95% CI 1.001, 1.026) for each year closer to diagnosis (Table 1). Compared to men without prostate cancer, men with prostate cancer were expected to increase the mean PSA by 10.2% (RR = 1.102, 95%CI 1.085–1.120) for each year closer to diagnosis (Figure 1). Factors associated with reduced mean PSA were body mass index, history of chronic conditions such as diabetes, liver disease, and heart disease. Older age, Black men, frequent urination (3 or more/night), benign prostatic hyperplasia (BPH), and prostate cancer were associated with increased mean PSA. The baseline PSA was associated with a 0.6% increase in prostate cancer hazards (HR = 1.006, 95% CI: 1.005–1.006), while incorporating longitudinal PSA measurements resulted in a greater 16.8% increase in prostate cancer hazards (OR = 1.168, 95% CI: 1.141–1.195).

Table 1.

The change in the mean of PSA from the GEE Gamma regression model.

			95% Confidence limits			95% Confidence limits
Variable		B	LCL	UCL	RR	LCL	UCL
Age		0.025	0.022	0.028	1.026	1.023	1.029
Race	Black	0.179	0.045	0.312	1.196	1.046	1.367
	White	Ref
College degree	Yes	0.028	0.001	0.056	1.028	1.001	1.057
	No	Ref
Married	Yes	−0.030	−0.063	0.003	0.970	0.939	1.003
	No	Ref
BMI Baseline		−0.012	−0.015	−0.008	0.988	0.985	0.992
Frequent urination	Yes	0.106	0.056	0.156	1.112	1.058	1.168
	No	Ref
Immediate Family	Yes	0.040	0.000	0.080	1.041	1.000	1.083
	No	Ref
BPH	Yes	0.246	0.217	0.275	1.279	1.242	1.317
	No	Ref
Bronchitis	Yes	−0.054	−0.109	0.000	0.947	0.897	1.000
	No	Ref
Diabetes	Yes	−0.167	−0.207	−0.127	0.846	0.813	0.881
	No	Ref
Hypertension	Yes	0.008	−0.019	0.035	1.008	0.981	1.035
	No	Ref
Liver disease	Yes	−0.068	−0.119	−0.018	0.934	0.888	0.982
	No	Ref
Heart disease	Yes	−0.096	−0.130	−0.063	0.908	0.878	0.939
	No	Ref
PCa status	Yes	1.234	1.084	1.383	3.433	2.956	3.988
	No	Ref
T		0.027	0.026	0.029	1.028	1.026	1.030
T × Race = Black		0.013	0.001	0.025	1.013	1.001	1.026
T × Race = White		Ref
T × PCa status =Yes		0.097	0.082	0.113	1.102	1.085	1.120
T × PCa status = No		Ref
PCa status = Yes* Race = Black		0.399	−0.287	1.084	1.490	0.751	2.956
PCa status = No* Race = White		Ref
T*PCa status = Yes* Race = Black		0.053	−0.026	0.131	1.054	0.975	1.140
T*PCa status = No* Race = White		Ref

Note: Boldface indicates statistical significance (p = 0.05).

Abbreviations: BPH, benign prostatic hyperplasia; PCa, prostate cancer; PSA, prostate‐specific antigen; T, Time in years before diagnosis.

Figure 1.

The change in the mean of PSA by time, race, and cancer status. PSA, prostate‐specific antigen. [Color figure can be viewed at wileyonlinelibrary.com]

4. Discussion

This study provides evidence that PSA levels change over time and differ significantly by prostate cancer status and race. Consistent with prior studies [2, 3, 4, 5], we found that the PSA levels are lower in men with non‐cancer conditions such as comorbidities and BPH. Long‐term follow‐up shows significant promise for assessing prostate cancer progression. With up to 60% of PSA‐detected prostate cancers being overdiagnosed [7], it is essential for effective screening to include serial PSA measurements and consider patient factors like race and pre‐existing conditions to prevent both overdiagnosis and subsequent overtreatment.

Disclosure

The contents of this publication are the sole responsibility of the authors and do not necessarily reflect the views, opinions, or policies of Uniformed Services University of the Health Sciences (USU), the Department of War, the Defense Health Agency, the Departments of the Army, Navy, or Air Force, nor the U.S. Government. Mention of trade names, commercial products, or organizations does not imply endorsement by the U.S. Government. This study was prepared by a military service member and an employee of the U.S. Government as part of the individuals' official duties. Therefore, it is in the public domain and does not possess copyright protection under Title 17, USC, § 101 (however, as a courtesy, it is requested that USU and the author be given an appropriate acknowledgment).

Conflicts of Interest

The authors declare no conflicts of interest.

Data Availability Statement

Requests to access the PLCO datasets should be directed to the National Cancer Institute (NCI) Cancer Data Access System (CDAS): https://cdas.cancer.gov/plco/.