The survival outcomes of Asian men with elevated prostate-specific antigen (PSA) levels at screening are largely unknown. We present the clinical outcomes of Taiwanese men based on their screening PSA levels. Between 1994 and 2006, 27,761 men aged over 40 years underwent PSA screening in a self-funded health examination. The clinical database was linked with the national cancer and death registry databases to generate prostate cancer incidence, prostate cancer mortality (PCM) and overall mortality (OM). Participants were followed until the end of 2009. Survival analyses were performed for the participants' outcomes, and were stratified by five 10-year age strata (age 40–<50, 50–<60, 60–<70, 70–<80 and ≥80), and six age-referenced PSA percentile groups, divided by the 50th, 75th, 90th, 95th and 99th percentile of PSA values for each 10-year age stratum. The median age of the 27,761 men was 54.7 years. The median PSA level at cancer diagnosis was 4.46 ngml−1. Specifically, the PSA levels for the five 10-year age strata in order of respectively increasing ages were 1.93, 3.50, 4.10, 6.94 and 12.4 ngml−1. After a median follow-up of 8.4 years, 2,463 men died and 337 were diagnosed with prostate cancer. Among the 337 patients, 29 (8.6%) died of prostate cancer. The prostate cancer incidence, PCM and OM rates were higher in men with higher age-referenced PSA percentile values. The 10-year PCM rate for men with ≥ the 99th age-referenced PSA percentile was 3.9%, which was significantly higher than the rate of ≤ 0.5% in the lower percentile groups.
Prostate-specific antigen (PSA) is a glycoprotein produced by the prostate gland.1 The serum PSA level is normally low in men with healthy prostates.1 An elevated serum PSA may be seen in men with benign prostatic hyperplasia, prostatitis and prostate cancer.2–4 Since its introduction into medical practice, PSA testing has been evaluated for the early detection of prostate cancer.5,6 Although PSA screening has been recognized to be effective for early diagnosis, it is still debated whether PSA screening reduces prostate cancer mortality.7
Recent results from the European randomized study of screening for prostate cancer (ERSPC) have shown that PSA screening was associated with reduction of prostate cancer mortality; however it may also be associated with the overdetection and overtreatment of potentially indolent prostate cancers.8 In contrast, the prostate, lung, colorectal and ovarian (PLCO) screening trial did not detect a significant difference in prostate cancer mortality of participants screened for PSA and those who were not screened.9 However, 52% of the control participants underwent unscheduled PSA testing within a follow-up duration of 6 years, and the biopsy compliance rate (40% in control group, 52% in PSA testing group) was low compared to the biopsy rate (86%) in the ERSPC study.10 Therefore, the results of these two large studies have not resolved the issue as to whether PSA screening really benefits men in general.
Although the number of prostate cancer cases has rapidly increased in Taiwan, the incidence of prostate cancer in Taiwanese men (30 per 100,000 person-year in 2010) is significantly lower than the incidence in American and European men.11–13 Because PSA screening in asymptomatic men is not reimbursed in Taiwan, most PSA screening is performed in the setting of the general health examination. Patients with elevated PSA levels are referred to urologists for further investigation, which may lead to prostate biopsy.
The more we know about the natural course of elevated PSA levels in asymptomatic men, the better we can inform our patients about risk stratification. To our knowledge, the survival outcomes of Asian men with elevated PSA levels have not been clearly elucidated. The aim of this study was to determine the survival outcomes of men with elevated PSA levels found on screening.
Material and Methods
Patient population and enrollment criteria
Between 1994 and 2006, 34,488 men underwent serum PSA screening as part of self-funded health examinations at various collaborating hospitals, including the National Taiwan University Hospital (NTUH, n = 20,091), Chang-Gung Memorial Hospital Linkou Campus (n = 12,830), St Martin De Porres Hospital (n = 968) and Lotung Poh-Ai Hospital (n = 599). All enrolled men were Taiwanese. Patient demographics and clinical information were extracted from the electronic databases of these hospitals. After the exclusion of men with known prostate cancer and ages younger than 40 years, 27,761 men were included in the study. All men who had a PSA > 4 ngml−1 were referred to urology for investigation and possible prostate biopsy. Normally, a PSA greater than 4 ngml−1 would trigger a recommendation for prostate biopsy. If a participant had had multiple health check-ups, only his initial PSA value was included in the analysis. This study was approved by the Research Ethics Committee of NTUH. The requirement of informed consent from each participant was waived.
Clinical data and outcome measurements
Participant databases were linked to the national cancer registry and death registry databases of the Bureau of Health Promotion, Department of Health, Executive Yuan, Taiwan. The extracted data included cancer diagnosis and date until the end of 2007, as well as all causes of death and date until the end of 2009. Because NTUH had the largest number of participants in the cohort and the in-hospital cancer registry at NTUH was generated earlier than the other registries, the cancer registry at NTUH was linked until the end of 2009.
The median duration of follow-up was defined as the median of the interval between the date of the PSA test and death or censoring. Prostate-cancer-free survival (PCFS) was defined as the interval between the date of the PSA test and prostate cancer diagnosis or censoring. The overall mortality (OM) and prostate cancer mortality (PCM) refer to the interval between the date of the PSA test and death due to all causes or to prostate cancer.
The Kaplan–Meier method was used to analyze survival outcomes. Because age is a significant confounding factor for survival, the study cohort was divided into five 10-year age strata, 40 to < 50 years, 50 to < 60 years, 60 to < 70 years, 70 to < 80 years and 80 years and above. We also divided the participants in each age stratum into 6 PSA percentile groups as follows: < the 50th (median), 50th to < 75th, 75th to < 90th, 90th to < 95th, 95th to < 99th and ≥ 99th percentiles of PSA levels. The log-rank test was used to compare survival outcomes between groups. All statistical tests were two-tailed and p < 0.05 was considered statistically significant. Statistical analyses were performed using PASW 18.0 for Windows.
Population demographics and PSA distribution
The median age of the 27,761 participants was 54.7 years. About two thirds were < 60 years of age. The median serum PSA value for the entire study cohort was 0.89 ngml−1. A total of 3,442 (12.40%), 1,717 (6.18%) and 327 men (1.84%) had PSA levels >2.5, 4.0 and 10 ngml−1, respectively. The median duration of follow-up was 8.4 years. There were 337 (1.21%) men who were diagnosed with prostate cancer. The median time to cancer diagnosis was 6.6 years. Overall, 2,463 (8.87%) men died during the follow-up period. However, among these deaths, only 29 men (1.18% of those who died or 0.1% of the entire cohort) died of prostate cancer.
The percentile PSA values
The percentile PSA values of each 10-year age group are listed in Table 1. All percentile values increased with increasing age. Less than 5% of men aged 40–60 years had a PSA > 4 ngml−1. In contrast, >25% of men over 80 had a PSA > 4 ngml−1.
Table 1. The percentile values for prostate-specific antigen among healthy Taiwanese men stratified by the age
Prostate cancer diagnosis
During the study period, 10, 75, 157, 84 and 11 men were diagnosed with prostate cancer in the age groups of 40 to < 50, 50 to < 60, 60 to < 70, 70 to < 80 and ≥ 80 years, respectively. The cumulative prostate cancer incidence rates are shown in Table 2. In general, the cumulative prostate cancer incidence rate increased with age and PSA percentiles, except for men aged over 80 years, where fewer cancers were diagnosed, probably because of the small numbers of subjects and/or fewer referrals for biopsy because of age.
Table 2. Diagnosis of prostate cancer and survival outcomes in healthy Taiwanese men stratified by PSA value and age
The median PSA value at cancer diagnosis for the entire cohort was 4.46 ngml−1. Specifically, the values were 1.93, 3.50, 4.10, 6.94, and 12.4 ngml−1 for the five strata in respective order of increasing age. The cumulative prostate cancer incidence rates were higher in men in the higher PSA percentile groups (Fig. 1, all p < 0.05 by the log-rank test). Specifically, 63.3% of cancers from men in the 99th percentile were diagnosed in the first year of screening. As expected, the time to prostate cancer diagnosis was shorter in participants with higher PSAs. A repeat statistical analysis after age stratification revealed similar findings.
Prostate cancer mortality
Of the 337 men with prostate cancer, only 29 (8.6%) died of prostate cancer. Of these 29 men, 4 (13.8%), 5 (17.2%), 2 (6.9%), 3 (10.3%), 6 (20.7%) and 9 (31.0%) were from the six percentile groups in order of respective increasing PSA level. The 5- and 10-year PCM rate for men of all ages in the 99th PSA percentile group was 1.9 and 3.9%, respectively (Table 2), which were significantly higher rates than seen in the other PSA groups (Fig. 2). Although there were significant differences in PCM rates between any two PSA groups lower than the 99th percentile, these differences were minimal and possibly clinically insignificant (Fig. 2, all pairwise p < 0.001).
Although participants in the older age groups had higher overall mortality rates, an elevated PSA level did not significantly impact the OS (Fig. 3). Although there were significant differences in OS between some PSA groups (p = 0.005), most of these differences were either clinically insignificant or too inconsistent to deduce meaningful conclusions. However, during the first 8 years, participants with PSAs ≥ the 99th percentile had poorer survival than those in the other percentile groups, suggesting that prostate cancer dominated the causes of death.
The study demonstrated that a single screening PSA measurement provided information regarding prostate cancer risk and PCM. The most significant finding was that men with an age-referenced PSA of ≥ 99th percentile not only had a significantly higher prostate cancer incidence but also a higher risk of dying from prostate cancer (3.9% at 10 years) compared to the lower percentile groups. Although an increase in prostate cancer incidence was also seen in men with lower PSA levels (< the 99th percentile), PCM rates remained low (0.5%) at 10 years. Therefore, the 99th percentile of age-referenced PSA may a valuable cut-off value for counseling patients with an elevated PSA at screening.
Men aged < 80 years and with a PSA < 99th percentile had a relatively low risk of PCM at 10 years, despite a cancer incidence of up to 37.9 in some subgroups. These prostate cancers may be low-to-intermediate-risk disease. More importantly, a combination of factors may have contributed to the low PCM rates but substantial OM rates, including the long natural history of prostate cancer, the lead time between abnormal PSA and cancer progression, and the presence of comorbidities in the elderly, which might have ultimately led to the death of participants from competing causes. To minimize the overtreatment of men with prostate cancer, investigators have proposed several selection criteria for active surveillance.10,14 Accordingly, subjects in lower PSA percentile groups may undergo close follow-up and be offered curative treatments should cancer progression become obvious.10 However, for younger men with life expectancies of 10 years or longer, prostate cancer may eventually impact survival and/or quality of life.
Because of the possibility of over- or underestimating the risk of developing prostate cancer, a single threshold PSA value is not considered a sufficient reason for biopsy.7 Instead, the decision for biopsy should also take into account age, ethnicity, family history, prior biopsy, comorbidities, findings of digital rectal examination, PSA velocity, and PSA density.7 In addition, our concept of age-referenced PSA percentiles may help to individualize risk stratification, which may have an impact on the decision to perform a transrectal ultrasound biopsy and on the type of follow-up regimen. For example, men aged 60–70 years have a cumulative risk of prostate cancer of 8.7% at 10 years despite an initial PSA between 2.20 and 4.08 ngml−1 (the 75th–90th percentile). Because the cumulative incidence of prostate cancer in this age-referenced PSA subgroup was only 2.2% at 5 years, most of the prostate cancer diagnoses (about 75%) were made between 5 and 10 years after PSA screening. These men, even with a low initial PSA level, may benefit from careful follow-up 5 years after the first PSA screening if they are still in good health. In contrast, young men aged 40–50 years had a very low 10-year cumulative incidence of prostate cancer unless their PSA was greater than the 99th percentile (4.07 ngml−1). Even with a PSA level greater than 99th percentile, these 40- to 50-year-old young men had zero risk of cancer-specific death at 10 years (Table 2), suggesting that there may be a lead time bias where cancer was revealed early by screening and there was a relatively short duration of follow-up in relation to their young age. Therefore, young men with low age-referenced percentile PSA levels may not benefit from frequent PSA testing after their initial baseline screening. This study therefore supports the conclusion that the interval between the first screening PSA and the next screening should be based on the initial baseline PSA level.1,10,15
The use of age-referenced PSA percentiles to stratify patients into groups is a method that can be used to adjust for age when making clinical decisions. Although we also analyzed the PSA percentile data separately for each age group, the pooled data that combined the subjects from all age groups by percentile PSA values confirmed the trend observed for the age-stratified data, and showed greater statistical significance.
The use of age-referenced PSA percentiles can eliminate the need for an arbitrary PSA cut-off value and facilitate clinical decision-making. In addition, the age-referenced PSA thresholds proposed by Oesterling et al. that had been previously recommended for clinical decision making, were derived from a cohort of cancer-free men, with the aim of improving diagnostic sensitivity and specificity.4 Because the reduction of prostate cancer mortality using PSA screening has not been established, a biomarker that leads only to enhanced cancer detection may not be a better clinical tool than a biomarker, such as the age-referenced PSA percentile used in this study that has both diagnostic and prognostic implications.
This study has limitations. First, the study cohort may not be representative of the entire population of Taiwan. It may instead be representative of men who are proactive about the early detection of disease and are able to afford a self-funded health examination. Second, the small number of men aged 80 years or older suggests that strong inferences about this group may be inappropriate. However, our results suggested that PSA screening may be of reduced importance among men aged 80 years or older. Third, our study did not consider the impact of subsequent PSA testing, PSA velocity, clinical stage of prostate cancer, and Gleason score on clinical decisions and outcomes. Nevertheless, a single age-referenced PSA percentile value appears to be efficient for assessing the risk of prostate cancer death in Taiwanese men. In addition, many of the cohort participants who underwent health examinations did not undergo subsequent assessments of prostate size or regular health examinations; therefore, PSA density measurements or PSA velocity data were not available for many participants. Analysis of the available data from participants who had three or more PSA determinations would very likely generate significant bias, because those who had had elevated PSA levels may have undergone diagnostic procedures and treatments for prostate cancer and would not undergo other health examinations. Fourth, the duration of follow-up was short, especially for younger men. Therefore we intend to continue following our cohort with the hope that additional data will further validate and develop our approach as a useful predictive method.
A single PSA measurement in a screening setting provides useful information for assessing the risk of prostate cancer and PCM. A greater than 99th percentile age-referenced PSA was associated with increased PCM. Further validation and comparisons with traditional criteria, such as PSA velocity, PSA density and DRE findings are required before concluding that a single PSA measurement can be used as a prognostic biomarker.
The authors thank the Bureau of Health Promotion, Department of Health, Executive Yuan, Taiwan for help with linking the cohort to the national cancer registry and death registry database.