Prostate-specific antigen velocity based risk-adapted discontinuation of prostate cancer screening in elderly men

Authors


Judd W. Moul, Division of Urologic Surgery and Duke Prostate Center, Duke University Medical Center, PO Box 3707, Durham, North Carolina 27710, USA. e-mail: judd.moul@duke.edu

Abstract

Study Type – Prognostic (cohort)
Level of Evidence 2b

What’s known on the subject? and What does the study add?

Currently, the U.S. Preventive Services Task Force (USPSTF) recommends against PSA screening for prostate cancer in men aged ≥75 years as it concluded that “the harm of screening for prostate cancer in men age ≥75 years may outweigh the potential benefits”. Our findings suggest that elderly men with a PSA velocity of ≥0.45 ng/ml/year have higher risk of death from prostate cancer. Continuing PSA testing may be beneficial for these men.

OBJECTIVE

• To evaluate weather prostate-specific antigen (PSA) velocity could be used to stratify patients at risk of death from prostate cancer (PCa) and be useful in aiding decision making regarding PSA screening in elderly men, as previous studies have shown that PSA velocity can predict PCa risk.

PATIENTS AND METHODS

• The cohort included 3,525 patients aged ≥ 75 years with two or more PSA tests before a diagnosis of PCa. Cox proportional hazard model was used to evaluate which variables at time of last PSA measurement were associated with death from PCa.

• The rates of death from PCa after diagnosis in different PSA velocity groups were calculated. Kaplan-Meier and log rank test were used to assess the significant difference in death from PCa after diagnosis, stratified by PSA velocity cutoff.

RESULTS

• On multivariate analysis, men with a PSA velocity of PSA velocity ≥0.45 ng/mL/year had a 4.8-fold higher risk of death from PCa as compared to men with a PSA velocity of <0.45 ng/mL/year (p value = 0.013). After a median 6.5 (up to 16.9) years of follow-up from diagnosis, 1.4% of the men with a PSA velocity <0.45 ng/mL/year had died of PCa as compared to 8.7% of those with a PSA velocity ≥0.45 ng/mL/year.

• The cumulative rate of death from PCa after diagnosis, stratified by a PSA velocity of 0.45 ng/mL/year, was statistically different (log rank test, P < 0.001).

CONCLUSION

• Men age ≥ 75 years old with a PSA velocity of <0.45 ng/mL/year are unlikely to die of PCa. It may be safe to discontinue PSA screening in these men.

INTRODUCTION

The role of prostate-specific antigen (PSA) screening for prostate cancer (PCa) is controversial. Recent long term, randomized, controlled studies produced conflicting results as to the benefit of PSA screening as it pertains to PCa mortality [1,2]. Nevertheless, PSA screening has been widely used to detect PCa worldwide for two decades, and remains common among elderly men [3,4]. Elderly men are more likely to be diagnosed with higher-risk tumors, but less likely to receive curative local therapy, commonly receiving primary androgen deprivation therapy instead [5,6]. Currently, the U.S. Preventive Services Task Force (USPSTF) recommended against PSA screening for PCa in men aged ≥ 75 years old as it concluded that ‘the harm of screening for PCa in men age ≥ 75 years, or those with a life expectancy less than 10 years, may outweigh the potential benefits’[7]. However, men of any age have varied health and functional status, and a simple cutoff such as ‘age 75’ years does not adequately reflect patient life expectancy. Walter et al. [8] reported that the life expectancy in the 75th percentile for American men aged 75 years is 14.2 years, 10.8 years for those 80 years of age, and 7.9 years for those age 85 years. Cooperberg et al. [5] suggested that decisions regarding both screening and treatment should consider patient co-morbidities, life expectancy, and treatment preferences rather than chronologic age. Schaeffer et al. [9] reported that men aged 75 to 80 years with a PSA less than 3 ng/mL were unlikely to die of, or experience aggressive PCa during their remaining life. They suggested that PSA testing might be safely discontinued for these men. We recently reported that African American men with a PSA less than 6 ng/mL at their 75 years of age could safely discontinue PSA screening [10].

It has been demonstrated that high PSA velocity before diagnosis is significantly associated with adverse pathologic factors, including high-grade cancer, seminal vesicle invasion, positive pelvic lymph nodes, time to recurrence, and PCa-specific mortality [11,12]. In particular, data from the Baltimore Longitudinal Study on Aging reported by Carter et al. [13] suggested that a PSA velocity above 0.35 ng/mL/year, measured 10–15 years ago, might identify life-threatening PCa at a PSA < 4.0 ng/mL, when it is often curable. No study has focused on the relationship of PSA velocity and the PCa risk in men age ≥ 75 years old.

We hypothesized PSA velocity could be used to stratify PCa risk in elderly men, and may be useful in decision making about PSA screening in these men.

PATIENTS AND METHODS

After obtaining institutional review board approval, a total of 125 696 PSA values collected between August 1987 and March 2009 were retrieved from the Duke Prostate Center database. These samples were from 48 128 men including 7157 African Americans (AA), 26 780 Caucasian Americans (CA), 868 of ‘other’ races, and 13 323 without race data. Of these men, 3525, including 577 AA, 2622 CA, 288 other races, and 38 missing races men, who had two or more PSA tests within their follow-up years and an age at last PSA measurement, before diagnosis, of ≥75 years old were included in this study.

The criteria for performing a prostatic biopsy was a positive digital rectal examination or a PSA that was considered elevated based on clinician practice.

PCa status was divided into three groups: non-PCa, PCa, and high risk PCa. The definition of high risk PCa was a PSA > 20 ng/mL or Gleason score ≥ 8 or clinical stage ≥ T2c (D’Amico criterion) [14]. Death from PCa was determined by a review of the death certificate or clinical records. Briefly, to record a death as being from PCa, patients with metastatic disease after castrate-resistant had to be documented, with evidence that the PSA level was increasing at the last follow-up visit before death.

PSA velocity was calculated as the slope of linear regression line of all PSA values collected over time before diagnosis. We chose the rationale cutoff point for PSA velocity which demonstrated adequate sensitivity and specificity, while emphasizing the specificity more than sensitivity in elderly men [15]. The PSA velocity values were divided into two groups, based on the rationale cutoff determined using the receiver operating characteristic (ROC) curve: <0.45 and ≥0.45 ng/mL/year.

ROC analyses were performed for high risk PCa (yes or no) using PSA velocity as a discriminator. The Mann-Whitey U test was used to compare parameters expressed as median between PCa and non-PCa groups and between PCa and high risk PCa groups.

Univariate and multivariate analysis with Cox proportional hazard model were used to determine which variables at time of last PSA measurement were associated with death from PCa. Kaplan-Meier and log rank test were used to assess the significant difference in death from PCa after diagnosis, stratified by a PSA velocity of 0.45 ng/mL/year. Statistical analyses were performed using SPSS program (version 17.0, Chicago, IL). All statistical tests were 2-side with P < 0.05 considered statistically significant.

RESULT

The rates of PCa and high risk PCa in the study cohort were 13.7% and 4.5%, respectively. Compared to men without a diagnosis of PCa, men with a diagnosis of PCa were statistically significantly older at the time of last PSA measurements (P= 0.018), had higher first PSA values (P < 0.001), higher last PSA values (P < 0.001), higher PSA velocities (P < 0.001), and longer follow-up in years from first PSA to last PSA measurement (P < 0.001). Conversely, men with a diagnosis of PCa were younger at the time of last PSA measurements (P= 0.022), had lower first PSA values (P < 0.001), lower last PSA values (P < 0.001), and lower PSA velocities (P < 0.001) when compared to men with a diagnosis of high risk PCa (Table 1).

Table 1.  The characteristics of the study population
VariablesTotalPCaNon-PCaHigh risk PCaComparisons (Mann-Whitney U test)
No PCa vs PCaPCa vs high risk PCa
  1. PCa indicates prostate cancer; PSA indicates prostate-specific antigen; SD indicates standard deviation; IQR indicates interquartile range; aindicates chi-square test.

No. of subjects (%)3525483 (13.7)3042 (86.3)157 (4.5)  
Age at last PSA (years)      
 Mean (SD)80.1 (4.1)79.7 (4.0)80.1 (4.1)80.4 (4.4)  
 Median (IQR)79.1 (76.8–82.4)78.7 (76.6–81.7)79.2 (76.8–82.6)80.0 (76.9–82.6)P= 0.018P= 0.022
Race    P < 0.001aP= 0.001a
 Other races (%)28814 (4.9)274 (95.1)3 (1.0)  
 African American (%)577106 (18.4)471 (81.6)51 (8.8)  
 Caucasian American (%)2622355 (13.5)2267 (86.5)101 (3.9)  
 Missing races388 (21.1)30 (78.9)2 (5.3)  
First PSA (ng/mL)      
 Mean (SD)6.5 (52.0)19.5 (108.4)4.5 (35.3)47.7 (186.4)  
 Median (IQR)2.0 (0.9–4.4)4.7 (2.1–11.2)1.8 (0.9–3.7)12.7 (4.4–26.6)P < 0.001P < 0.001
Last PSA (ng/mL)      
 Mean (SD)  11.1 (115.7)44.5 (285.1)5.8 (49.3)120.4 (491.7)  
 Median (IQR)2.2 (1.0–5.1)8.1 (3.1–18.5)1.9 (0.9–4.1)26.7 (12.2–59.6)P < 0.001P < 0.001
First to last PSA (years)      
 Mean (SD)6.0 (4.7)4.5 (3.9)6.3 (4.8)5.0 (4.1)  
 Median (IQR)5.2 (2.0–9.3)3.6 (1.1–7.1)5.5 (2.0–9.6)4.3 (1.3–8.1)P < 0.001P= 0.065
PSA velocity (ng/mL/year)      
 Mean (SD)3.5 (4.1)9.8 (4.8)2.5 (4.0)20.7 (6.1)  
 Median (IQR)0.04 (0.00–0.28)0.6 (0.0–2.6)0.02 (0.0–0.2)2.6 (0.5–9.2)P < 0.001P < 0.001
Counts of PSA measurement      
 Mean (SD)6.0 (4.8)6.0 (4.9)6.0 (4.8)7.0 (5.6)  
 Median (IQR)4.0 (2.0–8.0)4.0 (2.0–8.0)4.0 (2.0–8.0)5.0 (2.0–10.0)P= 0.552P= 0.015

The areas under the ROC curve (AUC) for PSA velocity in predicting high risk PCa was 0.841. The AUC was high enough to warrant considering PSA velocity sufficiently sensitive and specific for predicting high risk PCa. We emphasized specificity more than sensitivity in elderly men. A PSA velocity cutoff point of 0.45 ng/mL/year was chosen because it optimized sensitivity and specificity. The cut-point of 0.45 ng/mL/year corresponded to a sensitivity of 78% and a specificity of 83%, respectively (Table 2).

Table 2.  The performance characteristics (sensitivity, specificity) for select PSA velocity cutoff points in prediction of high risk PCa
PSA velocity (ng/mL/year)Sensitivity (%)Specificity (%)
  1. PSA indicates prostate-specific antigen. When specificity was emphasized more than sensitivity in elderly men, PSA velocity 0.45 ng/mL/year was found to be the optimal cutoff.

Cutoffs  
0.108662
0.208372
0.308179
0.457883
0.757189
1.006691
2.005795
3.004996

In the univariate model, men with a PSA velocity ≥ 0.45 ng/mL/year had a 6.8-fold relative risk of death from PCa compared to that of PSA velocity <0.45 ng/mL/year (P= 0.002). In the multivariate model (stratified be PSA velocity, race, last PSA, last PSA age, and last PSA year), PSA velocity ≥ 0.45 ng/mL/year remained an independent factor in predicting death from PCa (hazard ration 4.8, P= 0.013) (Table 3). The cumulative rate of death from PCa after diagnosis, stratified by a PSA velocity of 0.45 ng/mL/year, was statistically different (log rank test, P < 0.001) (Fig. 1).

Table 3.  Univariate and multivariate analysis of variables at time of last PSA measurement in predicting death from PCa with Cox proportional hazard model
VariableHazard ratio (95% CI)P
  1. PCa indicates prostate cancer; CI indicates confidential interval; PSA indicates prostate-specific antigen.

Univariate  
 PSA velocity (ng/mL/year)  
 <0.45reference 
 ≥0.456.8 (2.0–22.7)0.002
Multivariate  
 PSA velocity (ng/mL/year)  
 <0.45reference 
 ≥0.454.8 (1.4–16.6)0.013
 Race  
  Other racesreference 
  African American4.3 (0.0–6.4)0.936
  Caucasian American2.1 (0.0–3.1)0.942
  Missing races1.1 (0.0–4.6)0.998
  Last PSA (ng/mL)1.0 (1.0–1.0)0.957
  Last PSA Age1.1 (1.0–1.2)0.060
  Last PSA year > 2000reference 
  Last PSA year ≤ 20009.1 (1.3–62.7)0.001
Figure 1.

Kaplan-Meier estimate of cumulative rate of death from prostate cancer after diagnosis, as stratified by PSA velocity of 0.45 ng/mL/year (log rank test, P < 0.001).

Of 483 men diagnosed with PCa, 470 had follow up information. The rates of death from all-causes, metastasis disease, and death from prostate cancer were 32.6% (153/470), 8.7% (41/470), and 5.5% (26/470). After a median of 6.5 (up to 16.9) years of follow-up from diagnosis, only 1.4% (3/207) of the men with a PSA velocity < 0.45 ng/mL/year died of PCa whereas 8.7% (23/263) of those with a PSA velocity ≥ 0.45 ng/mL/year died of PCa.

DISCUSSION

The results of our study (area under the ROC curve) showed PSA velocity was effective in predicting high risk PCa and was in line with the previous study [16]. Morgan et al. [15] reported that sensitivity and specificity should not be weight equally in clinical practice. In a given patient, it is important to take into account associated risk factors and coexisting conditions. The cutoff points used with younger men should emphasized sensitivity, whereas the points in men more than 70 years old, men with a limited life expectancy, or both should emphasized specificity. When emphasizing specificity more than sensitivity in elderly men, we found a PSA velocity of 0.45 ng/mL/year was the rationale cutoff point for predicting high risk PCa. We then used this cutoff point to stratify patients at risk of death from PCa. As expected, we found that PSA velocity of ≥0.45 ng/mL/year was a strong and independent predictor of death from PCa in both univariate and multivariate analyses.

Recently, Carter et al. [13] reported that PSA velocity measured 10–15 years before diagnosis was associated with cancer-specific survival 25 years later. Furthermore, men with a PSA velocity above 0.35 ng/mL per year had a higher relative risk of PCa death than men with a PSA velocity of ≤0.35 ng/mL/year. The association between PSA velocity and PCa risk in our study was in agreement with the report by Carter et al. [13] There were some discrepancies between these two studies that resulted in different cutoffs being used to stratify PCa risk. In our study, we included men whose last PSA measurement was at ≥75 years of age and calculated the real time PSA velocity to predict high risk PCa. The 75th percentile for initial PSA value in our study cohort was 4.4 ng/mL. Carter et al. [13] however, used PSA velocity measurements from 10–15 years prior to predict PCa risk in men with an initial PSA value of <4.0 ng/mL. Furthermore, because of the high correlation between PSA and PSA velocity, including PSA level in the multivariate model may affect the independence of PSA velocity predicting PCa risk [17]. PSA level was not included in the multivariate model in the study by Carter et al. [13]. However, Etzioni et al. [18] reported that studies assessing independent prognostic information of markers should provide adjusted odds ratios for both markers and treat both markers in the same fashion (continuous or dichotomized at a cut point) for a fair and equally powerful comparison. We included PSA as a continuous variable point in the multivariate model, PSA velocity ≥ 0.45 ng/mL/year remained an independent predictor of high risk PCa. Lastly, in the study by Carter et al. [13], the total number of deaths from PCa was very small (only 12 deaths from PCa), they did not precise evaluate the risk of death from PCa based on the determined PSA velocity cutoff point of 0.35 ng/mL/year. We further evaluated death from PCa based on the threshold of 0.45 ng/mL/year. After a median of 6.5 (up to 16.9) years follow-up from diagnosis, we found that only 1.4% (3/207) of men with a PSA velocity < 0.45 ng/mL/year died of PCa whereas 8.7% (23/263) those of with a PSA velocity ≥ 0.45 ng/mL/year had died of PCa. To our knowledge, this is the first study evaluating high risk PCa and death from PCa in men age ≥ 75 years old based on PSA velocity. As expected, the results from our study indicate that a PSA velocity cutoff point of 0.45 ng/mL/year could be used to stratify PCa risk in elderly men. These results may be used to design a strategy for PSA screening in elderly men.

Our study confirmed that PSA velocity is associated with PCa risk as reported in previous studies [10,11,19,20]. However, data from the PCa Prevention Trial (PCPT) and European Randomized Study of Screening for PCa (ERSPC) (Rotterdam section) showed that PSA velocity was not useful for predicting PCa risk since it was not an independent predictor of PCa risk when PSA was included in the multivariate models [21,22]. There are several possible explanations for this disparity. First, all the men in these two trials underwent a biopsy whereas only a part of the cohorts underwent a biopsy in other reports. The remaining part of the cohorts is assumed to not have PCa, which is most likely not true. The verification bias that is inherent to this type of analysis will result in a different conclusion. Second, many men in these trials were diagnosed with indolent PCa as all men underwent a biopsy at the end of the PCPT regardless of PSA level. This is another possible reason that PSA velocity was not associated with PCa risk. Third, it is feasible that the PSA velocity was affected by different PSA levels. However, all subjects in the PCPT had an initial PSA value ≤ 3 ng/mL and subjects with PSA ≤ 3 ng/mL were excluded from the Rotterdam section of ERSPC [21,22]. However, unscreened cohorts in other studies always included the whole PSA range. Additionally, the study cohorts in the PCPT and ERSPC mostly consisted of men aged 55–75 years, whereas other studies included various age ranges. This may result in the discordant conclusions as age has been shown to be related to the PCa predictive value of PSA velocity [23]. Lastly, the distribution of PSA velocity based on PCa status in our study cohort was much different from that of the ERSPC, Rotterdam section, which further explained the discrepancy of PSA velocity in prediction of PCa risk.

The present study has several limitations. First, it is retrospective in nature. Prospective studies about the relationship between the PSA velocity and PCa risk in elderly men are warranted. Furthermore we used data from a large, tertiary medical center. This is partially counteracted by the size of the cohort, which must be taken into consideration and may mitigate this limitation. Next, we did not evaluate the impact of family history of PCa and digital rectal examination although an abnormal digital rectal examination was sometimes utilized to trigger a biopsy. However, the results from the ERSPC indicated that the overall characteristics of detected cases of PCa, differed very little based on how they were detected (elevated PSA, digital rectal examination or transrectal ultrasound) [24]. Finally, this study is not population-based. It is possible that our results overestimate the risk of death from PCa.

CONCLUSION

Our findings suggest that men age ≥ 75 years old who have PSA velocity of <0.45 ng/mL/year are unlikely to die of PCa. It may be safe to discontinue PSA screening in these men, which could reduce the costs associated with screening and the potential morbidity of additional treatment.

ACKNOWLEDGEMENTS

The authors acknowledge the Guangzhou First Municipal People’s Hospital, Guangzhou Medical College, Guangdong, China, for support of Ping Tang MD to conduct the research at Duke University Medical Center.

Financial support: Supported by research funds from the Committee for Urologic Research, Education, and Development (CURED) of Duke University (LS, and JWM).

CONFLICT OF INTEREST

None declared.

Ancillary