SEARCH

SEARCH BY CITATION

Keywords:

  • prostate cancer;
  • prostate-specific antigen;
  • predictor;
  • mortality

Abstract

  1. Top of page
  2. Abstract
  3. MATERIALS AND METHODS
  4. RESULTS
  5. DISCUSSION
  6. Acknowledgements
  7. CONFLICT OF INTEREST DISCLOSURES
  8. REFERENCES

BACKGROUND:

A diagnosis of prostate cancer is not often predictive of death from prostate cancer because of competing causes of mortality. Identification of the risk of death from prostate cancer and death from all causes using information available at the time of baseline prostate-specific antigen (PSA) measurement appears to be particularly pertinent.

METHODS:

The Duke Prostate Center database was used to identify men who had their PSA level measured over the past 20 years. The Cox proportional hazards model was used to assess whether baseline PSA, race, and age at baseline PSA could predict death from prostate cancer and death from all causes after baseline PSA measurement. The receiver operating characteristic (ROC) curve was performed to analyze the accuracy of baseline PSA as a continuous variable in predicting death from prostate cancer.

RESULTS:

A total of 4568 men diagnosed with prostate cancer after baseline PSA measurement were included. On multivariate analysis, baseline PSA levels of 4.0 to 9.9 ng/mL and ≥10 ng/mL were associated with significantly higher rates of death from prostate cancer compared with PSA levels <2.5 ng/mL. An advanced age at baseline PSA and African American race were associated with a higher death rate from prostate cancer and death from all causes. The area under the ROC curve for baseline PSA predicting death was 0.839. When a baseline PSA of 10 ng/mL was chosen to predict death from prostate cancer, the corresponding sensitivity and specificity were 77% and of 78%, respectively.

CONCLUSIONS:

Baseline PSA appears to be a reliable and independent predictor of death from prostate cancer. A baseline PSA of ≥4 ng/mL has been associated with higher risk of death from prostate cancer. Cancer 2010. © 2010 American Cancer Society.

Aside from skin cancer, prostate cancer is the most commonly diagnosed new cancer among men and the second most common cause of cancer-related death in men in the United States. In 2009, an estimated 192,280 new cases of prostate cancer and 27,360 deaths from prostate cancer were expected to occur.1 Although screening for prostate cancer using prostate-specific antigen (PSA) has been widely used for 2 decades, until recently, the benefits of screening had not been proven.2 The results of 2 large randomized, controlled trials were recently released,3, 4 but due to conflicting results, the utility of PSA screening in reducing prostate cancer mortality remains hotly debated.

Methods to assess the risk of prostate cancer, including tools that integrate multiple risk factors, are available and can improve the effectiveness of prostate cancer detection when PSA is used in these models.5 Gann et al first reported that a single PSA measurement was associated with a risk of a prostate cancer diagnosis within 10 years.6 Lilja et al demonstrated that a baseline PSA measurement at age 44 to 50 years was predictive of subsequent long-term prostate cancer detection for up to 25 years.7 The authors suggested that a baseline PSA measurement could be used for risk stratification for prostate cancer screening programs. However, the most important goal of PSA screening is to reduce prostate cancer-specific mortality. Only a small portion of the men diagnosed with prostate cancer will die of the disease, even when diagnosed with high-risk prostate cancer.2 The ability to assess the risk of death from prostate cancer based on baseline PSA values would be useful and strategically enhance the utility of PSA screening. Given the long natural history of prostate cancer and evidence that a diagnosis of prostate cancer or high-risk prostate cancer is not often predictive of death from prostate cancer because of competing causes of mortality,2 the evaluation of death from prostate cancer and death from all causes is particularly pertinent.

We hypothesized that baseline PSA can be used to stratify patients and predict risk of death from prostate cancer and death from all causes in the population targeted for PSA measurement. We then evaluated the relation between baseline PSA and death from prostate cancer and death from all causes. We also assessed whether other variables at baseline PSA measurement (race and age at baseline PSA) could be predictive of death from prostate cancer and death from all causes.

MATERIALS AND METHODS

  1. Top of page
  2. Abstract
  3. MATERIALS AND METHODS
  4. RESULTS
  5. DISCUSSION
  6. Acknowledgements
  7. CONFLICT OF INTEREST DISCLOSURES
  8. REFERENCES

Patient Selection

After obtaining Institutional Review Board approval, a total of 34,805 subjects had their baseline PSA measurement between August 1987 and March 2009 with information regarding race retrieved from the Duke Prostate Center database. Of these subjects, 15,093 (43.4% 3030 African American [AA] subjects, 11,595 Caucasian American [CA] subjects, and 468 “other” races) had a follow-up time of <6 months after baseline PSA measurement and 225 (1.5%) subjects were diagnosed with prostate cancer. These subjects were excluded from the study. This left 19,712 subjects (56.6% 4127 AA subjects, 15,185 CA subjects, and 400 subjects of “other” races) with a follow-up time of ≥6 months after baseline PSA measurement, and 4568 (23.2%) were diagnosed with prostate cancer. Only 4568 subjects diagnosed with prostate cancer with a follow-up time of ≥6 months after baseline PSA measurement were included in the study. Serum PSA was measured using the Tandem-R immunoradiometric assay (Tandem-E assay before 2000) and the Access assay (Hybritech-Beckman, San Diego, CA). The criteria for performing a prostatic biopsy was a positive digital rectal examination or a PSA level that was considered elevated based on clinician practice.

The clinical variables included in the study were race (AA, CA, and “other” races), age at baseline PSA (continuous), and baseline PSA as a continuous and categoric variable (0.0-2.4 ng/mL, 2.5-3.9 ng/mL, 4.0-9.9 ng/mL, and ≥10.0 ng/mL).

Follow-Up

Follow-up began from the date of baseline PSA measurement and was continued until death or date of last visit. Many subjects in our database had only 1 PSA measurement and no further follow-up, and the majority of the subjects were not diagnosed with prostate cancer. To eliminate the probable bias induced by these subjects, we only included those subjects who were diagnosed with prostate cancer and had a follow-up time of ≥6 months in the current study. Death from prostate cancer was determined by a review of the death certificate or clinical records. Briefly, to record a death as being from prostate cancer, patients with metastatic disease after castrate-resistant prostate cancer had to be documented, with evidence that the PSA level was increasing at the time of the last follow-up visit before death.

Statistical Analyses

The receiver operating characteristic (ROC) curve was used to analyze the accuracy of baseline PSA as a continuous variable in predicting death from prostate cancer. The Kaplan-Meier method and log-rank test were used to assess the 5-year, 10-year, and 15-year actuarial prostate cancer-specific and all-causes survival rates and significant difference in death from prostate cancer among various baseline PSA groups. Multivariate Cox proportional hazards models were used to determine which variables (baseline PSA as a continuous or categoric variable, age at baseline PSA as a continuous variable, and race as a categoric variable) were predictive of death from prostate cancer and death from all causes. Statistical analyses were performed using SPSS statistical software (version 17.0; SPSS Inc, Chicago, IL). All statistical tests were 2-sided with a P < .05 considered to be statistically significant.

RESULTS

  1. Top of page
  2. Abstract
  3. MATERIALS AND METHODS
  4. RESULTS
  5. DISCUSSION
  6. Acknowledgements
  7. CONFLICT OF INTEREST DISCLOSURES
  8. REFERENCES

The demographic characteristics of the study cohort are shown in Table 1. For the 4568 subjects, the mean and median age at the time of the baseline PSA measurement was 65.4 years and 65.8 years, respectively; the mean and median baseline PSA level was 36.3 and 4.5 ng/mL, respectively; and the mean and median follow-up time after baseline PSA measurement was 9.5 years and 9.9 years, respectively. The rates of death from prostate cancer and death from all causes were 3.5% (162 of 4568 subjects) and 20.2% (923 of 4568 subjects), respectively (Table 1). The rates of death from prostate cancer in subjects who had their baseline measurement taken before 1990, between 1991 and 1998, and after 1999 were 5.8% (22 of 381 subjects), 4.9% (130 of 2660 subjects), and 0.7% (10 of 1527 subjects), respectively. The rates of death from all causes in subjects who had their baseline measurement taken before 1990, between 1991 and 1998, and after 1999 were 31.0% (118 of 381 subjects), 25.2% (670 of 2660 subjects), and 8.8% (135 of 1527 subjects), respectively (Table 2).

Table 1. Clinical Characteristics of the Study Cohort (N = 4568)
CharacteristicNo.%
  1. PSA indicates prostate-specific antigen; SD, standard deviation; IQR, interquartile range.

Subjects4568 
Age at baseline PSA, y
Mean (SD)65.4 (9.6)
Median (IQR)65.8 (58.9-72.1)
Range21.2-93.1
Baseline PSA, ng/mL
Mean (SD)36.3 (275.1)
Median (IQR)4.5 (1.7-9.7)
Range0.0-7399.0
Race
Caucasian American343975.3
African American106223.2
Other races671.5
Year of baseline PSA
≤19903818.3
1991-1998266058.2
≥1999152733.5
Years after baseline PSA measurement
Mean (SD)9.5 (4.8)
Median (IQR)9.9 (5.3-13.5)
Range0.5-21.2
Death from prostate cancer1623.5
Death from all-causes92320.2
Table 2. Characteristics of Study Population Stratified by Year of Baseline PSA Measurement
 Total (n)≤19901991-1998≥1999
No.%No.%No.%
  1. PSA indicates prostate-specific antigen; SD, standard deviation; IQR, interquartile range; Tx, tumor stage of biopsy was unknown.

Prostate cancer4568381 2660 1527 
Death from prostate cancer162225.81304.9100.7
Death from all causes92311831.067025.21358.8
Baseline PSA, ng/mL
Mean (SD)36.3 (275.1)26.0 (118.3)38.2 (279.8)35.5 (294.3)
Median (IQR)4.5 (1.7-9.7)4.5 (1.6-11.7)4.6 (1.8-11.0)4.4 (1.7-7.8)
Range0.0-7399.00.0-1313.00.0-7399.00.0-6280.1
Tumor stage of biopsy
T1251318660.2150263.682566.5
T210818427.264627.435128.3
T3156154.91004.2413.3
T4160247.81124.7241.9
Tx65872 300 286 

On multivariate analysis, baseline PSA (either as a continuous or categoric variable), age at baseline PSA, and race were found to be predictors of death from prostate cancer. Subjects with a baseline PSA of 4.0 to 9.9 ng/mL and ≥10 ng/mL were found to have a 3.0-fold and 11.5-fold, respectively, higher rate of death from prostate cancer compared with subjects with a baseline PSA of <2.5 ng/mL. AA race and advanced age at the time of baseline PSA measurement was associated with a higher rate of death from prostate cancer and death from all causes (Table 3). The portion of all-causes mortality by prostate cancer substantially increased with baseline PSA value (Table 4).

Table 3. Multivariate Analysis of Variables at Time of Baseline PSA Measurement in Predicting Death from Prostate Cancer and Death From All Causes With the Cox Proportional Hazards Model (N = 4568)
 Death From Prostate CancerDeath From All-Causes
HR (95% CI)PHR (95% CI)P
  1. PSA indicates prostate-specific antigen; HR, hazard ratio; 95% CI, 95% confidence interval.

Variable
Baseline PSA as a continuous variable
Baseline PSA, ng/mL1.0 (1.0-1.1)<.0011.0 (0.9-1.0).254
Age at baseline PSA, y1.1 (1.1-1.2)<.0011.1 (1.1-1.2)<.001
Race
Caucasian AmericanReference Reference 
African American3.5 (2.5-4.7)<.0011.4 (1.2-1.6)<.001
Other races0.5 (0.1-3.9).5271.0 (0.5-2.0).887
Baseline PSA as a categoric variable
Baseline PSA, ng/mL
0.0-2.4Reference Reference 
2.5-3.91.6 (0.5-4.8).4360.8 (0.6-1.0).045
4.0-9.93.0 (1.4-6.7).0070.9 (0.8-1.1).502
≥1011.5 (5.5-23.8)<.0011.1 (0.9-1.3).420
Age at baseline PSA, y1.1 (1.0-1.2)<.0011.1 (1.1-1.2)<.001
Race
Caucasian AmericanReference Reference 
African American2.7 (2.0-3.7)<.0011.3 (1.2-1.6)<.001
Other races0.8 (0.1-5.9).8451.1 (0.6-2.1).843
Table 4. Twenty One-Year Actuarial Competing Risk Estimate of Number of Total Deaths, Prostate Cancer-Specific Deaths, and Deaths Not Caused by Prostate Cancer After Baseline PSA Measurement (n = 4568)
Baseline PSA, ng/mLNo. of Subjects (%)Estimated Deaths (%)
TotalCaused by PCaNot Caused by PCaPortion of All-Cause Mortality by PCa (%)
  1. PSA indicates prostate-specific antigen; PCa, prostate cancer.

0.0-2.41482 (32.4)259 (28.1)8 (4.9)251 (33.0)3.1
2.5-3.9575 (12.6)79 (8.6)5 (3.1)74 (9.7)6.3
4.0-9.91405 (30.8)240 (26.0)25 (15.4)215 (28.3)10.4
≥101106 (24.2)345 (37.4)124 (76.5)221 (29.0)35.9
All patients456892316276117.6

The 5-year, 10-year, and 15-year prostate cancer-specific survival rates were 99.8%, 98.7%, and 89.9%, respectively (Fig. 1A). The 5-year, 10-year, and 15-year overall all-causes survival rates were 97.4%, 88.8%, and 60.8%, respectively (Fig. 1B). The cumulative rate of death from prostate cancer, stratified by various baseline PSA levels, was found to be statistically different (log-rank test, P < .001) (Fig. 2). The area under the ROC curve (AUC) for baseline PSA as a continuous variable in predicting death from prostate cancer was 0.839 (Fig. 3). The sensitivity and specificity of different baseline PSA cutoffs on the ROC curve for predicting death from prostate cancer are shown in Table 5.

thumbnail image

Figure 1. Actuarial Kaplan-Meier (A) prostate cancer-specific and (B) all-causes estimated risk of death among patients are shown after baseline prostate-specific antigen (PSA) measurement.

Download figure to PowerPoint

thumbnail image

Figure 2. Kaplan-Meier estimate of the cumulative incidence of death from prostate cancer is shown stratified by baseline prostate-specific antigen (PSA) levels (log-rank test, P < .001).

Download figure to PowerPoint

thumbnail image

Figure 3. The area under the receiver operating characteristic curve (AUC) for baseline PSA as a continuous variable in predicting death from prostate cancer is shown.

Download figure to PowerPoint

Table 5. Performance Characteristics (Sensitivity and Specificity) for Select Baseline PSA Cutoffs in Predicting Death From Prostate Cancer (N = 4568)
Baseline PSA Cutoffs, ng/mLSensitivity, %Specificity, %
  1. PSA indicates prostate-specific antigen.

1.09818
2.09629
3.09239
4.09247
5.09156
6.08963
7.08568
8.08372
9.07875
10.07778
11.07580
12.07182
13.07083
14.06984
15.06885

DISCUSSION

  1. Top of page
  2. Abstract
  3. MATERIALS AND METHODS
  4. RESULTS
  5. DISCUSSION
  6. Acknowledgements
  7. CONFLICT OF INTEREST DISCLOSURES
  8. REFERENCES

The most important result from the current study was that baseline PSA was a reliable predictor of death from prostate cancer. The AUC was 0.839 for baseline PSA in predicting death from prostate cancer. This was high enough to warrant considering baseline PSA to be sufficiently sensitive and specific for predicting death from prostate cancer. When a baseline PSA of 10.0 ng/mL was chosen to predict death from prostate cancer, the corresponding sensitivity and specificity were 77% and 78%, respectively. Fall et al reported that baseline PSA is not an accurate predictor of death from prostate cancer. In their study, the best cutoff point of baseline PSA on the ROC curve was found to be 10 ng/mL, which produced a sensitivity and specificity of only 69% and 58%, respectively. Their conclusions were based on 267 patients with untreated localized prostate cancer who were managed by watchful waiting.8 However, the current study cohort included 4568 subjects with different tumor stages who were being treated with multiple different treatment regimens. These differences between the current study and the previous study may be the rationale for explaining the differences in results regarding the accuracy of baseline PSA in predicting prostate cancer-specific mortality.

The use of PSA measurement as a screening tool to detect prostate cancer has increased dramatically in the United States since its introduction in 1988.9 However, to the best of our knowledge, the benefit of using PSA, digital rectal examination, or any other screening test for detecting prostate cancer is still unclear. The results of 2 large randomized, controlled trials were recently released,3, 4 but due to conflicting results, the utility of PSA screening in reducing prostate cancer mortality remains hotly debated. Currently, the American Urological Association (AUA), American Cancer Society, and the US Preventive Services Task Force recommend PSA screening for men with a life expectancy of >10 years because the harms of screening for prostate cancer in men with a life expectancy of <10 years may outweigh the benefits.10-12 However, Cooperberg and Konety suggested that decisions regarding both screening and treatment should consider patient comorbidities, life expectancy, and treatment preferences.13 Recently, Schaeffer et al reported that baseline PSA could be used to stratify the patients with regard to PSA screening. They suggested that PSA measurement might be safely discontinued for men ages 75 to 80 years except for those men with a PSA ≥3 ng/mL, in whom PSA measurement should continue.14 In the current study, we found that, after approximately 20 years of follow-up, men with a baseline PSA <4 ng/mL had a very low risk of death from prostate cancer (0.63%), whereas those with a baseline PSA ≥4 ng/mL had a relatively much higher risk of death from prostate cancer (5.93%). These findings indicate that men with a higher baseline PSA are at higher risk of prostate cancer-specific mortality, which is in agreement with previous studies that decisions regarding discontinuation of screening for prostate cancer should consider patient comorbidities, life expectancy, and PSA levels rather than a simple chronologic age.13, 14

Preliminary studies have suggested that a baseline PSA is associated with the risk of subsequent prostate cancer detection.6, 7 Kuller et al reported that baseline PSA was a strong predictor of death from prostate cancer with 6 to 25 years of follow-up.15 However, the banked serum sample for analyzing PSA in that study was collected before the introduction of PSA. Degradation of PSA in archived serum can introduce measurement bias in predicting prostate cancer unless specific procedures are used for serum processing and storage.7 Furthermore, the subjects included in that study were between the ages of 35 to 57 years (mean age, 46 years), which is much younger compared with populations traditionally targeted for PSA measurement. Recently, Connolly et al evaluated a cohort of 68,354 men (mean age, 65.2 years), and demonstrated that baseline PSA was positively associated with risk of prostate cancer-specific mortality.16 However, the maximum follow-up time for their study was only 10 years, which limits the value of that report. Prostate cancer is a disease with a long natural history; a diagnosis of prostate cancer or high-risk prostate cancer is not often predictive of death from prostate cancer because of competing causes of mortality. However, the previous 2 studies did not evaluate the relation between baseline PSA and death from all causes.15, 16

An important result of the current study was that baseline PSA was an independent predictor of subsequent death from prostate cancer. Baseline PSA was found to be associated with death from prostate cancer, which was in agreement with previous preliminary studies.15, 16 Among this cohort targeted for PSA measurement with >20 years of follow-up, we found that a baseline PSA level ≥4 ng/mL was associated with higher prostate cancer-specific mortality. These results indicated that men with a baseline PSA of ≥4 ng/mL had a significant risk of death from prostate cancer. Men who entered the healthcare system with a low baseline PSA had a corresponding low rate of prostate cancer-specific mortality in long-term follow-up. The new AUA and National Comprehensive Cancer Network guidelines for the best use of PSA are in favor of men receiving a baseline PSA assessment at age 40 years, in addition to using this baseline PSA to dictate the intensity of subsequent screening.17, 18 Data from the current study suggest that the baseline PSA value may also be correlated with long-term prognosis. The results of the current study also provide new information regarding the association between the variables at the time of baseline PSA measurement and death from all causes. Multivariate evaluation indicated that age at baseline PSA and AA race were predictive of both death from prostate cancer and death from all causes. These results indicated that baseline PSA, age at baseline PSA, and race could be used to stratify patients at risk of death from prostate cancer. Furthermore, age at baseline PSA measurement and race also could be used to stratify patients at risk of death from all causes.

The current study is limited in that the longest follow-up in our study cohort was 21 years. It is unknown whether the association between baseline PSA and death from prostate cancer and death from all causes may change with an even longer follow-up. Second, we did not evaluate the impact of a family history of prostate cancer and digital rectal examination, although an abnormal digital rectal examination was occasionally used to trigger a biopsy. However, the results from the European Randomized Screening of Prostate Cancer study indicated that the overall characteristics of the prostate cancer cases differed very little with regard to how they were detected (PSA level, digital rectal examination, or transrectal ultrasound).19 In addition, the current study was a retrospective study. 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 the limitation. Lastly, this study was not population-based. It is possible that the results overestimate the risk of death from prostate cancer.

Conclusions

Baseline PSA is a reliable and independent factor in predicting death from prostate cancer. A baseline PSA level ≥4 ng/mL was found to be associated with higher prostate cancer-specific mortality. Men with a baseline PSA of <4 ng/mL were found to have a very low risk of death from prostate cancer. The results of the current study would appear to support a strategy of encouraging healthy men with a long life expectancy to obtain a baseline “risk assessment” PSA at an early age to avoid a high baseline PSA level.

Acknowledgements

  1. Top of page
  2. Abstract
  3. MATERIALS AND METHODS
  4. RESULTS
  5. DISCUSSION
  6. Acknowledgements
  7. CONFLICT OF INTEREST DISCLOSURES
  8. REFERENCES

We thank the Guangzhou First Municipal People's Hospital, Guangzhou Medical College, Guangdong, China, for its support of Ping Tang, MD, to conduct the research at Duke University Medical Center.

CONFLICT OF INTEREST DISCLOSURES

  1. Top of page
  2. Abstract
  3. MATERIALS AND METHODS
  4. RESULTS
  5. DISCUSSION
  6. Acknowledgements
  7. CONFLICT OF INTEREST DISCLOSURES
  8. REFERENCES

Supported by research funds from the Committee for Urologic Research, Education, and Development (CURED) of Duke University (to L.S. and J.W.M.); the Department of Defense (to S.J.F.); and the American Urological Association Foundation Astellas Rising Star in Urology Award (to S.J.F.).

REFERENCES

  1. Top of page
  2. Abstract
  3. MATERIALS AND METHODS
  4. RESULTS
  5. DISCUSSION
  6. Acknowledgements
  7. CONFLICT OF INTEREST DISCLOSURES
  8. REFERENCES