Fax: (617) 632-2270
Prostate-specific antigen velocity and the detection of gleason score 7 to 10 prostate cancer
Version of Record online: 20 SEP 2007
Copyright © 2007 American Cancer Society
Volume 110, Issue 9, pages 1973–1978, 1 November 2007
How to Cite
Punglia, R. S., Cullen, J., McLeod, D. G., Chen, Y. and D'Amico, A. V. (2007), Prostate-specific antigen velocity and the detection of gleason score 7 to 10 prostate cancer. Cancer, 110: 1973–1978. doi: 10.1002/cncr.23014
- Issue online: 18 OCT 2007
- Version of Record online: 20 SEP 2007
- Manuscript Accepted: 5 JUN 2007
- Manuscript Revised: 30 MAY 2007
- Manuscript Received: 25 APR 2007
- cancer detection;
- Gleason score;
- prostate cancer;
- prostate-specific antigen
An increasing prostate-specific antigen (PSA) velocity is associated with a shorter survival after local therapy for prostate cancer. In this study, the authors evaluated whether PSA velocity was associated with prostate cancer detection and grade at diagnosis after adjusting for established predictors.
Between January 1989 and December 2003, 914 men who had PSA levels ≥4 ng/mL were identified by using the Center for Prostate Disease Research (CPDR) multicenter national database, including 541 men who were diagnosed with prostate cancer. Multivariable logistic regression analyses were performed that included continuous variables (PSA velocity and level, number of prior negative biopsies, and age) along with categorical variables (ethnicity and family history) were used to identify the factors associated with prostate cancer detection and grade.
An increasing PSA velocity was associated with Gleason scores from 7 to 10 versus Gleason scores form 2 to 6 or no cancer (adjusted odds ratio [OR], 1.04 ng/mL per year; 95% confidence interval [95% CI], 1.003–1.085 ng/mL per year; P = .035). This finding was not evident in patients who had prostate cancers with Gleason scores between 2 and 6 or for any prostate cancer. PSA level was associated with the detection of any prostate cancer (OR, 1.06 ng/mL; 95% CI, 1.03–1.10 ng/mL; P = .004) and Gleason score ≤6 prostate cancer (OR, 1.06 ng/mL; 95% CI, 1.02–1.10 ng/mL; P = .0027); however, in the presence of PSA velocity, PSA no longer remained independently predictive for high-grade prostate cancer (OR, 1.01 ng/mL; 95% CI, 0.98–1.04 ng/mL; P = .45) after adjusting for established predictors.
PSA velocity enhanced the detection of high-grade cancer in men who had PSA levels >4 ng/mL. These findings, in conjunction with life expectancy, may be used when deciding which men should not be recommended for prostate biopsy despite a PSA level >4 ng/mL. Cancer 2007. © 2007 American Cancer Society.
Factors associated with an increased risk of detecting prostate cancer1 include advancing age, African-American ethnicity, positive family history, a palpable nodule on digital rectal examination (DRE), and an increasing prostate-specific antigen (PSA) level. Using the current paradigm of recommending a prostate biopsy when a specific PSA threshold is achieved (eg, 4 ng/mL) in some patients leads to a diagnosis of prostate cancer that, if left undiagnosed (and thus untreated), would not have resulted in a shortened survival, particularly in elderly men with comorbid illnesses. This phenomenon, known as “over diagnosis,”2 triggers the initiation of therapy, often in the form of surgery or radiation, leading to the subsequent risk of genitourinary and/or gastrointestinal side effects that may have an adverse impact on quality of life. Therefore, determining all clinical factors that are associated significantly with an increased risk of detecting high-grade prostate cancer that, if left undiagnosed (and thus untreated), would result in prostate cancer death and a shorter survival could lead to an improvement in the recommendations for prostate biopsy and, subsequently, less over diagnosis.2
There is a general consensus that disease progression and death from cancer are observed most commonly in men who are diagnosed with higher grade prostate cancers.3 In addition, there is some evidence to suggest that treatment of these cancers with a radical prostatectomy (RP) lengthens survival, particularly among men aged <65 years, compared with watchful waiting.4 Therefore, the detection and treatment of these clinically significant cancers could improve survival.
It has been demonstrated that a change in the serum PSA level during the year before diagnosis (PSA velocity) is associated significantly with adverse pathologic factors, including high-grade cancer identified at prostatectomy, seminal vesicle invasion, positive pelvic lymph nodes, time to PSA recurrence, and prostate cancer-specific and all-cause mortality after surgery5 or after external beam radiation therapy.6 However, although a change in the PSA level during the year before diagnosis had clinical utility as a prognostic factor, PSA velocity did not contribute independent information for the prediction of high-grade disease in men from the Prostate Cancer Prevention Trial, in which the PSA level at study entry study was required to be <3 ng/mL.7
These findings differ from early studies of PSA velocity in men with higher PSA levels, in whom higher PSA velocity was a strong predictor of newly diagnosed prostate cancer.8, 9 A recent report from the Prostate, Lung, Colorectal, and Ovarian Cancer Screening Trial indicated that there was a significant association between PSA velocity and the presence of cancers with Gleason scores from 7 to 10 among men who were diagnosed with cancer.10 However, the association of PSA velocity in men with higher PSA levels in conjunction with detection of high-grade cancer has not been assessed among men who have not yet been diagnosed with cancer. Therefore, the objective of the current study was to evaluate whether the PSA rise during the 1.5 years before biopsy was associated with prostate cancer detection and grade at diagnosis in men with entry PSA levels >4 ng/mL. We used data from the Center for Prostate Disease Research (CPDR) multicenter national database.
MATERIALS AND METHODS
Between January 1, 1989 and December 31, 2003, after excluding men who had a palpable nodule on DRE, 914 men with serum PSA levels ≥4 ng/mL and who had ≥2 PSA measurements were identified at 1 of 9 participating CPDR multicenter national database medical centers. All patient data were collected as part of ongoing CPDR database activities. This database has been described previously and includes men who were referred for urologic evaluation.11 Of the 914 biopsies, 541 cancers (59.2%) were identified. To be eligible for study entry, the Institutional Review Board at each clinical center approved the study, and all men gave written informed consent. All men had serial PSA testing, had at least a 6-core needle biopsy reviewed, and were assigned a Gleason score by a pathologist with expertise in genitourinary cancers.12 We determined that, among men with PSA levels >4 ng/mL, the vast majority (86.9%) underwent prostate biopsy, thereby limiting the influence of verification bias.13
Each participant in the study had at least 2 prebiopsy PSA levels obtained at least 6 months apart and within 18 months of diagnosis using the same assay. These PSA values permitted the calculation of the PSA velocity using linear regression.14 Therefore, all men had to have a minimum of 2 PSA values to be considered in our population, and each PSA value had to be obtained at least 3 months apart from the other values. Only values that were obtained before the first positive or last negative biopsy were considered, and PSA values that were obtained after the first positive or last negative biopsy were not considered in the velocity calculation. If the PSA level dropped during the 18 months before the most recent biopsy that was performed because of a nodule on the DRE, then the PSA velocity was assigned a value of 0 ng/mL per year. The “diagnositic PSA” was defined as the PSA value that was obtained closest to the first positive or last negative biopsy. In addition to the PSA velocity, the patient's ethnicity (Caucasian and other, African American, or missing), prostate cancer family history (yes, no, or missing), PSA level (ng/mL), number of prior negative biopsies (count), and the patient's age at the time of the most recent biopsy (years) were evaluated.
Three dependent outcomes were examined using multivariable logistic regression: 1) cancer versus no cancer; 2) cancer with biopsy Gleason scores from 2 to 6 versus no cancer; and 3) cancers with biopsy Gleason scores from 7 to 10 versus either Gleason scores ≤6 or no cancer.14 The covariates examined included the continuous variables of diagnostic PSA (ng/mL) and PSA velocity (ng/mL per year), number of prior negative biopsies (count), and age (years) at the time of the most recent biopsy; and the categorical variables included ethnicity and prostate cancer family history. The examined versus referent groups for the categorical covariates were as follows: for ethnicity, either 1) African American or 2) missing ethnicity information compared with Caucasian ethnicity; and prostate cancer in at least 1 first degree relative compared with 1) no family history of prostate cancer or 2) missing family history. The adjusted odds ratio (OR) with 95% confidence interval (95% CI) for each of the biopsy Gleason score endpoints (cancer, Gleason score 2–6 cancer, and Gleason score 7–10 cancer) were calculated for each covariate.
Analyses were conducted using SAS statistical software (version 9.1 for Windows, SAS Institute Inc., Cary, NC). Statistical significance was defined as a 2-sided P value <.05.
Prevalence of Cancer and Cancer-specific Grades
Of 914 men, 541 (59.2%) were diagnosed with prostate cancer; and, of those men, 167 (18.3%) had high-grade disease (Gleason 7–10). Prostate cancer was identified at the first biopsy attempt in 272 of the 541 affected men (29.8% of all 914 men) and in 150 (16.4%), 66 (7.2%), and 53 (5.8%) men after 1, 2, 3 or more prior negative biopsies, respectively. The distribution of Gleason scores at diagnosis in the 541 men who were diagnosed with prostate cancer was ≤6 in 374 men (69%), 7 in 128 men (24%), and 8 to 10 in 39 men (7%) (Table 1).
|Characteristic||Mean ± SD||Median [IQ range]||No. of patients (%)|
|Age at biopsy, y||64.6 ± 7.5||65 [60–70]|
|Diagnostic PSA, ng/mL||8.42 ± 6.2||6.55 [5.1–9.3]|
|PSAV, ng/mL/y||2.20 ± 5.8||1.64 [0.51–3.2]|
|No. of previous biopsies||1.04 ± 1.26||1 [0–2]|
|African American||195 (21.3)|
|Caucasian and other||652 (71.3)|
|Biopsy Gleason sum|
|No cancer||373 (40.8)|
|Cancer detection among men with cancer of any grade|
Adjusted Odds of Detecting any Prostate Cancer
After adjustment for established predictors, PSA velocity, when described as a continuous variable, was not associated significantly (OR, 0.99; 95% CI, 0.96–1.02; P = .60) with the detection of any prostate cancer. This OR represents the odds of detecting cancer (vs no cancer) per 1 unit increase in PSA velocity (ie, for each 1 ng/mL increase per year). In contrast, PSA level (OR, 1.06; 95% CI, 1.03–1.10; P = .004), age (OR, 1.02; 95% CI, 1.001–1.04; P = .04), number of prior negative biopsies (OR, 0.77; 95% CI, 0.68–0.86; P = .012), family history (OR, 1.27; 95% CI, 1.12–1.41; P = .001), and race (African American vs Caucasian: OR, 1.52; 95% CI, 1.05–2.19; P = .006; missing vs Caucasian: OR, 0.70; 95% CI, 0.41–1.21; P = .045) were associated significantly with the detection of prostate cancer. At a PSA velocity threshold of 0.75 ng/mL per year, sensitivity for any prostate cancer detection was 71.1%, and specificity was 41.8%. Sensitivity dropped to 64.5% and specificity increased to 41.8% at a threshold of 1.0 ng/mL per year.
Adjusted Odds of Detecting Gleason Score 2 to 6 Prostate Cancer
After adjustment for established predictors, PSA velocity (OR, 0.98; 0.94–1.02; P = .28) and age (OR, 1.01; 0.99–1.03; P = .39) were not associated significantly with the detection of Gleason score 2 to 6 prostate cancers versus no cancer. PSA level (OR, 1.06; 1.02–1.10; P = .0027), the number of prior negative biopsies (OR, 0.72; 0.62–0.83; P < .0001), family history (OR, 1.55; 1.01–2.38; P < .0001), and race (African American vs Caucasian: OR, 1.54; 95% CI, 1.04–2.28; P = .016; missing vs Caucasian: OR, 0.78; 95% CI, 0.43–1.42; P = .13) were associated significantly with the detection of low-grade prostate cancer. At a PSA velocity threshold of 0.75 ng/mL per year, sensitivity for any prostate cancer detection was 69.0%, and specificity was 32.7%. Sensitivity dropped to 60.2% and specificity increased to 36.7% at a threshold of 1.0 ng/mL per year.
Adjusted Odds of Detecting Gleason Score 7 to 10 Prostate Cancer
After adjustment for established predictors, PSA velocity (OR, 1.04; 95% CI, 1.003–1.08; P = .035) was associated significantly with the detection of Gleason score 7 to 10 cancers (vs Gleason score 2–6 cancer or no cancer) along with patient age (OR, 1.05; 95% CI, 1.02–1.07; P = .0003). After the addition of the variable describing PSA velocity, the PSA level itself was no longer a significant predictor of high-grade prostate cancer (OR, 1.01; 95% CI, 0.98–1.04; P = .45). In addition, the number of prior negative biopsies (OR, 0.90; 0.78–1.05; P = .17), family history (OR, 1.09; 95% CI, 0.93–1.27; P = .31), and race (African American vs Caucasian: OR, 1.12; 95% CI, 0.73–1.71; P = .39; missing vs Caucasian: OR, 0.79; 95% CI, 0.37–1.69; P = .45) were not associated significantly with the detection of high-grade cancer. The area under the curve for this logistic model was 0.636 and was 0.578 for PSA velocity alone. At a PSA velocity threshold of 0.75 ng/mL per year, sensitivity for any prostate cancer detection was 77.8%, and specificity was 31.9%. Sensitivity dropped to 74.3% and specificity increased to 38.3% at a threshold of 1.0 ng/mL per year. For illustrative purposes, the sensitivity and specificity of PSA velocity at other thresholds for the detection high-grade prostate cancer are shown in Table 2.
|PSAV threshold, ng/mL/y||Sensitivity, %||Specificity, %||PPV, %||NPV, %|
In addition, Table 3 shows the distribution of men with high-grade cancer and men without high-grade cancer separated into 3 PSA groups and PSA velocity groups so that the diagnostic tests can be compared more readily. Using a PSA velocity threshold of 1.25 ng/mL per year, 114 of 167 cancers were detected (sensitivity, 68.3%), but it also led to biopsy for 421 men with Gleason score 2 to 6 cancer or without cancer (specificity, 43.6%). To achieve the same level of sensitivity for high-grade cancers using PSA, the threshold would have to be set at 5.75 ng/mL. However, at this threshold, 468 men without high-grade disease would undergo biopsy (specificity, 37.4%) or 47 more men without high-grade disease would have undergone biopsy in our dataset.
|PSAV category||Diagnostic PSA category: No. of patients (%)|
|Total PSA, ≥4 and <7 ng/mL||Total PSA, ≥7 and <10 ng/mL||Total PSA, ≥10 ng/mL|
|Gleason score 7–10||27 (14.5)||8 (14.8)||2 (5.7)|
|Gleason score <6/no cancer||159 (85.5)||46 (85.2)||33 (94.3)|
|Between ≥0.75 and <1.25||86||14||4|
|Gleason score 7–10||15 (17.4)||0 (0)||1 (25)|
|Gleason score <6/no cancer||71 (82.6)||14 (100)||3 (75)|
|Gleason score 10–7||44 (18.6)||27 (20)||43 (26.4)|
|Gleason score <6/no cancer||193 (81.4)||108 (80)||120 (73.6)|
It has been demonstrated that the change in PSA level during the year before diagnosis (PSA velocity) is associated significantly with lymph node and seminal vesicle invasion, high-grade cancers at prostatectomy, and a shorter time to PSA recurrence in addition to cancer-specific and all-cause mortality after either surgery5 or external beam radiation therapy.6 Given the prognostic significance of PSA changes before diagnosis, we evaluated whether this factor could aid in the identification of high-grade prostate cancer when considering factors that have known, significant associations with prostate cancer detection.1 We observed that, as the PSA level before diagnosis increased, there was a significant increase in the detection of Gleason score 7 to 10 cancers after adjusting for the known predictors of prostate cancer detection in men with PSA values >4 ng/mL. It is noteworthy that, once PSA velocity was added to the model predicting high-grade cancers, the absolute PSA level no longer was predictive.
The potential clinical implications of this study are especially important for men with shorter life expectancies in whom a biopsy may not be considered despite a PSA level >4 ng/mL. For men of advanced age (eg, >75 years) with benign prostatic hyperplasia or in younger men with significant comorbid illness (eg, medically inoperable, unstable angina), diagnosing and treating a low-grade prostate cancer may have minimal impact, if any, on life expectancy but could lead to a decrement in quality of life because of the potential side effects of prostate cancer therapy. Indeed, a stark reminder of the problem of over diagnosis comes from a recent study showing that older men who were diagnosed with prostate cancer had a risk of suicide that was 4.24 times that of men in an age- and sex-matched control group.15 Those results help to provide the basis on which to enhance the identification of men at high risk for high-grade cancers using PSA velocity in addition to age, DRE status, and PSA level.
A few considerations warrant further discussion. Current standards of practice advocate recommending prostate biopsy when PSA velocity exceeds 0.75 ng/mL per year.8 However, we have demonstrated that using a threshold of 1.0 ng/mL per year or 1.25 ng/mL per year sacrifices little sensitivity in terms of detection of high-grade disease and reduces the number of biopsies in which no cancer or low-grade disease is identified. However, to determine the ideal threshold for PSA velocity above which biopsy should be recommended to optimize the detection of high-grade disease, the performance characteristics of the PSA velocity would need to be estimated using a receiver operator characteristic methodology adjusted for possible verification bias and interaction with PSA level.11 In addition, because the patients in our study came from a urologic referral population, we limited our sample to men who had PSA levels >4 ng/mL to avoid the presence of verification bias confounding our results. Therefore, consequently, our power to detect interactions with PSA level was limited concomitantly. Future studies of PSA in larger databases could help determine the optimal PSA velocity threshold for biopsy and its interaction with patient age and PSA level. Finally, we have demonstrated that PSA velocity enhances the detection of high-grade cancer in men with PSA levels that exceed 4 ng/mL. Whether PSA velocity can enhance the detection of high-grade cancer in men with a normal DRE and a PSA level <4 ng/mL will require further study.
In conclusion, the current results indicated that PSA velocity enhances the detection of high-grade cancer in men with a PSA level >4 ng/mL more than the absolute PSA level alone. Because low-grade disease is likely to be insignificant in men with shorter life expectancy, these findings can be used in conjunction with life expectancy when deciding which men should not be recommend for prostate biopsy, despite a PSA level >4 ng/mL, helping to mitigate the problem of over diagnosis in prostate cancer.
- 12Histologic grading and staging of prostatic carcinoma. In: TannenbaumM, ed. Urologic Pathology. Philadelphia, Pa: Lea & Febiger; 1977: 171–187., .
- 14Simultaneous inferences and other topics in regression analyses-1. In: KutnerM, ed. Applied Linear Regression Models. Homewood, Ill: Richard D. Irwin; 1983: 150–153.