- Email: [email protected]
Prostate Size Does Not Predict High Grade Cancer
Oncology: Prostate/Testis/Penis/Urethra
Prostate Size Does Not Predict High Grade Cancer Tin C. Ngo,* Simon L. Conti, Rajesh Shinghal and Joseph C. Presti, Jr.† From the Department of Urology, Stanford University School of Medicine, Stanford and Division of Urology, Santa Clara Valley Medical Center (RS), San Jose, California
Purpose: Several radical prostatectomy series have linked small prostates with high grade cancer based on the hypothesis that a small prostate results from a low androgen milieu that selects for less hormone dependent, more aggressive tumors. We previously reported that this association resulted from ascertainment bias from the performance characteristics of prostate specific antigen rather than from tumor biology in our radical prostatectomy cohort. In this study we analyzed this association in a more generalized population of men who underwent prostate needle biopsy. Materials and Methods: The prostate needle biopsy database at our institution was queried for all initial biopsies. Included patient characteristics were age, race, family history of prostate cancer, prostate specific antigen, abnormal digital rectal examination and prostate volume in ml on transrectal ultrasound. Multivariate logistic regression was used to determine the influence of prostate volume on the odds of high grade cancer. Results: The study population included 1,295 patients during 2000 to 2010, of whom 582 (44.9%) had prostate cancer and 398 (30.7%) had high grade cancer. When all patients were pooled, the OR for high grade cancer was 0.85 (95% CI 0.78 – 0.92) for each 10 ml increase in prostate volume. When patients were divided by clinical T stage, the corresponding ORs for those with T1c disease was 0.83 (95% CI 0.74 – 0.93) and for those with T2 or greater disease it was 0.99 (0.98 –1.00). Conclusions: The association between small prostates and high grade cancer exists only in men with clinical T1c (normal digital rectal examination) prostate cancer. It likely resulted from ascertainment bias due to the performance characteristics of prostate specific antigen rather than tumor biology.
Abbreviations and Acronyms DRE ⫽ digital rectal examination PSA ⫽ prostate specific antigen SPND ⫽ Stanford Prostate Needle Biopsy Database Submitted for publication June 3, 2011. Study received Stanford University School of Medicine institutional review board approval. * Correspondence: Department of Urology, Stanford University School of Medicine, 300 Pasteur Dr., S-287, Stanford, California (telephone: 408-813-7049; FAX: 408-503-0050; e-mail: tin. [email protected]). † Financial interest and/or other relationship with PrognosDx and Myriad.
For another article on a related topic see page 707.
Key Words: prostate, organ size, prostate-specific antigen, prostatic neoplasms, prognosis SEVERAL groups have reported an association between small prostates and an increased risk of adverse outcomes such as high grade cancer, extraprostatic extension, positive surgical margins and biochemical recurrence in men undergoing radical prostatectomy.1–3 To explain this observation some investigators advanced the theory that a small prostate acts as a surrogate marker for an androgen depleted milieu that selects for more aggressive,
less hormone dependent tumors. To support this reasoning they cited prior studies linking low pretreatment serum testosterone to poor outcomes in men with newly diagnosed metastatic prostate cancer.4 In a recently published study of 1,404 patients from the radical prostatectomy database at our institution we reported that the association between small prostates and adverse pathological features, particularly the
0022-5347/12/1872-0477/0 THE JOURNAL OF UROLOGY® © 2012 by AMERICAN UROLOGICAL ASSOCIATION EDUCATION
Vol. 187, 477-481, February 2012 Printed in U.S.A. DOI:10.1016/j.juro.2011.10.042
AND
RESEARCH, INC.
www.jurology.com
477
478
PROSTATE SIZE DOES NOT PREDICT HIGH GRADE CANCER
volume and percent of high grade cancer, exists only in patients with clinical T1c prostate cancer.5 This association was not observed in men undergoing radical prostatectomy for clinical T2 disease. This disparity most likely represented ascertainment bias between the 2 groups due to PSA performance characteristics. If tumor biology drove the size-grade association, one would expect to see it in each group. A limitation of making inferences from the results of radical prostatectomy series is selection bias. Cohorts that undergo surgery represent a subset of men who have cancer on prostate needle biopsy, who represent a larger subset than those who undergo surgery. To address this limitation of radical prostatectomy series we studied the association of prostate size and high grade cancer in a contemporary group of patients from SPND.
MATERIALS AND METHODS SPND is an institutional review board approved, prospectively maintained database of all prostate needle biopsies performed at our institution. We reviewed SPND for all extended scheme (12 cores or more) biopsies since database inception. Patients with an existing diagnosis of prostate cancer, eg those on active surveillance, those evaluated for local recurrence after radiotherapy and those undergoing biopsy to confirm a diagnosis made elsewhere, were excluded from analysis. Those with prior negative biopsies were also excluded since the detection of cancer and high grade cancer in this patient group differs significantly from that in men undergoing initial biopsy. Patient characteristics used in our analysis included age in years, race (white, black, Asian or other), family history of prostate cancer (yes/no), abnormal DRE (yes/no) and prostate volume in ml ascertained by volume (V) estimates based on transrectal ultrasound measurements using the formula, V ⫽ /6 ⫻ length ⫻ width ⫻ height in cm. High grade prostate cancer, defined as a positive biopsy with a Gleason score of 7 or higher, was modeled as a dichotomous variable (yes/no). Table 1. Demographics of patients with prostate cancer on prostate needle biopsy
Median age (IQR) No. race (%): White Asian Black Other No. prostate Ca family history (%): Yes No Median ng/ml PSA (IQR) Median ml prostate vol (IQR) No. high grade Ca (%): Yes No
Normal DRE
Abnormal DRE
65
(59–71)
67
(62–75)
257 42 30 22
(75.4) (12.3) (5.9) (6.5)
171 40 9 21
(70.1) (16.6) (3.7) (8.7)
p Value ⬍0.0001 0.21
0.65 82 (24.1) 259 (76.0) 5.8 (4.5–8.4) 40 (27–53)
54 (22.4) 187 (77.6) 6.9 (4.3–15.0) 34 (26–49)
197 144
201 40
(57.8) (42.2)
(83.4) (16.6)
0.02 0.02 ⬍0.0001
Biopsy result by DRE status. Light blue indicates normal, medium blue low grade cancer and dark blue high grade cancer.
The Wilcoxon rank sum test was used to compare medians between different groups. The chi-square test was used to compare frequencies between different groups. Multivariate logistic regression was done to model the odds of high grade prostate cancer as a function of patient prostate size, PSA, age, race and family history of prostate cancer. Since we did not having missing variables, a complete case analysis was performed. To completely control for the effect of clinical T stage we generated separate models for men with normal (clinical T1c) and abnormal (clinical T2 or higher) DRE. The OR of each predictor is reported with the Wald 95% CI. Statistical analysis was done with SAS® 9.2. Tests for significance were 2-sided with p ⱕ0.05 considered significant.
RESULTS Table 1 lists the demographics of our study cohort. A total of 1,435 patients who were never diagnosed with prostate cancer underwent an extended prostate biopsy scheme during 2000 to 2010. We excluded 140 men who underwent a prior negative biopsy, leaving 1,295 as the study population. The figure shows biopsy results stratified by DRE findings. As expected, the overall positive biopsy rate and the risk of finding high grade cancer were higher in men with abnormal DRE and these observations were statistically significant. Of these men 582 (44.9%) had cancer and 398 (30.7%) had high grade cancer. Of the 582 men with cancer 341 (58.6%) had normal DRE and 241 (41.4%) had abnormal DRE. Although there was a statistically significant difference in median age between the 2 groups, a difference of 2 years likely had no clinical significance. Differences in the distribution of race and family history of prostate cancer between the 2 groups were not statistically significant. Median PSA was lower in men with normal DRE in a statistically significant manner. Conversely median prostate volume was higher in those with normal
PROSTATE SIZE DOES NOT PREDICT HIGH GRADE CANCER
DRE in a statistically significant manner, likely since these patients underwent prostate needle biopsy due only to increased PSA. Thus, it was more likely that benign enlargement of the prostate partially contributed to the PSA increase. To analyze the association between prostate size and the odds of high grade prostate cancer we first determined the unadjusted impact of prostate size on high grade prostate cancer using univariate logistic regression. The crude OR of each 10 ml increase in prostate volume was 0.84 (95% CI 0.79 – 0.89). Multivariate logistic regression was then used to control for potential covariates, including age, PSA, race and family history. The adjusted OR of each 10 ml increase in prostate volume was 0.85 (95% CI 0.78 – 0.92). Thus, when all patients with prostate cancer were analyzed together, there was an association between smaller prostates and high grade cancer on univariate and multivariate analysis. To completely control for the effect of clinical T stage we then created separate models for men with clinical T1c and T2 or greater prostate cancer. Age, PSA and prostate size were statistically significant predictors of high grade prostate cancer in men with clinical T1c prostate cancer (table 2). However, of those with clinical T2 or greater prostate cancer the only statistically significant predictor of high grade prostate cancer was PSA. Prostate size lost significance in this patient subset (table 2).
DISCUSSION The controversy over the association of a small prostate and high grade prostate cancer goes back more
Table 2. Predictors of high grade in patients with clinical T1c and T2 or greater prostate cancer
Prostate size/10 ml Log PSA Age Race: White (referent) Asian Black Other Family history Prostate size/10 ml Log PSA Age Race: White (referent) Asian Black Other Family history
OR (95% CI)
p Value
T1c 0.83 (0.74–0.93) 3.25 (1.82–5.82) 1.07 (1.03–1.10)
0.001 ⬍0.0001 ⬍0.0001
1.00 0.91 (0.41–1.99) 1.73 (0.62–4.83) 0.91 (0.35–2.34) 1.10 (0.62–1.95) T2 or greater 0.99 (0.98–1.00) 2.24 (1.44–3.49) 1.02 (0.97–1.06)
— 0.81 0.30 0.84 0.75
1.00 0.54 (0.19–1.50) 0.55 (0.09–3.42) 0.84 (0.21–3.28) 0.47 (0.21–1.04)
— 0.24 0.52 0.80 0.06
0.09 0.0003 0.51
479
than a decade to a report by Karakiewicz et al, who noted no association between prostate volume and Gleason grade in men undergoing sextant biopsy.6 The Prostate Cancer Prevention Trial reignited this debate with the finding that high grade cancer was more prevalent in men treated with finasteride and who consequently had a smaller prostate.7 To address this concern Kulkarni et al compared the effect of prostate volume on Gleason grade in the prostate needle biopsy and final surgical specimen of men undergoing radical prostatectomy.8 They observed that the size-grade association existed in the biopsy cohort but not in the radical prostatectomy cohort, suggesting that the effect may be due to sampling bias. To counter this Kassouf et al reported that among men undergoing radical prostatectomy at their institution those with prostate volume 50 ml or greater had a higher incidence of low grade cancer.9 In our study the association between a small prostate and high grade prostate cancer was observed in men with clinical T1c but not T2 or greater prostate cancer. These results agree with findings from the previously published study from our institution that analyzed the size-grade association in men undergoing radical prostatectomy.5 The characteristics of patients with clinical T1c and clinical T2 or greater prostate cancer are quite different. The clinical T1c group underwent prostate needle biopsy only due to increased PSA. Thus, we believe that the size-grade association previously reported in multiple radical prostatectomy series resulted from ascertainment bias rather than from tumor biology. If smaller prostate volume were a surrogate marker for tumor aggressiveness or led to higher grade cancer, one would expect this relationship to exist regardless of clinical stage. Since this association is quite strong, when patients with clinical T1c disease were commingled with those with a higher clinical T stage, the association was statistically significant on univariate and multivariate analysis. However, when the effect of clinical stage was completely controlled for by stratifying our analysis by clinical T stage, it became clear that the driver of the size-grade association was the subset of patients with clinical T1c disease. The question then becomes what is responsible for this ascertainment bias. An explanation may be the improved performance characteristics of PSA for high grade disease. For any given PSA a smaller prostate will have higher PSA density than a larger prostate and, thus, be more likely to harbor a high grade tumor.10,11 On separate analysis PSA density was a strong independent predictor of high grade cancer in men with palpable and with nonpalpable disease (p ⫽ 0.0002). We interpret this to mean that when controlling for prostate size, PSA is the
480
PROSTATE SIZE DOES NOT PREDICT HIGH GRADE CANCER
real driver behind high grade disease. Thus, observing that smaller prostates are more likely to have high grade cancer may simply be a consequence of the performance characteristics of PSA instead of tumor biology. Another explanation may lie with the observation that the clinical T1c group has a larger prostate on average and so the group may be enriched with men who primarily have benign prostatic hyperplasia but also an incidental low grade tumor. Our findings directly contradict the results of several published radical prostatectomy series. However, since this cohort was based on a group of men undergoing prostate needle biopsy, it is more representative of the general population. Furthermore, upon review of the results of those studies most patients had clinical T1c prostate cancer. Freedland et al reported that 78% of their Shared Equal Access Regional Cancer Hospital Database cohort had a normal DRE before biopsy.1 Likewise, 75% of the cohort analyzed by Newton et al had clinical T1 prostate cancer.3 Finally, Briganti et al noted that 66% of their series had clinical stage T1c.2 The aim of these groups was to characterize prostate volume as a predictor of a worse outcome and in that sense they were successful. Small prostate volume is associated with high grade prostate cancer but this association is likely due to the way that we diagnose prostate cancer and does not reflect the underlying biology of the disease. Furthermore, the association is only valid in men with clinical T1c prostate cancer. This study has some important limitations. It is a single institution, retrospective series and, thus, subject to problems inherent to that study design, namely selection bias, and the misclassification of exposures and outcomes. Although transrectal ultrasound volume correlates well with prostate volume with a reported Pearson correlation coefficient
of greater than 0.8, it is not an exact measurement of prostate size. Also, our multivariate logistic regression model for men with palpable disease may have been limited by a smaller proportion of men with low grade disease (18%). However, our parsimonious selection of predictors for our models may have mitigated this. Furthermore, there are always inherent inaccuracies when assigning a Gleason score to prostate biopsy cores. In our series pathological upgrading occurs between biopsy cores and the final specimen 13% of the time. Nevertheless, a biopsy population represents a superset of a radical prostatectomy population and, thus, can be said to be more generalizable to the population at large. Conclusions drawn from a radical prostatectomy series inherently show selection bias since the series does not include patients who elected other treatment modalities. The current findings and the results of our prior report from our institutional radical prostatectomy database provide a more nuanced picture of the prostate cancer size-grade association. Among men undergoing prostate needle biopsy and men undergoing radical prostatectomy smaller prostate size only predicts high grade cancer in those with nonpalpable disease, ie men who underwent biopsy for increased PSA. Together these findings imply that this association is the result of a statistical artifact rather than true tumor biology. Therefore, if confirmed, the findings of this study may change the way that urologists counsel and risk stratify patients. It may be incorrect to tell a patient that the prognosis is worse because his prostate is smaller. Rather, the prognosis may be worse because he has more aggressive cancer independent of prostate size. In conclusion, in a contemporary series of men undergoing extended scheme prostate needle biopsy a small prostate did not predict high grade cancer.
REFERENCES 1. Freedland SJ, Isaacs WB, Platz EA et al: Prostate size and risk of high grade, advanced prostate cancer and biochemical progression after radical prostatectomy: A SEARCH database study. J Clin Oncol 2005; 23: 7546. 2. Briganti A, Chun FKH, Suardi N et al: Prostate volume and adverse prostate cancer features: fact not artifact. Eur J Cancer 2007; 43: 2669. 3. Newton NR, Phillips S, Chang SS et al: Smaller prostate size predicts high grade prostate cancer at final pathology. J Urol 2010; 184: 930. 4. Chodak GW, Vogelzang NJ, Caplan RJ et al: Independent prognostic factors in patients with
metastatic (stage D2) prostate cancer: the Zoladex Study Group. JAMA 1991; 265: 618. 5. Liu JJ, Brooks JD, Ferrari M et al: Small prostate size and high grade disease— biology or artifact? J Urol 2011; 185: 2108. 6. Karakiewicz PI, Bazinet M, Aprikian AG et al: Outcome of sextant biopsy according to gland volume. Urology 1997; 49: 55. 7. Thompson IM, Goodman PJ, Tangen CM et al: The influence of finasteride on the development of prostate cancer. New Engl J Med 2003; 349: 213. 8. Kulkarni GS, Al-Azab R, Lockwood G et al: Evidence of biopsy derived grade artifact among larger prostate glands. J Urol 2006; 175: 505.
9. Kassouf WK, Nakanishi H, Ochiai A et al: Effect of prostate volume on tumor grade in patients undergoing radical prostatectomy in the era of extended prostatic biopsies. J Urol 2007; 178: 111. 10. Benson MC, Whang IS, Pantuck A et al: Prostate specific antigen density: a means of distinguishing benign prostatic hypertrophy and prostate cancer. J Urol 1992; 147: 815. 11. Elliott CS, Shinghal R and Presti JC: The performance of prostate specific antigen, prostate specific antigen density and transition zone density in the era of extended biopsy schemes. J Urol 2008; 179: 1756.
PROSTATE SIZE DOES NOT PREDICT HIGH GRADE CANCER
481
EDITORIAL COMMENT The ratio of PSA to prostate size (PSA density) is an important predictor of biopsy results (reference 10 in article), pathological characteristics at radical prostatectomy and biochemical failure (reference 3 in article).1–3 Several theories have been postulated to explain this but the most likely is the simplest. If a lot of PSA is produced by a small prostate, it is probably due to relatively aggressive cancer. If a lot of PSA is produced by a large prostate, it is probably due to the bulk of the benign prostate. In this light PSA and prostate size are important to determine risk. These authors successfully identified this association between PSA density and high grade cancer in patients with palpable and nonpalpable disease but they draw the incorrect conclusion that PSA is the real driver while controlling for prostate size. In fact, the nature of entering these 2 terms as a ratio is to acknowledge that their relationship to each other is important. PSA density increases when PSA is higher or when prostate size is lower. The fact that they identified the relationship between a small prostate and high grade disease among patients with nonpalpable disease and not among those with palpable disease probably reflects a limitation in statistical power rather than an absent relationship. Of 241 patients with palpable disease 201 (83.4%) had high grade disease. In a logistic regression model power is determined by the outcome with a lower number of events, in this
instance 40 patients with nonhigh grade disease. Most biostatisticians recommend limiting the model to 1 covariate per 10 events. The authors overshot this rule of thumb, resulting in an overfit model and a lower chance of finding a difference based on size. The authors conclude that clinicians need not consider prostate size when counseling a patient with palpable disease. Despite concern over the model on which this recommendation is based they are probably correct. A palpable nodule is usually a sufficiently poor prognostic sign to guide management. On the other hand, these factors should be considered carefully in patients undergoing biopsy with normal DRE or in patients with low risk prostate cancer, in whom a small prostate and/or high PSA density portends a higher likelihood of any cancer and high grade cancer on biopsy, and predicts the likelihood of aggressive pathological characteristics in those undergoing surgery. Since men with normal DRE comprise a substantial proportion of those with biopsy and prostate cancer whom we see, clinicians should be encouraged to pay more attention to prostate size rather than less. Beware the small prostate. Daniel A. Barocas Department of Urologic Surgery Vanderbilt University Nashville, Tennessee
REFERENCES 1. Kundu SD, Roehl KA, Yu X et al: Prostate specific antigen density correlates with features of prostate cancer aggressiveness. J Urol 2007; 177: 505.
2. Radwan MH, Yan Y, Luly JR et al: Prostate-specific antigen density predicts adverse pathology and increased risk of biochemical failure. Urology 2007; 69: 1121.
3. Davies JD, Aghazadeh MA, Phillips S et al: Prostate size as a predictor of Gleason score upgrading in low-risk prostate cancer patients. J Urol 2011; 187: 2221.
REPLY BY AUTHOR Our point was to evaluate the relationship between prostate size and high grade cancer, specifically to address the claim that a small prostate is a surrogate marker for aggressive disease. In men with palpable disease there was no such association on multivariate logistic regression. The concern about model overfit is well taken. However, overfit models typically result in falsely significant p values and not the other way around. The concern about statistical power is also valid. To address each concern we reran analysis with a more parsimonious model that included only 3 predictors (prostate size, PSA and age). It yielded results similar to those of the full model.
The influence of prostate size is subtle. We contend that prostate size alone does not drive aggressive disease. Rather, the index of PSA and prostate volume (or PSA density) predicts adverse features. Thus, a higher PSA for any given prostate volume portends a worse prognosis. In that sense PSA is the predictor of aggressive disease after one controls for prostate size. It may be a mistake to tell a patient that he is going to have a worse outcome only because he has a small prostate while ignoring the bigger picture, that is high PSA density, high Gleason score or many positive cores.
Prostate Size Does Not Predict High Grade Cancer Tin C. Ngo,* Simon L. Conti, Rajesh Shinghal and Joseph C. Presti, Jr.† From the Department of Urology, Stanford University School of Medicine, Stanford and Division of Urology, Santa Clara Valley Medical Center (RS), San Jose, California
Purpose: Several radical prostatectomy series have linked small prostates with high grade cancer based on the hypothesis that a small prostate results from a low androgen milieu that selects for less hormone dependent, more aggressive tumors. We previously reported that this association resulted from ascertainment bias from the performance characteristics of prostate specific antigen rather than from tumor biology in our radical prostatectomy cohort. In this study we analyzed this association in a more generalized population of men who underwent prostate needle biopsy. Materials and Methods: The prostate needle biopsy database at our institution was queried for all initial biopsies. Included patient characteristics were age, race, family history of prostate cancer, prostate specific antigen, abnormal digital rectal examination and prostate volume in ml on transrectal ultrasound. Multivariate logistic regression was used to determine the influence of prostate volume on the odds of high grade cancer. Results: The study population included 1,295 patients during 2000 to 2010, of whom 582 (44.9%) had prostate cancer and 398 (30.7%) had high grade cancer. When all patients were pooled, the OR for high grade cancer was 0.85 (95% CI 0.78 – 0.92) for each 10 ml increase in prostate volume. When patients were divided by clinical T stage, the corresponding ORs for those with T1c disease was 0.83 (95% CI 0.74 – 0.93) and for those with T2 or greater disease it was 0.99 (0.98 –1.00). Conclusions: The association between small prostates and high grade cancer exists only in men with clinical T1c (normal digital rectal examination) prostate cancer. It likely resulted from ascertainment bias due to the performance characteristics of prostate specific antigen rather than tumor biology.
Abbreviations and Acronyms DRE ⫽ digital rectal examination PSA ⫽ prostate specific antigen SPND ⫽ Stanford Prostate Needle Biopsy Database Submitted for publication June 3, 2011. Study received Stanford University School of Medicine institutional review board approval. * Correspondence: Department of Urology, Stanford University School of Medicine, 300 Pasteur Dr., S-287, Stanford, California (telephone: 408-813-7049; FAX: 408-503-0050; e-mail: tin. [email protected]). † Financial interest and/or other relationship with PrognosDx and Myriad.
For another article on a related topic see page 707.
Key Words: prostate, organ size, prostate-specific antigen, prostatic neoplasms, prognosis SEVERAL groups have reported an association between small prostates and an increased risk of adverse outcomes such as high grade cancer, extraprostatic extension, positive surgical margins and biochemical recurrence in men undergoing radical prostatectomy.1–3 To explain this observation some investigators advanced the theory that a small prostate acts as a surrogate marker for an androgen depleted milieu that selects for more aggressive,
less hormone dependent tumors. To support this reasoning they cited prior studies linking low pretreatment serum testosterone to poor outcomes in men with newly diagnosed metastatic prostate cancer.4 In a recently published study of 1,404 patients from the radical prostatectomy database at our institution we reported that the association between small prostates and adverse pathological features, particularly the
0022-5347/12/1872-0477/0 THE JOURNAL OF UROLOGY® © 2012 by AMERICAN UROLOGICAL ASSOCIATION EDUCATION
Vol. 187, 477-481, February 2012 Printed in U.S.A. DOI:10.1016/j.juro.2011.10.042
AND
RESEARCH, INC.
www.jurology.com
477
478
PROSTATE SIZE DOES NOT PREDICT HIGH GRADE CANCER
volume and percent of high grade cancer, exists only in patients with clinical T1c prostate cancer.5 This association was not observed in men undergoing radical prostatectomy for clinical T2 disease. This disparity most likely represented ascertainment bias between the 2 groups due to PSA performance characteristics. If tumor biology drove the size-grade association, one would expect to see it in each group. A limitation of making inferences from the results of radical prostatectomy series is selection bias. Cohorts that undergo surgery represent a subset of men who have cancer on prostate needle biopsy, who represent a larger subset than those who undergo surgery. To address this limitation of radical prostatectomy series we studied the association of prostate size and high grade cancer in a contemporary group of patients from SPND.
MATERIALS AND METHODS SPND is an institutional review board approved, prospectively maintained database of all prostate needle biopsies performed at our institution. We reviewed SPND for all extended scheme (12 cores or more) biopsies since database inception. Patients with an existing diagnosis of prostate cancer, eg those on active surveillance, those evaluated for local recurrence after radiotherapy and those undergoing biopsy to confirm a diagnosis made elsewhere, were excluded from analysis. Those with prior negative biopsies were also excluded since the detection of cancer and high grade cancer in this patient group differs significantly from that in men undergoing initial biopsy. Patient characteristics used in our analysis included age in years, race (white, black, Asian or other), family history of prostate cancer (yes/no), abnormal DRE (yes/no) and prostate volume in ml ascertained by volume (V) estimates based on transrectal ultrasound measurements using the formula, V ⫽ /6 ⫻ length ⫻ width ⫻ height in cm. High grade prostate cancer, defined as a positive biopsy with a Gleason score of 7 or higher, was modeled as a dichotomous variable (yes/no). Table 1. Demographics of patients with prostate cancer on prostate needle biopsy
Median age (IQR) No. race (%): White Asian Black Other No. prostate Ca family history (%): Yes No Median ng/ml PSA (IQR) Median ml prostate vol (IQR) No. high grade Ca (%): Yes No
Normal DRE
Abnormal DRE
65
(59–71)
67
(62–75)
257 42 30 22
(75.4) (12.3) (5.9) (6.5)
171 40 9 21
(70.1) (16.6) (3.7) (8.7)
p Value ⬍0.0001 0.21
0.65 82 (24.1) 259 (76.0) 5.8 (4.5–8.4) 40 (27–53)
54 (22.4) 187 (77.6) 6.9 (4.3–15.0) 34 (26–49)
197 144
201 40
(57.8) (42.2)
(83.4) (16.6)
0.02 0.02 ⬍0.0001
Biopsy result by DRE status. Light blue indicates normal, medium blue low grade cancer and dark blue high grade cancer.
The Wilcoxon rank sum test was used to compare medians between different groups. The chi-square test was used to compare frequencies between different groups. Multivariate logistic regression was done to model the odds of high grade prostate cancer as a function of patient prostate size, PSA, age, race and family history of prostate cancer. Since we did not having missing variables, a complete case analysis was performed. To completely control for the effect of clinical T stage we generated separate models for men with normal (clinical T1c) and abnormal (clinical T2 or higher) DRE. The OR of each predictor is reported with the Wald 95% CI. Statistical analysis was done with SAS® 9.2. Tests for significance were 2-sided with p ⱕ0.05 considered significant.
RESULTS Table 1 lists the demographics of our study cohort. A total of 1,435 patients who were never diagnosed with prostate cancer underwent an extended prostate biopsy scheme during 2000 to 2010. We excluded 140 men who underwent a prior negative biopsy, leaving 1,295 as the study population. The figure shows biopsy results stratified by DRE findings. As expected, the overall positive biopsy rate and the risk of finding high grade cancer were higher in men with abnormal DRE and these observations were statistically significant. Of these men 582 (44.9%) had cancer and 398 (30.7%) had high grade cancer. Of the 582 men with cancer 341 (58.6%) had normal DRE and 241 (41.4%) had abnormal DRE. Although there was a statistically significant difference in median age between the 2 groups, a difference of 2 years likely had no clinical significance. Differences in the distribution of race and family history of prostate cancer between the 2 groups were not statistically significant. Median PSA was lower in men with normal DRE in a statistically significant manner. Conversely median prostate volume was higher in those with normal
PROSTATE SIZE DOES NOT PREDICT HIGH GRADE CANCER
DRE in a statistically significant manner, likely since these patients underwent prostate needle biopsy due only to increased PSA. Thus, it was more likely that benign enlargement of the prostate partially contributed to the PSA increase. To analyze the association between prostate size and the odds of high grade prostate cancer we first determined the unadjusted impact of prostate size on high grade prostate cancer using univariate logistic regression. The crude OR of each 10 ml increase in prostate volume was 0.84 (95% CI 0.79 – 0.89). Multivariate logistic regression was then used to control for potential covariates, including age, PSA, race and family history. The adjusted OR of each 10 ml increase in prostate volume was 0.85 (95% CI 0.78 – 0.92). Thus, when all patients with prostate cancer were analyzed together, there was an association between smaller prostates and high grade cancer on univariate and multivariate analysis. To completely control for the effect of clinical T stage we then created separate models for men with clinical T1c and T2 or greater prostate cancer. Age, PSA and prostate size were statistically significant predictors of high grade prostate cancer in men with clinical T1c prostate cancer (table 2). However, of those with clinical T2 or greater prostate cancer the only statistically significant predictor of high grade prostate cancer was PSA. Prostate size lost significance in this patient subset (table 2).
DISCUSSION The controversy over the association of a small prostate and high grade prostate cancer goes back more
Table 2. Predictors of high grade in patients with clinical T1c and T2 or greater prostate cancer
Prostate size/10 ml Log PSA Age Race: White (referent) Asian Black Other Family history Prostate size/10 ml Log PSA Age Race: White (referent) Asian Black Other Family history
OR (95% CI)
p Value
T1c 0.83 (0.74–0.93) 3.25 (1.82–5.82) 1.07 (1.03–1.10)
0.001 ⬍0.0001 ⬍0.0001
1.00 0.91 (0.41–1.99) 1.73 (0.62–4.83) 0.91 (0.35–2.34) 1.10 (0.62–1.95) T2 or greater 0.99 (0.98–1.00) 2.24 (1.44–3.49) 1.02 (0.97–1.06)
— 0.81 0.30 0.84 0.75
1.00 0.54 (0.19–1.50) 0.55 (0.09–3.42) 0.84 (0.21–3.28) 0.47 (0.21–1.04)
— 0.24 0.52 0.80 0.06
0.09 0.0003 0.51
479
than a decade to a report by Karakiewicz et al, who noted no association between prostate volume and Gleason grade in men undergoing sextant biopsy.6 The Prostate Cancer Prevention Trial reignited this debate with the finding that high grade cancer was more prevalent in men treated with finasteride and who consequently had a smaller prostate.7 To address this concern Kulkarni et al compared the effect of prostate volume on Gleason grade in the prostate needle biopsy and final surgical specimen of men undergoing radical prostatectomy.8 They observed that the size-grade association existed in the biopsy cohort but not in the radical prostatectomy cohort, suggesting that the effect may be due to sampling bias. To counter this Kassouf et al reported that among men undergoing radical prostatectomy at their institution those with prostate volume 50 ml or greater had a higher incidence of low grade cancer.9 In our study the association between a small prostate and high grade prostate cancer was observed in men with clinical T1c but not T2 or greater prostate cancer. These results agree with findings from the previously published study from our institution that analyzed the size-grade association in men undergoing radical prostatectomy.5 The characteristics of patients with clinical T1c and clinical T2 or greater prostate cancer are quite different. The clinical T1c group underwent prostate needle biopsy only due to increased PSA. Thus, we believe that the size-grade association previously reported in multiple radical prostatectomy series resulted from ascertainment bias rather than from tumor biology. If smaller prostate volume were a surrogate marker for tumor aggressiveness or led to higher grade cancer, one would expect this relationship to exist regardless of clinical stage. Since this association is quite strong, when patients with clinical T1c disease were commingled with those with a higher clinical T stage, the association was statistically significant on univariate and multivariate analysis. However, when the effect of clinical stage was completely controlled for by stratifying our analysis by clinical T stage, it became clear that the driver of the size-grade association was the subset of patients with clinical T1c disease. The question then becomes what is responsible for this ascertainment bias. An explanation may be the improved performance characteristics of PSA for high grade disease. For any given PSA a smaller prostate will have higher PSA density than a larger prostate and, thus, be more likely to harbor a high grade tumor.10,11 On separate analysis PSA density was a strong independent predictor of high grade cancer in men with palpable and with nonpalpable disease (p ⫽ 0.0002). We interpret this to mean that when controlling for prostate size, PSA is the
480
PROSTATE SIZE DOES NOT PREDICT HIGH GRADE CANCER
real driver behind high grade disease. Thus, observing that smaller prostates are more likely to have high grade cancer may simply be a consequence of the performance characteristics of PSA instead of tumor biology. Another explanation may lie with the observation that the clinical T1c group has a larger prostate on average and so the group may be enriched with men who primarily have benign prostatic hyperplasia but also an incidental low grade tumor. Our findings directly contradict the results of several published radical prostatectomy series. However, since this cohort was based on a group of men undergoing prostate needle biopsy, it is more representative of the general population. Furthermore, upon review of the results of those studies most patients had clinical T1c prostate cancer. Freedland et al reported that 78% of their Shared Equal Access Regional Cancer Hospital Database cohort had a normal DRE before biopsy.1 Likewise, 75% of the cohort analyzed by Newton et al had clinical T1 prostate cancer.3 Finally, Briganti et al noted that 66% of their series had clinical stage T1c.2 The aim of these groups was to characterize prostate volume as a predictor of a worse outcome and in that sense they were successful. Small prostate volume is associated with high grade prostate cancer but this association is likely due to the way that we diagnose prostate cancer and does not reflect the underlying biology of the disease. Furthermore, the association is only valid in men with clinical T1c prostate cancer. This study has some important limitations. It is a single institution, retrospective series and, thus, subject to problems inherent to that study design, namely selection bias, and the misclassification of exposures and outcomes. Although transrectal ultrasound volume correlates well with prostate volume with a reported Pearson correlation coefficient
of greater than 0.8, it is not an exact measurement of prostate size. Also, our multivariate logistic regression model for men with palpable disease may have been limited by a smaller proportion of men with low grade disease (18%). However, our parsimonious selection of predictors for our models may have mitigated this. Furthermore, there are always inherent inaccuracies when assigning a Gleason score to prostate biopsy cores. In our series pathological upgrading occurs between biopsy cores and the final specimen 13% of the time. Nevertheless, a biopsy population represents a superset of a radical prostatectomy population and, thus, can be said to be more generalizable to the population at large. Conclusions drawn from a radical prostatectomy series inherently show selection bias since the series does not include patients who elected other treatment modalities. The current findings and the results of our prior report from our institutional radical prostatectomy database provide a more nuanced picture of the prostate cancer size-grade association. Among men undergoing prostate needle biopsy and men undergoing radical prostatectomy smaller prostate size only predicts high grade cancer in those with nonpalpable disease, ie men who underwent biopsy for increased PSA. Together these findings imply that this association is the result of a statistical artifact rather than true tumor biology. Therefore, if confirmed, the findings of this study may change the way that urologists counsel and risk stratify patients. It may be incorrect to tell a patient that the prognosis is worse because his prostate is smaller. Rather, the prognosis may be worse because he has more aggressive cancer independent of prostate size. In conclusion, in a contemporary series of men undergoing extended scheme prostate needle biopsy a small prostate did not predict high grade cancer.
REFERENCES 1. Freedland SJ, Isaacs WB, Platz EA et al: Prostate size and risk of high grade, advanced prostate cancer and biochemical progression after radical prostatectomy: A SEARCH database study. J Clin Oncol 2005; 23: 7546. 2. Briganti A, Chun FKH, Suardi N et al: Prostate volume and adverse prostate cancer features: fact not artifact. Eur J Cancer 2007; 43: 2669. 3. Newton NR, Phillips S, Chang SS et al: Smaller prostate size predicts high grade prostate cancer at final pathology. J Urol 2010; 184: 930. 4. Chodak GW, Vogelzang NJ, Caplan RJ et al: Independent prognostic factors in patients with
metastatic (stage D2) prostate cancer: the Zoladex Study Group. JAMA 1991; 265: 618. 5. Liu JJ, Brooks JD, Ferrari M et al: Small prostate size and high grade disease— biology or artifact? J Urol 2011; 185: 2108. 6. Karakiewicz PI, Bazinet M, Aprikian AG et al: Outcome of sextant biopsy according to gland volume. Urology 1997; 49: 55. 7. Thompson IM, Goodman PJ, Tangen CM et al: The influence of finasteride on the development of prostate cancer. New Engl J Med 2003; 349: 213. 8. Kulkarni GS, Al-Azab R, Lockwood G et al: Evidence of biopsy derived grade artifact among larger prostate glands. J Urol 2006; 175: 505.
9. Kassouf WK, Nakanishi H, Ochiai A et al: Effect of prostate volume on tumor grade in patients undergoing radical prostatectomy in the era of extended prostatic biopsies. J Urol 2007; 178: 111. 10. Benson MC, Whang IS, Pantuck A et al: Prostate specific antigen density: a means of distinguishing benign prostatic hypertrophy and prostate cancer. J Urol 1992; 147: 815. 11. Elliott CS, Shinghal R and Presti JC: The performance of prostate specific antigen, prostate specific antigen density and transition zone density in the era of extended biopsy schemes. J Urol 2008; 179: 1756.
PROSTATE SIZE DOES NOT PREDICT HIGH GRADE CANCER
481
EDITORIAL COMMENT The ratio of PSA to prostate size (PSA density) is an important predictor of biopsy results (reference 10 in article), pathological characteristics at radical prostatectomy and biochemical failure (reference 3 in article).1–3 Several theories have been postulated to explain this but the most likely is the simplest. If a lot of PSA is produced by a small prostate, it is probably due to relatively aggressive cancer. If a lot of PSA is produced by a large prostate, it is probably due to the bulk of the benign prostate. In this light PSA and prostate size are important to determine risk. These authors successfully identified this association between PSA density and high grade cancer in patients with palpable and nonpalpable disease but they draw the incorrect conclusion that PSA is the real driver while controlling for prostate size. In fact, the nature of entering these 2 terms as a ratio is to acknowledge that their relationship to each other is important. PSA density increases when PSA is higher or when prostate size is lower. The fact that they identified the relationship between a small prostate and high grade disease among patients with nonpalpable disease and not among those with palpable disease probably reflects a limitation in statistical power rather than an absent relationship. Of 241 patients with palpable disease 201 (83.4%) had high grade disease. In a logistic regression model power is determined by the outcome with a lower number of events, in this
instance 40 patients with nonhigh grade disease. Most biostatisticians recommend limiting the model to 1 covariate per 10 events. The authors overshot this rule of thumb, resulting in an overfit model and a lower chance of finding a difference based on size. The authors conclude that clinicians need not consider prostate size when counseling a patient with palpable disease. Despite concern over the model on which this recommendation is based they are probably correct. A palpable nodule is usually a sufficiently poor prognostic sign to guide management. On the other hand, these factors should be considered carefully in patients undergoing biopsy with normal DRE or in patients with low risk prostate cancer, in whom a small prostate and/or high PSA density portends a higher likelihood of any cancer and high grade cancer on biopsy, and predicts the likelihood of aggressive pathological characteristics in those undergoing surgery. Since men with normal DRE comprise a substantial proportion of those with biopsy and prostate cancer whom we see, clinicians should be encouraged to pay more attention to prostate size rather than less. Beware the small prostate. Daniel A. Barocas Department of Urologic Surgery Vanderbilt University Nashville, Tennessee
REFERENCES 1. Kundu SD, Roehl KA, Yu X et al: Prostate specific antigen density correlates with features of prostate cancer aggressiveness. J Urol 2007; 177: 505.
2. Radwan MH, Yan Y, Luly JR et al: Prostate-specific antigen density predicts adverse pathology and increased risk of biochemical failure. Urology 2007; 69: 1121.
3. Davies JD, Aghazadeh MA, Phillips S et al: Prostate size as a predictor of Gleason score upgrading in low-risk prostate cancer patients. J Urol 2011; 187: 2221.
REPLY BY AUTHOR Our point was to evaluate the relationship between prostate size and high grade cancer, specifically to address the claim that a small prostate is a surrogate marker for aggressive disease. In men with palpable disease there was no such association on multivariate logistic regression. The concern about model overfit is well taken. However, overfit models typically result in falsely significant p values and not the other way around. The concern about statistical power is also valid. To address each concern we reran analysis with a more parsimonious model that included only 3 predictors (prostate size, PSA and age). It yielded results similar to those of the full model.
The influence of prostate size is subtle. We contend that prostate size alone does not drive aggressive disease. Rather, the index of PSA and prostate volume (or PSA density) predicts adverse features. Thus, a higher PSA for any given prostate volume portends a worse prognosis. In that sense PSA is the predictor of aggressive disease after one controls for prostate size. It may be a mistake to tell a patient that he is going to have a worse outcome only because he has a small prostate while ignoring the bigger picture, that is high PSA density, high Gleason score or many positive cores.