What risk of prostate cancer led urologist to recommend prostate biopsies?

25 décembre 2015

Auteurs : M. Roumiguié, J.-B. Beauval, B. Bordier, T. Filleron, F. Rozet, A. Ruffion, N. Mottet, O. Cussenot, B. Malavaud
Référence : Prog Urol, 2015, 16, 25, 1125-1131




 




Introduction


Prostate cancer is the most common solid cancer in France (71,500 patients/year) and the second cause of death by cancer (8791 patients/years) in males [1]. The French Association of Urology supports individual screening by Prostatic Specific Antigen (PSA) testing and Digital Rectal Examination (DRE) [2].


In a recent report, PSA with a threshold of 4ng/mL exhibited excellent sensitivity (93%) but mediocre specificity (24%) [3]. In the PCPT study, random systematic sextant biopsies demonstrated cancer in 15.2% patients with PSA values <4ng/mL and unsuspicious DRE while in the Catalona report the rate of positive biopsies in the so-called "grey zone" (PSA: 4-10ng/mL) was in the range of 25% [4, 5]. Thus, the large majority of prostate biopsies do not show cancer [6]. According to the French Parliamentary Office for the evaluation of Health Care Policies (OPEPS), up to 250,000 biopsies are taken every year in France suggesting that a large number of patients are subjected to biopsy with a little or no benefit. (r08-3181.pdf). Pre-biopsy evaluation of the individual risk of positive biopsies would help control the negative aspects of prostate biopsy (morbidity, negative results) with minimal impact on the benefits of early prostate cancer diagnosis (Schroeder). Many predictive instruments such as PCPT-CRC and SWOP-PRI were designed and validated to estimate the risk of positive biopsy [7, 8]. They were used in the present paper to estimate in a cohort of patients submitted to primary and repeat biopsies the risks of showing cancer that led the referring urologists to recommend biopsies.


Patients and methods


Eight hundred and eight patients biopsied in five French university centres were recruited in this retrospective multicenter study [9]. Age, digital rectal examination, prostate volume measured with transrectal ultrasonography (TRUS), biopsy history, and serum PSA values were obtained for all patients and used to estimate the individual risk of cancer using two validated instruments of prediction: PCPT-CRC (calcs.jsp) and SWOP-PRI (seven-prostate-cancer-risk-calculators). For each patient, life expectancy (less or more than 10 years) was estimated by the referring urologist based on patient's age, personal history and concurrent morbidity.


PCPT-CRC nomogram was developed from the placebo arm (5519 patients) of Prostate Cancer Prevention Trial where six random cores were taken for PSA values higher than 4ng/mL or suspicious DRE. Patients not found positive for prostate cancer at initiation of the study or during the seven-year follow-up were also offered end-of-study random systematic cores (6 cores) [10]. This nomogram was later validated with 12-core biopsy protocol [11]. Of note, as family history of prostate cancer was not available in our database, the weight of this information was modeled by the bootstrap technique.


The SWOP-PRI "riskcalculator 4+ DRE" was developed from the data obtained within the Rotterdam arm (6288 men) of the European Randomized Study of screening of the Prostate Cancer (ERSPC) [12, 13].


These nomograms also provided the estimate of the risk of high-grade prostate cancer (Gleason score ≥ 7 for PCPT-CRC; Gleason score≥7 and/or >T2b for SWOP-PRI).


Statistical analysis


Mean and standard deviation are presented for quantitative variables, percentage and 95% confidence interval for qualitative variables. Statistical analysis was performed using GraphPad Prism® version 5.0 (GraphPad Software, Inc, La Jolla, CA, USA). Differences between the quantitative and qualitative variables were sought by Mann-Whitney test, Chi2 test and Fisher's exact test, respectively. P values <0.05 were considered significant. Receiver Operating Characteristic (ROC) curves were plotted to illustrate the performance of PSA, PCPT-CRC and SWOP-PRI and AUCs of the ROC curves compared by the C-statistics. Finally, calibration curves were plotted to illustrate the relationship existing between the estimated risks and the actual results.


Results


The files of 808 consecutive patients submitted to prostate biopsies centres in 2010, were collected in 5 academies. One hundred and eighty-three files did not fit the requisites of the predictive models (PSA>50ng/mL, age<55 years) or did not cover some key information (TRUS-measured prostate volume) and were excluded from the analysis.


Characteristics of the 625 patients included in the final analysis are presented in Table 1. Five hundred and sixty-eight (90.8%) patients had a life expectancy>10 years. PCa was diagnosed in 291 patients (46.6%) who were older (66.7±6.8 vs 64.3±5.6 years, P <0,001) and showed higher PSA values (10±7.9 vs 7.7±4.3ng/mL, P <0.0001) and lower prostate volumes (43.8±19.8 vs 51.3±20.7mL, P <0.0001) than healthy subjects. Gleason score was respectively≤6, 7, and >7 in 47.1%, 41.6% and 11.3% of the cases.


PCPT-CRC risk estimates were 56.2±12.8% and 50.6±14% in patients with and without family history of prostate cancer, respectively. SWOP-PRI showed a lower figure (31.2±17.3%). The estimates to detect high-grade cancer were 22.4±16.9% for PCPT-CRC and 14.8±18.2% for SWOP-PRI (Table 2). For the two risk calculators, the calculated estimates increased with age and prior history of negative biopsy (Table 3). AUCs of the ROC curves were 0.58±0.02, 0.67±0.02 and 0.72±0.02 for PSA, PCPT-CRC and SWOP-PRI, respectively. No significant differences were demonstrated in the diagnostic performances of the two nomograms (P =0.09).


Conversely, for the high-grade cancer risk estimation (PCPT-CRC) and for the high-risk cancer prediction (SWOP-PRI), the AUC was 0.72±0.03 (CI 95% 0.67-0.78; P <0.0001) and 0.73±0.03 (0.68-0.79; P <0.0001) respectively. The calibration graphs demonstrated that the SWOP-PRI was closer to the ideal prediction model than the PCPT-CRC (Figure 1).


Figure 1
Figure 1. 

a: ROC curves PSA (black), PCPT-CRC (blue) and SWOP-PRI (red); b: calibration of PCPT-CRC; c: calibration of SWOP-PRI.





Discussion


Currently, overdiagnosis of PCa is a major concern for health high authorities. This study shows that the urologists of tertiary centres recommended prostate biopsy only in high risk of PCa (50.1% PCPT-CRC with negative family history, 55.8% PCPT-CRC with positive family history and 31.2±17.3% SWOP-PRI).


Before, to perform biopsy, urologists considered the life expectancy and the morbidities. Over 90% of patients had life expectancy higher than 10 years. For the other 10%, the risk estimated of PCa was very high (PCPT-CRC 65.6%; SWOP-PRI 45.1%). Indeed, the Farmington group study reported that the risk of death increases in all clinical situations along with time and comorbidities. For example, the overall mortality at 5-years follow-up of a man with comorbidities and a well-differentiated PCa was 10 times upper than PCa mortality (42.5% and 4.3%) [14].


The risk estimate difference in patients with or without negative previous biopsy (49% vs 58.2%, P <0.0001, PCPT-CRC and 14.8 vs 35.8%, P <0.0001, SWOP-PRI) showed that a prior negative biopsy disturbed the strategy of PCa diagnosis. Indeed, negative biopsies create uncertainty among the urologists because of false negative rate of 40-25%, leading them to repeat biopsies even if the estimated risk of cancer is. Ploussard et al. reported 25 to 35% of re-biopsies in 5 years after a first negative biopsy with a rate of cancer at 17% in the first two re-biopsies, then between 12 and 14% for the next series [15]. This uncertainty related to the false negative numbers of prostate biopsies can be overcome by providing new biopsy techniques (MRI-guided) and with a better selection of patients [6].


In our study, there was no difference between patients with either one or several series of negative biopsies. However, the incidence of PCa decreases as the number of previous negative histology increases. Moreover, the threshold to re-biopsy should increase in proportion to the number of previous biopsies. In addition, the cancer detection decreases with the increasing number of biopsies. Gann et al. showed that the rate of positive biopsy decreased to 46% at the first re-biopsy, to 64% at the second and to 88% after 3 series. Thus, they concluded that there is a lack of benefit obtaining more than two biopsies when the clinical and biological criteria are stable [16].


The risk calculators were based on pertinent clinical and biological dataset to specify the benefit of undergoing or repeating systematic biopsies in the overall population. However, the threshold of PCa risk that should lead the urologist to perform a prostate biopsy was not yet defined. In their study, Roobol et al. showed that biopsies could be avoided if estimated risk with SWOP-PRI nomogram was less than 12.5%. In our study, with this threshold value, 66 biopsies (10.6%) were avoided with 15 PCa (5.2%) missed. According to Roobol et al. study, when the risk was between 12.5 and 20%, the authors recommended biopsy only if the high-grade cancer risk was>3%. Thus, in our cohort, 45 biopsies would have been avoided for only 8 missed cancers among which only one had a Gleason score≥7 [17]. Steyerberg and Vickers. had determined two risk thresholds, 10 and 40% [18]. In our study, the SWOP-PRI model, the use of the lower threshold (10%) would have avoided 46 biopsies (7.4%) and neglected 8 (2.7%) cancers. The latter value (40%) is also controversial since 17.5% and 58.4% of cancer wouldn't have been diagnosed among 24.2% and 72.5% avoided biopsies according to the PCPT-CRC and SWOP-PRI respectively. Finally, Van Vugt et al. defined a SWOP-PRI threshold value of 20% in a cohort of 320 patients without previous negative biopsy. They reported 11 over 63 (17%) missed cancers with only two having a Gleason score≥7 [19]. In our population, this threshold would avoid 27.8% of biopsies but 14.1% of cancer would not detect 14.1% of which 29.3% with a Gleason score≥7. The threshold concept is a matter of choice and some have suggested values between 12.5 and 25%. We believe that the risk must be integrated into a general approach of the patient including his comorbidities, his choice and his tolerance to the risk, which is an individual characteristic. However, according to the Youden's J statistic, we could set a threshold with the greater match between sensitivity and specificity. In our study, the reasonable threshold risk to perform biopsy was 49.6% (sensitivity: 60.1%; specificity: 68.6%) for PCPT and 30% (sensitivity: 61.7%; specificity 70.4%) for the nomogram ERSPC.


Several studies assessed the diagnostic accuracy of risk calculators in PCa. A review of 23 studies which evaluated 36 predictive models determined that nomogram were 2 to 26% more accurate than PSA in diagnosis of prostate cancer [20]. In these papers, the risk of PCa for patients varied between 22.8 and 51% for the PCPT and 22 to 26% for the SWOP-PRI and was lower than those observed in our study [21, 22, 23, 24, 25] (Table 4). Cussenot et al. characterized the overall population enrolled in this study as high risk for PCa with a high number of suspicious DRE and a high rate of positive biopsies. They concluded that the increasing reluctance of French general practitioners toward screening could explain this population at high risk of PCa [9]. Finally, our population was into the real life of clinical practice whereas patients enrolled in the clinical trials originated from two main populations, screening for the ERSPC and chemoprevention for the PCPT.


On the other hand, this study has limitations. First of all, the lack of data about family history of prostate cancer had led to the use of random sampling tests (bootstrap). In addition, all the patients enrolled in our study are Caucasians. Moreover, the risk calculators were based on protocol at 6 prostate-cores whereas our average biopsy number was higher (12.4 [3-42]). This represents an additional bias because the method of risk verification changes the discriminant function of the nomogram. For instance, Ploussard et al. reported the increase of prostate cancer detection rate, which depends on the number of cores: from 32.5% to 40.4% and 43.3% if 6, 12 or 21 cores were taken [26].


Finally, the risk calculators aim to help the physician in the biopsy decision. Recently, the integration of biomarkers (PHI index, PCA3 score) in these programs improved their accuracy [27]. Roobol et al. reported that the urologists may be able to select the individual screening schedule based on the risk threshold calculated in the last visit [28].


Conclusion


If the prostate cancer overdiagnosis and the increased number of prostate biopsies in France are debatable issues, the patients biopsied in this study had a higher risk of prostate cancer, calculated by the nomogram, than other studies. In this study, urologists take into consideration the patient's age and his life expectancy before performing prostate biopsy. Contrariwise, negative previous biopsy didn't change the threshold risk value prompting to perform biopsy, as though negative result didn't reassure urologists.


Disclosure of interest


The authors declare that they have no conflicts of interest concerning this article.




Table 1 - Characteristics of the overall population and ofthe both subgroups of patients with and without prostate cancer (Mann-Withney test, Fisher exact test or &khgr; 2).
  Overall population  Prostate cancer  No prostate cancer  P value 
n (%)   625 (100)  291 (46.6)  334 (53.4)   
Age (years)   65.4±6.3 (55-86)  66.7±6.8 (55-86)  64.3±5.6 (55-85)  P <0.0001 
Life expectancy          
>10 years (%)  568 (90.9)  249 (85.6)  319 (95.5)  P <0.0001 
<10 years (%)  57 (9.1)  42 (14.4)  15 (4.5)   
Total PSA (ng/mL)   8.8±6.3 (0.7-46)  10.0±7.9 (2-46)  7.7±4.3 (0.7-30)  P =0.001 
DRE          
Normal (%)  435 (69.6)  164 (56.4)  271 (81.1)  P <0.0001 
Suspicious (%)  190 (30.4)  127 (43.6)  63 (18.9)   
Prostate volume (mL)   47.8±20.6 (10-120)  43.8±19.8 (10-116)  51.3±20.7 (15-120)  P <0.0001 
Prior negative biopsy          
486 (77.8)  253 (86.9)  233 (69.8)  P <0.0001 
> 139 (22.2)  38 (13.1)  101 (30.2)   



Légende :
DRE: Digital Rectal Examination.



Table 2 - Estimated risk of prostate cancer with PCPT-CRC and SWOP-PRI (Mann-Whitney test).
  Overall population  Prostate cancer  No prostate cancer  P value 
Risk of positive biopsy (%)          
PCPT-CRC         
No family history  50.6±14 (14.9-75)  55.2±14.9 (17.7-75)  46.5±11.8 (14.9-75)  P <0.0001 
Family history  56.2±12.8 (18.6-75)  60.3±13.2 (24-75)  52.6±11.2 (18.6-75)  P <0.0001 
SWOP-PRI  31.2±17.3 (7-88)  38.4±18.8 (7-88)  24.9±12.9 (7-70)  P <0.0001 
Risk of high risk prostate cancer (%)          
PCPT-CRC  22.4±16.9 (2-75)  28.2±19.9 (2.6-75)  17.3±11.6 (2-75)  P <0.0001 
SWOP-PRI  14.8±18.2 (0.37±93)  22±22.7 (0.37-93)  8.4±9.3 (1-61)  P <0.0001 





Table 3 - Mean estimated risk according to age and negative prior biopsies (test One Way ANOVA, Mann-Withney test).
  n   PCPT-CRC  P   SWOP-PRI  P  
Age (years)            
<65  n =314  54.1±12.0  P <0.0001  28.9±15.9  P <0.0001 
65≤age<75  n =255  56.7±12.8    30.9±16.3   
≥75  n =56  65.6±12.4    45.1±22.1   
Previous negative biopsies ( n)            
No  n =486  58.2±12.3  P <0.0001  35.8±16.4  P <0.0001 
Yes           
Total  n =139  49.0±11.6    14.8±7.6   
n =87  41.3±12.9  P =0.01  15.5±7.6  P =0.3 
> n =52  45.6±10.2    13.9±7.6   





Table 4 - Estimated risk of prostate cancer with PCPT-CRC and SWOP-PRI in several clinical studies.
Authors  Population (n Prostate cancer (%)  PCPT-CRC risk (%)  AUC  SWOP-PRI risk (%)  AUC 
Parekh et al., 2006 [22 446  33.2  30.4  0.66 
Kaplan et al., 2009 [23 624  10.2  22.8 
Hernandez et al., 2009 [21 1280  35.6  43.6  0.67 
Cavadas et al., 2010 [25 545  35.2  51  0.74  22  0.80 
Trottier et al., 2011 [24 982  46.2  45  0.63  26  0.71 
Our series, 2013  625  46.6  50.6  0.67  31.2  0.72 




References



Rebillard X., et al. Projected incidence and mortality from urologic cancer in France in 2010 Prog Urol 2010 ;  20 (Suppl. 4) : S211-S214
Soulie M., et al. Prostatic cancers Prog Urol 2007 ;  17 (6) : 1159-1230
Salomon L., et al. Recommendations Onco-Urology 2010: prostate cancer Prog Urol 2010 ;  20 (Suppl. 4) : S217-S251 [inter-ref]
Thompson I.M., et al. Prevalence of prostate cancer among men with a prostate-specific antigen level<or =4.0ng per milliliter N Engl J Med 2004 ;  350 (22) : 2239-2246 [cross-ref]
Smith D.S., Humphrey P.A., Catalona W.J. The early detection of prostate carcinoma with prostate specific antigen: the Washington University experience Cancer 1997 ;  80 (9) : 1852-1856 [cross-ref]
Postma R., Schroder F.H. Screening for prostate cancer Eur J Cancer 2005 ;  41 (6) : 825-833 [cross-ref]
Ouzaid I., et al. A direct comparison of the diagnostic accuracy of three prostate cancer nomograms designed to predict the likelihood of a positive initial transrectal biopsy Prostate 2012 ;  72 (11) : 1200-1206 [cross-ref]
Oliveira M., et al. Head-to-head comparison of two online nomograms for prostate biopsy outcome prediction BJU Int 2011 ;  107 (11) : 1780-1783 [cross-ref]
Cussenot O., et al. The management of prostate cancer: a retrospective chart review in 808 French men undergoing biopsy Prog Urol 2013 ;  23 (5) : 347-355 [inter-ref]
Thompson I.M., et al. Assessing prostate cancer risk: results from the Prostate Cancer Prevention Trial J Natl Cancer Inst 2006 ;  98 (8) : 529-534 [cross-ref]
Ngo T.C., et al. The prostate cancer risk calculator from the Prostate Cancer Prevention Trial underestimates the risk of high grade cancer in contemporary referral patients J Urol 2011 ;  185 (2) : 483-487
Roobol M.J. The use of nomograms in the detection of prostate cancer Prostate 2006 ;  66 (12) : 1266-1267 [cross-ref]
Kranse R., Roobol M., Schroder F.H. A graphical device to represent the outcomes of a logistic regression analysis Prostate 2008 ;  68 (15) : 1674-1680 [cross-ref]
Albertsen P.C., et al. Impact of comorbidity on survival among men with localized prostate cancer J Clin Oncol 2011 ;  29 (10) : 1335-1341 [cross-ref]
Ploussard G., et al. Risk of repeat biopsy and prostate cancer detection after an initial extended negative biopsy: longitudinal follow-up from a prospective trial BJU Int 2013 ;
Gann P.H., et al. Risk factors for prostate cancer detection after a negative biopsy: a novel multivariable longitudinal approach J Clin Oncol 2010 ;  28 (10) : 1714-1720 [cross-ref]
Roobol M.J., et al. A risk-based strategy improves prostate-specific antigen-driven detection of prostate cancer Eur Urol 2010 ;  57 (1) : 79-85 [cross-ref]
Steyerberg E.W., Vickers A.J. Decision curve analysis: a discussion Med Decis Making 2008 ;  28 (1) : 146-149
Van Vugt H.A., et al. Prospective validation of a risk calculator which calculates the probability of a positive prostate biopsy in a contemporary clinical cohort Eur J Cancer 2012 ;  48 (12) : 1809-1815 [cross-ref]
Schroder F., Kattan M.W. The comparability of models for predicting the risk of a positive prostate biopsy with prostate-specific antigen alone: a systematic review Eur Urol 2008 ;  54 (2) : 274-290 [cross-ref]
Hernandez D.J., et al. Predicting the outcome of prostate biopsy: comparison of a novel logistic regression-based model, the prostate cancer risk calculator, and prostate-specific antigen level alone BJU Int 2009 ;  103 (5) : 609-614 [cross-ref]
Parekh D.J., et al. External validation of the Prostate Cancer Prevention Trial risk calculator in a screened population Urology 2006 ;  68 (6) : 1152-1155 [inter-ref]
Kaplan D.J., et al. Evaluation of the Prostate Cancer Prevention Trial Risk calculator in a high-risk screening population BJU Int 2010 ;  105 (3) : 334-337 [cross-ref]
Trottier G., et al. Comparison of risk calculators from the Prostate Cancer Prevention Trial and the European Randomized Study of Screening for Prostate Cancer in a contemporary Canadian cohort BJU Int 2011 ;  108 (8 Pt 2) : E237-E244
Cavadas V., et al. Prostate Cancer Prevention Trial and European Randomized Study of Screening for Prostate Cancer Risk Calculators: a performance comparison in a contemporary screened cohort Eur Urol 2010 ;  58 (4) : 551-558 [cross-ref]
Ploussard G., et al. Prospective evaluation of an extended 21-core biopsy scheme as initial prostate cancer diagnostic strategy Eur Urol 2012 ;
Perdona S., et al. Prostate cancer detection in the "grey area" of prostate-specific antigen below 10ng/mL: head-to-head comparison of the updated PCPT calculator and Chun's nomogram, two risk estimators incorporating prostate cancer antigen 3 Eur Urol 2011 ;  59 (1) : 81-87 [cross-ref]
Roobol M.J., et al. A calculator for prostate cancer risk 4 years after an initially negative screen: findings from ERSPC Rotterdam Eur Urol 2013 ;  63 (4) : 627-633 [cross-ref]






© 2015 
Elsevier Masson SAS. Tous droits réservés.