Mortality at 120 days after prostatic biopsy: A population-based study of 22,175 men

Authors


Abstract

Trans-rectal ultrasound guided biopsy of the prostate represents the diagnostic standard for prostate cancer, but its mortality rate has never been examined. We performed a population-based study of 120-day mortality after prostate biopsy in 22,175 patients, who underwent prostate biopsy between 1989 and 2000. The control group consisted of 1,778 men aged 65–85 years (median 69.5), who did not undergo a biopsy. Univariable and multivariable logistic regression analyses were performed in 11,087 of 22,175 (50%) men subjected to prostate biopsy, to identify predictors of 120-day mortality. Variables were age at biopsy, baseline Charlson comorbidity index and cumulative number of biopsy procedures. We externally validated the model's predictors in the remaining 50% of men. Overall 120-day mortality after biopsy was 1.3% versus 0.3% (p < 0.001) in the control group. Of men aged ≤60 years, 0.2% died within 120 days versus 2.5% aged 76–80. Zero Charlson comorbidity score yielded 0.7% mortality versus 2.2%, if 3–4. First ever biopsy procedures carried a higher mortality risk than subsequent procedures (1.4 vs. 0.8 vs. 0.6%). In the multivariable model, first ever biopsy, increasing age and comorbidity predicted higher mortality. Overall, the model's variables were 79% accurate in predicting the probability of 120-day mortality after biopsy. In conclusion, our data suggest that prostate biopsy might predispose to higher mortality rate. The certainty of this association remains to be proven. © 2008 Wiley-Liss, Inc.

Prostate cancer (PCa) is one of the most common male malignancies in the United States.1 Trans-rectal ultrasound (TRUS) guided biopsy of the prostate represents the cornerstone diagnostic procedure in patients at risk for PCa. The indications for performing a TRUS biopsy have substantially broadened over the past 2 decades.2 An increasing number of men without digital rectal examination (DRE) abnormalities and with increasingly lower prostate specific antigen (PSA) blood levels are subjected to TRUS biopsy and may result in higher absolute number of complications.3–8 Infectious complications are most common and may result in septic shock or other life-threatening conditions. Instances of severe hemorrhage,9, 10 osteitis pubis,11 abscess formation,12–14 endocarditis,15, 16 multi-drug resistant infections,17 sepsis or septic shock have been reported.12, 18 These complications are rare but nonetheless can result in biopsy related mortality. To the best of our knowledge, except for isolated case reports, no previous large-scale analysis addressed the rate of mortality after TRUS prostate biopsy. To address this void, we performed a population-based study of 120-day mortality after TRUS biopsy in 22,175 men subjected to TRUS-guided prostate biopsy. Moreover, we tested whether patients at a negligible risk of 120-day mortality can be accurately identified.

Material and methods

The Quebec Health Plan represents the exclusive insurer in the Province of Quebec and its database allows ascertainment of all health services covered by the Plan. These include prostate biopsy. Moreover, the Health Plan relies on ICD-9 diagnostic codes and their respective dates allow defining individual Charlson comorbidity index scores.19, 20 Finally, the Health Plan provides crude survival data for all enrollees.

The Health Plan data allowed us to identify all men who underwent a prostate biopsy between 1989 and 2000. Each record included age at biopsy, number of previous biopsy sessions if applicable, vital status at 120 days after prostate biopsy, as well as the Charlson comorbidity index score.20 None of included patients received definitive treatment for PCa before the end of the 120days interval. The analyses targeted 22,175 evaluable patients.

For comparison purposes, we also used a convenience sample of 1,778 men aged 65–85 years (mean 71, median 69), who were unexposed to a biopsy procedure. These men were entered into the Health Plan records between 1989 and 1995. The follow-up period of 120 days consisted of the 120-day interval starting on the 1st day of registration with the Health Plan. Since diagnostic codes used in the definition of the Charlson Comorbidity Index are entered prospectively in the Health Plan database, the calculation of the Charlson Comorbidity Index was not possible for the control group and only overall and age-stratified comparisons with the biopsy-exposed cohort could be made. Since no new enrollees were identified in the 86–90 years of age category, mortality comparisons could not be made for this age stratum.

Statistical analyses

The proportion of deaths from any cause at 120 days was compared using the chi-square statistic, between individuals exposed to prostate biopsy and the controls who did not have a biopsy. Subsequent statistical analyses relied of logistic regression models, which targeted overall mortality at 120 days in a 50% random sample (n = 11088) from the original cohort of 22,175 patients exposed to a prostate biopsy. This cohort will be referred to as the development cohort. Predictors of 120-day mortality consisted of age at biopsy (coded as cubic spine to allow non-linear effects), Charlson comorbidity index at biopsy and the cumulative number of biopsy procedures that each individual was exposed to between 1989 and 2000, including the last index biopsy procedure. These variables were then used within a multivariable logistic regression model designed to predict the individual probability of 120-day mortality after prostate biopsy. In multivariable models, fast backward variable selection relied on Akaike's information criteria to identify independent predictors of biopsy mortality.21

To confirm the validity of our model and the significance of its predictor variables we tested the accuracy of the model (quantified with the area under the receiver operating characteristics curve [AUC]) within the remaining 50% of our population (n = 11087). The model was then displayed graphically to illustrate the multivariable effects of the predictors on the probability of 120-day mortality after biopsy. All statistical tests were performed using S-PLUS Professional, version 1 (MathSoft, Seattle, WA). All tests were two-sided with a significance level set at 0.05.

Results

Patient characteristics are shown in Table I. Mean age at biopsy was 69.3 years (median 69). The distribution of patients across 5 year intervals was as follows: ≤60 years 12.1%, 61–65, 18.9%; 66–70, 27.0%; 71–75, 21.6%; 76–80, 12.8%; 81–85, 5.8%; 86–90, 1.5% and >90, 0.2%. Charlson comorbidity index score was 0 in 40.7%, 1–2 in 40.4%, 3–4 in 14.4%, 5–6 in 3.7% and 7 or more in 0.9% of patients. Most (78.8%) underwent 1 prostate biopsy session. Two or more biopsy procedures were performed in 15.9 and 5.3% of patients, respectively. The nomogram development cohort and the external validation cohort failed to demonstrate statistically significant differences in the distribution of known variables, such as age at biopsy, Charlson comorbidity index score, total number of biopsy procedures, year of biopsy and the region of residence.

Table I. Descriptive characteristics of patients included in the analyses (n = 22175) and of the control population (n = 1778)
 OverallNomogram developmentExternal validationp-valueControl population
  1. Values in paranthesis indicates percentages.

Number of patients22,17511,08811,087 1,778
Age (years)   0.8 
 Mean (Median)69.3 (69)69.3 (69)69.3 (69) 71.1 (69.5)
 Range36–10136–10141–97 65–85
    1.0 
  ≤602,693 (12.1)1,345 (12.1)1,348 (12.2) 
  61–654,184 (18.9)2,091 (18.9)2,093 (18.9) 
  66–705,987 (27.0)3,013 (27.2)2,974 (26.8) 967 (54.4)
  71–754,795 (21.6)2,377 (21.4)2,418 (21.8) 431 (24.2)
  76–802,838 (12.8)1,429 (12.9)1,409 (12.7) 250 (14.1)
  81–851,294 (5.8)648 (5.8)646 (5.8) 130 (7.3)
  86–90333 (1.5)160 (1.4)173 (1.6) 
  >9051 (0.2)25 (0.2)26 (0.2) 
Charlson comorbidity index score   1.0 
 09,020 (40.7)4,510 (40.7)4,510 (40.7) 
 1–28,949 (40.4)4,475 (40.4)4,474 (40.4) 
 3–43,183 (14.4)1,591 (14.3)1,592 (14.4) 
 5–6822 (3.7)411 (3.7)411 (3.7) 
 7 or more201 (0.9)101 (0.9)100 (0.9) 
Number of biopsy session   0.4 
 117,471 (78.8)8,693 (78.4)8,778 (79.2) 
 23,531 (15.9)1,797 (16.2)1,734 (15.6) 
 3 or more1,173 (5.3)598 (5.4)575 (5.2) 
Year or biopsy   0.9 
 1989–901,608 (7.3)802 (7.2)806 (7.3) 636 (35.8)
 19911,047 (4.7)546 (4.9)501 (4.5) 302 (17.0)
 19921,689 (7.6)852 (7.7)837 (7.5) 230 (12.9)
 19932,303 (10.4)1,160 (10.5)1,143 (10.3) 214 (12.0)
 19942,351 (10.6)1,167 (10.5)1,184 (10.7) 184 (10.3)
 19951,947 (8.8)980 (8.8)967 (8.7) 212 (11.9)
 19961,910 (86)947 (8.5)963 (8.7) 
 19972,180 (9.8)1,065 (9.6)1,115 (10.1) 
 19982,218 (10.0)1,103 (9.9)1,115 (10.1) 
 19992,618 (11.8)1,324 (11.9)1,294 (11.7) 
 20002,304 (10.4)1,142 (10.3)1,162 (10.5) 
Region of residence   0.6 
 Montreal5,632 (25.4)2,797 (25.2)2,835 (25.6) 
 Bas Saint Laurent845 (3.8)441 (4.0)404 (3.6) 
 Saguenay-Lac Saint Jean721 (3.3)360 (3.2)361 (3.3) 
 Quebec City1,959 (8.8)984 (8.9)975 (8.8) 
 Mauricie et Centre du Quebec1,775 (8.0)884 (8.0)891 (8.0) 
 Estrie1,078 (4.9)539 (4.9)539 (4.9) 
 Chaudiere-Appalches1,386 (6.3)701 (6.3)685 (6.2) 
 Laval1,079 (4.9)556 (5.0)523 (4.7) 
 Lanaudiere921 (4.2)440 (4.0)481 (4.3) 
 Laurentides1,474 (6.6)755 (6.8)719 (6.5) 
 Monteregie3,611 (16.3)1,821 (16.4)1,790 (16.1) 
 Others1,694 (7.6)810 (7.3)884 (8.0) 
Mortality at 120 days279 (1.3)144 (1.3)135 (1.2)0.66 (0.3)

At 120 days, 279 of 22175 men died, which resulted in an overall mortality rate of 1.3% versus 6 of 1778 men (0.3%, p < 0.001) in the control group. The mortality rate in individuals aged 85 years of age or less was 1.1% in the cohort of men exposed to prostate biopsy versus 0.3% in the control group (p = 0.0018). Age-specific mortality was respectively 0.2, 0.2, 0.8 and 0.8%, for age categories 65–70, 71–75, 76–80 and 81–85 years. Figure 1a graphically displays age-specific overall mortality rate at 120 days after prostate biopsy in the entre cohort of 22,175 men versus the control group of 1,778 men (data not shown for men older than 85). Despite a difference in the recorded mortality rate between the 2 groups, across all age ranges, the confidence intervals around the survival estimates of biopsy exposed and biopsy unexposed men overlap. Figure 1b illustrates the difference in the rate of mortality in controls relative to men exposed to prostate biopsy, after restricting the biopsy population to those men without any comorbidity, according to the Charlson index. Figures 1c1f respectively illustrate the mortality rates according to the time interval after biopsy, the Charlson comorbidity index score, the number of biopsy session and the year of biopsy during the study period in the population of 22,175 men.

Figure 1.

Mortality at 120 days after prostate biopsy according to age category (Panel 1a). Mortality at 120 days after prostate biopsy according to age category in men with no comorbidities (Panel 1b). Mortality at 30, 60, 90 and 120 days after biopsy (Panel 1c). Mortality at 120 days after prostate biopsy according to Charlson comorbidity index score (Panel 1d), number of biopsy session (Panel 1e) and year of biopsy (Panel 1f). Panels 1a–1c also show the mortality rate in the control population. [Color figure can be viewed in the online issue, which is available at www.interscience.wiley.com.]

For example, the mortality in patients younger than 60 years old was 0.2 versus 2.5% for men aged 76–80 years (Table II). Similarly, the mortality in patients without comorbidities was 0.7 vs. 2.2% in patients with 3–4 comorbidities. Finally, the mortality in patients subjected to 3 or more biopsy procedure was 0.6% versus 1.4% in patients subjected to only 1 biopsy procedure. During the study period, a decreasing albeit non-significant (p = 0.1) trend in 120-day mortality rate was recorded.

Table II. Mortality at 120 days after biopsy, according to age, comorbidity, number of biopsy session and year of biopsy
 Biopsy populationControl population
n (%)95% confidence intervalsn (%)95% confidence intervals
  1. Data for controls are only stratified according to age.

Total279 (1.3)1.11–1.416 (0.3)0.12–0.73
Age    
 ≤605 (0.2)0.06–0.43
 61–6512 (0.3)0.15–0.50
 66–7037 (0.6)0.44–0.852 (0.21)0.03–0.75
 71–7560 (1.3)0.96–1.611 (0.23)0.01–1.29
 76–8072 (2.5)1.99–3.182 (0.8)0.09–2.86
 81–8557 (4.4)3.35–5.671 (7.7)0.02–4.21
 86–9025 (7.5)4.92–10.88
 >9011 (21.6)11.29–35.32
Charlson comorbidity index score    
 065 (0.7)0.5–0.9
 1–2107 (1.2)0.9–1.4
 3–469 (2.2)1.7–2.7
 5–626 (3.2)2.1–4.6
 7+12 (6.0)3.1–10.2
Number of biopsy session    
 1242 (1.4)1.2–1.5
 230 (0.8)0.5–1.2
 3+7 (0.6)3.1–10.2
Year of biopsy    
 1989–9028 (1.7)1.1–2.5
 199118 (1.7)1.0–2.7
 199232 (1.9)1.3–2.6
 199331 (1.3)0.9–1.9
 199430 (1.3)0.8–1.8
 199517 (0.9)0.5–1.4
 199627 (1.4)0.9–2.0
 199724 (1.1)0.7–1.6
 199827 (1.2)0.8–1.7
 199928 (1.1)0.7–1.5
 200017 (0.7)0.4–1.2

Table III shows univariable and multivariable logistic regression models addressing to the mortality at 120 days after prostate biopsy in the nomogram development cohort. In univariable analyses, age (p < 0.001), Charlson comorbidity index score (p < 0.001) and the total number of biopsy procedures (p = 0.02) were highly statistically significant, whereas the year of biopsy (p = 0.1) and the region of origin (p = 0.6) failed to reach statistical significance. In multivariable analyses, age, Charlson comorbidity index score and the total number of biopsy procedures represented independent predictors of 120-day mortality (all p ≤ 0.03), were not removed from the final multivariate model and were included in the nomogram.

Table III. Univariable and multivariable logistic regression analyses predicting mortality at 120 days after prostate biopsy in the nomogram development cohort (n = 11,088)
PredictorsUnivariable logistic regressionMultivariable logistic regression
OR95% CIp-valueOR95% CIp-value
  1. OR, Odds ratio; CI, confidence interval.

Age at biopsy<0.001<0.001
 Age1.060.87–1.290.51.060.87–1.280.6
 Age1.230.78–1.910.41.210.77–1.880.4
 Age0.520.47–7.700.40.550.13–2.200.4
Charlson comorbidity index score<0.001<0.001
 1–2 vs. 01.771.15–2.740.011.430.92–2.230.1
 3–4 vs. 03.332.07–5.37<0.0012.201.35–3.570.002
 5–6 vs. 04.572.38–8.78<0.0012.651.36–5.170.004
 7+ vs. 08.843.61–21.64<0.0014.521.80–11.320.001
Cumulative number of prostate biopsy procedures0.020.03
 2 vs. 10.490.28–0.870.0150.550.31–0.960.02
 3 or more vs. 10.450.17–1.230.10.220.05–0.890.2
Year of biopsy0.1   
 1991 vs. 1989–901.360.62–3.010.4   
 1992 vs. 1989–901.380.68–2.820.4   
 1993 vs. 1989–900.900.44–1.870.8   
 1994 vs. 1989–900.900.43–1.860.8   
 1995 vs. 1989–900.500.21–1.210.1
 1996 vs. 1989–900.650.28–1.490.3   
 1997 vs. 1989–900.690.31–1.520.4   
 1998 vs. 1989–900.840.40–1.770.6   
 1999 vs. 1989–900.560.25–1.220.1   
 2000 vs. 1989–900.480.21–1.130.1   
Region of residence0.2   
 Bas Saint Laurent vs. Montreal1.080.48–2.430.8   
 Saguenay – Lac Saint Jean vs. Montreal1.330.59–2.990.5   
 Quebec City vs. Montreal1.320.76–2.290.3   
 Mauricie – Center du Quebec vs. Montreal0.850.43–1.660.6   
 Estrie vs. Montreal0.380.12–1.220.1   
 Chaudiere – Appalache vs. Montreal1.370.74–2.530.3
 Laval vs. Montreal0.610.24–1.550.3   
 Lanaudiere vs. Montreal1.090.48–2.440.8   
 Laurentides vs. Montreal0.720.34–1.540.4   
 Monteregie vs. Montreal0.560.31–1.010.055   
 Others vs. Montreal0.590.26–1.310.2   

Figure 2a shows the graphical representation of the effect of the predictor variables on the rate of mortality at 120 days after biopsy. The model's overall predictive accuracy was confirmed in the independent cohort of 11,087 patients, where the model demonstrated 79.3% accuracy (AUC p < 0.001). Moreover, the testing of our model's performance characteristics confirmed excellent calibration, which was evidenced by virtually perfect correlation between predicted probabilities and observed rates of 120-day mortality (Fig. 2b).

Figure 2.

Nomogram predicting 120-day mortality after prostate biopsy (Panel 2a) and its calibration plot (Panel 2b). *Including the planned biopsy; Nomogram instructions: To obtain nomogram predicted probability of 120-day mortality after prostate biopsy, locate patient values at each axis. Draw a vertical line to the “Point” axis to determine how many points are attributed for each variable value. Sum the points for all variables. Locate the sum on the “Total Points” line to be able to assess the individual probability of 120-day mortality after biopsy on the “Probability of mortality at 120 days” line. Calibration plot: This calibration plot shows the performance of the nomogram in an external validation cohort. Specifically, nomogram predicted probabilities are compared to the observed proportions of 120-day mortality after biopsy. X-axis represents nomogram predicted probability of 120-day mortality after biopsy. Y-axis shows observed proportion of 120-day mortality after biopsy. Perfect prediction would correspond to a slope of 1 (diagonal 45° broken line). Solid line indicates nomogram performance in the external cohort.

Discussion

PCa is one of the most common non-cutaneous male malignancies in North America.1 TRUS guided prostate biopsy is the diagnostic gold standard for men at risk of PCa. Despite its overall proven safety profile, prostate biopsy has been associated with life-threatening complications.9–18 However, we are unaware of large scale, population-based studies that reported short-term mortality rates after biopsy. On the basis of this limitation, we decided to address 120-day mortality in a large population-based administrative database. We hypothesized that prostate biopsy may be related to non-negligible rate of 120-day mortality and we tested our hypothesis in a cohort of 22,175 men subjected to a prostate biopsy. We used a convenience sample of 1,778 men unexposed to prostate biopsy to assess 120-day mortality in a control group.

Our analyses demonstrated a 1.3% overall mortality rate at 120-days after biopsy vs. 0.3% in the control group. The mortality rate according to 5-year age strata from 66–85 years ranged from 0.6 to 4.4% in the cohort of 22,175 biopsied patients versus from 0.2 to 0.8% in the control group of 1,778 men. This difference between controls and biopsy-exposed individuals persisted, when we restricted the analysis to men without any comorbidities according to the Charlson index and to 85 years of age or younger.

In the univariable logistic regression analyses predicting 120-days mortality in the development cohort of 11,088 men, age, Charlson comorbidity and the total number of biopsy procedure were statically significant predictors (Table II). In the multivariable logistic regression models, age, Charlson comorbidity index score and the total number of biopsy procedures represented independent predictors. We confirmed the accuracy of this model in the remaining 50% of the population, where the model achieved 79.3% accuracy (AUC p < 0.001).

All diagnostic procedures are associated with complications. Infectious complications represent the main adverse outcomes of prostate biopsy. Severe infections and septic shock have been reported in several publications.12, 15, 17, 18 The mortality of septic shock is roughly 50%.22 Therefore, prostate biopsy mortality rate is non-negligible. However, this most feared complication has never been reported in large scale studies. In consequence, there are no validated risk factors for prediction of short-term mortality after prostatic biopsy. An explanation for the unavailability of prostatic biopsy mortality data is related to lack of inclusion of this endpoint in the standard institutional urologic databases, where focus is placed on detection rates and on less severe complications. Alternatively, severe septic complications may be treated at different institution and patients may be lost from urological follow-up, as most will not volunteer for another biopsy session. Therefore, administrative population-based databases represent a valid alternative for capturing these relatively rare events.

Our findings have several crucial clinical implications. PCa is a slowly progressive disease. Bill-Axelsson et al. demonstrated very low disease specific mortality rates at 10 years, either in the watchful waiting or in the radical prostatectomy arm, especially in patients older than 65 years.23–25 Others suggested that treatment benefits are not negligible up to 75 years of age.26, 27 This implies disease detection up to that age. Our data support a more conservative approach to diagnostic prostate biopsy, as advanced age represented the most informative predictor of prostate biopsy mortality at 120 days. For example a 70 years old man, without any comorbidities who is undergoing the first biopsy has a <1% risk of 120-day mortality. However, the risk increases to 5%, if an 85-year-old without comorbidities is subjected to his first biopsy. Interestingly, the number of biopsy session exerts a protective effect on the recorded rate of mortality. This effect is highly likely attributable to a selection of men among those exposed to more than 1 biopsy session.

Taken together, our findings indicate that the 120-day mortality rate is higher in men exposed to prostate biopsy than in control subjects. Our nomogram-based multivariable analysis demonstrates the specific effect of studied variables on the individual risk of mortality after prostate biopsy. It suggests that the indications for biopsy may warrant a reappraisal. For example, older and less healthy men should be carefully prescreened to weigh the risks and benefits of a biopsy. These men represent the category of individuals who benefit the least from the diagnosis of early localized PCa.23, 28 Moreover, a more restrictive approach to prostate biopsy may reduce the cost related to this diagnostic procedure. Indirectly, a more restrictive approach to biopsy may result in lower detection rate of indolent PCa, especially in older individuals. Conversely, a more restrictive approach will also result in a stage shift at diagnosis, whereby some of the older and sicker individuals will be diagnosed with more advanced stages of PCa. These men may require systemic therapies for more locally advanced disease, if they do not succumb to old age or comorbid conditions.

Our findings are not devoid of limitations. Lack of cause-specific mortality data represents the foremost limitation. It may be argued that the overall health status of men in whom a biopsy was performed should not be so poor that they succumb to comorbid conditions within 120 days after biopsy. Individuals whose health status is so poor would not be considered for biopsy by any urologist. Therefore, it might be argued that the deaths that occurred within 120 days after biopsy were attributable to the prostate biopsy. Our control group at least partly validates this observation. However, we do not have cause-specific mortality to ensure that the deaths were not related to other causes, in exceptional cases. Unfortunately, a controlled trial of the effect of biopsy on mortality cannot be considered, due to ethical issues. However, only such design would definitively prove the causality effect. In its absence our findings need to be considered as partial proof, or more specifically as an association without causation. The nature of our database also prevented adjusting for antibiotic prophylaxis. Differences in antibiotic regimens may contribute to differences in 120-day mortality rates. Moreover, the extent of the biopsy procedure also could not be included. However, more extensive biopsies, such as 12 or more cores schemes, may contribute to small absolute but important relative differences in septic complication rates. These differences may not be interpreted as significant in underpowered small scale negative studies. Absence of indication for biopsy represents another limitation of the current study. For example, 20–30% of the patients underwent a biopsy despite being over 75 years of age. It is conceivable that the indication for biopsy was to confirm clinically evident disease prior to initiation of androgen deprivation therapy. Such practice is no longer required, especially when reversible LHRH agonist therapy is delivered. Our data also indicate that the median age at biopsy has decreased over time (from 71 to 69 years). It might be postulated that younger age at biopsy contributed to the decreasing mortality rate over time (from 1.7 to 0.7%). This is particularly encouraging given that the absolute numbers of men subjected to biopsy increased from less than a 1,000 to over 2,000 per year over time. Despite these limitations, our data provide the first population-based assessment of biopsy mortality and indicate that in some, relatively rare instances mortality might be attributable to the biopsy procedure. The novelty of our findings requires corroboration in other population-based registries and limits our report to a hypothesis generating manuscript

In conclusion, prostate biopsy is associated with higher mortality than that observed in controls unexposed to biopsy. The risk of death at 120-days after biopsy increases with age and Charlson comorbidity index. Conversely, increasing number of previous biopsy sessions exerts a protective effect. Moreover, the causality of this association has not been proven.

Acknowledgements

Dr. Pierre I. Karakiewicz is partially supported by the University of Montreal Urology Associates, Fonds de la Recherche en Santé du Québec, the University of Montreal Department of Surgery and the University of Montreal Foundation.

Ancillary