The original database consisted of 1,447 men who were scheduled for open retropubic RP at Sahlgrenska University Hospital in Göteborg, Sweden, during the study period 2001–2008, when standardized data recording into a quality assurance database was performed.
The quality control program was approved by the Ethical Committee at Göteborg University in 2001. Patients were mailed questionnaires, a cover letter with information regarding the quality assurance program, and a statement of voluntary participation in the study.
Of the 58 patients excluded, 38 were enrolled in another clinical RP trial (LAPPRO) , 4 underwent surgery outside the university hospital, 7 did not undergo surgery (with 4 switching to radiotherapy and 3 receiving hormonal therapy), and 9 did not undergo surgery but may have had surgery elsewhere for unknown reasons (moved or loss-to-follow-up). This left 1,389 patients in the cohort. Surgery dates ranged from Jan 2, 2001 through July 16, 2008.
Biochemical recurrence (BCR) post-prostatectomy was defined as a prostate-specific antigen (PSA) value > 0.2 ng/mL with at least one confirmatory rise. Measurements were obtained from medical charts.
Of 1,389 patients, an additional 109 were excluded from analysis since they were never administered any questionnaires, leaving 1,280 patients available for analysis of functional outcomes. Patients responded to questionnaires regarding continence and potency on 4 occasions, approximately 2 weeks preoperatively and at 6, 18, and 36 months after surgery.
Patients were administered questionnaires assessing urinary continence using pads . Continence was defined as no leakage or occasional leakage associated with physical activity requiring sporadic use of pads (score 0–1 on a 0–4 point scale, where 2–4 implies incontinence with regular use of pads).
Erectile function was assessed by the standardized International Index of Erection Function questionnaire IIEF-5 . The IIEF-5 score consists of 5 items with 6 responses. The total score is the sum of the 5 items and ranges from 5–25, the higher the score the better potency. For patients missing responses for an item, the sum of the remaining items was used. Potency was defined as an IIEF-5 score of ≥ 17, which corresponds to Rosen et al.’s categorization of mild ED (17–21) and no ED (22–25) . This also makes our study comparable to functional outcomes reported in an earlier study by Vickers et al. where postoperative potency was defined as 1 = normal, full erections and 2 = full, but diminished erections .
Missing data on potency status (potent/impotent) and continence (incontinent/continent) at 18 months were imputed following an algorithm assuming that few men regain and then lose function as well as that recovery of function can occur beyond 18 months. If a patient was missing a questionnaire at 18 months but reported potency at 6 months, he was assumed to be potent at 18 months. Comparably, a patient reporting impotence at 6 and 36 months was considered impotent at 18 months. A similar approach was used for the continence endpoint.
Patients reporting use of alprostadil injections for erectile aid were categorized as impotent (3 men preoperatively, 222 men at 18 months, and 198 men at 36 months). Men reporting use of PDE5-inhibitors were included in the analysis.
Data on a total of 25 individual surgeons working at the hospital during the study period were included in the database. We analyzed patient-reported functional outcomes for 9 surgeons who performed ≥ 20 surgeries during the study period. Data for these surgeons were entered both as random and fixed effects.
Logistic regression models were adjusted for age at surgery, PSA at diagnosis, pathologic stage (pT0, pT2, pT3), pathologic Gleason score (≤ 6, 7, ≥ 8), year of surgery, and surgical experience. We defined surgical experience as a variable that took into account both the surgeon’s prior experience, i.e., number of RPs before 2001, and the annual number of prostatectomies performed during the study period . To statistically test for heterogeneity, a random effect following an inverse Gaussian distribution was included in the model.
The logistic regression model used to predict the probability of potency at 18 months was restricted to men who were potent preoperatively and also adjusted for the IIEF-5 score as a continuous variable. The model for continence at 18 months was restricted to men who were continent preoperatively. Since there have been changes in patient characteristics as well as operative technique, we also included year of surgery as a covariate. Nerve sparing status was not included in the model since it is a surgical decision. Take the case of two surgeons, one of whom only spared nerves if the cancer was very low risk and accordingly resected far from the neurovascular bundles, the other of whom only resected nerves for advanced disease. Overall, the former surgeon would have far lower potency rates, but adjusting for nerve sparing would lead to higher apparent rates.
A log-logistic distributed parametric regression survival model was used to model BCR rates following adjustment for the same covariates as for functional outcomes. To statistically test for heterogeneity, a shared frailty survival model was fitted.
For each surgeon studied, forest plots were created for the adjusted predicted probability of potency and continence, respectively, using the mean value for the covariates from the fixed regression model with a 95% confidence interval.
A scatter plot was created for adjusted rates of continence and potency with the size of each surgeon’s data point proportional to the number of RPs (prior experience and during the study period) performed by each surgeon. Spearman’s rank correlation was applied to test the correlation between surgeons’ adjusted probabilities of potency and continence at 18 months as well as BCR.
Statistical analyses were performed using Stata v. 12.0 (Stata Corp, College Station, TX, USA).