We conducted a retrospective cohort study of patients who underwent pelvic floor reconstructive surgery with 5 surgeons at a tertiary referral centre from January 2015 to October 2017. This project was reviewed and approved by the Foothills Medical Centre Research Ethics Boards (IDs# CHREB150706). Surgeries included for analysis were obliterative procedures (with or without MUS) and reconstructive procedures that addressed the vaginal apex (including vaginal hysterectomy with sacrospinous or uterosacral vault suspension, with or without anterior and posterior colporrhaphy, sacrospinous vault suspension, sacrocolpopexy (SCP), with or without MUS). Cases were identified in the Section of Pelvic Medicine & Reconstructive Surgery Surgical Booking Database. Peri-operative data was entered as part of routine clinical practice in the city-wide inpatient EMR (Sunrise Clinical Manager) and abstracted by a FPMRS/Urogynecology fellow to a database designed for research purposes.
Single site prolapse surgeries such as isolated anterior or posterior repairs were not captured by the dataset. All MUS procedures at time of pelvic organ prolapse (POP) surgery were performed with the use of vaginal xylocaine 1% with 1:100,000 Epinephrine mixed 1:1 with sterile water, in the range of 10–40 ml for hydro dissection during placement. Patients were excluded from our study if they underwent hysterectomies that did not include an apical suspension procedure (such as those performed for non-prolapse indications), isolated incontinence surgeries, vesicovaginal and rectovaginal fistula repairs, as well as day surgery cases (such as dilation & curettage, laparoscopic resection of endometriosis). Patients who had pre-operatively elevated post-void residuals ≥ 150 ml were excluded from this dataset. Data were extracted from the patient’s post-operative electronic chart. Variables extracted included: age, body mass index (BMI), American Society of Anesthesiologists (ASA) score as a marker of health, previous pelvic floor surgery and type of procedure performed. Perioperative variables extracted included: length of hospital stay (in days), number of voiding trials in hospital and whether the patient was sent home with a catheter or self-catheterizing.
At the tertiary centre where the study was conducted, two bladder protocols are administered in the post-operative period at the attending surgeon’s discretion. The first (Retrograde Protocol) is carried out by retro-filling the bladder with 300 ml of normal saline or sterile water through the foley catheter which was left in situ overnight. The voided volume is measured in a voiding hat by nurses and post void residual (PVR) is determined based on the volume voided. The second and subsequent (Spontaneous Filling Protocol) bladder protocols involve removing the foley catheter and allowing the bladder to spontaneous fill. The patient must void within 4 hours of catheter removal, and PVR is measured by bedside bladder scanner. Voided volumes are measured in a voiding hat by nurses and the PVR measured with a bladder scanner. If patients have two consecutive voids > 200 ml with a PVR ≤ 150 ml, then they are considered to have passed the bladder protocol and monitoring of voiding stops. If the first TOV is failed, then two further consecutive voids must be “passed” in order to pass the voiding protocol. If the PVR is > 250 ml, an in & out catheter is placed to both confirm PVR and decompress the bladder. With both of these protocols, the post-operative indwelling catheter is removed at 6 AM on Day 1 in compliance with Early Recovery After Surgery (ERAS) guidelines for minimal duration of catheterization [19]. Patients must pass a minimum of two consecutive TOVs in order to pass the voiding protocol. This is based on the unreliability of PVR measurements a the need for repetition to confirm consistency [20]. A study by Dunsmuir et. al., showed that only one-third of patients had approximately constant PVRs (variation in range < 120 mL), and so repeated transabdominal bladder ultrasound is required [21]. A patient may have a falsely low PVR and may represent with urinary retention if only one PVR is done as part of a TOV. Type of bladder protocol administered and results are then documented in the electronic patient chart.
If the patient has met all other criteria for discharge, but have not passed their TOV, they are given the option to perform self-catheterization after each measured void at home or to be discharged with an indwelling foley catheter. Those patients who elect to self-catheterize upon discharge can discontinue self- catheterization once their voiding pattern demonstrates voids of > 200 ml with a PVR ≤ 150 ml for 3 consecutive voids. Those who elect to discharge with a catheter are reviewed in clinic for a retrograde TOV performed 4–7 days after discharge. This is in keeping with a prospective randomized controlled trial by Schachar et al. which showed that women with POUR after prolapse surgery had a 7-fold higher risk of failed repeat office TOV if performed on postoperative day 4 compared to postoperative day 7 [22].
Statistics were conducted using Stata 16 (College Station, Texas). Descriptive statistics were used to describe the study sample, calculating proportions, mean, and median values for demographic characteristics. Elevated PVR was defined as having a PVR of greater than 150 cc. We calculated descriptive statistics, chi-square tests, and crude odds ratios for elevated PVR on the second void, stratifying the results by POP procedure type (reconstructive vs. obliterative) and presence of MUS procedure. Reconstructive procedures were then further stratified by concomitant versus previous hysterectomy to explore for effect of hysterectomy on POUR risk factors.
Sample size for regression-based analyses is difficult to compute a-priori. However, a general rule of 10 cases/events per regression variable is accepted for logistic regression. Our original model included over 10 variables and multiple interaction terms. For this reason, we estimated we would need 200 events for our logistic regression model. Recognizing that some cases may have missing information due to charting errors, we increased our sample size by 20% to 240 events. Estimating the prevalence of post-operative urinary retention to be 40% this would be a sample size of 600 women. However, recognizing that other forms of regression analysis would require samples larger than logistic regression due to multiple group comparisons, the same size was again increased by 50% to 900. Based on our average surgical volume, it was estimated that a review of all cases over 34 months would provide this volume (January 2015-October 2017).
We conducted several regression analyses, recognizing that the concept of POUR can be defined multiple ways. Binary logistic regression evaluated for the effect of patient age, BMI, ASA Score, bladder protocol type, anesthesia type (general vs regional), surgeon, concomitant hysterectomy, anterior and posterior vaginal wall repairs on the outcome of passing the TOV at the second post-operative void. Age and BMI were explored in linear and non-linear fashion. Interaction terms between age and BMI and combinations of surgical procedures were explored. We also explored the potential effect of surgeon on the voiding outcomes through mixed effects logistic regression and the value of the variance reported by McFadden’s R-squared (not reported).
Association with the absolute number of voids taken to pass the bladder protocol and peri-operative variables were explored by zero-truncated Poisson regression where the dependent variable is an observed, non-zero count, assumed to follow a Poisson distribution. The Poisson regression modelling examined whether any measured clinical variable (including age, BMI, MUS, ASA score, anaesthesia type, hysterectomy, uterosacral or sacrospinous suspension, and laparoscopic SCP) predicted the number of voids to pass the bladder protocol. Both crude odds ratios and adjusted odds ratios were calculated.
Plotting a histogram of the number of voids required to pass the TOV allowed us to determine where natural clustering of the data occurred, and to classify pattern groups of the number of voids needed to pass the TOV were performed which were ordinal in nature. We assessed the proportional odds assumption using the Brant test. Results of the Brant test indicated that the proportional odds assumption was not violated and that we can assume that the relationship between each pair of outcome groups is the same (i.e., being in class 2 or 3 compared to 1 is the same as being in group 3 compared to 2 and 1). We calculated both crude models (using the single predictor variable and the outcome of void class) and adjusted models (using all predictor variables and the outcome of void class).
For all of the regression modelling, we assessed for significance at the p < 0.05 level. For the binary outcome of passing TOV on the second void, sensitivity analyses using a cutpoint of a 3rd void was also performed. Use of multiple regression methodologies acted as sensitivity analysis to ensure our results were consistent across different ways of classifying POUR.