Predictors for Outcome and Complications Related to Urinary Diversion

Discussion

Choosing the right diversion method for each individual patient is mainly dependent on clinical factors. Thus, the decision process lacks objective parameters and is biased by empirical factors. One important factor is highlighted in the study of Hart et al. (4). The authors emphasize the good outcome in the overall quality of life for patients in all analyzed types of urinary diversions. Postoperative quality of life is consequently not critical for the choice of the diversion method. This factor was therefore not implemented in our analysis. Their study, however, as that of others, lacks specific differentiation in subgroups according to their pre- and postoperative data. The proportional distribution of diversion methods in our cohort did not vary from those of other high-volume centers, allowing the assumption of a similar decision process throughout urological clinics.

One goal of our study was to identify an independent specific preoperative parameter for choosing the right diversion method. We focused on a major complication, the AI, presuming it to be critical, as it is generally associated with high mortality and a high morbidity. This presumption could be proven in our data: AI leads to a high rate of morbidity (p=0.002). Interestingly, AI was not associated with a high rate of mortality (p=0.159). Accordingly, these results show a good management of postoperative complications in the two departments and that AI is not as a critical complication as expected.

Our baseline population characteristics, such as mean age, gender distribution or smoking status did not differ much from the average population of high-volume studies (5, 6). However, the number of diabetics was higher in our cohort (20%) than in the average matched population (12%) (7). Diabetes mellitus impairs among other processes the physiological pathways of wound healing (8). This could bias the general outcome and survival of our patients. Especially the physiology and odds of healing in bowel anastomoses could be affected as well (9). A further aspect is that obesity causes a significant increase in costs due to longer hospital stay and complications (10). However, our analysis shows no impact of diabetes on the event of AI or CDS (p=0.265 and p=0.871). AI was associated with severe complications (CDS >IIIa; p<0.001), a high ASA-score (p=0.047) and a longer hospital stay (p<0.001). Thus, it did not correlate with any specific diversion method. Neither AI (p=0.226) nor CDS (p=0.780) showed statistical significance regarding this aspect, so that the event of complications is not affected by the extent of the surgical procedure. Apparently, the operation method is not critical for the patient’s outcome. Furthermore, smocking and diabetes are risk factors for longer operation times, and with longer operation times the risk for another surgery also increases (11).

Our subgroup analysis revealed preoperative vascular comorbidities as an independent variable for the event of a severe postoperative complication (p=0.007). However, the analysis correlating the mortality of diversion method with vascular comorbidities did not reach any statistical significance (p=0.320). Patients with severe vascular comorbidities should be informed and monitored postoperatively regarding potential postoperative complications, independently of the diversion method. Furthermore, it is also known that a significant increase in complications is found in the leal conduit group (12). In the study of Monn et al., where a comparable cohort of patients was investigated, 50% of the patients required blood transfusion (13). The increase in blood transfusions was also investigated in the analyzed cohort, with a 57% overall transfusion rate. The analysis of the correlation between transfusion rate and a high CDS did not show any statistical significance (p=0.806). The mortality in the cohort was too small for a correlation analysis with transfusion rates. However, transfusion rates did not have a predictive value for postoperative outcome in cystectomies.

Nieuwenhuijzen et al. correlated urinary diversions with ASA-score, and minor and major complications (14). Patients who died within the first 30 postoperative days were omitted from Nieuwenhuijzens’ study. Unfortunately, the authors did not specify why this important aspect was left out. Intestinal AIs were assessed in the study of Nieuwenhuijzen et al., but not specifically evaluated. Only a high ASA-score (3 or more) was associated with early or late complications. In our study, we found a significant correlation between AI and ASA-score (p=0.047), but not for CDS >IIIa (p=0.968) despite the fact that our study was performed in a comparable extent and setting as in Nieuwenhuijzens’. A possible explanation may be a bias in patient selection, exclusion criteria and the different distribution of diversion methods.

As expected, anastomosis-related complications (AI) were higher (14.8%) and most of AIs occurred within the IC group (15.9%). However, there was no significant impact related to the operation method and the event of an AI (p=0.695). This shows that the operation technique and operation time do not affect postoperative outcome.

A high CDS (>IIIa) was detected in 41.6% of all patients. This represents a higher incidence than in comparable cohorts as shown by De Nunzios’ et al. (15). The difference can be explained by our high rate of preoperative comorbidities, diabetics and subsequently with our 30-day-mortality rate of 4.9% vs. 1.7%. In the subgroup analysis, vascular comorbidities showed a good predictive value for suffering more severe complications (CDS >IIIa, p=0.007). At that time, we expected vascular comorbidity to be associated with a high mortality and AI. This could not be verified in our statistical analysis. Furthermore, ACCI, diversion method, surgeon’s experience, and diabetic status had no direct statistical impact on postoperative complications. Interestingly, positive nodal status (pN1; p=0.007) and positive resection margins (R1; p=0.009) were associated with a poor postoperative outcome. The morbidity of the operation is obviously affected by the oncological status of the patient. We were not able to find a similar correlation specifically for urothelial carcinoma, but for other tumor entities such as ovarian cancers (16). Kahl et al. underline the association of ACCI and oncological status with postoperative mortality. This confirms our findings and those of Chang et al. (17). However, we did not assess and evaluate the preoperative oncological status of these patients, so that the predictive value of the preoperative clinical oncological assessment of the patients remains unclear. This has to be evaluated in further analysis. AI and a high CDS were consistently associated with a prolonged hospitalization (p=0.002). Chang et al. explored the causes of prolonged hospitalization in 304 patients with radical cystectomy (18). Their results showed no correlation with the diversion type, major complications or preoperative comorbidities. Postoperative ileus showed to be the most important cause of a prolonged hospital stay in their study. These findings are consistent with ours. However, AI was not explicitly analyzed. Our 30-day postoperative mortality rate was 4.9%. The highest mortality was within the IC group (6.5%). A statistically significant impact could not be shown, thus this form of diversion is congruently associated with the highest mortality rate in other studies such as in Monn et al. (2%) (13). Accordingly, our mortality rate showed a weak association with ACCI (p=0.054) and a strong association with CDS (p=0.002). This shows that AI does not have a higher impact on the clinical outcome of patients, than other severe complications, underlining the clinical value of the CDS.

When comparing the two departments, pN-stage, CDS and age of the patients showed a significant difference. One explanation for this bias is that a university center (Bonn) will treat more frail patients and advanced disease stages. In order to evaluate the impact of this variance, a separate analysis of the two departments was performed. Accordingly, the results did show divergence, however, not in the critical parameters. We therefore renounced on performing a matched-pair analysis. The limitations of our study are mainly due to its retrospective nature. A selection bias of this specific patient group, which has commonly undergone some operations prior to urinary diversion, does have an effect on the preoperative comorbid status. The differences between both departments are few; however, they could induce a selection bias.

In conclusion, urinary diversion persists having a high rate of severe complications. We were able to elucidate specific and independent predictors for postoperative complications: vascular diseases, smoking, a high ASA-score, and advanced oncological status. These were independent of the diversion method. Surgeon’s experience, blood loss or the extent of the operation were not found to be relevant to the outcome or survival of the patients. However, the clinical value of CDS could be confirmed. The clinical relevance of vascular comorbidities needs to be confirmed in further clinical studies. Especially oncological status has not been described to date to have an effect on patients’ outcome in the context of urinary diversion.