Pre-op variables may predict post-cystectomy complication risk

The risk of high-grade postoperative complications after radical cystectomy for bladder cancer might be predicted using a model that incorporates several routinely collected preoperative variables, a researcher reported at the Society of Urologic Oncology annual meeting in Phoenix.

Using data collected during the course of preoperative history, physical examination, and staging, Sida Niu, MD, and colleagues at the University of Kansas Medical Center, Kansas City developed a series of prediction models using a cohort of patients who underwent radical cystectomy for bladder cancer at their institution.

The best performing prediction model utilized four variables-body mass index, creatinine, preoperative chemotherapy, and robotic approach, said Dr. Niu, of the University of Kansas, working with Eugene Lee, MD, and colleagues.

Although further study is needed, the current data suggest these models might help identify certain patients who can benefit from further patient education, counseling, and development of risk reduction strategies, Dr. Niu said in an interview with Urology Times.

Also see:Noninvasive bladder Ca test performs favorably compared to cystoscopy

“We tried to include mostly modifiable variables, but some non-modifiable variables are also included because they provide information in helping patients understand their postoperative risks, which can be informative when counseling patients contemplating radical cystectomy,” he said.

Their retrospective study included 508 patients, of whom 336 (66%) had at least one complication of any grade. In looking at 28 preoperative variables collected for those patients, they identified 14 that were associated with increased risks of high-grade postoperative complications.

Out of those 14 variables, they chose 11 to design a series of prediction models, the accuracy of which they assessed using area under the receiver operating characteristic curve (AUROC) and a cross-validation procedure.

The variables with the highest prediction accuracy in univariable analysis were BMI (AUROC 0.643; odds ratio, 1.09), creatinine (0.609, 2.43), preoperative chemotherapy (0.597, 1.22), and robotic approach (0.590, 2.33), Dr. Niu reported.

The authors described a total of 20 prediction models, the best performing of which utilized those four variables and had an AUROC of 0.727.

The complication rates in this retrospective cohort were in line with previous studies showing that up to 60% of patients undergoing radical cystectomy will experience complications, such as postoperative ileus or infections, within 90 days of the procedure, according to Dr. Niu.

“Many of these complications will prolong the hospital stay, increase total cost of the treatment, and decrease the overall quality of life for these patients,” he said.

Continue to the next page for more.While this predictive model is not ready for use on a wide scale, Dr. Niu said investigators hope that with further evaluation of interactions between individual variables, they’ll be able to further improve the accuracy of their prediction models and eventually create a prediction calculator that could be used to counsel patients in the preoperative setting.

That might be a substantial improvement over current practice, he added.

Read: Neoadjuvant PD-L1 inhibitor therapy found safe, effective

“When we identify patients who may be candidates for radical cystectomy, we often use a vague picture of their health that we get from their charts-do they have comorbidities like heart conditions or lung conditions? Have they had previous abdominal surgeries or radiation therapy?” he said. “ While these questions help us assess the patient as a surgical candidate, the process can certainly be improved to include a feedback component that allow patients to make lifestyle changes and potentially decrease their risk of postoperative complications.”

The authors reported on complications from 31- to 90-days post-procedure in part because their earlier investigations showed risk models in this window performed better than in the 30-day postoperative period, Dr. Niu said.

The authors hypothesized that complications in that earlier 30-day time frame may be more related to the surgery itself, rather than due to individual characteristics or preoperative factors, he added.