Diabetes and kidney cancer outcomes: a propensity score analysis

Abstract

There is conflicting evidence whether diabetes is associated with survival outcomes in patients undergoing a nephrectomy for renal cell carcinoma. We performed a retrospective review of 1034 patients undergoing nephrectomy for unilateral, M0, renal cell carcinoma between 2000 and 2016 at a tertiary academic center. Inverse probability of treatment weights were derived from a propensity score model based on various clinical, surgical, and pathological characteristics. We used Cox proportional hazard models to evaluate the association between diabetes and disease-free survival, cancer-specific survival, and overall survival in the sample weighted by the inverse probability of treatment weights. Furthermore, to evaluate whether severity of diabetes was associated with survival outcomes, we performed separate analyses where inverse probability of treatment weights were computed based on the probability of having diabetes that was controlled by medication. Of the 1034 patients, 180 (17 %) had diabetes. Of these, 139 (77 %) patients required medications for diabetes control while the remaining 41 (23 %) had diet controlled diabetes. Median follow-up was 50 months (IQR 17-86). Diabetes at the time of surgery was not significantly associated with disease-free survival (HR 1.11, 95 % CI 0.64 -1.91), cancer-specific survival (HR 0.96, 95 % CI 0.49-1.91), or overall survival (HR 1.28, 95 % CI 0.84-1.95). We found similar results when we compared diabetics controlled with medication vs. non-diabetics or diet controlled diabetics. In summary, we found no significant association between diabetes and survival outcomes in patients undergoing nephrectomy for M0 renal cell carcinoma. These results suggest that diabetics should be treated and followed in a similar manner to non-diabetics.

Critical appraisal of the application of propensity score methods in the urology literature

“…Interestingly, we noted that the majority of studies using PS methods were in oncology, despite our search strategy including general urology journals. Although PS methods have typically been used in studies comparing treatments [18][19][20], they can also be used in non-intervention studies [21,22]. While many methods to estimate the PS have been proposed including boosting, recursive partitioning, random forests, and neural networks [1], we found that logistic regression was the most commonly used method in urology.…”

“…Interestingly, we noted that the majority of studies using PS methods were in oncology, despite our search strategy including general urology journals. Although PS methods have typically been used in studies comparing treatments , they can also be used in non‐intervention studies .…”

Does diabetes mellitus play an independent prognostic role in kidney cancer?

“…Another interesting finding of the analysis of Nayan et al is that there is a correlation between DM and clear cell histology, which has been described earlier (7) and deserves a closer look on the underlying mechanisms in future studies.…”

“…But there was no association to carcinoma-specific survival (CSS) in this metaanalysis. In a single-center propensity score analysis of over 1,000 patients with RCC, no correlation of DM with survival of RCC could be shown (7). In a retrospective analysis on the influence of metabolic syndrome on the outcome of localized RCC, DM as single component of metabolic syndrome had no independent influence on survival (8).…”

The value of complementing administrative data with abstracted information on smoking and obesity: A study in kidney cancer

“…Although we have previously included patients undergoing nephrectomy after this date, [12][13][14] comorbidity information in administrative data was only available until March 2015.…”