Predictors of the regional variation of prostatectomy or radiotherapy: evidence from German cancer registries.

To assess the association of public health parameters with the regional variation in the initial treatment for prostate cancer. We used data from German epidemiologic cancer registries for the years 2009-2013. Presence of a certified cancer center, a radiotherapy and/or urology institution, the district-specific GDP, and population density were used as predictors. Patients with indication for adjuvant treatment were excluded (T3b). Only districts with defined quality criteria were eligible. We used general linear mixed models (equivalent to logistic regression) with a covariance matrix weighted by the Euclidean distances between districts. Models were adjusted for age, grading, and TNM stage. We performed sensitivity analyses by imputing missing data with multiple imputation and considering extreme case scenarios. We applied inverse probability weighting to account for missing values. When radiotherapy/surgery is compared to neither treatment, the probability for the latter was higher in East than in West Germany (OR 1.7, 95% CI 1.43-2.02). The same was true for districts with both, a radiotherapy and urologic treatment facility (OR 1.43, 1.19-1.72). Analyzing radiotherapy vs. surgery, the probability for prostatectomy was inversely associated with the presence of a radiotherapy unit when compared to districts with neither treatment facility (OR 0.52, 95% CI 0.38-0.73). Patients treated in East Germany were more likely to receive a surgical treatment (OR 1.34, 95% CI 1.08-1.66). Sensitivity analyses revealed no relevant change of effect estimates. Treatment differs between East and West Germany and is associated with the presence of a radiotherapy or urology clinic.