Background
For the treatment of colorectal cancer, the S3-Guideline of the German Guideline Program in Oncology supports clinical decision-making. Centers certified by the German Cancer Society are required to implement the guideline recommendations as comprehensively as possible. When guidelines provide insufficient or ambiguous evidence, heterogeneity of treatment preferences is likely to emerge across individual centers. The aim of this study is to describe the preferences of the centers’ tumor boards for treatment decisions when clear, evidence-based guideline recommendations are lacking.

Methods
To investigate the tumor boards’ preferences for different treatment options, an anonymous online survey was conducted among 314 certified colorectal cancer centers. The survey included seven visceral oncological and nine visceral surgical case vignettes. Centers were asked to discuss the vignettes in the tumor board or, alternatively, delegate them to an appropriate physician representatively speaking for the tumor board. The responses were analyzed descriptively and normalized entropy estimates (NE) were calculated for each vignette.

Results
A total of 123 centers (39%) responded to the survey. For oncological cases without clear guideline recommendations, substantial heterogeneity (NE: 0.39–0.71) in treatment preferences was observed. For instance, opinions varied widely for UICC II pT4a colon cancer. In this situation, 28% of centers would prefer fluoropyrimidine monotherapy, 39% oxaliplatin-based combination therapy and 33% no adjuvant chemotherapy at all. Surgical vignettes showed a preference for laparoscopic and robotic approaches, with variations based on tumor location (NE: 0.46–0.67). Importantly, in case of a clear evidence-based guideline recommendation, treatment preferences did not differ considerably between hospital sites.

Conclusions
In prototypical case vignettes without evidence-based guideline recommendations, pronounced heterogeneity of treatment preferences between centers was found. Reconstructing these treatment preferences can contribute to enhancing the quality of evidence derived from observational studies. This seems especially important in the context of clinical questions that cannot be assessed in randomized trials; clinical routine data represent an adequate resource for evidence generation in these scenarios.