Erratum

Oncology 2010;79:303–305 305 dient; (6) biological plausibility; (7) coherence with existing knowledge; (8) evidence, and (9) analogy. The above criteria (Hills criteria of causation [10] ) outline the minimal conditions needed to establish a causal relationship between 2 items. So, for instance, the strength of our association between guideline adherence and recurrence-free survival and overall survival rules out hypotheses that the association is entirely due to 1 weak unmeasured confounder or other source of modest bias. The consistency refers to the repeated observation of our association in different patient groups under different circumstances (results of studies done around the world on guideline adherence [11, 12] have consistently shown similar results; furthermore, there was no difference in results between Universitätsfrauenklinik Ulm and the corresponding 16 network hospitals). In the absence of random assignment, guideline-adherent and guideline non-adherent patients are in general not comparable at baseline. Therefore, we had to remove baseline differences in important prognostic covariates by model-based adjustments.