gms | German Medical Science

Background/research question: Guidelines of the German Guideline Program in Oncology (GGPO) follow a predefined methodology, resulting in recommendations or statements categorised as evidence- or consensus-based. Every recommendation is assigned a grade (GoR), ranging from strong via weak to open recommendations. For evidence-based recommendations, the underlying level of evidence (LoE) is also provided. Herein, we examine the assumption that lower quality evidence is associated with weaker GoR and vice versa based on the recommendations from oncological guidelines.

Methods: The digital guidelines published by the GGPO were extracted from a relational database. All evidence-based recommendations were used to construct count matrices with the respective GoR and corresponding LoE. This was performed separate for each underlying LoE grading system, i. e. Oxford 2009, Oxford 2011, SIGN, and GRADE. To formulate conclusions across all grading systems, recommendations were regrouped into a condensed system projecting all systems except GRADE into a single system. Results were visualised using bubble graphs. Pearson’s correlation coefficients were calculated between GoR and LoE.

Results: From 28 published oncological guidelines, 2.178 evidence-based recommendations were extracted. Of these, 946 (43.4%) were strong, 750 (34.4%) were weak, and 482 (22.1%) were open. 859 recommendations (39.4%) were graded based on Oxford 2009, 152 (7.0%) on Oxford 2011, 969 (32.0%) on SIGN, and 471 (21.6%) on GRADE. For each LoE system, significant correlations were identified for Oxford 2009 vs. strong recommendations (R²=0.655, p=0.015), Oxford 2009/2011 combined vs. weak recommendations (R²=0.883, p=0.018) and vs. open recommendations (R²=0.999, p<0.0001), condensed LoE vs. open recommendations (R²=0.978, p=0.011). There was a negative correlation between GRADE and open recommendations (R²=0.994, p=0.003).

Conclusion: There were only inconsistent correlations between LoE and GoR. The positive correlations identified for strong recommendations align with the working assumption: high quality evidence was linked predominantly to strong recommendations. In contrast, for open recommendations one would assume that the driving force was low quality evidence. However, this was only found for GRADE, but not for Oxford 2009/2011 and the condensed system. A more detailed breakdown of recommendations by topic and by evidence-to-decision reasoning might help to understand these conflicting findings.