PHP169 - Good Cop, Bad Cop; G-BA and Iqwig in Clinical Outcome Evaluations

Abstract

G-BA and IQWiG state the same guidelines for assessing clinical efficacy data on Mortality, Morbidity, and Quality of Life (QoL). This analysis compares the two agencies’ approaches to clinical data evaluation to determine if they interpret the guidelines and standards similarly in practice, and accept the same outcomes for demonstrating benefit. This analysis included 98 G-BA and 153 IQWiG evaluations for 88 different drugs across 41 disease conditions. 1,312 clinical outcomes were evaluated to determine the rates of outcome rejection. Rejection was defined as instances where the manufacturer submitted information and G-BA or IQWiG concluded that the data was not usable; instances of no data or inconclusive data were excluded. 15% of outcomes were rejected. QoL outcomes were more likely to be rejected than Morbidity and Mortality outcomes (27%, 15%, and 2%, respectively). Intra-agency disagreements were identified; IQWiG rejected more outcomes than G-BA (16% vs 12%; p<0.083). G-BA rejected zero Mortality outcomes (n=121), while IQWiG rejected six (n=125). IQWiG never accepted PFS, whereas G-BA rejected it outright 52% (n=25) of the time, and reported it with a disclaimer and commentary on internal disagreement on its validity 48% (n=23) of the time. The most common reasons for rejection were using a non-validated outcome or having too small of a proportion of patients in the analysis. During the presentation, further case studies on intra-agency (and internal) disagreement will be presented. IQWiG and G-BA appear to use differing standards in tandem, with IQWiG’s more critical assessment bolstering G-BA’s subsequent assessment of additional benefit. G-BA accepts more outcomes than IQWiG in every category, and reports internal disagreement on outcomes that IQWiG rejects outright. Overall, 15% of outcomes assessed in Germany are disregarded, a concern for manufactures developing clinical trials. Use of validated outcomes and sufficient population size, especially for QoL, is essential.