Requirements for benefit assessment in Germany and England: Overview and comparison

Background: This study compared the methodological requirements for early health technology appraisal (HTA) by the Federal Joint Committee/Institute for Quality and Efficiency in Health Care (G-BA/IQWiG; Germany) and the National Institute for Health and Care Excellence (NICE; England). Methods: The following aspects were examined: guidance texts on methodology and information sources for the assessment; clinical study design and methodology; statistical analysis, quality of evidence base, extrapolation of results (modeling), and generalisability of study results; and categorisation of outcome. Results: There is some degree of similarity regarding basic methodological elements such as selection of information sources (e.g. preference of randomised controlled studies, RCTs) and quality assessment of the available evidence. Generally, the approach taken by NICE seems to be more open and less restrictive as compared with G-BA/IQWiG. Any kind of potentially relevant evidence is requested, including data from non-RCTs. Surrogate endpoints are also accepted more readily, if they are reasonably likely to predict clinical benefit. Modeling is expected to be performed wherever possible and appropriate, e.g. for study duration, patient population, choice of comparator, and type of outcomes. The resulting uncertainty is quantified through sensitivity analyses before making a recommendation regarding reimbursement. By contrast, G-BA/IQWiG bases its assessment and quantification of the additional benefit largely, if not exclusively, on evidence of the highest level and quality and on measurements of "hard" clinical endpoints. This more conservative approach rather firmly dismisses evidence from non-RCTs and measurements of surrogate endpoints that have not or only partly been validated. Moreover, neither qualitative extrapolation nor quantitative modeling of data is done. Conclusions: Methodological requirements differed mainly in the acceptance of low-level evidence, surrogate endpoints, and data modeling. Some of the discrepancies may be explained, at least in part, by differences in the health care system and procedural aspects (e.g. timing of assessment).