Missing information because of death in time-to-event analyses of clinical and epidemiological studies

• Applicant: Professor Dr. Martin Schumacher
• Subject Area: Epidemiology and Medical Biometry/Statistics
• Term: from 2012 to 2019
• Project identifier: Deutsche Forschungsgemeinschaft (DFG) - Project number 220403157

In most clinical and epidemiological studies information on disease status is collected at regular follow-up times. Often, this information can only be retrieved in those individuals who are alive at follow-up and will be missing in those who died before follow-up. The arising complications are of particular relevance in long-term studies and/or when studying elderly populations. There is some indication in the biomedical literature that clinical and epidemiological studies have excluded individuals with missing information because of death and restricted the analysis to the surviving ones. These naïve analyses can lead to serious bias in incidence estimates that is then transferred to bias in estimates of hazard ratios corresponding to potential risk or prognostic factors. Within the project we did investigate the prevalence of this bias in highly-cited medical and epidemiological journals. We did further elaborate on a formula for the resulting bias in order to estimate its direction and magnitude based on individual study characteristics. We did investigate several statistical approaches based on multi-state models and related regression methodology to which extent and under which circumstances they allow for an adequate and unbiased analysis. In a subsequent step, we will extend these investigations by adding a comparison to alternative approaches thereby also considering the issue of loss of follow-up. This should lead to the development of a comprehensive analysis strategy for studies where missing information because of death constitutes a problem. This strategy will then be used for the reanalysis of some existing studies. Finally, we will work out recommendations for the design of studies in order to avoid this bias or to keep it in negligible magnitude.

• DFG Programme: Research Grants