Grant: Operating Grant

Individual patient data are considered the gold-standard among meta-analytic methods, which attempt to combine evidence from many studies into one message. An IPD-MA uses line-by-line data for each patient, in contrast to a typical, aggregate-data meta-analysis which uses only averaged or estimated information from studies (e.g. odds ratios, risk differences, rate ratios, means, proportions, etc.). IPD meta-analyses permit the possibility of detailed statistical analyses including subgroup analyses and the ability to adjust for confounders within and between studies. However, strikingly, a standardized data analytic approach for IPD meta-analysis does not exist, and has been noted as one of the main drawbacks for IPD meta analyses (after the time and expense needed to assemble the data). In this work, we propose to two objectives: 1. To evaluate the effectiveness of surgery as a treatment for multiple and extensively drug resistant tuberculosis while accounting for prior patient bias, and the substantial confounding that may exist between and within studies using individual patient data from over 30 studies comprising over 9000 patients. To address this aim well, we first need to improve on the statistical methodology - which we tackle in Objective 2. 2. To advance biostatistical methodology for the analysis of IPD meta- analyses.