Meta-analysis of failure-time data with adjustment for covariates.

The objective of this study was to present and illustrate a technique for combining failure-time data from various sources, adjusting for differences in case-mix among studies. Based on the proportional-hazards model and the actuarial life-table approach, the method used assumes that the variation across studies is in part due to heterogeneity of the case-mix and adjusts for the case-mix before pooling results. As an example, the technique is applied to life-table data from six selected papers reporting patency of affected arteries following femoropopliteal angioplasty. Published 4- and 5-year patency results ranged from 25% to 58%, with a pooled five-year cumulative patency rate (without adjustment for case-mix) of 45% (+/- 2%). The populations in these studies, however, differed markedly in the prevalence of factors with prognostic value: type of lesion and distal runoff vessels. After adjustment for these differences in case-mix, the pooled five-year patency rates ranged from 60% (+/- 2%) for patients with stenotic lesions and good runoff to 24% (+/- 9%) for those with occlusion and poor runoff. The authors conclude that pooling studies without considering the effect of case-mix yields an average result with inappropriately narrow confidence intervals that does not reflect the variability across subgroups. The presented technique provides a method for combining failure-time data, adjusting for case-mix.