While randomized control trials (RCT) are the gold standard for drug
approval, there is often a lack of data directly comparing different treatment options.
An indirect comparison of the treatment effects may serve as a proxy for a head-to-
head RCT, however, naively comparing treatments using published trial data without
adjusting for distribution differences in patient characteristics and prognostic factors
can result in misleading conclusions. a novel matched-adjusted approach to indirectly
compare absolute survival estimates (median overall survival (OS) or progression free
survival (PFS)) for competitive treatment options is presented.
proposed approach requires patient-level data for one of the treatments and summary
data of patient characteristics and survival outcomes for the comparator of interest.
Using this proposed method, the researcher would fi
rst decide on one or two matching
variables that are prognostic for survival, and apply a program involving an extension
of a common SAS 9.2 procedure, Proc Surveyselect, to select 1000 random repeated
sub-samples from the original population with the same distribution of matched
variables. The analysis also requires programming statements using ODS and survival
analysis procedures. The median OS or PFS estimates are computed for each boot-
strapped sample and a 95% confi
dence interval (CI) is inferred around the mean of
the sampled survival estimates. These absolute survival estimates, based on the
adjusted population, can then be compared to the absolute survival estimates reported
in published literature of the comparator treatment.
In the absence
of head-to-head RCT data, an adjusted indirect comparison accounts for observed
differences between populations making them more comparable and results in an effect
of treatment exposure on survival outcomes that is less likely due to confounders.