Evaluating the benefit of therapies for malignant mesothelioma is a quite
difficult task. The difficulty lies in identifying patients who might have benefited from a
given treatment but knowingly did not receive it. Unless these 'suitable' patients have been
identified in advance of any treatment and the process of allocation to one treatment or
another is without bias. This is best achieved by randomization and in our view, for
relatively small trials with heterogeneous patients, that is best done with the additional step
of minimization.
Keywords: Observational data, outcome, sample size, selection bias, statistical
analysis.
