In reply: We agree with Johnson and Lassere about the value of longitudinal studies, especially clinical trials, in assessing healthcare benefits and costs. They are critical to informing clinical practice and health policy.

If it were possible to conduct these studies across a wide variety of settings, representing the range of "real life" practice, then there would be little need for epidemiological modelling. However, this is not possible. Clinical trials (with or without cost components) will only ever be conducted over relatively short durations, on circumscribed populations and under tightly controlled conditions.

A key, but often overlooked, issue is whether the results of studies are externally valid (generalisable). Indeed, the evidence base that dictates clinical practice and health policy should comprise data that are both internally and externally valid.

We do not suggest that epidemiological modelling replace longitudinal studies (in fact, modelling depends critically on robust prospective data); rather, it complements these studies by providing a means to assess their external validity.

We are also mindful of the limitations of epidemiological modelling, as outlined in our article,1 and acknowledge the importance of ensuring rigour in the methods.

Our article dealt with generating the data needed for sound economic evaluation, by taking into account the long-term benefits, risks and costs of treatment strategies, and "real-life" health service conditions. This is distinct from the issue of whether "conditional listing" on the Pharmaceutical Benefits Scheme should be implemented for drugs that are yet to be proven cost-effective.