Commentary

The review by Bollini et al shows that as the dose of an antidepressant is increased, so does the rate of adverse effects and dropouts from treatment. The question is what happens to treatment response rates with increasing dose? Basic pharmacology teaches us to expect a dose response relation. Unfortunately, patients who drop out of treatment cloud what should be a simple issue, and alternative analyses lead to conflicting conclusions. Conventionally, the percentage of patients showing improvement to a fixed dose of antidepressant is expressed for patients who complete the protocol and do not drop out. In this review, the response rate for protocol completers rose from 60% at low dose to 76% for the highest doses. Emphasis, however, was placed on an “intention to treat” analysis. Response rates were calculated as the proportion responding after randomisation: all dropouts counted as treatment failures. This analysis resulted in reasonably similar response rates for all doses of drug. The authors conclude that no dose effect relation exists and that low doses of drug are as effective as higher doses. This is inconsistent with much current clinical practice and teaching.

The paradox concerns how the factors that cause patients to drop out of clinical trials relate to their treatment response. If patients who drop out tend not to respond anyway, then an intention to treat analysis would be the only way to overcome the potential bias of a per protocol analysis. However, if no relation, or even a positive one, exists between side effects and treatment response, the per protocol analysis will not be entirely misleading, and a dose effect relation for antidepressants may exist. It is unclear which of these possibilities is the case: my hunch is that it is the second, and that a dose effect relation exists.

In a real clinical situation, of course, we do not have to be so inflexible as to choose and stick to a given dose. Recovery may occur at low doses as a placebo response or an unusual sensitivity to the drug. If it does not, we can increase the dose and expect a greater probability of response if the drug is tolerated. Efforts to get patients to accept higher doses of drugs where necessary may be an important part of both the therapeutic and pharmacological processes.

Most clinical trials in depression that are done, as in this review, for regulatory purposes provide an inadequate evidence base for many clinical decisions. Randomised comparisons of protocols of stepped care that are more representative of real clinical practice are needed to help us to understand what strategies of treatment are best for patients. Bollini et al are correct, however, in drawing our attention to a law of pharmacology even more fundamental than the dose response curve. If patients do not take their drugs, we can hardly expect them to derive a benefit from them.