Gracious & Falodun find that our study of mortality in affective disorder patients commenced on lithium (Reference Brodersen, Licht and VestergaardBrodersen et al, 2000) paints an unfairly negative picture of lithium's efficacy. They object to our intention-to-treat analysis of all patients commenced on lithium irrespective of compliance, which showed a significantly elevated standardised mortality ratio (SMR) of 2.5. They suggest that we should have compared compliant with non-compliant patients and with the general population, as did Kallner et al (Reference Kallner, Lindellius and Petterson2000).

Kallner et al actually reported — even in patients compliant with lithium — that mortality in general (SMR=1.6) and suicide in particular (SMR=14.0) were significantly elevated. They also found that mortality was even higher in non-compliant patients, a result which may very well be valid. However, comparison of compliant with non-compliant patients introduces a considerable selection bias, since patients are not randomly allocated to the two groups. Rather, patients with comorbidity, such as drug and alcohol misuse and other predictors of negative outcome (Reference Vestergaard, Licht and BrodersenVestergaard et al, 1998) select themselves to the non-compliant patient group. Therefore, a finding that non-compliant patients fare worse than compliant patients may testify only to the existence of negative predictor variables among patients who were non-compliant, instead of supporting the efficacy of lithium treatment. Neither our study nor Kallner et al's allow conclusions as to whether or not lithium has specific antisuicidal effects exceeding what can be inferred from its ability to prevent recurrent illness episodes in affective disorder patients.

The efficacy of long-term prophylactic treatment with lithium has been questioned frequently (Reference MoncrieffMoncrieff, 1995). We believe, as apparently do Gracious & Falodun, that despite its shortcomings lithium is a very helpful tool in the psychiatric armamentarium. Arguments that support the efficacy (or inefficacy) of long-term lithium treatment should, however, rest on sound scientific evidence.