Background. Though previous studies have clearly shown that lithium affords prophylaxis in bipolar affective disorder, these studies have not demonstrated the persistence of this prophylactic effect beyond the first year of recovery.

Methods. One hundred and eighty-one patients with bipolar affective disorder recovered during 5 years of semi-annual follow-up. After 8 weeks of recovery, 139 were taking lithium prophylaxis and 42 were not. Analyses used drug status (lithium v. no-lithium) as a censoring variable to compare these two groups by interval-specific probabilities of recurrence.

Results. Recurrence was initially less likely in the lithium group but interval-specific probabilities of recurrence did not consistently favour either group after the first 32 weeks of recovery.

Conclusions. Biases in treatment decisions may have both reduced the size and altered the specificity of the lithium effects seen here. Nevertheless, the apparent transience of lithium prophylactic effects is unexplained and may reflect important, physiological differences between relapse and recurrence. This possibility invites a controlled lithium discontinuation study, with gradual taper, of patients who have had at least 8 months of sustained euthymia.