The role of lithium in treating bipolar disorder has recently been the focus of renewed interest; this may be attributable to preclinical data that suggest lithium may have potent neuroprotective effects and to several large-scale clinical trials of new agents that used lithium as the internal comparator and reconfirmed its utility in both the acute and maintenance phases of bipolar disorder. Thus the report in this issue by Cavanagh et al. is a timely investigation of two key treatment issues in the treatment of bipolar disorder: the risk of lithium-withdrawal phenomena and the risk of discontinuation-induced refractoriness (1). Although there are a number of previous reports of discontinuation-induced relapse, in a meta-analysis previously published in this journal, the authors declared ‘We also failed to find sufficient evidence to prove that the lithium-withdrawal relapse phenomenon exists’ (2) leaving this issue open to further investigation. The cases in this report were patients who had previously participated in a trial involving acute withdrawal of lithium after a period of at least 1 year of stability on lithium (3). Following this trial, all except three patients recommenced lithium treatment. The cases who restarted treatment with lithium were each matched with two controls. The controls were selected from the lithium card index database at the Royal Edinburgh Hospital, to be the next card in the alphabetical database following the case that was of the same gender, within 5 years of the case's age and had been treated with lithium for a duration approximating that of the case within 2.5 years. Although two controls were selected for each case, the data were averaged and a mean control subject score was used to compare against each case. There were no prediscontinuation differences between cases and controls when clinical outcomes were examined over an observation period of approximately 15 years, but cases had more illness episodes than did controls in the 7–8 years following the discontinuation study. These data thus add weight to the previous reports that abrupt discontinuation of lithium can result in illness recurrence. Interestingly in this paper, the risk of recurrence was not limited to manic recurrence but extended to an increased rate of depressions following discontinuation as well. Some investigators have recently suggested that the risk of lithium withdrawal associated recurrence may be limited to ‘atypical bipolar patients’ compared with ‘classic manic-depressive patients’ (4), and although there appears to be little in the prediscontinuation data reported on these patients to suggest that they have an atypical pattern of illness it would have been interesting to know specifically whether this hypothesis was supported or refuted by the data from these cases and controls. Perhaps an even more controversial area of debate than the issue of lithium-withdrawal recurrence is the issue of whether discontinuation-induced refractoriness occurs. Cavanagh et al. report here that most of the long-term difference in outcome between cases and controls was attributable to the early occurrences following withdrawal but that there was little evidence for an overall change in the long-term responsiveness of patients to lithium (1). The authors acknowledge, however, that their power to detect anything but a very large effect size was minimal and so the absence of evidence for lithium-refractoriness in this sample cannot be taken as evidence for its absence. From a clinical perspective, it appears common practice now to caution patients against abrupt lithium withdrawal and to implement a gradual weaning schedule when a planned withdrawal occurs. We remain less certain about whether to caution patients that lithium withdrawal might result in decreased responsiveness to the agent over time, but these issues must be addressed in the development of best practices for the use of lithium. Whether other psychotropic agents used in the treatment of bipolar disorder produce effects similar to these described for lithium also requires further investigation in clinical and preclinical studies, so that we may make optimal clinical decisions regarding not only when and how to begin psychotropic medication for bipolar disorder but also when and how to withdraw such psychotropic agents so as to minimize risk of short and long-term adverse patient outcomes.