British Journal of Psychiatry (1993), 163(suppl. 21), 16-19 The Challenge of Predicting Response to Stabilising Lithium Treatment The importance of patient selection P. GROF, M. ALDA, E. GROF, D. FOX and P. CAMERON Lithium treatment, an approach with well documented efficacy, has recently been losing its treatment value. Lithium continues working, however, for those patients for whom it was proven efficacious; that is, most patients with primary episodic affective disorders. Such responders to lithium prophylaxis can be reliably identified beforehand by a comprehensive clinical assessment. The explanation for the paradox of lithium's lost efficacy lies mostly in the educational bias against a comprehensive patient assessment, and in the shift in diagnostic fashion favouring affective disorders and the treatment methods associated with them in the clinicians' minds. In recent years, neuropsychopharmacology has faced an interesting paradox. In terms of its efficacy, lithium is the best-studied drug - certainly in psychiatry and possibly in all medicine - and yet several recent studies have indicated that lithium offers little help in current clinical practice and experimental studies. These recent reports are in sharp contrast with earlier research. Between 1967 and 1974, the results of a series of open and double-blind studies on the prophylactic action of lithium were published (Schou & Thomsen, 1975). These studies included over 650 patients and clearly demonstrated that lithium was a highly efficacious prophylactic agent, markedly superior to placebo, and preventive of both manic and depressive episodes. A large majority (70-80070) of patients benefited either by freedom from recur- rences or their marked reduction. Studies carried out in the 1980sand early 1990sreflect a striking change: in several of these only 20-30070 of patients treated with lithium responded (Prien & Gelenberg, 1989). For whom does lithium still work, and why has lithium 'stopped working' for other patients? For whom does lithium still work? Consensus is emerging from the literature that attempts to predict a lithium responder have failed. Many variables have been proposed as successful predictors, only to be proven ineffectual later. The outcome of research, however, always depends on the chosen strategy. In fact, the response to lithium 16 prophylaxis is predictable, provided that an appro- priate methodology is employed: the desired type of benefit defined; the study correctly designed to allow an interpretation of findings; and proper statistical analysis applied (Grof et al, 1983a). The benefit must be defined as mood stabilisation (the reduction of expected affective morbidity), as little is known about the prediction of other benefits: e.g. anti-aggressive effects on lithium augmentation or a non-specific reduction of overactivity (Grof & Grof, 1990). The experimental design needs to involve elements which will allow for linking lithium treatment more directly with change in the clinical course, for example by using lithium as the only medication between acute episodes. Statistical analyses should be commensurate with the complexity of the issue, which is multivariate in nature. We have studied a number of clinical, biochemical, neuroendocrinal, neurophysiological, familial, psych- ological and pharmacological variables in connection with lithium response (Grof et al 1979a, 1983a). These variables were selected because of various reports in the literature promising predictive power for each of them. The allotted space will allow a few examples of the variables which have been particularly popular at some time in the past. Based on the assumption that the red blood cell concentration of lithium reflects the lithium concentration in the brain better than the plasma lithium level, researchers studied erythrocyte lithium (Mendels et al, 1976). Many concluded, on the basis of their findings, that the lithium ratio (lithium