The aim of this study was to systematically integrate the available evidence on response prediction to prophylactic lithium based on clinical factors.Methods
Each clinical variable that was related to lithium response in at least one prior study was examined with respect to response prediction. If several studies were located for the same variable, results were integrated using the meta-analytic approach as suggested by DerSimonian and Laird which was developed for substantial heterogeneity in primary studies.Results
Of 42 potential clinical predictors investigated, five variables were identified as possible response predictors of prophylactic lithium:  An episodic pattern of mania-depression-interval, and  a high age of illness onset were identified as potentially protective against a recurrence under lithium.  A high number of previous hospitalizations,  an episodic pattern of depression-mania-interval, and  continuous cycling were identified as potential risk factors. Six further variables were found to be significantly related to lithium response, though calculation of fail-safe numbers indicates that current evidence is not sufficient to hold these variables as reliable predictors of lithium response. All effect-sizes relating clinical predictors to response were small to moderate.Conclusions
Although a few variables are quite robustly supported as response-predictors in this review, a more in-depth analysis of each potential predictor is needed. As none of the potential predictors had a very strong impact on response, prediction of lithium response should be based on a multitude of variables.