The main finding of El-Sayeh et al's systematic review that ‘compared with placebo, aripiprazole treatment was associated with a significant decrease in relapse rates, increased compliance with the study protocol, and a decrease in prolactic levels below the expected values’ is overshadowed by a background of complaint about lack of data. What one wants to know is what was the spectrum of activity with respect to symptoms? A substantial body of data was collected with standard rating scales on 4125 patients in ten separate trials. With no single exception, El-Sayeh et al record that these data were either ‘unusable’ or that standard deviations were not available (Table 1); they therefore conducted no analysis.

It seems incredible that after contacting relevant authors and the manufacturers of aripiprazole El-Sayeh et al came up with such a barren yield. There surely are data available and a systematic reviewer has a duty to obtain them and make them available in comparative form.

A more serious deficiency relates to the authors' clear innuendo that reports of deaths which are possibly drug related have not been widely disseminated. They further argue that ‘not disseminating clear information regarding these people's outcomes… breaks that unspoken contract that occurs between researchers and trial participants at the point of gaining informed consent'. In a poster presentation at the Winter Workshop on Schizophrenia Research in February 2006 the authors were even more explicit, ‘In two studies 8 people allocated aripiprazole died. Even if the mortality of people with schizophrenia is 2-3 times that of the general population, the age-standardised death rate in these studies exceeds even that pessimistic estimate by 400-500 percent… Mortality data are concerning'. To make the point crystal clear the poster included a representation of a coffin.

It appears that El-Sayeh et al made a simple mistake - they thought that a number of deaths recorded in trials on the Food and Drug Administration (FDA) website (http://www.fda.gov) related differentially to patients on aripiprazole, whereas in fact these deaths were in the uncontrolled follow-up phase and were neither selective to aripiprazole, nor in excess relative to age and gender norms. The data were accessible, and were known to the FDA and to the relevant companies. Authors of systematic reviews no doubt have a duty to draw attention to deficiencies of trial data as they see them, but they also have a responsibility to marshall all the findings in a scientifically revealing way. If they make an error they have a duty to correct it.