Authors' reply

BMJ 1996;312:513 (24 February)

Letters

Authors' reply

EDITOR,--The assertion that depression has a better prognosisif detected by general practitioners¹ is refuted in our sample.Our study was not designed to test a treatment effect, and itis therefore not relevant to criticise it, as David Goldbergdoes, for failing to show one. Goldberg iterates our conclusionthat a prevalence sample may show apparently worse outcomesthan an incidence sample, though why he is worried about thisis unclear. There are methodological hazards in defining incidencesamples in primary care settings, particularly if incidenceis confused with first recognition by a general practitioner.We are surprised that Goldberg suggests that our study duplicatesHoeper et al's,² since their study did not assess outcomes.In fact, our initial methodological model was his own study,which, like ours, reported data on treatment patterns as outcomerather than process measures.³

We are interested in Alan Currie and Ann Ryman's speculationson the different response rates in the study groups (data thatGoldberg seems not to have noticed). We would also value commentsabout the difference in response rate between the sexes.

We estimated that a minimum sample size of 31 per group wasnecessary to detect a difference of 7 in the median score onthe Beck depression inventory with 85% power and a 5% probabilityof detecting a false effect on a background standard deviationof 9. We chose 7 as the minimum change in the score that couldbe clinically important; this was influenced by previous studies.³We therefore do not accept Paulo Menezes and colleagues' criticismof our sample size. We accept that the decision to include onlyfull responders could introduce selection bias. Inclusion ofpartial responders, however, introduces another source of variationthat could blur the main comparison of the study. In practiceit does not: reanalysis of our data to include partial responders,as Menezes and colleagues suggest, shows a minimal effect onthe statistical results, with no changes in the inferences originallydrawn. Menezes and colleagues' preference for follow up at sixweeks does not take account of the delay between disclosureto and potential action by a general practitioner. In this studythe effect of disclosure on diagnosis was greatest after sixmonths.⁴

We agree that comprehensive interventions may be more effectivethan simple disclosure in undetected cases and welcome the proposalfor further research into the best treatment of mild depressionin primary care. However, we consider ourselves to be realists(not nihilists) and would not wish to encourage undue expectationsof general practitioners' abilities to influence the outcomeof depression. The frequency and duration of untreated depressiveepisodes were similar to those of treated episodes in a recentstudy in the United States,⁵ whose authors argue that the differingcharacteristics of people who seek help and those who do notmake it difficult to measure the value of treatment in mitigatingthe impact of depressive disorders.