258 CORRESPONDENCE use the full doses at which the efficacy of dothiepin was established. There is also inadequate discussion of the reasons for the more frequent withdrawals from dothiepin during the prophylactic phase. Is it possible that patients themselves recognised that dothiepin was no more effective than placebo? There are studies that provide clear-cut evidence of the prophylactic efficacy of antidepressants in reducing the risks of new episodes of depression (reviewed by Montgomery & Montgomery, 1992). This is not one of them. CASSIDY, S. & Hn@uty, J. (1987) Fatal toxicity of antidepressant drugs in overdose. British MedicalJournal, 295, 1021—1024. MINDHAM, R. H. S., HOWLAND, C. & SHEPHERD, M. (1973) An evaluationofcontinuation therapywith tricyclic antidepressants in depressive illness. Psychological Medicine, 3,5—17. MONTGOMERY, S. A.(l990) The methodology necessary to establish the long-term efficacy of antidepressants. In Hu,nan Psycho pharmacology: Methods and Measures (eds 1. Hindmarch & P. D. Stonier). Chichester: Wiley. —¿ &MONTGOMERY, D. B. (1992) Prophylactic treatment in recur rent unipolar depression. In Long-term Treatment of Depression (eds S. A. Montgomery & F. Rouillon). Chichester: Wiley. STUART A. MONTGOMERY Academic Department of Psychiatry St Mary's Hospital Praed Street London W2 INY AUThoRs' REPLY: If Montgomery thinks we “¿ have not entirely understood [the] arguments― for separ ation of continuation therapy from prophylaxis, he has not read our paper carefully enough. We do dis cuss these arguments in some detail (p. 180; para graphs 2 & 3) and explain why we did not apply the distinction strictly to our elderly patients, as might have been appropriate for a younger group. To reiterate and amplify the point, we were dealing with a population which has been consistently shown by research to contain only about 20% of people, or less, who will not experience a further episode of major depression; whereas the remaining 80% or so are liable either to suffer from chronic affective symp toms or one or more rapid relapses. The key ques tion, therefore, is whether —¿ the patient's physical state permitting —¿ treatment should ever be stopped. Having made the presumption that our study should have been conducted according to the design he prescribes, Montgomery employs inappropriate statistical methods toanalyseitasifwe had con ducted it that way. In spite of our explanation (p. 177; first paragraph, second column), he does not appear to have understood the error of using cross sectional analysis (x2 tests) on longitudinal data. His statements aboutsignificance havenovalidstatisti cal foundation. In fact, his assertion that “¿ analysis at the six-month point makes sense― makes no sense at all. Our Table 2 clearly shows how relapses occurred over a two-year period, and the higher incidence in the placebo group is evident. We had hoped to make it abundantly clear that P values obtained from x2 tests were mentioned only to underline how mislead ing they can be when incorrectly applied (as, alas, they so frequently are in studies of this kind). The analysis whichisappropriate tolongitudinal data, suchasthese, istofit asurvival curve.Inthisinstance a proportional hazards model was fitted and the sur vival curves speak for themselves. Clearly, it would have been preferable to have a larger sample since the consequent increase in power would have made it possible to detect a smaller difference between the treatment groups but, in the event, the difference was found to be large enough for the sample size to be acceptable. Montgomery accuses us of giving an eccentric definition of relapse of one point more than the recommended cut-off. Where does he get this idea from? There is no inconsistency between having a criterion MADRS response <11 and a definition of relapse >10 because, provided the individual scores are integers (which they are), the two sets are mutually exclusive and exhaustive. The mean MADRS score on relapse was 24.01. There was no discussion of suicide or serious side effects in the interests of brevity. There were no suicides duringthetrial. Norwerethereseriouside effects, which were carefully monitored throughout in accordance with international standards. How ever, if it were to give Dr Montgomery some peace of mind, none of the psychiatrists taking part in this trial have an ‘¿ axe to grind' about tricyclics in general, or dothiepin in particular. The important point is the maintenance of effective pharmacological treatment, without which the great majority of this group of patients can expect their relatively short remaining life-span to be blighted by continuous or recurrent illness. The Maudsley Hospital London SE5 8AZ Worcester College Oxford OXJ 2HB Sexandschizophrenia: viveLadifference ROBIN JAcOBY DANIEL Lur@m@ Sm: Lewis provides an interesting review of the literature investigating the sex differences in