of treatment and prophylaxis. Our cxperience of such therapy during pregnancy is very limited because most HIV infected pregnant patients have not been severely immunocompromised. However it seems reasonable to draw on information from other patient groups, as Dr Kell has done, demonstrating the benefit of PCP prophylaxis and antiretrovirals. Our own experience of death in acute PCP is that all patients dying during the first episode have neglected medical contact for the previous six months, even when they became unwell. Early recognition of medical condition is probably the most important factor in survival rather than risk activity, gender or pregnancy. In our series, the three women who developed PCP duringpregnancy were not taking prophylaxis (one disclosed her pregnancy only when she presented with PCP, one did not attend, and one was not known to be infected with HIV). We advise offering PCP prophylaxis in pregnancy at a similar CD4 level to nonpregnant individuals ( s200/mm3) and discussing Zidovudine on a case by case basis. Many individuals, especially those with injection drug use as a background risk activity, have poor compliance with such therapy, but we have previously demonstrated that it is possible to treat drug users with Zidovudine (Cowan et al. 1989). Treatment and prophylaxis during pregnancy have been excellently reviewed (Schoenbaum et al. 1992).We disagree slightly with some of Dr Kell's interpretations of the quoted papers:Kooninethl. (1989) did not describe deaths only in pregnant women but in women dying within a year of prcgnancy. Sperling et al. (1992) did not 'support the use of AZT' in pregnancy, but simply reported a retrospective review of 43 pregnancies where this drug was taken (12 in the first trimester) without serious immediate maternal or neonatal problems. The authors are careful to emphasise the limited number of cases, and also that early adverse outcomes of pregnancy, including teratogenic effects resulting in termination of pregnancy, could have been unintentionally under-reported. Although we agree with the overall framework of discussion suggested by Minkoff and Moreno (1990) we would advise caution in the literal acceptance of a CD4 count of 500/mm3 as an action line. This level is the upper limit of a band where an effect of ZDV prophylaxis has been demonstrated but it is not certain that individuals at the upper end of this band do in fact show long-term benefit. In addition, pregnancy results in a decrease in absolute CD4 count of about 100/mm3 in healthy, non-HIV infected women, and nearly 50% of a small group of such non-infected women had a CD4 count <500/mm3 at some time during pregnancy (unpublished work). There are also the confounding variables of smoking, cxcrcise, meals, timing of transport to the laboratory and. most importantly, diurnal variation which seem rarely mcntioned in advice about basing trealment on CD4 counts. Whcther anti-retroviral therapy may reduce vertical transmission rate is a key question but one which will not be answered for some...