Haiti is by all conventional criteria the poorest country in the western hemisphere and one of the poorest in the world: 13 per capita gross national product (GNP) is around US $400; unemployment exceeds 70%; and fewer than one in 50 Haitians have regular employment. 14 Not coincidentally, Haiti is also the hemisphere’s most HIV-burdened country. 15 In 1999, UNAIDS reported national HIV seroprevalence as 5% among women attending antenatal clinics—and rates were twice as high in urban slums. 11 The latest estimates of life expectancy at birth are 47· 5 years for men and 49· 2 years for women, with HIV considered the chief contributor to premature adult death. 16 Initially an urban epidemic, HIV prevalence is lower in rural Haiti, where we have worked for more than 15 years. Most of the local inhabitants in the lower Central Plateau are peasant farmers working small plots of infertile land. Many are sharecroppers. Local health indicators are worse than national estimates.

Our clinical facility, founded in 1985 in the middle of a settlement of individuals displaced by a hydroelectric dam, documented its first case of HIV disease in 1986. Following international convention, prevention efforts were tightly linked to education and condom promotion. 17 These efforts have been hampered by political violence and resulting migration, and by gender inequality and poverty, which conspire to make the male condom an imperfect prevention measure. Thus, HIV transmission continued in spite of aggressive prevention campaigns. 18 Our modest therapeutic efforts have been aggressive when compared with other clinics in poor, rural regions of the less-developed world. Shortly after the publication of the ACTG-076 trial, 19 we began offering zidovudine to pregnant women to block mother-to-child transmission. More than 90% of women offered HIV testing accepted it after zidovudine was made available free of charge; dramatic declines in vertical HIV transmission ensued. In 1997, we began offering post-exposure prophylaxis with a three-drug regimen (usually zidovudine, 3TC, and a protease inhibitor) to victims of rape or professional injury. 20 Beginning in late 1998, a small number of patients with long-standing HIV disease who no longer responded to syndromic treatment of opportunistic infections were offered directly observed HAART. Inclusion criteria for HAART have not been codified rigidly, but follow a certain logic in the absence of CD4 lymphocyte counts and viral-load testing. Patients assessed for HAART are those with chronic enteropathies or other forms of HIV-associated wasting; patients with presumed neurological complications of HIV (encephalopathy, distalsensory, or other polyneuropathies); those with repeated opportunistic infections unresponsive to antibacterials and antifungals; and patients with severe leukopaenia, anaemia, or thrombocytopaenia (panel 1). Assessments are done by two physicians, one with infectious-disease training. Patients diagnosed with active tuberculosis are not offered HAART because most respond to antituberculous