Trends in HIV incidence in India from 2000 to 2007

Abstract

Indias HIV epidemic is of global interest. 2 years ago we showed that HIV prevalence in young women declined by about a third between 2000 and 2004 in the southern states of Andhra Pradesh Karnataka Maharashtra and Tamil Nadu. HIV prevalence at young ages (15-24 years) is a useful proxy for trends in HIV incidence. We now present trends up to 2007. Among 423842 women aged 15-24 years tested nationally at antenatal clinics prevalence declined by 54% (95% CI -45 to -63; p<0.0001) between 2000 and 2007 in south India and there was no significant change in north India (3% -47 to 53; p=0.73) where HIV is less prevalent. Declines in south India were similar if we analysed individual age-groups if we excluded Tamil Nadu or restricted the analyses to each individual state or to the sites tested continuously for at least 4 years. Women who use antenatal clinics differ from those who do not in education residence and migration but these demographic factors remained similar from year to year. More research is needed to understand why incidence has fallen in south India. The most probable reason is reduced contacts with female sex work by the husbands of tested women or increased condom use in sex work. Although useful for estimating trends in HIV incidence data from antenatal clinics cannot estimate community prevalence reliably. The National Family Health Survey of 2005-06 (NFHS-3) yielded lower HIV prevalence nationally in adults (0.28% 95% CI 0.25-0.31 at ages 15-49 years) than seen among women at antenatal clinics in our study (0.60% 0.57-0.63 at ages 15-49 years). A study in one district suggested that women with HIV were over-represented in public antenatal clinics but we found that HIV infection was associated with lower use of public antenatal clinics within the NHFS-3. Among 8743 eligible women survival analyses with Coxs regression of time since last antenatal clinic use yielded a hazard ratio for HIV of 0.44 (0.22-0.90; p=0.02) after adjustment for age and sampling unit. The halving of new infections in south India and the lack of demonstrable increases in the north would at first glance seem to be consistent with Indias downward revision of HIV prevalence in 2006 from 5.1 million to 2.5 million (range 2.0-3.1 million). However the revised prevalence estimates are based largely on hybrid analyses that combine antenatal clinic and NFHS-3 data whereas earlier estimates were based on antenatal clinic data. The NFHS-3 has biases also including the under-representation of high-risk groups. In conclusion although the estimation of HIV trends is reasonably robust we caution that prevalence estimates remain uncertain. Reliable estimation of prevalence requires combining various sources of data including information on AIDS mortality. (full-text)