India's HIV prevalence rates have been grossly overestimated and require substantial downwards revision, according to a late-breaker presentation at the Sixteenth International Conference in Toronto on Thursday August 17th. Population-based HIV prevalence estimates in Andhra Pradesh, the south Indian state thought to have the highest prevalence of HIV, were found to be two-and-half times lower than estimates based on sentinel screening at antenatal and sexual health clinics.
At the end of 2005, India's National AIDS Control Organisation (NACO) estimated that there were 5.21 million people living with HIV, although this figure had been criticised by some as an underestimate.
Indeed, UNAIDS estimated that the number of people living with HIV in India at the end of 2005 was closer to 5.7 million - higher than South Africa's estimated 5.5 million - although it suggested a lower and upper range of 3.4 million and 9.4 million, respectively.
NACO extrapolates data from the public health system – primarily women testing for HIV during antenatal care, as well as anonymous surveillance from sexual health clinics – in order to estimate HIV prevalence in India, and until now no population-based study had systematically examined the validity of this extrapolation.
Investigators from the Administrative Staff College of India and the Institute of Medical Sciences in Hyderabad sampled 13,838 adults aged 15-49 from 66 rural and urban clusters that represented the 4.5 million adults in the Guntur district of south India's Andhra Pradesh state, which is thought to have the highest HIV prevalence in India.
Demographic data were obtained from a total of 12,617 individuals (91.2% of those sampled), and dried blood spots obtained at the same time were tested for HIV antibodies, p24 antigen and HIV RNA. The results were then compared with the HIV estimates from NACO's sentinel surveillance data.
The investigators found that the adjusted HIV rate in adults was 1.72% (95% CI, 1.35-2.09%), with a slightly higher prevalence in men (1.74%; 95% CI, 1.27-2.21%) than women (1.70%; 95% CI, 1.36-2.04%). Prevalence was higher in urban settings (1.89%; 955 CI, 1.39-2.39%) compared with rural settings (1.64%, 95% CI, 1.10-2.18%).
Using a standard of living index (SLI), the investigators found that HIV prevalence rates were twice as high in people who were in the lower half of the SLI (i.e. poorer), compared with those in the upper half (2.58% vs 1.20%, respectively).
Around 23% of pregnant pregnant women had accessed antenatal care through the public health system in the previous two years. Significantly, they discovered that very poor women were over-represented, which resulted in an unusually high HIV prevalence rate (3.61% versus 1.08% of pregnant women who did not use the public health system). This would have skewed the surveillance-based data which found antenatal clinic HIV prevalence to be 3%.
The investigators calculated that their population-based estimate for the Guntur district is 45,925 adults after adjusting for underrepresented high-risk groups. In contrast, NACO's surveillance-based estimate was 112,635 adults. In other words, population-based HIV prevalence was two-and-a-half times lower than the sentinel surveillance data.
By applying the investigator's population-based estimates to the entire state of Andhra Pradesh, this results in a reduction from 1.44 million adults living with HIV to just over half a million.
Lalit Dandona, of the Centre for Human Development in Hyderabad, presenting, concluded that NACO's method for estimating HIV prevalence grossly overestimates HIV in all south Indian states – home to around three-quarters of India' HIV-positive population – and which have a similar pattern of public health system use by the poorest people. Therefore, NACO's current HIV estimate of 5.21 million for India should be revised downwards by a substantial amount.
Dandona L et al. A population-based study suggests that the HIV estimate for India needs major revision. Sixteenth International AIDS Conference, Toronto, abstract ThLB0107, 2006.