Study Finds Advances In Estimating HIV Incidence In Epidemics

An article published in the open-access journal PLoS Medicinereports on the development and testing of two user-friendly methodsthat use changes in cross-sectional HIV prevalence (the fraction of thepopulation infected with HIV) to estimate HIV incidence (the number ofnew infections occurring during a specific time period). TimothyHallett (Imperial College London) and colleagues suggest that theincidence of HIV can be estimated from repeat surveys of prevalencewith enough accuracy to monitor the epidemic.

Currently, 33million people are infected with HIV (the virus that causes AIDS), andAIDS has already killed more than 25 million people. Working to thwartthis epidemic, governments and international agencies have beenassessing the impact of interventions by keeping track of how the virusspreads. Usually, agencies monitor generalized epidemics (ones thathave spread to the whole population) by determining the prevalence ofHIV infection among women who attend antenatal clinics. More accuratemeasures of HIV prevalence are being acquired by testing blood forantibodies related to the AIDS virus (serological testing).

Researchers,though still interested in prevalence, are also concerned with theincidence of the virus in order to gage how the epidemic changes overtime and how the virus is transmitted. However, measures of incidenceare generally more costly - researchers would have to identifyindividuals, test their blood and then repeatedly follow up the sameindividuals. Hallett and colleagues tackle this problem by developingmathematical methods that allow one to use prevalence data to estimatethe incidence of HIV in generalized epidemics.

Realizing that changes in HIV incidence and mortality rates contributeto changesin HIV prevalence, the researchers developed models that includemethods to disentangle these features. The first method combinesinformation on mortality rates collected incohort studies of HIV infection, while the second method uses datacollected in long-running cohort studies that focuses on survival afterHIV infection. Computer-simulated data and real data on HIV prevalenceand incidence from cohort studies in Zimbabwe, Uganda, and Tanzaniawere used to assess the accuracy of the two methods. Both estimationmethods resulted in accurate predictions of HIV incidence from thesimulated data. When the data from Africa were used, the averagedifference between actual incidence measurements and estimates was19% from the first method 14% from the second method.

“Neither method tends to systematically over- or underestimateincidence,” write the authors.

These findings indicate that repeat surveys of prevalence can helpestimate HIV incidence rates with sufficient accuracy to monitor theepidemic. One potential weakness of the study is that it may bedifficult to generalize to other parts of Africa where HIV epidemicsare restricted to subsets of the populations. As the availability ofprevalence measures from blood tests increases, the new methodspresented for estimating HIV incidence from HIV prevalence could bevery useful for tracking the progress of national epidemics and forinforming HIV control programs.

Estimating incidence from prevalence in generalised HIVepidemics: Methods and validation
Hallett TB, Zaba B, Todd J, Lopman B, Wambura M, et al.
PLoS Medicine (2008). 5(4): e80.
doi:10.1371/journal.pmed.0050080
ClickHere to View Article