How many persons in Canada have been infected with human immunodeficiency virus? An exploration using backcalculation...

We have used the technique of backcalculation to estimate the number of persons in Canada who have been infected with human immunodeficiency virus type 1 (HIV) as of July 1989. We first corrected national AIDS surveillance data in Canada for reporting delay and for underreporting. We then used standard Weibull natural history models as well as an alternative progression model in which the hazard of AIDS was dampened in keeping with observed data from large cohorts. Maximum likelihood techniques were then used to derive the infection curve most consistent with these data. Our best estimate based on this alternative progression model, an underreporting rate of 20%, and a logistic infection curve, was that approximately 29,000 persons in Canada had ever had HIV infection as of July 1989. In a sensitivity analysis utilizing less likely assumptions, the estimates ranged from 17,243 to 48,277. Restricting the same backcalculation process to females under the same assumptions, we estimated that approximately 2,900 females in Canada had ever had HIV infection as of July 1989. The best fitting step function infection curve in females appears to be continuing to rise. Given these estimates, it follows that approximately one in ten infected persons in Canada is female. However, females have only accounted for about one in seventeen AIDS cases. These data are in accord with the widespread impression that transmission of HIV to women has occurred more recently and is on the rise across Canada. The current estimate is lower than previous estimates which placed the number of infected Canadians in the 50,000 to 100,000 range. This lowering should not be taken to mean the situation is improving; rather, the early estimates were simply too high having been based on inadequate data and a rudimentary understanding of the natural history of HIV. Backcalculation is an excellent technique for modelling the incidence of HIV infection several years in the past, but it is not reliable for the most recent few years. A significant increase in HIV infection rates may have occurred in the past few years and it would be beyond the capacity of backcalculation to detect. Backcalculation and unlinked cross-sectional specimen surveys together have the potential to provide an effective means of monitoring the HIV epidemic.