Navigating the Health Care System
Understanding Research Results: The Difference between Prevalence and Incidence
This article was first published in Global Campaign News, the newsletter of the Global Campaign for Microbicides. It explains the difference between prevalence and incidence by examining recent research results on microbicides.
Several recent articles in Global Campaign News have made reference to “lower than anticipated HIV incidence” during effectiveness trials. Here, we take a closer look at what incidence means, why so many trials are seeing lower incidence rates than expected, and what the implications are for current and future trials.
Prevalence and incidence are two related, but different measures that describe the distribution of disease in a particular population. Prevalence is a measure of the number of total cases of a disease in a population at a certain moment in time. For example, among women presenting for screening during a feasibility study at the Mtubatuba site in South Africa, the prevalence of pre-existing HIV infection was 50% (number of HIV positive women/number of total women being screened).
The incidence of disease is the number of new cases occurring in a population over a defined time interval. Incidence measures how quickly one sees new cases of infection or disease, whereas prevalence describes how many people total in a population are affected, regardless of when they become infected or sick. At the Mtubatuba site quoted above with a prevalence of 50% among screened women, the HIV incidence among those women enrolled was 12.6 infections per 100 person years. This means, among every 100 women they followed, 12 to 13 people became infected in the course of one year. [Note: the term "person-years" is a convention from epidemiology that allows researchers to annualize estimates of infection from individuals followed up for different lengths of time].
Thus it is possible to have situations of high prevalence but low incidence and vice versa. For example, the incidence of new cases of diabetes in a population may be only 1 per 1000 people per year (or 0.01% annually) but the prevalence of diabetes in a population could be 8 percent. The .01% incidence estimate includes only people who were newly diagnosed with diabetes this year whereas the prevalence estimate includes these people as well as those already living with diabetes who were diagnosed in the past.
Likewise is it possible to have pockets of high HIV incidence (e.g. high rates of new HIV infections among intravenous drug users) in settings with an overall lower rate of HIV prevalence (e.g. less than 1 percent overall). Significantly, most measures of infection quoted by UNAIDS, National AIDS control authorities, and in the media are measures of HIV prevalence, not incidence.
Regrettably, there is no cheap, easy way to derive accurate estimates of HIV incidence. The most reliable way to establish incidence is to enroll HIV negative women in a cohort study and evaluate, using repeat HIV tests, the number of HIV infections that occur over time. For example, the Microbicide Development Programme conducted cohort “feasibility” studies to determine the incidence of HIV in the different populations being considered for inclusion in their current phase III trial of two concentrations of Pro2000. They found HIV incidence rates ranging from a low of 3.5 per 100 person-years in Mwanza, Tanzania to a high of 12.6 per 100 person-years in Mtubatuba, South Africa.
Cohort studies, however, are expensive to run and delay the start date of a potential trial for at least 6 months to a year while incidence data is being collected. In the interest of speed, some trial sponsors have tried to estimate the likely incidence in their participant population based on past prevalence estimates or data from previous studies about the observed ratio of prevalent to incident cases. As recent trial closures demonstrate, however, such approaches can yield misleading results. The field is currently discussing the pros and cons of different strategies for estimating the incidence of HIV.
Researchers in the Ghana Savvy trial estimated, for example, that there would be at least five infections per 100 person years of follow up in the placebo group (an HIV incidence rate of 5%), and that they would observe at least 66 incident infections during the trial. However, halfway through the study, an interim analysis found that only 17 total sero-conversions had occurred: nine on placebo and eight on Savvy. This translates into an HIV incidence of only 1.0% for Savvy and 1.1% for the placebo.
This incidence was dramatically lower than anyone anticipated, and the trial was closed on the recommendation of the Data Safety and Monitoring Board. Given the low rate of incident HIV infection observed, the DSMB concluded it would not be possible to recruit enough participants to answer the question of Savvy’s effectiveness.
There are several possible explanations for the lower than expected incidence rates in effectiveness trials. The first is that the original estimate of HIV incidence could have been inaccurate, especially if not based on cohort data. As noted above, accurate estimates of incidence are difficult to come by. Also incidence can shift dramatically over time, especially in populations where men and women migrate frequently.
High rates of pregnancy among trial participants may also have contributed to lower rates of HIV acquisition. According to a presentation by Dr. Wes Rountree at M2006, women in the Ghana Savvy trial who discovered they were pregnant changed their sexual behavior in ways that reduced their risk of HIV. They engaged in sex less often, had fewer unprotected sex acts, and fewer partners. These behaviour changes together with high rates of pregnancy could partially account for the low rate of HIV incidence observed in the Savvy study.
Finally, just by participating in a prevention trial of this sort, a participant’s risk of HIV acquisition may be diminished. Participants are getting the best available safer sex counseling and support, which is reinforced with every clinic visit. Participants also receive treatment for other sexually transmitted infections, which in turn indirectly decreases their risk of acquiring HIV. Thus, a good HIV prevention study itself can dramatically lower HIV incidence among participants. This is great news for the trial communities, but makes it more difficult to determine whether the candidate product is effective.
To address this issue, recruitment strategies are being modified to increase the likelihood of enrolling women at highest risk of HIV infection. Since younger women are often at the highest risk for new HIV infection, current efficacy trials are focusing on recruiting younger participants. Trial groups and sponsors are also working to explore the use of new assays and surveillance techniques to better estimate HIV incidence during screening and/or through pilot studies, in order to arrive at estimates of HIV incidence that are as accurate as possible.
For more information, and sources for this article, see:
- Smart, T. Microbicides 2006: Are the microbicide clinical efficacy studies big enough? NAM’s aidsmap.org Wednesday, May 10, 2006.
- Skoler, S., Peterson, L., Cates, W. Our Current Microbicide Trials: Lessons Learned and To Be Learned, The Microbicide Quarterly, Jan-March 2006, Vol 4. No. 1.
Written by: The Global Campaign for Microbicides
Last revised: Nov 2006
< Return to Navigating the Health Care System Overview