Eileen Stillwaggon, Gettysburg College, Department of Economics

Larry Sawers, American University, Department of Economics, lsawers@american.edu

The recent PLoS One article by Morris, Epstein, and Wawer [1] seeks to support the notion that a single form of heterosexual behavior—concurrency, or multiple concurrent partnering—explains the 50-fold difference in HIV prevalence between Africa and countries with the modal prevalence. In late 2009, Mark Lurie and Samantha Rosenthal [2, 3] sharply criticized the concurrency hypothesis. Our article in September 2010 in the Journal of the International AIDS Society [4] extended their critique and included a systematic review of the quantitative and qualitative evidence offered by the proponents of the hypothesis. In their November 2010 PLoS One article, Morris, Epstein, and Wawer present survey data on concurrency from Thailand, Uganda, and the United States from the early 1990s, but they do not address any of the criticisms of the concurrency hypothesis that were published in 2009 and 2010.

The core of their article is a discussion of the data in Table 1 on page 4, comparing sexual behavior in the three countries. There are obvious errors in Table 1 that muddle all the subsequent discussion and undermine confidence in the rest of their data. Morris et al. report three different racial/ethnic categories in the United States with 1003, 259, and 161 women, respectively. They report on three locations in Uganda with 1003, 259, and 161 women. Since the total number of Uganda respondents reported in Table 1 far exceeds the sample size reported on page 7, the Ugandan data, at the least, are incorrectly reported. In addition, in Table 1 (page 4) they report the median duration of partnership overlap for Ugandan rural women as 36 months, but they report it as 6 months in the text (page 4).

Morris et al. argue that higher concurrency rates in Uganda help to account for the higher HIV prevalence there. But even Morris’s own modeling (with Kretzschmar), using data from the same survey in Uganda, shows that concurrency cannot make much of a difference. Despite assuming transmission rates and coital frequencies far higher than the evidence supports, they found that concurrency could generate only 26 percent more HIV than serial monogamy (page 124 in [5]). Morris’s own modeling, even with exaggerated parameters, explains less than 1 percent (= 26 percent/3000 percent) of the difference between Uganda and the United States at that time.

Morris, Epstein, and Wawer also point to the longer duration of partnership overlap in Uganda compared with the United States, arguing that it is more likely to guarantee contact during acute infection. For one ethnic/gender group in the United States, mean partnership overlap was 7.5 months, and it was more than a year for all the other groups. The length of those partnerships would be more than sufficient to utilize the high transmission efficiency of primary infection. The longer partnership overlap in Uganda is not important in understanding its higher HIV prevalence.

Morris, Epstein, and Wawer make the argument that the reason concurrency spreads HIV is not one’s behavior, but the behavior of one’s partner. Therefore (page 2), “estimating this effect requires a study design that enrolls both partners in a sexual relationship.” They go on to assert that studies using that design “have consistently supported the concurrency hypothesis.” They cite five studies to support that assertion, but none of them corroborates their statement. One article studied people who acquired HIV, but not from their primary partner. The study provides no information about whether the secondary partners were long-term or heterosexual or if infection could have occurred through blood exposures [6]. Another cited study is about HIV risk for women whose partners are bisexuals and/or patronize commercial sex workers [7]. One study did not enroll both partners [8], but reports people’s guesses as to how their partner had become infected (for more information, see [4]). A fourth article [9] shows that very few people whose spouse had a concurrent partner became infected. And finally, one article presents findings about the gender of the person who infects his or her partner that sheds no light on the concurrency hypothesis [10]. In sum, their assertion that they cite five studies that enroll both partners and that those studies “have consistently supported the concurrency hypothesis” is incorrect on both counts.

As the authors admit, their data are not suitable for drawing statistical inferences, since there are only three observations (Thailand, Uganda, and the United States) that leave no degrees of freedom. Nevertheless, they repeatedly appeal to the reader to see that the behavioral data “align with” [abstract], “lines up with” [page 4], “lined up with” [page 4], or even “line up perfectly with” [page 5] HIV prevalence in the three countries, and that therefore the study “supports the hypothesis” [page 5]. Researchers use statistics for a reason. Lining up data in a table and then eyeballing them is not an appropriate methodology for determining the validity of the concurrency hypothesis.

We have written a more detailed criticism (nearly three times the length of this comment) of the Morris, Epstein, and Wawer article, but PLoS One refused to consider publishing it, saying that the journal published only the results of primary scientific research. Interested readers are invited to contact us for the paper.

Works Cited
1. Morris, M., H. Epstein, and M. Wawer, Timing Is Everything: International Variations in Historical Sexual Partnership Concurrency and HIV Prevalence. PLoS One, 2010. 5(11): p. 1-8.
2. Lurie, M.N. and S. Rosenthal, Concurrent partnerships as a driver of the HIV Epidemic in sub-Saharan Africa? The evidence is limited. AIDS and Behavior, 2010. 14(1): p. 17-24; discussion 25-8.
3. Lurie, M.N. and S. Rosenthal, The Concurrency Hypothesis in Sub-Saharan Africa: Convincing Empirical Evidence is Still Lacking. Response to Mah and Halperin, Epstein, and Morris. AIDS and Behavior, 2010. 14(1): p. 17-24.
4. Sawers, L. and E. Stillwaggon, Concurrent sexual partnerships do not explain the HIV epidemics in Africa: a systematic review of the evidence. Journal of the International AIDS Society, 2010. 13(34): p. 13-34.
5. Morris, M. and M. Kretzschmar, A Microsimulation Study of the Effect of Concurrent Partnerships on the Spread of HIV in Uganda. Mathematical Population Studies, 2000. 8(2): p. 109–133.
6. Celum, C., et al., Acyclovir and transmission of HIV-1 from persons infected with HIV-1 and HSV-2. N Engl J Med, 2010. 362(5): p. 427-39.
7. Johnson, K., et al., Sexual networks of pregnant women with and without HIV infection. AIDS, 2003. 17: p. 605-612.
8. Guwatudde, D., et al., Relatively low HIV infection rates in rural Uganda, but with high potential for a rise: a cohort study in Kayunga District, Uganda. PLoS One, 2009. 4(1): p. e4145.
9. Mermin, J., et al., Risk Factors for Recent HIV Infection in Uganda. Journal of the American Medical Association, 2008. 300(5): p. 540-549.
10. Hugonnet, S., et al., Incidence of HIV Infection in Stable Sexual Partnerships: A Retrospective Cohort Study of 1802 Couples in Mwanza Region, Tanzania. Journal of Acquired Immune Deficiency Syndromes, 2002. 30: p. 73-80.