Vermund, Sten H.
Departments of Epidemiology, Medicine, and Pediatrics, Schools of Public Health and Medicine, University of Alabama at Birmingham, Birmingham, Alabama, U.S.A.
Address correspondence and reprint requests to Dr. S. H. Vermund, UAB School of Public Health, 720 20th Street South (TH203), Birmingham, AL 35294-0008, U.S.A.
The control and ultimate elimination of sexually transmitted infections (STIs) depend on the social dynamics of sexual behavior (1-3) and on its biological features. STIs are often asymptomatic, may be chronic, and are diagnosed with some difficulty using syndromic, microbiologic, and/or serologic assessment. To address all of the features that modulate STI risk, particularly for HIV, is enormously complex. In this issue, an internationally recognized study is the data source for a series of models attempting to quantify the average risk of heterosexual HIV transmission from an HIV infected person.
Downs, De Vincenzi, and the European Communities Study Group on Heterosexual Transmission of HIV analyzed 563 heterosexual couples from a total of 13 centers in Belgium, Britain , France , Germany, Greece, Italy , the Netherlands, Portugal and Spain  (4). Fourteen couples were excluded due to incomplete data, and 24 couples were excluded when they reported having had no unprotected (i.e., without using male condoms) sexual encounters since January 1982. Interview data were used to estimate the number of unprotected sexual contacts for a decade or more for 377 couples in which the male was the known seropositive subject and the female had no other known risk exposure for HIV, and for 148 couples in whom the female was the known seropositive subject and the male had no other known risk exposure for HIV. Estimates of heterosexual transmission risk were calculated from the serostatus and sexual behavior data; no efforts to control for history of prior STI, circumcision status of male, or other transmission-relevant factors are reported.
Two mathematical models were used to estimate the risk of transmission, highlighting 90 couples in which both partners were seropositive (concordant) at study entry compared to the discordant couples (4). Each model was then expanded to include an additional 12 couples who began as discordant for HIV status but whose seronegative partner seroconverted to HIV over the course of the study. The two models were based on nonparametric principles (the isotonic regression model) and on parametric assumptions (the Bernoulli model), respectively. A constant per-contact infectivity was assumed, i.e., every sexual contact has the same likelihood of HIV transmission. Finally, a third model was introduced as a “simple” alternative model to the Bernoulli parametric model. It was applied only to the data at study entry (90 concordant couples), and it was built on the assumption that only a fraction of cases were at all infectious (i.e., 17%) but that these cases all transmitted at the first contact.
Parametric assumptions rely on data being normally distributed, as with blood pressure, or serum biochemical findings, or a host of biological variables that vary symmetrically around a mean. Data that are distributed in a non-normal (skewed) fashion may be analyzed using parametric techniques if logarithmic or other biostatistical transformations result in a reasonable approximation of normality. When raw or transformed data are not close to being normally distributed or when the understanding of the distribution of the data depends on unknown, unmeasurable parameters, then nonparametric statistical techniques are appropriate.
Of 377 couples in which the male was the seropositive subject, 74 female partners without other known risk were seropositive at study entry and 8 more seroconverted over the life of the study (82/377 or 21.8%). Of 148 couples in whom the female was the seropositive subject, 16 male partners without other known risk were seropositive at study entry and 4 more seroconverted over the life of the study (20/148 or 13.5%). When applying the two models, there were substantial differences between them, suggesting that the parametric model poorly estimated transmission risk when the number of unprotected contacts was either at its highest or lowest values. Overall risk of transmission was estimated at 1.2 per 1,000 unprotected contacts (95% C.I.: 0.7, 2.1 per 1000) with male-to-female risk somewhat less than double the female-to-male risk. However, the fit of the data in the “simple” parametric model in which infectiousness was assumed in only 17% of subjects and risk was judged 100% at the first sexual contact (both admittedly wrong assumptions) was superior to the constant infectivity parametric model. In the judgment of the authors and of this reviewer, this comparative failure of the constant infectivity model to fit the data casts serious doubt on the validity of the model to quantitate risk.
How can one interpret? This reviewer would like to highlight two sentences by Downs and De Vincenzi: “The present findings accord well with a hypothesis of variable infectivity...,” and, “Only by taking into account as fully as possible the distribution of infectivity in given populations at given times can we hope to better elucidate the true risks... of heterosexual intercourse (4).” The authors are frank in their explicit acceptance of the limitations of epidemiological data based on HIV serology and sexual behavioral data from interview alone (5).
Factors of highest relevance for heterosexual HIV transmission can include viral, host, and environmental aspects (Table 1), thereby expanding substantially the complexities of a heterosexual couples study research agenda (6). For a heterosexual inoculation to occur, the viral load in the seminal or cervicovaginal fluid must be of adequate dose to infect a partner, and the target cells that are infectable must come in contact with viable HIV. Even then, infection may not be successful. Viral load corresponds to the time course of the HIV-infected person's illness, highest immediately prior to seroconversion and, again, late in the clinical course (7-10). Use of antiretroviral chemotherapy may alter substantially risk of transmission (11). Infectable cells are recruited when a sexually transmitted infection is present and may be more susceptible to infection when activated by systemic co-infections. Increased probability of successful infection may occur in the face of increased bleeding (e.g., cervical ectopy, traumatic sex), increased contact time (e.g., lack of male circumcision), susceptible host genetic profile, suboptimal epithelial integrity (e.g., due to abnormal vaginal or urethral colonization profiles or nutritional compromise), or other factors. The next generation of epidemiological studies will be illuminating only insofar as these factors can be measured and their impact on transmission judged. Given the improbability of adequate measurement of these factors in an observational context, large-scale clinical trials that attempt to intervene on one or more risk variables and that can measure other factors and randomize to attempt to balance them between groups may be the best approach to future work in this arena (12,13).
1. Laumann EO, Gagnon JH, Michael RT, Michaels S. The social organization of sexuality: practices in the United States. Chicago: University of Chicago Press, 1994:1-742.
2. Johnson AM, Wadsworth J, Wellings K, Field J, Bradshaw S. Sexual attitudes and lifestyles. Cambridge: Blackwell Scientific Publications, 1994:1-499.
3. Horn JS. Away with all pests. New York: Monthly Review Press, 1969:81-93.
4. Downs AM, De Vincenzi I. Probability of heterosexual transmission of HIV: Relationship to the number of unprotected sexual contacts. J Acquir Immune Syndr Hum Retrovir 1996;11:388-95.
5. Taubes G. Epidemiology faces its limits. Science 1995;269:164-9.
6. Vermund SH. Transmission of HIV-1 among adolescents and adults. In: DeVita VT, Jr., Hellman S, Rosenberg SA, Curran JW, Essex M, Fauci AS, eds. AIDS. Philadelphia: Lippincott, 1996 (in press).
7. Pantaleo G, Graziosi C, Fauci AS. New concepts in the immunopathogenesis of human immunodeficiency virus infection. N Engl J Med 1993;328:327-35.
8. Clumeck N, Taelman H, Hermans P, Piot P, Schoumacher M, de Wit S. A cluster of HIV infection among heterosexual people without apparent risk factors. N Engl J Med 1989;321:1460-2.
9. Koopman JS, Longini IM Jr., Jacquez JA, et al. Assessing risk factors for transmission of infection. Am J Epidemiol 1991;133:1199-209.
10. Koopman JS, Jacquez JA, Simon CP, et al. The importance of primary HIV infection in sustaining the HIV epidemic. Program and Abstracts of the 3rd Conference on Retroviruses and Opportunistic Infections. Alexandria, VA: IDSA, 1996, Abstract S30:176.
11. Anderson DJ, O'Brien TR, Politch JA, et al. Effects of disease stage and zidovudine therapy on the detection of human immunodeficiency virus type 1 in semen. JAMA 1992;267:2769-74.
12. Vermund SH. The role of prevention research in HIV vaccine trials. AIDS Res Hum Retrovir 1994;10:S303-5.
13. Grosskurth H, Mosha F, Todd J, et al. Impact of improved treatment of sexually transmitted diseases on HIV infection in rural Tanzania: Randomised controlled trial. Lancet 1995;346:530-6.
© Lippincott-Raven Publishers.