Estimating the HIV transmission risk in the presence of combined antiretroviral treatment (cART) is fundamental for patient counseling, design of prevention interventions, and predicting, at the population scale, the impact of the treatment as prevention strategy, based on widespread HIV testing and early cART. Results from 3 recent major studies, HPTN 052,1 PARTNER (phases 1 and 2),2,3 and Opposites Attract,4 confirmed earlier observations5–9 that cART, leading to undetectable viral load, is highly effective as a prevention strategy for HIV-serodiscordant couples.
Before the completion of the 3 aforementioned studies, data were lacking to quantify the HIV transmission risk in the presence of effective cART.10 Under these circumstances, Wilson et al11 pioneered the use of mathematical modeling in estimating the risk of HIV transmission under cART. Specifically, in 2008, Wilson et al11 extrapolated a dose–response relationship between viral load and the risk of HIV transmission among untreated individuals12 to low values for the plasma viral load (pVL below 400 copies/mL), to simulate the effect of cART and estimate the risk of HIV transmission per unprotected sex act under effective cART. Then, they used these estimates to predict how many seroconversions would occur in a population of 10,000 monogamous serodiscordant couples over 10 years, assuming that effectively treated HIV-infected individuals may have pVL as low as 10 copies per milliliter. In particular, they found that abandonment of condom use by effectively treated people could lead to substantial increase in the cumulative number of new seroconversions among serodiscordant couples, up to 35% for serodiscordant couples of men who have sex with men (MSM).11 Their conclusions rely heavily on the assumption that the dose–response relationship between viral load and risk of HIV transmission, observed among untreated individuals, can be extrapolated to treated individuals. In subsequent studies, others performed the same extrapolation,13,14 using the same or similar dose–response relationships developed for untreated individuals,15 for other purposes; eg, to estimate risks of transmission across stages of HIV disease and the proportion of new HIV infections due to effectively treated individuals.13
Presently, data on serodiscordant couples have accumulated for a direct validation of this extrapolation.11 A systematic review16 identified 6 studies,5–9,17 published between 2008 and 2011, on HIV transmission within heterosexual HIV-serodiscordant couples, where the infected partner was on cART, in comprehensive care including viral load monitoring. Since then, 3 other studies on HIV transmission within HIV-serodiscordant heterosexual and MSM couples were published2–4,18 as well as additional results for the HPTN 052 study.1,19 Overall, 9 clinical studies have gathered data on the risk of HIV transmission when the infected partner was on cART in comprehensive care1–9,17–19; this includes studies monitoring HIV transmission since cART initiation1,5–9,17–19 or only when HIV-infected partners were virally suppressed.2–4 Using a statistical framework and estimates of the risk of HIV transmission under effective cART proposed by Wilson et al,11 it is possible to compute the expected number of HIV cases that should have occurred across the 9 clinical studies and compare it with the number of cases actually observed. Thus, in the light of available data, we may conclude on whether or not the extrapolation of the dose–response relationship between viral load and risk of HIV transmission should still be used to estimate the risk of HIV transmission under effective cART. This is the main objective of this study. In addition, using a data-driven approach,16 we estimate the maximum value for the risk of HIV transmission per unprotected sex act under cART; we update previous estimates for heterosexual couples16 and provide, for the first time, estimates for MSM. In each case, we excluded seroconversions that occurred within the first 6 months of cART or were diagnosed after treatment failure. Thus, for studies monitoring HIV transmission since cART initiation, we dropped follow-up corresponding to the first 6 months of cART.
We first present the dose–response relationship linking the risk of HIV transmission per unprotected sex act to the pVL of the untreated HIV-positive partner.12 Then, we describe a statistical framework for both calculating the expected number of seroconversions in a cohort of serodiscordant couples and also directly estimating the per sex-act risk of HIV transmission from clinical data on serodiscordant couples.
Dose–Response Relationship Between pVL and the Risk of HIV Transmission
Among untreated individuals, the probability of HIV transmission per unprotected sex act was shown to increase by a factor of 2.45 (95% confidence interval: 1.85 to 3.26) for each log10 increment in pVL,12 according to the equation:where β0 is the per-act probability of HIV transmission from a person with the baseline pVL V0 and β1 is the per-act transmission probability corresponding to any other pVL V1, whether above or below V0. Following Wilson et al,11 we assumed that V0 was 4.5 log10 copies per milliliter, β0 was equal to 0.0005 for female-to-male transmission, 0.001 for male-to-female transmission, and 0.01 for male-to-male transmission. We used equation (1) to estimate β1 for effectively treated HIV-infected individuals with pVL as low as 10 copies per milliliter (V1 = 1 log10), for each type of transmission. Uncertainty bounds (UB) for these estimates can be obtained based on the confidence interval for the empirical factor.
Starting from the standard formula for the cumulative risk of HIV transmission on cART over n sex acts (eg, Ref. 11 and references therein)a complete formalism can be constructed using binomial statistics (Ref. 16 and its supplementary material). Briefly, consider a cohort of c serodiscordant couples, where a fraction k use condoms all the time and the rest never use condoms. The probability of observing s seroconversions among (1 − k)c serodiscordant couples not using condoms is given by the binomial formula
Similarly, if [1 − (1 − ε) β1] represents the probability of remaining uninfected after 1 condom-protected sex act, where condom efficacy is ε, then the probability of observing s seroconversions among kc serodiscordant couples, using condoms 100% of the time, is given by
Hence, the likelihood of observing a total of s seroconversions in a study where (1 − k) c serodiscordant couples are not using condoms, and kc serodiscordant couples are using condoms can be written as
The likelihood for observing a certain pattern of seroconversions over N studies is given by the product of the N likelihoods corresponding to the studies.
The total likelihood can be used in 2 different ways. First, given all needed parameters, including β1, one can estimate the likely number of seroconversions that occurred in the studies, where the number of seroconversions expected in a single study is
We used the values of β1 obtained by Wilson et al11 for effectively treated HIV-infected individuals (ie, β1 obtained from equation (1) assuming that V1 = 1 log10) to estimate the expected number of HIV transmissions under cART that should have been observed across the 9 clinical studies.1–9,17–19 This implies that all treated HIV-infected individuals participating in the clinical studies reached viral suppression at 10 copies per milliliter. It may be a reasonable assumption for studies only monitoring virally suppressed patients.2–4 For studies where patients were monitored since cART initiation,1,5–9,17–19 we dropped follow-up corresponding to the first 6 months of cART for each patient and then considered that each patient was virally suppressed beyond 6 months of cART; in comprehensive care, most patients achieve viral suppression within the first 6 months of cART. Overall, these are optimistic assumptions (Table 1) leading to underestimates for the expected number of HIV transmissions under cART.
Second, given the number of seroconversions that actually occurred in the clinical studies, the total likelihood can be engaged in Bayesian formalism to numerically estimate β1.16 Here, we used data collected across the 9 studies on couples where the HIV-positive partner was on cART to estimate the upper bound of the 95% credibility interval of β1. As above, for studies where patients were monitored since cART initiation,1,5–9,17–19 we dropped follow-up corresponding to the first 6 months of cART.
In both analyses, we excluded seroconversions that occurred within the first 6 months of cART or were diagnosed after treatment failure; condom efficacy was considered to be 75%.16 For MSM, only anal intercourse was considered, whereas for heterosexuals, data did not allow to distinguish the type of intercourse.
Across the 9 clinical studies, 3181 heterosexual and 1126 MSM couples provided, respectively, 7409 and 1778.2 couple-years of follow-up, where the HIV-positive partner was on cART for >6 months or virally suppressed (Table 1). The reported sexual behavior characteristics and number of linked cases of HIV transmission that occurred during follow-up are listed in the table.
Equation (1) provides estimates of the risk of HIV transmission per unprotected sex act with effectively treated individuals with pVL as low as 10 copies per milliliter11: 2.2:100,000 (UB: 0.8–5.8:100,000) for female-to-male transmission, 4.3:100,000 (UB: 1.6–11.6:100,000) for male-to-female transmission, and 43:100,000 (UB: 16–116:100,000) for male-to-male transmission. Using these estimates, data on sexual behavior (Table 1), and the statistical framework described in the Methods section, we computed the number of cases of HIV transmission that should have been expected across the 9 clinical studies.
We obtained that 4.7 (UB: 1.7–12.6) cases of HIV transmission should have been observed among heterosexual serodiscordant couples and 35.1 (UB: 13.2–92.0) cases of HIV transmission among MSM serodiscordant couples. It is important to note that these figures are underestimates; they assume that all HIV-positive partners reached a pVL of 10 copies/mL across the clinical studies (either after 6 months of cART or when individuals were virally suppressed), while the fraction of virally suppressed individuals was 70%–100%, and the viral suppression threshold was 40–400 copies per milliliter (Table 1). Still, the underestimated predictions for the number of HIV transmissions are not validated by observations, which reported at most 1 HIV transmission among heterosexual serodiscordant couples and 0 among MSM serodiscordant couples (Table 1). For heterosexual couples, we considered that at most 1 HIV transmission occurred after 6 months of cART while the HIV-positive partner had an undetectable viral load. This concerns the transmission reported by Apondi et al9 that occurred within the first year of cART while the HIV-positive partner took 6 months to achieve viral suppression. Data were insufficient to determine whether this infection occurred before or after 6 months of cART because the partner, found HIV-negative at baseline, was retested only at 12 months. All other genetically linked seroconversions occurred either within the first 6 months of cART (4 in Ref. 19, 1 in Ref. 8, 3 in Ref. 18) or were diagnosed after cART failure (4 in Ref. 19), and they were excluded. Thus, the per-act risk of HIV transmission under effective cART, estimated by extrapolating the dose–response relationship between viral load and the risk of HIV transmission, seems too high, in conflict with the data accumulated over the past decade.
Rather than using an extrapolation, a direct estimation of the risk of HIV transmission per unprotected sex act, when the HIV-positive partner is on cART, is possible, as described in the Methods section. However, given the very low number of observed transmissions (at most 1 seroconversion among heterosexual couples and 0 for MSM), it is more meaningful to estimate an upper bound for the risk (ie, maximum risk); see Ref. 16 for technical details. Using the data listed in the table, we found that the upper-bound risk for heterosexual couples is 2.6:100,000 (3.9:100,000) when no (one) transmission occurred. Similar analyses of data from the 2 studies of MSM couples (Table 1) yield an upper-bound risk of 4.4:100,000.
Our findings clearly show that, with the completion of major studies on serodiscordant couples,1–4 data have accumulated to render obsolete the extrapolation of the dose–response relationship between pVL and risk of HIV transmission observed among untreated individuals for estimating the risk of HIV transmission on cART. A direct estimation of the risk of HIV transmission per unprotected sex act under cART is possible, using data from the studies on serodiscordant couples and a minimum of modeling assumptions.
Direct estimates of the upper-bound risk are 1.5- to 26.4-fold lower than those obtained by the extrapolation11; ie, 3.9:100,000 versus 5.8:100,000 for female-to-male risk of HIV transmission, 3.9:100,000 versus 11.6:100,000 for male-to-female risk of HIV transmission, and 4.4:100,000 versus 116:100,000 for male-to-male risk of HIV transmission. The extrapolation of the dose–response relationship significantly overestimates the risk of HIV transmission for low pVL in treated individuals. This implies that the actual relationship is vastly different from equation (1) and could display a pVL threshold below which transmission is highly improbable.10 Hence, extrapolation constitutes a strong assumption, to be avoided. Rather, the dose–response relationship remains useful for pVL values similar to those in the original study.12
Although our direct estimates seem close to 1:100,000, the figure proposed by the Swiss statement,20 the decline in upper-bound risk with the amount of data is highly nonlinear,21 and validating the Swiss statement remains far of reach. It is important to note that our direct estimates were obtained by excluding seroconversions that occurred within the first 6 months of cART or were diagnosed after treatment failure; thus, they hold for individuals who have been on cART for more than 6 months and who are virally suppressed. Furthermore, our data-driven approach provides a conservative picture of the transmission risk, by estimating upper-bound or maximum risks. Because no transmission cases have been specifically documented from individuals receiving cART for more than 6 months and having an undetectable viral load, the risk could theoretically be zero and the upper-bound risk just measures the monitoring effort to estimate this risk. It is possible that cART has a more profound impact on reducing HIV transmission than reflected by our conservative approach.
Some limitations should be acknowledged. In the direct estimation approach, we could not distinguish between male-to-female and female-to-male transmission risks because of lack of data. However, these 2 risks were found to be similar after adjustment for pVL.15 We did not account for specific factors that can impact risk, such as HIV genotype, host genetics, circumcision status, type of intercourse, sexually transmitted infections,22 pregnancy, or postpartum status.23 In these respects, our results remain averaged over the populations participating to the clinical studies.
In conclusion, the transmission risk under effective cART could theoretically be zero. We estimated the upper-bound risk, which measures the monitoring effort, at 3.9:100,000 for heterosexuals and 4.4:100,000 for MSM. It is very important to use direct estimates of upper-bound risk, rather than those obtained by extrapolation, to avoid underestimating the impact of treatment as prevention at the population scale. The information that effective cART prevents HIV transmission may reduce stigmatization and discrimination against HIV-infected individuals, reduce unjustified fears of transmitting HIV, and help with family planning of serodiscordant couples, and should thus be broadly disseminated.
The authors thank Andrew Mujugira and Rose Apondi for providing additional data and Susan Eshleman for clarifications.
1. Cohen MS, Chen YQ, McCauley M, et al. Antiretroviral therapy for the prevention of HIV-1 transmission. N Engl J Med. 2016;375:830–839.
2. Rodger AJ, Cambiano V, Bruun T, et al. Sexual activity without condoms and risk of HIV transmission in serodifferent couples when the HIV-positive partner is using suppressive antiretroviral therapy. JAMA. 2016;316:171–181.
3. Rodger AJ, Cambiano V, Bruun T, et al. Risk of HIV transmission through condomless sex in MSM couples with suppressive ART: the PARTNER2 Study extended results in gay men. J Int AIDS Soc. 2018; 21(S6): e25148.
4. Bavinton BR, Pinto AN, Phanuphak N, et al. Viral suppression and HIV transmission in serodiscordant male couples: an international, prospective, observational, cohort study. Lancet HIV. 2018;5:e438–e447.
5. Melo MG, Santos BR, De Cassia Lira R, et al. Sexual transmission of HIV-1 among serodiscordant couples in Porto Alegre, southern Brazil. Sex Transm Dis. 2008;35:912.
6. Del Romero J, Castilla J, Hernando V, et al. Combined antiretroviral treatment and heterosexual transmission of HIV-1: cross sectional and prospective cohort study. BMJ. 2010;340:c2205.
7. Reynolds SJ, Makumbi F, Nakigozi G, et al. HIV-1 transmission among HIV-1 discordant couples before and after the introduction of antiretroviral therapy. AIDS. 2011;25:473–477.
8. Donnell D, Baeten JM, Kiarie J, et al. Heterosexual HIV-1 transmission after initiation of antiretroviral therapy: a prospective cohort analysis. Lancet. 2010;375:2092–2098.
9. Apondi R, Bunnell R, Ekwaru JP, et al. Sexual behavior and HIV transmission risk of Ugandan adults taking antiretroviral therapy: 3 year follow-up. AIDS. 2011;25:1317–1327.
10. Wilson DP. Data are lacking for quantifying HIV transmission risk in the presence of effective antiretroviral therapy. AIDS. 2009;23:1431–1433.
11. Wilson DP, Law MG, Grulich AE, et al. Relation between HIV viral load
and infectiousness: a model-based analysis. Lancet. 2008;372:314–320.
12. Quinn TC, Wawer MJ, Sewankambo N, et al. Viral load
and heterosexual transmission of human immunodeficiency virus type 1. Rakai Project Study Group. N Engl J Med. 2000;342:921–929.
13. Skarbinski J, Rosenberg E, Paz-Bailey G. Human immunodeficiency virus transmission at each step of the care continuum in the United States. JAMA Intern Med. 2015;175:588–596.
14. Goodreau SM, Carnegie NB, Vittinghoff E, et al. What drives the US and Peruvian HIV epidemics in men who have sex with men (MSM)? PLoS One. 2012;7:e50522.
15. Hughes JP, Baeten JM, Lingappa JR, et al. Determinants of per-coital-act HIV-1 infectivity among African HIV-1-serodiscordant couples. J Infect Dis. 2012;205:358–365.
16. Supervie V, Viard JP, Costagliola D, et al. Heterosexual risk of HIV transmission per sexual act under combined antiretroviral therapy: systematic review and bayesian modeling. Clin Infect Dis. 2014;59:115–122.
17. Cohen MS, Chen YQ, McCauley M, et al. Prevention of HIV-1 infection with early antiretroviral therapy. N Engl J Med. 2011;365:493–505.
18. Mujugira A, Celum C, Coombs R, et al. HIV transmission risk persists during the first 6 months of antiretroviral therapy. J Acquir Immune Defic Syndr. 2016;72:579–584.
19. Eshleman SH, Hudelson SE, Redd AD, et al. Treatment as prevention: characterization of partner infections in the HIV Prevention Trials Network 052 Trial. J Acquir Immune Defic Syndr. 2017;74:112–116.
20. Vernazza P, Hirschel B, Bernasconi E, et al. Les personnes séropositives ne souffrant d'aucune autre MST et suivant un traitement antirétroviral efficace ne transmettent pas le VIH par voie sexuelle. Bull des Médecins Suisses. 2008;89:165–169.
21. Supervie V, Viard JP, Costagliola D, et al. Risk of HIV transmission under combined antiretroviral therapy: toward risk zero? J Acquir Immune Defic Syndr. 2015;68:e41–e42.
22. Jarzebowski W, Caumes E, Dupin N, et al. Effect of early syphilis infection on plasma viral load
and CD4 cell count in human immunodeficiency virus-infected men: results from the FHDH-ANRS CO4 cohort. Arch Intern Med. 2012;172:1237–1243.
23. Thomson KA, Hughes J, Baeten JM, et al. Increased risk of female HIV-1 acquisition throughout pregnancy and postpartum: a prospective per-coital act analysis among women with HIV-1 infected partners. J Infect Dis. 2018;218:16–25.