Study design

We performed a model-based decision analysis to compare the following 5 reproductive strategies: (1) unprotected sexual intercourse (treatment as prevention strategy); (2) unprotected sexual intercourse limited to fertile days determined by urine ovulation testing (targeting fertile days); (3) treatment as prevention and preexposure prophylaxis, using a continuous tenofovir/emtricitabine combination; (4) treatment as prevention and preexposure prophylaxis limited to fertile days; and (5) MAP: sperm washing and 6 IUI attempts at most per couple. We did not consider in vitro fertilization among our strategies because the couples were supposed to be fertile. Outcomes included cumulative risk of HIV heterosexual transmission to HIV– partner and mother-to-child transmission, risk of birth defects, life expectancy of both woman and child, costs, and the incremental cost-effectiveness ratios, measured in €/life year of both woman and baby saved. We applied the guidance of the World Health Organization Commission on Macroeconomics and Health, which suggests that an intervention be labeled “cost-effective” in a given country if its incremental cost-effectiveness ratio is <3 times the nation’s gross domestic product per capita. Since the French gross domestic product per capita is €30,000, programs delivering life year saved for <€90,000 were deemed “cost-effective.” Because our goal was to inform decision making at the level of the French National Health Service, we conducted our analyses from a modified societal perspective (excluding indirect costs), and discounted costs, and life expectancy, at 4%/annum. We report lifetime medical costs in 2013 euros; €1 was equivalent to US$1.32 in January 2013.

Model and population

We constructed a Markov model to simulate a hypothetical cohort of 10,000 fertile HIV-serodiscordant couples desiring a child over 1 year with the male partners being treated for HIV with undetectable blood viral load and the female partners (HIV–) being 33 years old. The model was run over 12 cycles (1 year), and simulated the HIV– women’s trajectory when the couple initiated the reproductive strategy until they became pregnant and/or they tested HIV+, or at 12 months. Figure 1 describes the decision process for each cycle. For each cycle (month), women could be pregnant or not. If pregnant, women were routinely tested for HIV, as were their babies if the test was positive. For each scenario, babies could be born with birth defects according to different probabilities (ie, if the mother was on highly active antiretroviral therapy during pregnancy or not) and if she received preexposure prophylaxis in treatment as prevention using a continuous tenofovir/emtricitabine combination and limited to fertile days. For nonpregnant women, HIV testing was performed every 3 cycles for natural insemination strategies and at each cycle for the MAP strategy. Therefore, 5 terminal stages were defined: (1) woman uninfected and baby uninfected; (2) woman infected and baby uninfected; (3) woman infected and baby infected; (4) woman uninfected and no baby; and (5) woman infected and no baby. The model was constructed and analyses were performed using software (TreeAge Pro 2014; TreeAge Software Inc, Williamstown, MA).

Decision tree

For each cycle, insemination strategy was performed and women became pregnant according to respective probabilities. Women were routinely tested for HIV if pregnant, and tested for each cycle of medically assisted procreation or every 3 months in other strategies. Infected tested nonpregnant women stopped to inseminate (exited model), and noninfected (or nontested) women started new cycle. Pregnant women also stopped to inseminate and babies became infected (if women were HIV+) and/or had birth defects according to respective probabilities.

Mabileau. Procreation in HIV-serodiscordant couples. Am J Obstet Gynecol 2015 .

Input parameters

Input parameters of the model are presented in Table 1 . In this model-based analysis, we compared the reproductive strategies considering an 85% cumulative probability of live births, which was assumed to be the same for the different strategies. This assumption is based on the pregnancy success rate in fertile couples after 12 cycles when sexual intercourse is unprotected, the women being between 32-35 years old. The pregnancy rate per cycle was 14.6% for natural insemination (85% after 12 cycles), instead of 27.1% for MAP, as we considered a maximum of 6 IUI attempts per woman (85% after 6 cycles). For the unprotected vaginal intercourse strategy, the risk of HIV transmission to the woman was an estimated 0.0083%/mo, with a 1% subsequent HIV mother-to-child transmission rate in the case of HIV transmission to the female partner, and a 4.4% birth defect risk related to combination antiretroviral therapy use in the female during the pregnancy, the risk being 2.3% in the French general population. We hypothesized that the infected mother was treated with combination antiretroviral therapy during the first trimester of pregnancy. Ovulation generally lasts from 2-5 d/cycle, but we assumed a large range of values for the relative risk reduction in targeting fertile days to take the uncertainty into account, decreasing the heterosexual transmission risk by 0-90% (base case = 80%) when compared with the unprotected vaginal intercourse strategy. We estimated that preexposure prophylaxis decreased the risk of heterosexual HIV transmission by 67% (base case) to 90%. In the case of preexposure prophylaxis during targeting fertile days, the risk of sexual HIV transmission was assumed to decrease by 93.4% (the relative risk of transmission considering the combination of both strategies assuming to be (1-80%)*(1-67%) = 16.6% in basecase). Preexposure prophylaxis was assumed to be discontinued as soon as mothers knew they are pregnant within the first 2 weeks of gestation. Consequently, no additional birth defect risk associated with preexposure prophylaxis was considered since during this early period the developing embryo is not susceptible to teratogenesis. IUI was performed after sperm washing and HIV testing of the washed sample, so the risk of transmission related to this strategy was an estimated 0.0001%. Input variables considered were derived from the international literature and/or were discussed by an expert group composed of infectious diseases specialists, gynecologists, health economists, and MAP specialists. Life expectancies for HIV– women and babies were derived from 2010 estimates of the French National Institute of Statistics and Economic Studies, assuming a 36% reduction in life expectancy for HIV+ patients as compared with the general population ; this was varied over a wide range in the sensitivity analysis.

Costs

Before each strategy, we considered that couples were tested for sexually transmitted infections (eg, HIV, hepatitis B virus, hepatitis C virus, and syphilis) and that each woman’s fertility was confirmed with usual tests ( Table 1 ). However, expenses related to these costs being strictly the same for each strategy, we did not include them in the analysis.

Costs for MAP were derived from the French nomenclature for clinical pathology procedures (Nomenclature des Actes Biologiques Médicale), which fixes the cost of such procedures in France. For MAP we considered 2 preconception visits (gynecologist and medical technologist–€23 each) for both women and men, and 1 additional visit for men for semen collection with sperm washing (a sperm bundle being used for a mean of 3 attempts, which was to be repeated after 3 unsuccessful cycles). The total mean cost of the preconception process for MAP was €405.30, including transportation to the referral center for both women (once) and men (twice) (€100/100-km round-trip).

For each MAP cycle, we considered additional procedures related to the IUI, including laboratory procedures for semen preparation and medical procedures for ovulation induction, insemination, and ultrasound. Treatments for ovarian stimulation (a low dose of Puregon given from day 3 to the day of ovulation induction) and to trigger ovulation (human chorionic gonadotropin 5000) were also included in costs. Costs of Puregon were included because for a part of fertile women, ovarian stimulation has been reported to provide best results than without it in HIV-infected women; it mostly brings monoovulation stimulation when it is initiated over the time of dominant follicle selection (over day 6 or 7) and facilitates its monitoring; and had not been suspected to increase the risk of breast cancer and birth defects, contrary to clomiphene citrate. Total mean cost for each MAP cycle was an estimated €561.57 (procedures and treatments). We also considered transportation of the woman to a referral center for each cycle.

Costs for preexposure prophylaxis included drug costs (€536.40/mo for continuous tenofovir/emtricitabine com bination) and visit costs (infectious diseases specialist and gynecologist at the beginning, then one to an infectious diseases specialist every 3 months–€23 each). Targeting fertile days without preexposure prophylaxis included visit costs at the beginning (infectious diseases specialist and gynecologist), then urine ovulation test costs (€17/mo). Costs for urine pregnancy testing (€8.10/U/mo) and for HIV testing (€18.90/U/mo for MAP, and /trimester for others) were also included for each strategy. Total lifetime costs related to HIV care for HIV+ women and their babies were added to costs associated with each strategy (€641,000 for a life expectancy of 33 years and €100,000,000 for a life expectancy of 52 years, respectively). We based our calculations on costs related to HIV published by Sloan et al who estimated a lifetime cost of €534,800 for a life expectancy of 27.5 years for patients presenting to care early.

Analyses

Sensitivity analyses were performed for the parameters with the greatest impact, considering a wide range of values ( Table 1 ), to validate the robustness of our outcomes. We first varied parameters in 1-way sensitivity analyses and then performed 2-way sensitivity analyses with major and uncertain parameters. We considered the baseline probability of sexual transmission from male to female to be between almost 0-0.058%/mo, as estimated by Cohen et al. Considering the uncertainty of the risk reduction associated with targeting fertile days, we varied it from 0.1-1 since targeting fertile days does not necessarily mean that couples have no unprotected sexual intercourses outside the ovulation period (generally lasting 2-7 days). Moreover, couples could decide to have more sexual intercourse during this period to multiply their chances of pregnancy, resulting in a decrease in relative risk reduction of transmission. Although preexposure prophylaxis was assumed by Baeten et al to reduce the risk of transmission by 67%, we increased the relative risk reduction of preexposure prophylaxis to 90% considering recent estimations by Dai et al. Costs related to preexposure prophylaxis were also varied to take into account that prices could be reduced in the future (eg, due to new generic drugs). Finally, we decreased the reduction of life expectancy for people infected by HIV to 0% considering that the life expectancy of patients treated early and adherent to care and treatment is similar to that of the general population.