Perinatal HIV-1 Transmission: Interaction Between Zidovudine Prophylaxis and Mode of Delivery in the French Perinatal Cohort

Context.— It is unclear whether elective cesarean delivery may have a protective effect against the transmission of human immunodeficiency virus 1 (HIV-1).

Objective.— To investigate whether mode of delivery has an impact on perinatal HIV-1 transmission in the presence of zidovudine prophylaxis.

Design.— A prospective cohort study.

Setting.— The 85 perinatal centers in the French Perinatal Cohort, from 1985 to 1996.

Patients.— A total of 2834 singleton children born to mothers with HIV-1 infection.

Main Outcome Measure.— Human immunodeficiency virus 1 infection of the infant.

Results.— No zidovudine was used in 1917 pregnancies and zidovudine prophylaxis was used in 902 pregnancies. Cesarean deliveries were performed in 10.9% on an emergent basis and in 8.3% electively, prior to labor or membrane rupture. In 1917 mothers who did not receive zidovudine, of 1877 with information on mode of delivery, 17.2% transmitted HIV-1 to their child. Risk factors statistically significantly associated with transmission were maternal p24 antigenemia, cervicovaginal infections during pregnancy, amniotic fluid color, and rupture of membranes 4 hours or more before delivery. Mode of delivery was not related to transmission. In 902 mothers receiving zidovudine, transmission was 6.4% in 872 with information on mode of delivery, and elective cesarean delivery (n=133) was associated with a lower transmission rate than emergent cesarean or vaginal delivery (0.8%, 11.4%, and 6.6%, respectively; P=.002). In a multivariate analysis of all mother-child pairs, including obstetrical risk factors, maternal p24 antigenemia, and zidovudine prophylaxis, interaction between mode of delivery and zidovudine prophylaxis was significant (P=.007). In the multivariate analysis of pregnancies with zidovudine prophylaxis, factors related to transmission rate were maternal p24 antigenemia, amniotic fluid color, and mode of delivery. Adjusted odds ratios (95% confidence intervals) were 1.6 (0.7-3.6) for emergent cesarean delivery and 0.2 (0.0-0.9) for elective cesarean delivery (P=.04) in comparison with vaginal delivery.

Conclusions.— We observed an interaction between zidovudine prophylaxis and elective cesarean delivery in decreasing transmission of HIV-1 from mother to child. This observation may have clinical implications for prevention.

THE IMPORTANCE of labor and delivery in mother-to-child transmission of human immunodeficiency virus 1 (HIV-1) has been suspected for years. A large body of data indicates that transmission occurs predominantly late in pregnancy and intrapartum.^1-4 However, the role of obstetrical management remains unclear. An increased risk of transmission was observed in several studies in association with rupture of membranes before the onset of labor⁵ or more than 4 hours before delivery,⁶ and in a few studies with events such as hemorrhage and bacterial infection.^5,7,8 Conversely, conditions of delivery itself, such as length of labor, instrumental delivery, or episiotomy, had little relation to transmission.^5,6,9 The impact of mode of delivery is still controversial.¹⁰ Transmission risk was decreased following cesarean delivery in some studies,^11-13 but not in others.^5,6,14-17

Elective cesarean delivery is a potential strategy for reducing perinatal HIV transmission. Elective cesarean delivery must be distinguished from emergent cesarean delivery, which is performed because of obstetrical complications. Surprisingly, in the European Collaborative Study¹¹ transmission rate was actually lower following emergent cesarean delivery compared with elective cesarean delivery. In the French Perinatal Cohort,⁵ neither elective nor emergent cesarean delivery was observed to be associated with a lower transmission rate than with vaginal delivery in the cohort prior to 1993.

Since introduction of zidovudine prophylaxis,¹⁸ mother-to-child transmission rates have decreased by about two thirds in various clinical settings.^19,20 We previously reported no difference in transmission rates in the presence of zidovudine prophylaxis between children delivered vaginally or by cesarean²⁰; however, we lacked statistical power to distinguish rates with emergent and elective cesarean delivery compared with vaginal delivery. The goal of this study was to investigate whether use of zidovudine changed the relationship between obstetrical factors and HIV-1 transmission.

The French Perinatal Cohort is a study of HIV transmission resulting from fusion of the French Pediatric Cohort, established in September 1985, and the SÉROGEST Cohort, established in 1989. The study was approved by the institutional review board of the Hôpital Necker, Paris, France, and by the French computer database watchdog commission (Commission Nationale de l'Informatique et Libertés), Paris. Follow-up from birth of all children born to consenting mothers with HIV was done in 85 obstetrical services. Protocols for enrollment, data collection, and anonymous data transmission to the coordinating center were reported previously.^5,21 Maternal data included age, geographic origin, suspected route of HIV transmission, clinical symptoms, lymphocyte subsets, viral load markers (p24 antigenemia until mid-1996 and routine plasma HIV RNA levels thereafter²²), and detail regarding antiretroviral prescriptions. Cervicovaginal infections were defined as any lower genital tract infection diagnosed by clinical examination and vaginal or endocervical sampling with wet mount and culture as appropriate. Genital herpes, gonorrhea, chlamydial endocervicitis, condyloma, gardnerella vaginosis, and trichomoniasis were classified as sexually transmitted diseases. Detailed obstetrical information was recorded as previously described.⁵ In particular, cesarean delivery was defined as elective when performed before the onset of labor and with intact membranes, and defined as emergent in all other cases. Neonatal bathing (mostly with benzalkonium chloride) and breast-feeding were also recorded.

The HIV infection status was determined for children born at least 3 months prior to April 1, 1997. A child was considered infected at 18 months if HIV-1 antibodies persisted or in the case of death from HIV-related disease, and considered uninfected if the findings of 2 HIV antibody tests were negative. Antibody testing was carried out with both of 2 commercial enzyme-linked immunosorbent assays, and positive results were confirmed by immunoblot. Immunoblot findings were considered negative when no antibody was detected directed against the HIV envelope glycoproteins. For children aged 3 to 18 months, infection status was determined using DNA-polymerase chain reaction and/or HIV culture.^20,23 The child was considered infected when 2 different sample test findings were concordantly positive, and was considered uninfected when 2 different sample test findings were concordantly negative (at least 1 of which was administered at or beyond age 3 months).

Because of the 3-fold decrease in transmission rate for mothers receiving zidovudine,²⁰ obstetrical factors were considered separately for these mothers and mothers not treated. Transmission rates were compared using the χ² test or the Fisher exact test when necessary. A logistic regression was performed on the overall cohort, which included in the model zidovudine prophylaxis and maternal p24 antigenemia,²⁴ as well as all obstetrical variables found to be statistically significant in either of the univariate analyses, except duration of membrane rupture. Because by definition there was no membrane rupture in elective cesarean delivery, these 2 variables could not be included in the same model. Interactions between zidovudine use and each obstetrical factor were tested. Only the interaction between zidovudine use and the mode of delivery was statistically significant. Logistic regressions were then performed separately for the untreated and treated groups. Analysis was performed with SAS, Version 6.12 (SAS Institute Inc, Cary, NC).

From September 1, 1985, through December 31, 1996, the participating sites reported 3474 deliveries of HIV-seropositive women. The study group comprised 2834 mother-child pairs after excluding 174 mothers infected with HIV-2, 44 breast-fed children, 114 twins, 43 children who died before HIV status could be determined, 61 children not enrolled because of lack of parental consent, 84 children lost to follow-up, and 120 children with indeterminate status at the time of analysis. The nonevaluable children did not differ statistically significantly from the study group for any of the variables investigated. All deliveries occurring through 1993 (n=1632) were included in a previous publication⁵ and 354 pregnancies with prophylaxis in 1994 and 1995 were included in another publication.²⁰ Mother's mean age at delivery was 28 years, 31% were past or present intravenous drug users, and 40% were born in sub-Saharan Africa or the Caribbean. Undissociated p24 antigenemia at delivery, available in 83.6% of treated mothers and 76.7% of untreated mothers, was found in 7%. Median maternal CD4 cell count at delivery was 0.47×10⁹/L (474/µL), and 13% of mothers had counts under 0.20×10⁹/L. Infected children were followed up for a mean duration of 51 months and a median of 48 months.

Zidovudine use increased sharply as of March 1994, when an official policy was issued following the French-American AIDS Clinical Trials Group 076 results.^20,25 The proportion of women treated was 15% in 1993, reaching 90% by mid-1994, and was stable thereafter. Data on zidovudine use were missing for 15 mother-child pairs. Transmission occurred in 58 (6.4%) of 902 treated mothers and 329 (17.2%) of 1917 untreated mothers, a 3-fold decrease as previously reported.²⁰

In the 1917 children not exposed to zidovudine, transmission rate was higher (Table 1) in the case of cervicovaginal infections during pregnancy, membrane rupture more than 4 hours before delivery, preterm delivery, and bloody or meconium-stained vs clear amniotic fluid, as reported.⁵ In 35 cases with bloody amniotic fluid, 6 were caused by placental abruption or placenta previa. We also observed a higher transmission rate in the case of maternal temperature higher than 38°C during labor (28.3% vs 16.6%, P =.003). As previously reported, none of the other obstetrical factors were related to transmission risk. Transmission rates did not differ by mode of delivery. Human immunodeficiency virus 1 was transmitted from mother to child in 17 (17.5%) of 97 elective cesarean deliveries (95% confidence interval [CI], 9.9%-25.1%), 31 (15.6%) of 186 emergent cesarean deliveries (95% CI, 10.4%-20.8%), and 279 (17.5%) of 1594 vaginal deliveries (95% CI, 15.6%-19.4%).

Transmission rates were significantly higher in the case of p24 antigenemia (40.0% vs 15.9% when negative, P =.001).

A total of 902 children had mothers who received zidovudine prophylaxis. Contrary to the findings in the absence of zidovudine, mode of delivery was related to the rate of transmission (Table 1), which was 6.6% (95% CI, 4.7%-8.8%) with vaginal delivery (41 of 625), 11.4% (95% CI, 6.2%-18.7%) with emergent cesarean delivery (13 of 114), and 0.8% (95% CI, 0.2%-4.0%) with elective cesarean delivery (1 of 133). For a global test for the 3 categories, P was .002.

The only other obstetrical factor that was statistically significantly associated with transmission risk in treated women was the color of the amniotic fluid.

Similar to findings in subjects without zidovudine use, the proportion of children with HIV infection was higher in the case of p24 antigenemia (16.7% vs 5.3% for women whose test findings were negative for the antigen, P=.003).

Of deliveries with detailed data available, cesarean deliveries were performed in 283 (15.2%) of 1877 cases in the untreated group, of which 10% were emergent and 5.2% were elective, and in 247 (28.3%) of 872 cases in the zidovudine-treated group, of which 13% were emergent and 15.3% were elective. Repeat section, fetal indications (principally anomalous fetal heart rate tracings or intrauterine growth retardation), maternal indications (such as diabetes mellitus, thrombocytopenia, or cervical intraepithelial neoplasia), fetopelvic disproportion, and breech presentation were the principal obstetrical indications for elective cesarean delivery in both groups. The proportion of elective cesarean deliveries performed without classic obstetrical indications, as prevention for HIV transmission, was higher in the zidovudine-treated group than in the untreated group (30% and 12%, respectively).

Median duration of gestation did not differ between treated and untreated groups. However, gestational age at delivery in both groups was 1 week less for elective cesarean delivery (37.8 weeks) than for emergent cesarean delivery (39 weeks) or vaginal delivery (39 weeks).

Durations of labor and membrane rupture were longer for emergent cesarean delivery than for vaginal delivery. Median length of active labor was 390 minutes for emergent cesarean delivery and 300 minutes for vaginal delivery in the untreated group and 410 minutes and 315 minutes, respectively, in the treated group. Median duration of membrane rupture was 405 minutes for emergent cesarean delivery and 190 minutes for vaginal delivery in the untreated group, and 330 minutes and 180 minutes, respectively, in women receiving zidovudine.

The proportion of deliveries with intravenous zidovudine infusion during delivery did not differ significantly by mode of delivery (85.4% for vaginal delivery, 81.8% for emergent cesarean delivery, and 90.8% for elective cesarean delivery, P =.12). The proportion of children receiving zidovudine postnatally also did not differ significantly by mode of delivery (98.1% for vaginal delivery, 98.0% for emergent cesarean delivery, and 98.4% for elective cesarean delivery, P=1.00). Mean duration of zidovudine prophylaxis during pregnancy was also similar for all modes of delivery (4.2 months, 4.4 months, and 4.3 months, respectively). In 30 (3.3%) of 902 pregnancies with prophylaxis, antiretroviral drugs were prescribed in addition to zidovudine. The proportion of those receiving combination therapy did not differ statistically significantly by mode of delivery.

We first tested the interactions between each obstetrical variable and zidovudine prophylaxis in a logistic regression model including the factors statistically significantly linked to transmission risk in either of the univariate analyses, as well as zidovudine prophylaxis. Only the interaction between mode of delivery and zidovudine prophylaxis was significant (P=.007).

Multivariate analyses were performed separately for mothers receiving zidovudine and for those not treated (Table 2). To adjust for unknown confounders that may be involved in decrease in transmission rate over time, we added a variable for the child's year of birth.

Because untreated women showed decreased transmission risk, from 20% from 1986 through 1991 to 15% from 1992 through 1996, we also tested in the model interaction between period of birth and mode of delivery. This interaction term was not significant (P=.97). In the final multivariate analysis for the untreated group, 4 variables remained significantly related to transmission rate: p24 antigenemia (P<.001), cervicovaginal infections (P=.008), amniotic fluid appearance (P<.001), and birth year (P=.008).

In the treated group, only 3 variables remained significantly associated with transmission: p24 antigenemia (P=.02), amniotic fluid color (P=.04), and mode of delivery (P=.04). To assess the impact of membrane rupture on difference between elective cesarean and other deliveries, we constructed a new 3-class variable for the logistic regression: elective cesarean delivery, other modes of delivery with membrane rupture 4 hours or less before delivery, and other modes of delivery with membrane rupture more than 4 hours before delivery. Transmission rate remained significantly lower following elective cesarean delivery vs other deliveries with membrane rupture 4 hours or less before delivery (adjusted odds ratio, 0.16; 95% CI, 0.01-0.84).

In the presence of zidovudine prophylaxis, we observed only 1 child infected with HIV-1 in 133 delivered by elective cesarean. In a multivariate analysis, transmission risk remained 5-fold lower following elective cesarean delivery than in vaginal or emergent cesarean delivery.

These findings contrast with those observed in our cohort in the absence of zidovudine prophylaxis, in which case the transmission rate was the same irrespective of the mode of delivery, as reported.⁵ Similarly, several studies performed prior to the introduction of zidovudine prophylaxis showed no relationship between mode of delivery and transmission.^5,6,14-17 However, data from other cohort studies had suggested a protective effect for cesarean delivery.^11-13 Observational studies may show divergent results because of differences in populations or because they lack statistical power to differentiate between elective and emergent cesarean delivery. Whereas both types of cesarean delivery may have an impact on intrapartum transmission,²⁶ planned cesarean delivery may avoid physiologic and pathologic processes related to labor, thereby decreasing late in utero transmission as well. Whether cesarean delivery has a protective effect independent of zidovudine prophylaxis can be further investigated by a large, international, individual patient data meta-analysis of observational studies. However, a definitive answer to the question will require a randomized clinical trial,²⁷ which is the only method to ensure that women who undergo an elective cesarean delivery do not differ from those with other types of delivery for any known or unknown confounding factor.

The mother-to-child HIV transmission rate started to decrease in France before the introduction of zidovudine prophylaxis.²⁰ A decrease in transmission independent of zidovudine prophylaxis has also been reported in the United States.¹⁴ Changes in clinical management over a decade may have played a role, possibly regarding risk factors such as cervicovaginal infections during pregnancy,²⁸ chorioamnionitis,^7,8 fever in labor, preterm delivery,⁹ and above all, prolonged membrane rupture,^5,6 all of which may involve subclinical bacterial infection.²⁹ Improved management of these complications cannot, however, account for the protective efficacy of elective cesarean delivery, which was by definition performed with intact membranes, whereas emergent cesarean delivery had longer duration of membrane rupture than vaginal delivery. When a variable combining mode of delivery and membrane rupture duration was introduced into a multivariate analysis, elective cesarean delivery remained significantly associated with lower transmission rates. We must raise the possibility that cesarean deliveries became more protective over time because of changes in indications for cesarean delivery as well as changes in the operative procedure itself. The proportion of elective cesarean deliveries tripled over a decade, largely because obstetricians increasingly performed "prophylactic" cesarean deliveries in low-risk cases with the hope that they might be protective.³⁰ Although none of the centers reported implementing "bloodless cesarean" techniques, obstetricians may also have taken increasing care to prevent contact between the fetus and maternal blood. However, in the multivariate analysis, we found no significant interaction between the mode of delivery and the time period in reducing transmission risk.

The only significant interaction was between elective cesarean delivery and zidovudine use. Our findings were based on a large cohort, but require confirmation from other studies, since only 133 of the women received zidovudine and were delivered by elective cesarean. In a smaller study, zidovudine and elective cesarean delivery appeared to have cumulative protective effects.³¹ Several hypotheses may be advanced to suggest why zidovudine prophylaxis would be more effective regarding elective cesarean delivery. First, we checked that women who had planned cesarean deliveries did not receive intravenous zidovudine infusions more frequently than those who were delivered vaginally or by emergent cesarean. Nonetheless, according to the published guidelines,²⁵ zidovudine infusion is started 4 hours before a planned cesarean delivery, whereas in other deliveries it is started later, when active labor is diagnosed.

A second hypothesis is that zidovudine is less preventive in the case of vaginal delivery than cesarean delivery because it has less effect on cervicovaginal excretion of HIV than on systemic viral load. Contrary to this hypothesis, zidovudine has been reported to decrease cervicovaginal HIV excretion.³² Furthermore, the efficacy of zidovudine prophylaxis is due only partially to a reduction in plasma viral load.³³ Finally, the transmission rate was actually higher with emergent cesarean delivery compared with vaginal delivery, indicating that transmission occurred up to and during labor, rather than during exposure of the fetus during the passage through the birth canal.

The impact of zidovudine prophylaxis may differ according to timing and mechanism of exposure. Although relative contributions of each possible mechanism to overall transmission rate are not known, they are likely to differ by mode of delivery. For instance, incidence of maternal-fetal microtransfusions has been reported to be 5-fold lower during elective cesarean delivery vs emergent cesarean or vaginal delivery.³⁴ Conversely, the procedure itself may expose the fetus to maternal blood, leading to transmission via the oropharynx or the digestive tract, as has been suggested for postnatal transmission via breast-feeding.³⁵ Prevalence of the presence of HIV RNA in neonatal gastric aspirates has been found to be higher in the case of exposure to maternal blood.³⁶ We speculate that zidovudine may be more effective as preexposure and postexposure prophylaxis regarding oral route exposure than regarding parenteral exposure, such as transplacental microtransfusions. If zidovudine is effective regarding oral exposure, it may also reduce transmission via the ascending route. It may also be more effective in the case of placental transfer of small amounts of virus or infected cells, as may occur with inflammatory placental changes in late gestation, than in the case of maternal-fetal microtransfusions.

The role of elective cesarean delivery in the prevention of mother-to-child HIV-1 transmission remains to be defined for clinical practice. Its preventive impact may differ in women who receive various combination antiretroviral therapies.³⁷ Most importantly, the risk-benefit ratio may differ according to the baseline transmission rate. Assuming that a planned cesarean delivery with zidovudine monotherapy can decrease the risk of transmission from 6% to 1%, 20 operations would be necessary to prevent 1 case of transmission. Maternal mortality with cesarean delivery is increased 5-fold (from 7 to 40 per 100000 deliveries in Sweden³⁸) compared with vaginal delivery, and postoperative complications have been reported in 31% of HIV-infected women, 3 times more than in HIV-negative subjects.³⁹ In most of the developing world, where preventive measures are desperately needed, the mortality associated with cesarean delivery is much higher.⁴⁰

Investigation of the mechanisms whereby obstetrical factors and antiretroviral therapy are protective, as well as their interaction, may offer insights into the pathophysiology of perinatal HIV transmission, which should be useful for developing new strategies for prevention worldwide.