Recognizing that there is a high risk of HIV transmission during parturition, the potential protective effect of cesarean delivery was examined in two major studies. The International Perinatal HIV Group conducted a meta-analysis of 15 prospective cohort studies including 8533 mother-infant pairs delivering between 1982 and 1996, most of whom did not receive any antiretroviral prophylaxis. Among 7840 mother-infant pairs with complete data, a multivariate analysis adjusting for receipt of antiretroviral therapy, advanced maternal disease, and low birth weight (all known to correlate with risk of infection), elective cesarean delivery was associated with a lower risk of vertical transmission (odds ratio, 0.43, 95% confidence interval 0.33-0.56).

Vaginal delivery and non-elective cesarean delivery were associated with similar risks of transmission of HIV. However, when the time between rupture of membranes and delivery in non-elective cesarean deliveries was examined, the risk of transmission was lower than with vaginal delivery if cesarean delivery was accomplished within 4 hours of rupture of membranes (odds ratio 0.53, 95% confidence ratio 0.4-0.7).

The authors found that cesarean delivery lowered the risk of transmission among mother-infant pairs receiving antiretroviral prophylaxis (mostly zidovudine monotherapy) during pregnancy, labor, and the neonatal period (2.0% transmission, 95% confidence interval 0.1-4.0% with antiretroviral therapy and cesarean delivery, compared to 7.3%, 95% confidence interval 5.9-8.8, with antiretroviral therapy alone).

The European Mode of Delivery Collaboration conducted a randomized controlled trial of cesarean versus vaginal delivery for prevention of vertical HIV infection. Among the 370 mother-infant pairs for whom outcome data were available, 64% received antiretroviral prophylaxis. Among women assigned to cesarean delivery, 11.7% gave birth vaginally; among women assigned to vaginal delivery, 26.8% had operative deliveries, 54% of which were emergent. Infants of 1.8% of women assigned to the cesarean delivery group became infected with HIV, and 10.5% of infants of women assigned vaginal delivery became infected (p < 0.001). By actual mode of delivery, 3.5% of cesarean births and 10.2% of vaginal births produced infected infants. Among 119 women taking zidovudine and undergoing cesarean delivery, only 1 infant (0.8%) was infected.

There is a clear conclusion from these studies: elective cesarean delivery decreases the risk of HIV transmission in women receiving either no antiretroviral prophylaxis or zidovudine monotherapy alone. However, the response among clinicians differed depending generally on which side of the Atlantic they were located. In the U.S., caution about generalization of the results to women receiving HAART was urged and cesarean delivery rates did not increase as much as they did in Europe, where between 1999 and 2003 elective cesarean delivery was performed for about 65% of HIV-infected women.

The European experience in the HAART era (data from 1997 through mid-2004) was reviewed in the European Collaborative Study report on 1983 mother-child pairs. During this period, the use of HAART increased from about 5% of pregnancies to about 85%. This study confirmed key findings of the WITS — that a lower viral load and potent antiretroviral therapy dramatically reduces risk of transmission.

In addition, the authors stated that among women receiving HAART whose most recent viral load was undetectable, cesarean delivery was associated with an additional reduction in the risk of transmission. Yet the confidence interval for this association in the data was 0.08-5.37 (p = 0.7). Only 1 woman receiving HAART and delivering vaginally transmitted the virus, and her viral load at term and the duration of HAART therapy were unclear. We do know that of the 11 infected children of mothers receiving HAART (10 of whom had cesarean deliveries, half of which were emergent), the median duration of HAART was only 38 days and most had advanced HIV disease. While the authors conclude that “offering an elective Caesarean delivery to all HIV-infected women, even in areas where HAART is available, is appropriate clinical management, especially for persons with detectable viral loads”, their data, along with the data from WITS and other case series, show that early, effective HAART is sufficient to prevent HIV transmission and additional benefit from cesarean delivery could not be demonstrated.

The Antiretroviral Pregnancy Registry (APR), managed by Charles River Laboratories Clinical Services and supported financially by antiretroviral manufacturers, reports data on teratogenic effects from two large European registries and the published literature. The APR’s most recent semi-annual review of the outcome of 6893 instances of antiretroviral use in pregnancy (prospectively reported directly to the APR) describes birth defects in 2.9% of the 2117 reported first trimester antiretroviral exposures. This is comparable to the rate reported by the CDC’s population-based surveillance system – 2.2% for defects diagnosed by the first day of life and 3.1% for defects discovered at any time.

There were enough reports in the APR to provide ample power for the detection of a two-fold increase in birth defects associated with first trimester exposure for 9 antiretrovirals. For 8 of these there was no increase over the expected prevalence. For didanosine, 6.0% of first trimester exposures were associated with birth defects (95% confidence interval 3.3-9.8%), although there was no specific pattern to the reported defects.

Efavirenz exposure was of special theoretical concern because in a preclinical study exposing 20 pregnant cynomolgous monkeys to levels of drug comparable to human use, 3 offspring had serious defects: anencephaly, microphthalmia, and cleft palate (described in the efavirenz product labeling). The APR recorded the outcome in 244 cases of first trimester efavirenz exposure and found 6 cases with birth defects, which is within the expected natural rate. Importantly, the defects seen were similar to common defects in the general population and did not resemble the defects seen in exposed monkeys. Although these data are reassuring, retrospective reports of 4 cases of neural tube defects after first trimester efavirenz exposure prompted the U.S. Food and Drug Administration to reclassify efavirenz as a pregnancy category D drug.

The data on birth defects after antiretroviral exposure are generally reassuring. There is enough data in humans to say that efavirenz may cause birth defects in humans – however, at a much lower rate or with more subtle manifestations than the disturbing outcomes seen in monkeys. Pending further data, we must also remain concerned about the increased prevalence of all birth defects associated with early didanosine exposure.

In 1999 Blanche and colleagues reported 8 cases of mitochondrial dysfunction with abnormally low absolute or relative activities of respiratory-chain complexes months or years after the end of antiretroviral treatment. The cases were identified from 1754 infants exposed to antiretrovirals in a prospective French cohort study. Two of the infants were asymptomatic and five of the infants had seizures, 2 of whom had progressive brain disease followed by death at around one year of age. Various other abnormalities such as lactic acidosis, increased serum levels of hepatic, pancreatic, or muscle enzymes, cardiomyopathy, brain MRI images, or electroretinography were seen inconsistently. In the three infants in whom it was evaluated, mitochondrial DNA content was normal. Potential causes, including known genetic diseases affecting mitochondria, were excluded.

In 2003, the same group described the retrospective search for mitochondrial disease in a prospective cohort study of 2644 antiretroviral-exposed and 1748 non-antiretroviral-exposed infants, in addition to cases included in a national registry of HIV-exposed infants. The study relied primarily on enzymatic and ultrastructural studies of muscle biopsies from children with suspicious symptoms. After systematic review of the records, 14 cases of possible and 7 cases of established mitochondriopathy were identified among the antiretroviral-exposed infants(these included the 8 original cases). In contrast, no instances of possible or established mitochondrial disease were found among the infants of HIV-infected mothers who were not exposed to antiretrovirals.

The initial French report was alarming, and in response the Perinatal Safety Review Working Group was formed in the U.S. to examine deaths among subjects into 5 cohort studies that included 16,313 HIV-exposed, uninfected infants, of which 75% definitely or probably were exposed in utero to antiretrovirals. A total of 252 deaths were categorized in 5 groups: 1) unrelated to mitochondrial disease; 2) consistent with mitochondrial disease but unlikely; 3) mitochondrial disease “might reasonably be included in the differential diagnosis”; 4) mitochondrial disease suggestive or proven; or 5) SIDS.

There were 30 deaths among uninfected children, 14 of whom were definitely or probably antiretroviral-exposed, and 16 of whom were not or probably not antiretroviral-exposed. SIDS was the cause of death for 4 of the uninfected children. All of the other 26 HIV-uninfected, antiretroviral-exposed infants died of causes deemed to be unrelated to mitochondrial disease. Death from all causes was several-fold more common in children whose HIV status was indeterminate. These included 3 for whom mitochondrial disease might reasonably be included in the differential; however, 2 of these children were definitely not exposed to antiretrovirals and the other’s prenatal exposure was unknown.

There were no deaths among HIV-uninfected children exposed to both zidovudine and lamivudine. Among infants of negative or indeterminate HIV status, there were 14 SIDS deaths among about 10,000 zidovudine-exposed infants and 14 deaths among about 3,000 antiretroviral-unexposed infants. Considering the demographics of SIDS at the time of the deaths under review, the authors did not find evidence of an increased incidence of SIDS over what was expected.

So how do we resolve the apparent discrepancy between these two studies? While the 2 fatal French cases are striking, it seems unlikely that the U.S. study would have missed similar events. It is more plausible to assume that cases similar to the more subtly affected French infants could be missed without active surveillance.

In addition, there is an important distinction between the antiretroviral exposure in four of the French cases (including both fatal cases) and the practice in the U.S. In the U.S., combination antiretroviral therapy for pregnant women commonly includes zidovudine and lamivudine – but only zidovudine is given to the newborn if the mother received prenatal treatment. In France, however, it was common practice to give zidovudine and lamivudine to both the mother and to the infant. But even if we accept that this difference in exposure pattern may account for the absence of fatal mitochondrial disease in the U.S., legitimate concern must remain over the potential for more subtle effects.

Alimenti et al. prospectively measured plasma lactate levels during the first 6 months of life in 38 British Columbian infants, 35 of whom were exposed to antiretroviral regimens containing 2 nucleoside analogues starting a mean of 17 weeks prior to delivery. A lactate level greater than the upper limit of normal (2.1 mM) was found on at least one measurement in 92% of infants and 26% had levels >5 mM. Among 33 infants followed beyond 28 weeks of life, lactate levels normalized in all but 2 (94%). Other than two infants with irritability and vomiting in the first 2 weeks of life, no infant had symptoms suggestive of mitochondrial dysfunction. There was no association between lactate level and duration of antiretroviral therapy or exposure to stavudine (a nucleoside analogue with more potent inhibition of mitochondrial DNA polymerase).

Noguera and colleagues in Spain prospectively followed 127 infants exposed to antiretrovirals, mostly zidovudine or stavudine plus lamivudine and nevirapine or nelfinavir, for a mean duration of 31 weeks during pregnancy, followed by intrapartum and neonatal zidovudine. Results were considered abnormal only when both lactate and alanine were elevated. Control levels were obtained from aged-matched infants undergoing venipuncture for routine presurgical testing. The mean lactate level in antiretroviral-exposed infants was 2.88 mM at 6 weeks and decreased steadily to 1.71 mM by 12 months, significantly different (p <0.0001) at each of the 4 time points from controls, which was 1.61 mM at 6 weeks and 1.24 mM by 12 months. Three of the exposed infants had transient neurologic abnormalities, one of whom had especially high lactate levels (about 7 mM through the first 6 months of life).

The authors point out that the prevalence of hyperlactatemia in HIV-uninfected, antiretroviral-exposed infants they describe is higher than the prevalence among HIV-infected children receiving antiretroviral therapy at their institution (17%).

Divi and colleagues found moderate to severe damage to umbilical cord arterial endothelium mitochondria by transmission electron microscopy in blinded specimens from 6 of 9 infants exposed in utero to zidovudine and lamivudine, whereas specimens from unexposed controls were normal. The severity of damage correlated with duration of prenatal exposure. The ratio of copies of the mitochondrial mtD loop gene to copies of the nuclear 18S RNA gene in both cord blood mononuclear cells and umbilical cord was higher in infants of HIV-uninfected controls than in antiretroviral-exposed infants. Review of medical records did not reveal associations between maternal health, other exposures, maternal age, or morphologic and genetic findings.

Poirier and colleagues from the WITS Group compared mitochondrial DNA content (using the same genes as Divi et al) and telomere length in peripheral blood leukocytes from control infants of 20 HIV-uninfected mothers, 10 antiretroviral-unexposed infants of HIV-infected mothers, and 10 infants with prenatal and neonatal zidovudine exposure. The mitochondrial:nuclear copy ratio was 146 at birth for the zidovudine-exposed infants and 442 for the controls, and the difference remained significant at 1 and 2 years (p <0.05 at all 3 time points). Interestingly, the ratio was intermediate in the zidovudine-unexposed, HIV-exposed infants; significantly less than controls at birth, 1 year, and 2 years; and significantly greater than zidovudine-exposed infants at birth and 2 years. There were no differences in telomere length between the three groups. The authors did not offer speculation on how HIV status in the mother, without antiretroviral exposure, could affect mitochondrial DNA content in the infant.

Taken together, these studies of plasma lactate levels and mitochondrial DNA content suggest that antiretroviral-exposed infants have measurable effects on mitochondria. The clinical significance of any metabolic, ultrastructural, or genetic abnormalities is unknown. Obtaining plasma lactate levels free of artifact from hypoxia or ischemia related to phlebotomy is difficult. The Alimenti study purports to have taken precautions to prevent artifact; the Noguera study describes the control population and confirmed high lactate levels with elevated alanine. The finding of Poirier and colleagues that infants of HIV-infected mothers who were not exposed to antiretrovirals have decreased mitochondrial DNA copy number in peripheral leukocytes is surprising.

Various metabolic abnormalities, including evidence of mitochondrial damage, can be observed in antiretroviral-naïve adults infected with HIV. Potential mechanisms for a chronic systemic infection to cause such abnormalities can be envisioned, but how an uninfected fetus might be affected by a mother’s HIV infection is less clear. It should be noted that the Poirier study was performed in the era when only zidovudine was given to pregnant women. The routine use of HAART for prevention of mother-to-child transmission of HIV might conceivably ameliorate effects of maternal HIV infection on the fetus.