Among women randomized to receive either PI-based or NRTI-based HAART during pregnancy, we found an association between PI use and preterm delivery. We noted less weight gain among women initiating PI-based HAART in late pregnancy, using change in BMI 1 month after HAART initiation as a surrogate for weight gain. The overall rate of prematurity in our study was comparable to rates identified in government hospitals in Botswana [21
The results of this randomized trial help clarify previous observational studies related to PI use in pregnancy. Although some previous studies have reported an association between PI use and preterm delivery [7
], others have reported no association [10
] or have been underpowered to detect a modest increase in preterm deliveries [12
]. We believe that previous conflicting results are related to limited power, as well as potential confounders in observational studies [11
]. However, as Townsend et al [23
] noted in their pooled analysis of 3 large cohorts, the differences may reflect substantial differences in the study populations and the data collected. Our randomized design allowed us to isolate the effect of PIs.
Despite association with preterm delivery, we were reassured by the lack of association between the type of HAART used in pregnancy and excess infant mortality. Although we noted increased odds of hospitalizations and mortality among preterm infants overall in the first 6 months of life, there was no significant increase in the PI group. These findings should be interpreted with caution, as the number of infant deaths was small, limiting statistical power to detect a true difference. Women and infants in our cohort may have had better access to medical care because of their participation in a randomized clinical trial, and this may have attenuated differences in infant mortality by preterm status. Although clinical care in our study was comparable to that available for the general population in Botswana, even small increases in the incidence of preterm deliveries in low-resource settings may have important consequences for infant mortality. Our ability to evaluate very preterm deliveries was limited by our study design, which enrolled women in the third trimester of pregnancy. However, the high mortality risk in very preterm infants, particularly in resource-limited settings [24
], suggests that further study of very preterm deliveries by HAART regimen is warranted.
Our study found that women in the PI group had reduced weight gain in late pregnancy compared with the NRTI group, an association that to our knowledge has not been reported previously. We do not have a biological explanation for this finding, and although severe adverse events did not differ by study arm, we cannot exclude nausea, vomiting, or poor oral intake as an explanation. Women who delivered preterm were noted to have a lower mean change in BMI 1 month after initiating HAART, and each 1 kg/m2 increase in BMI 1 month after initiation of HAART was associated with a 19% decline in odds of preterm delivery. However, this difference was not significant in the univariate analysis, likely owing to the multifactorial etiology of preterm delivery. We believe that additional studies are warranted to determine whether poor weight gain in late pregnancy may, in part, explain preterm deliveries among PI-treated women, and whether BMI can be followed during pregnancy to predict risk for preterm delivery.
We did not detect associations with preterm delivery for many established risk factors, in part because of limited power. However, the provision of regular antenatal care, including management of sexually transmitted infections, hypertension, and preeclampsia, may have contributed to the lack of association with these important potential risks for preterm delivery [17
]. We could not evaluate the association between infant HIV infection and preterm delivery because there were only 4 infants (all born at term) included in this analysis with in utero HIV infection. Similarly, analysis of risk factors associated with very preterm deliveries via logistic regression was not possible because there were only 12 very preterm deliveries. We were also unable to analyze the association between hypertension, pregnancy induced hypertension, or preeclampsia as no mothers who delivered preterm were diagnosed with these conditions. While these diagnoses did exist among women in the Mma Bana study who experienced stillbirths, our study excluded all stillborn deliveries.
Our study had several limitations. Power was limited to detect an association with less prevalent potential risk factors, and limited risk factor data were available. For example, we did not collect data on maternal use of tobacco, alcohol. or illicit substances. However, the prevalence of these risk factors among women in Botswana is reported to be low (Botswana-Harvard Partnership, unpublished data, 2010) and would be unlikely to differ by HAART randomization group. We also had a limited ability to assess very preterm delivery, as discussed previously. To make use of the HAART regimen randomization, our analyses were limited to women with baseline CD4+ cell counts ≥200 cells/mm3
, limiting generalizability. However, we detected no confounding or collinearity from baseline CD4+ cell count or baseline HIV-1 RNA among the women who were included. Our study focused on 2 HAART regimens, with Kaletra as our PI-based regimen. While our results may not be generalizable to all PI and non-PI HAART regimens, we would not expect PI within class differences. Fiore et al [19
] have proposed antiretroviral immunomodulation as a potential mechanism triggering preterm deliveries. The Mma Bana study did not measure change in maternal CD4+ cell counts at frequent intervals, limiting our ability to compare the rate of change in CD4+ cells between women experiencing preterm deliveries and those experiencing term deliveries. We believe that further study of this potential mechanism is important. The major strength of our study was the randomization by HAART regimen, which eliminated several potential confounders that limited the interpretability of previous studies.
In summary, PI-based HAART initiated in the third trimester of pregnancy was associated with a 2-fold higher odds of a preterm delivery compared with triple NRTI-based HAART, and with reduced weight gain in late pregnancy. Although PIs were not associated with increased infant morbidity or mortality in our study population, additional mortality data for PI-exposed infants are needed from other settings. PI-based HAART is a critical component of both PMTCT and treatment programs in the developed and developing world, and offers proven benefits to maternal and infant health [6
]. However, skilled obstetrical and neonatal care may be required to manage preterm deliveries to maximize the benefits of PI-based HAART use during pregnancy.