Research in sub-Saharan Africa (SSA) predating the introduction of antiretroviral treatment (ART) showed strong associations between HIV infection, adverse pregnancy outcomes, and increased morbidity and mortality.1,2 Similar to multiple studies, the recent PROMISE (Promoting Maternal and Infant Survival Everywhere) multicountry randomized trial of HIV-infected women and their infants provided conclusive evidence that perinatal HIV transmission can be dramatically reduced (∼0.5%) when triple ART is used during pregnancy.3 Coupled with postpartum continuation of maternal ART or infant ART prophylaxis during breastfeeding,4 these strategies can nearly eliminate transmission by reducing viral load, slowing disease progression, and improving maternal and infant health.
Despite progress made toward eliminating mother to child transmission of HIV with ART, low birth weight (LBW), preterm birth (PTB), and small for gestational age (SGA) continue to have negative impact on child health in Africa.5,6 Other demographic, behavioral, nutritional, and clinical risk factors also contribute to the burden of adverse pregnancy outcomes in these settings,7 potentially including exposure to ART in utero. Data from SSA suggest that maternal triple ART use may be associated with adverse pregnancy outcomes. The PROMISE trial found statistically significant higher rates of PTB and LBW in mother–infant pairs randomized to triple ART3; the regimens studied included zidovudine (ZDV), nevirapine (NVP), tenofovir (TDF)/emtricitabine (FTC), lamivudine (3TC), and lopinavir/ritonavir (LPV-r). However, other ART studies suggest that the ART effect on adverse pregnancy outcomes is not limited to protease inhibitors, and exposure to efavirenz (EFV)-based ART or other combinations may increase risk.8–11 Although recent data from Botswana showed reassuring findings for TDF/FTC/EFV compared with other ART regimens,12,13 previous reports showed conflicting findings.14–16 Several mechanisms have been hypothesized for the occurrence of adverse outcomes such as ART may create an imbalance in the Th1/Th2 immune response, chronic placental insufficiency, and lowering of progesterone levels.17–21
We conducted a study comparing 3 adverse pregnancy outcomes (PTB, LBW, and SGA including the subclasses of proportionate and disproportionate SGA) among HIV-infected women on ART during pregnancy and HIV-uninfected women concurrently recruited in Blantyre, Malawi. The Blantyre research site participated in the PROMISE trial and has conducted maternal and child HIV research for over 20 years. The aim of the current analysis of the study, known as POISE (Pregnancy Outcomes and Infant Survival in the Era of Universal HAART in Africa), was to compare PTB, LBW, and SGA between healthy HIV-infected women on ART and HIV-uninfected women. We hypothesized that ART has eliminated differences in pregnancy outcomes between HIV-infected and HIV-uninfected women. We specifically included HIV-uninfected women and HIV-infected women with high CD4 (≥350 cell count) in this analysis to allow for comparison with general population of women, focusing on HIV ART effect as well as other underlying risk factors for adverse pregnancy outcomes. We excluded women with low CD4 because they are recommended to receive ART for their own health, irrespective of any other underlying factor or health of their children. Other aims of the POISE study are differences in mortality and morbidity among infants born to HIV-infected and HIV-uninfected women.
Setting and Study Population
This study was conducted in Blantyre, Malawi, from January 2016 to September 2017 at the main teaching hospital, the Queen Elizabeth Central Hospital (QECH), and at 4 major health centers representing the catchment area for QECH. Women were screened and enrolled at delivery if eligible. Based on the 2015–2016 Malawi Demographic and Health Survey, 96% of births are delivered in a health facility in urban areas including Blantyre.22
Study Design and Enrollment
This was a prospective, observational study. Women were enrolled at delivery and followed for 1 year with their infants. This analysis is limited to data obtained at delivery. Eligibility criteria were: confirmed HIV status; providing written informed consent; and singleton births. For HIV-infected women, additional inclusion criteria were: CD4 ≥350 cells/mm3, no WHO stage 3 or 4 HIV disease, and on ART for at least 1 week before delivery (standard-of-care ART, not provided by the study). The exclusion criteria were: inability to understand or sign informed consent; multiple births; and for HIV-infected women: CD4 cell count <350 cells/mm3, WHO stage 3 or 4 HIV disease, and less than 1 week on ART. Women were consented to enroll with their infants, and no infant was excluded if mother provided consent. Women were counseled and consented either before or after delivery depending on time of arrival at the facility and the circumstances of each woman.
In Malawi, all women attending antenatal care are tested for HIV and those with HIV infection are immediately counseled and started on the national ART regimen (TDF, 3TC, and EFV) within 7 days. This information is available in a “Health Passport” that women present at health facilities. For HIV-infected women, we chose a cutoff of CD4 ≥350 cells/mm3 to be consistent with the PROMISE study to allow for comparison with previous studies. Also, the transition to ART initiation at CD4 <500 cells/mm3 (vs 350) was not completed in Malawi at the time when the study started. For enrollment, we used the HIV clinical stage to determine eligibility because CD4 results were not immediately available. Once CD4 test results became available, women with low CD4 were exited from the study. In this study, eligible HIV-infected women were enrolled at delivery. Eligible HIV-uninfected woman delivering at the same health facilities within a similar time period as an HIV-infected woman was concurrently enrolled.
BW was measured immediately after birth by trained nurses and midwives using calibrated standard scales. GA assessment using the Ballard score23 was performed by a trained study nurse within 36 hours of birth. GA was also estimated using date of the last menstrual period (LMP) that was available for all women in this study. At the enrollment visit (delivery), the following maternal procedures were completed: counseling and consenting for the study, verification of HIV status, physical examination, completion of questionnaires, and collection of blood samples. Maternal blood samples were used to verify HIV status based on 2 rapid tests, to measure CD4 cell count and HIV viral load for HIV-infected women, and to assess hematology and chemistry among all women. These tests were conducted at the Johns Hopkins Research Project laboratory in Blantyre, Malawi. For the newborn, a physical examination and anthropometric measurements were completed. Trained study nurses used study-specific structured questionnaires and case report forms to collect outcomes and exposures, including risk factors and potential confounders. Data were collected on socioeconomic and demographic factors, reproductive history (outcome of previous pregnancies and survival of children), time of initiation, and adherence to ART, regimen, and physical examination findings. Based on our previous research in this setting in Malawi, we identified women who reported having electricity in the house as having high socioeconomic status and those with no electricity in the house as low socioeconomic status. We also collected data on body mass index (BMI) at delivery as a measure of nutritional status. Although this study did not aim to study perinatal HIV transmission, infants found to be infected (N = 5) were referred to start ART per country guidelines. Similarly, HIV-uninfected women who seroconverted on verification of their HIV status at delivery (N = 8) were referred for appropriate clinical assessment.
Data Management and Analyses
Data were entered, cleaned, anonymized, and aggregated before analysis. Three main outcomes were evaluated in this study: PTB, LBW, and SGA. We defined PTB as GA <37 completed weeks and LBW as BW <2.5 kg. SGA was defined as BW below the 10th percentile for GA based on a reference population developed by Oken et al.24 We classified SGA <third percentile of the reference population as very SGA (VSGA), an indication of IUGR. SGA was further classified using ponderal index [BW relative to birth length of infant: (100 × BWT in grams)/(birth length in centimeters)3] to proportionate (symmetric or stunted) SGA and disproportionate (asymmetric or wasted) SGA.7 We used a cutoff of ≥2.25 and <2.25 of ponderal index to classify SGA infants to proportionate and disproportionate SGA, respectively.25
Descriptive [frequencies, mean values, medians, and interquartile ranges (IQRs)] and bivariate analyses of sociodemographic, economic, and other factors were conducted first, overall, and stratified by key factors. We compared rates of the 3 major outcomes between HIV-infected and HIV-uninfected women. Data from HIV-infected women were stratified by duration of exposure to ART (initiated ART before pregnancy, during first trimester, during second trimester, or during third trimester of pregnancy). Multivariable analyses assessed associations between pregnancy outcomes and HIV infection after adjusting for potential confounders. Multivariable logistic (binary outcome of PTB, LBW, and SGA) regression analysis was used and included covariates based on biological, epidemiological, and statistical importance. Crude and adjusted odds ratios (ORa) [and 95% confidence intervals (CIs)] are presented. Multivariable linear (continuous BW outcome) regression analysis was conducted to assess the association of HIV with BW distribution after adjusting for other risk factors. Quantile regression was also used to examine differences in BW outcome at various quantiles (eg, at the tails of the distribution).26
Per WHO recommendation and national guidelines, ART for all pregnant and breastfeeding women is the standard of care; therefore, we did not enroll an ART-naive cohort. However, we conducted several analyses restricted to HIV-infected women [eg, to assess effect of adherence using detectable or undetectable HIV viral load at delivery (<40 copies)]. In these analyses, we performed multivariable models using dummy variables for HIV status (with HIV-uninfected being the reference) to explore the effect of HIV on pregnancy outcome. We compared those with undetectable viral load, as proxy for ART exposure alone, with HIV-uninfected; and detectable to undetectable viral load to show the effect of HIV exposure alone.
We originally based sample size estimates of this study on the PTB outcome taking into account a stratified approach based on 3 groups of ART initiation before/during pregnancy by the trimester. Earlier data from Malawi suggested the overall frequency of PTB was ∼8%27 and a Botswana study reported a frequency of 26.5% PTB among women who continued triple ART from before pregnancy and 22.7% among all other HIV-infected women.10 We estimated a sample size of 225 HIV-infected women for each of the 3 groups and included an equal number of HIV-uninfected women for each HIV-infected strata of ART exposure duration. Therefore, the total required sample size was 1350 (675 HIV-infected and 675 HIV-uninfected; 90% power, 2-sided α of 0.05, 1:1 ratio of exposed to unexposed, the Fisher exact test). We considered a P value ≤0.05 to be statistically significant. SAS version 9.4 (Cary, NC) was used for all analyses.
This study was approved by institutional review boards in Malawi (College of Medicine Research and Ethics Committee P10/15/1814) and in the United States (Johns Hopkins Bloomberg School of Public Health IRB00006708). Study staff completed human subjects' protection and good clinical practice certification before participant contact.
A total of 5423 women were approached at delivery across all study sites. Figure 1 shows a flow chart of the study with exclusions and enrollments. Approximately equal proportions of HIV-uninfected and HIV-infected women were eligible and successfully consented (48.1% vs 51.9%) per protocol. Overall, 1299 women were included in this analysis: 614 HIV-infected and 685 HIV-uninfected. Compared with HIV-uninfected women, HIV-infected women were older, had more pregnancies/children, and reported more previous pregnancy losses or adverse outcomes (eg, spontaneous abortions, stillbirths, premature births, and infant deaths); these differences were statistically significant. The median GA based on Ballard score was 38 weeks for both HIV-infected and HIV-uninfected women, and similarly, the median GA based on LMP estimates was 39.29 weeks for both HIV-infected and HIV-uninfected women. See Supplemental Digital Content 1, https://links.lww.com/QAI/B223 for a full list of baseline characteristics of HIV-infected and HIV-uninfected women in Blantyre, Malawi.
Table 1 shows rates and distributions of adverse pregnancy outcomes. The overall rates of adverse pregnancy outcomes were 10.0% for PTB, 6.0% for LBW, and 17.8% for SGA. The rates of these outcomes were similar between HIV-infected women on ART and HIV-uninfected women and not statistically different (P > 0.05). To further compare the distribution of these outcomes by HIV status, we analyzed multiple cross-tabulations separately for HIV-infected women using ART and HIV-uninfected women. As described by others,5,28 we created 4 mutually exclusive groups for PTB and SGA: (1) Term and Appropriate for GA (Term AGA); (2) Term SGA; (3) PTB and AGA; and (4) PTB and SGA. Similarly, 4 mutually exclusive groups were created for SGA and LBW, and PTB and LBW. Table 1 shows summaries of these rates among HIV-infected women on ART and HIV-uninfected women. The rates for all combinations of outcomes are comparable between HIV-infected and HIV-uninfected women. In HIV-uninfected women, the rate of Term SGA or in those with normal BW (NBW SGA) was 16.4% and 14.3%, respectively. Similar rates were observed in HIV-infected women on ART (16.3% and 12.5%, respectively). The rate of Term LBW was 2.0% in HIV-uninfected women and 3.8% in HIV-infected women. The rates of PTB SGA, LBW SGA, and PTB LBW were 2.0%, 4.1%, and 2.9%, respectively, in HIV-uninfected women and 0.8%, 4.6%, and 3.4%, respectively, in HIV-infected women.
There were no associations between adverse pregnancy outcomes and HIV infection after controlling for other variables. Table 2 shows that in multiple logistic regression models, the associations between HIV infection and PTB, LBW, and SGA were not statistically significant (models 1, 2, and 3). In these models, the main predictors of the 3 adverse pregnancy outcomes were BMI and socioeconomic status: higher BMI was associated with lower risk and low socioeconomic status was associated with higher risk. Gravidity was associated with PTB and LBW, and previous pregnancy loss/adverse outcome was associated with LBW. Similar results were obtained when a multiple linear regression model is used with BW on a continuous scale, with no statistically significant association observed (See Supplemental Digital Content 2, https://links.lww.com/QAI/B223: Association of HIV infection with BW controlling for other factors, a linear regression model). The main predictors of increased BW were higher gravidity, no previous pregnancy losses/adverse outcomes, higher BMI, and higher socioeconomic status.
Of the 231 mother–infant pairs with SGA, 182 (78.8%) were proportionate SGA and 49 (21.2%) were disproportionate SGA (See Supplemental Digital Content 3, https://links.lww.com/QAI/B223: Outline of SGA infants, Blantyre, Malawi). Most of the proportionate SGA infants were born to HIV-uninfected women (N = 109, 59.9%) and most of the disproportionate SGA were born to HIV-infected women (N = 32, 65.3%) (P = 0.002). We stratified the overall multivariable model of SGA in Table 2 (model 3) by proportionate and disproportionate SGA outcomes. Because proportionate SGA is generally assumed to occur due to early insults during pregnancy, whereas disproportionate SGA likely occurs due to late pregnancy insults, and stratification by these 2 subcategories has the potential to explore time effects controlling for HIV infection and other risk factors. Table 3 shows that among 1250 women (182 proportionate SGA and 1068 appropriate for GA), HIV infection and proportionate SGA were not associated (P = 0.30) after controlling for other factors; only low socioeconomic status was significantly associated with proportionate SGA (adjusted OR = 1.39; P = 0.05). However, among 1117 women (49 disproportionate SGA and 1068 appropriate for GA), the association between HIV infection and disproportionate SGA was statistically significant (ORa 2.28; P = 0.01); no other factors were significantly associated with disproportionate SGA (Table 3). We further classified the proportionate and disproportionate SGA to VSGA or not VSGA (see Supplemental Digital Content 3, https://links.lww.com/QAI/B223). Table 4 shows that there was no association between treated HIV infection and proportionate VSGA. However, the adjusted association between treated HIV infection and disproportionate VSGA (N = 11 HIV-infected and 4 HIV-uninfected) was statistically significant (ORa = 4.95, 95% CI: 1.43 to 17.19) (Table 4). Lower gravidity was also significantly associated with increased risk of disproportionate VSGA.
We examined the rate of PTB outcome stratified by timing of ART initiation before or during pregnancy. The rate of PTB did not differ by time of ART initiation; 11.4% (34/299) if initiated before pregnancy, 9.3% (24/257) during the first or second trimester, and 12.1% (7/58) during the third trimester. Adjusting for the same variables as in Table 2 showed no significant association between timing of ART initiation and adverse pregnancy outcomes, both in HIV-infected-only models or models with HIV-uninfected as a reference dummy variable (data not shown). Of the 614 HIV-infected women who were on ART at time of delivery, 459 (74.8%) had suppressed viral load (<40 copies) and 155 (25.2%) had detectable viral load (10.9% ≥40 to <400; 8.1% 400 to <10,000; and 6.2% >10,000 copies). The median duration of exposure to ART for the 155 women with detectable viral load was 129 days (IQR 76–493). Baseline CD4 cell count among HIV-infected women was high: median 658.00 (IQR 503.00–833.00); 75.7% had CD4 count >500 cells/mm3.
In this relatively healthy cohort of HIV-infected women using ART, adverse pregnancy outcomes were comparable with that of HIV-uninfected women recruited from the same setting and at the same time in Blantyre, Malawi. There were no statistically significant differences in rates of PTB, LBW, or SGA between the 2 cohorts. These results are consistent with our original hypothesis and suggest that ART (particularly TDF/3TC/EFV) can eliminate differences potentially attributed to HIV without introducing additional risk of adverse pregnancy outcomes. These findings are consistent with the findings from Botswana.12,13 In addition, this study shows that certain risk factors, specifically maternal nutrition and overall socioeconomic status, are persistent risk factors for adverse pregnancy outcomes assessed in this population. These statistically significant associations were independent of HIV/ART and excluding women with low CD4 or severe HIV disease.
The lack of association of HIV infection with PTB, LBW, and SGA is likely due to consistent ART use during pregnancy. Approximately 75% of women had suppressed virus at time of delivery, and ∼76% had CD4 count >500 cells/mm3 at delivery, which suggest high adherence to ART during pregnancy. The lack of association is unlikely due to confounding; we adjusted for several potential risk factors known to influence these outcomes using multiple methods. Duration of ART exposure before or during pregnancy did not impact results. We excluded HIV-infected women with low CD4 count and severe clinical disease to target healthy HIV-infected women and minimize the effect of HIV to better understand the effect of ART. In addition, HIV-infected women with low CD4 cell count and clinical disease are prescribed ART for their own health, regardless of pregnancy status. In the large multicountry PROMISE trial, women with CD4 <350 cells were also excluded.3
Our findings also highlight the benefits of incorporating other measures of newborn health in Malawi, and the region. Relying heavily on BW data alone may lead to underestimation of the magnitude of adverse pregnancy outcomes and related morbidity. Although BW is routinely available in health facilities, GA is not consistently or reliably assessed, and estimation of SGA requires calculations based on an appropriate reference population. However, newborns can be preterm or SGA without being LBW.7 Data from this study show important findings regarding the distribution of pregnancy outcomes: the rate of term SGA was ∼16% and the rate of infants with NBW but still SGA was ∼13% (Table 1). This information could not be known through BW alone.
Use of BW percentiles and ponderal index in this study allowed for robust analysis of SGA as proportionate and disproportionate. These analyses could highlight the timing of insult during pregnancy and may reveal the potential effect of risk factors other than HIV/ART. Although all infants with intrauterine growth restriction (IUGR) are also SGA, not all SGA infants have IUGR.7 Among women with proportionate SGA infants (early insults during pregnancy), HIV infection was not associated with proportionate SGA. However, among women who had disproportionate SGA (late insults during pregnancy), there was a strong association between HIV infection and disproportionate SGA. Further stratification by severity of SGA (<third percentile of BW) showed an association between HIV infection and disproportionate VSGA, although numbers were small (see Supplemental Digital Content 3, https://links.lww.com/QAI/B223). Examination of the 11 HIV-infected mothers with disproportionate VSGA infants showed that all had suppressed viral load, except 2 women with high viral load and one woman with PTB, and as expected, all had LBW. All women reported using ART for relatively long periods and had high CD4 cell counts. We have no clear mechanism to explain why HIV is associated with this severe adverse outcome in the presence of ART. Potential interactions and residual confounding (eg, genetic or environmental) may play a role. Future studies should consider the effect of ART on these subcategories of adverse outcomes. The causes of many adverse pregnancy outcomes are complex and multifactorial.7
A major strength of this study is the inclusion of HIV-uninfected cohort to obtain background rates of adverse pregnancy outcomes and risk factors, and provided a comparison group for the HIV-infected cohort from the same setting, allowing for greater generalizability of study findings. We considered the potential for selection bias. In this study, all women were recruited from the same health facilities and approximately around the same calendar time. All women had the same procedures conducted by a trained research team. It is likely that any selection would have affected both cohorts similarly and therefore introduce minimal bias. We compared rates of pregnancy outcomes using both Ballard score and LMP to assess misclassification. Although there was no observed difference in PTB between cohorts, PTB was overestimated if date of LMP was used in the analysis, similar to a study in South Africa.29 We also adjusted for potential confounders using a multiple linear regression model with BW as a continuous measure to avoid a specific cutoff (eg, 2.5 kg); again, there was no association with HIV status. We repeated the logistic regression models using log-binomial models to assess whether the estimates were overestimated; the ORs and relative risks obtained were comparable, even for SGA which was the most frequent outcome. Residual confounding is a possibility and the list of variables we included in the models did not include infections such as malaria and genital tract infections. We could not classify PTB into very PTB or other subgroups because the numbers were too small, partly due to the exclusion criteria.
Despite the encouraging results of HIV treatment effects, expanding use of ART across the life span requires continuous monitoring to assess safety. Also, these results show that several conventional risk factors such as maternal nutrition and socioeconomic status are still taking a toll on pregnancy outcomes, independent of HIV infection, and ART use. This study showed that the rates of adverse pregnancy outcomes are high in Malawi: with the exception of LBW (which could be due to the exclusion of unhealthy women from the study), the rates of PTB and SGA in this study are similar to what have been reported in the general population of African women. In the era of universal ART and expansion of ART services in SSA, the value of improving maternal health beyond HIV treatment remains. This is a priority for both HIV-infected and HIV-uninfected women in these resource-constrained settings.
The authors thank the study participants and their families for being part of this study. The authors are grateful for the study teams in Malawi and the community stakeholders for the excellent collaboration during the conduct of the study.
1. Brahmbhatt H, Kigozi G, Wabwire-Mangen F, et al. Mortality in HIV-infected and uninfected children of HIV-infected and uninfected mothers in rural Uganda. J Acquir Immune Defic Syndr. 2006;41:504–508.
2. Taha TE, Dallabetta GA, Canner JK, et al. The effect of human immunodeficiency virus infection on birthweight, and infant and child mortality in urban Malawi. Int J Epidemiol. 1995;24:1022–1029.
3. Fowler MG, Qin M, Fiscus S, et al. Benefits and risks of antiretroviral therapy for perinatal HIV prevention. N Engl J Med. 2016;375:1726–1737.
4. Flynn P, Taha TE, Cababasay M, et al. Prevention of HIV-1 transmission through breastfeeding: efficacy and safety of maternal antiretroviral versus infant nevirapine prophylaxis for duration of breastfeeding in HIV-1-infected women with high CD4 cell count (IMPAACT PROMISE): a randomized, open label, clinical trial. J Acquir Immune Defic Syndr. 2018;77:383–392.
5. Katz J, Lee AC, Kozuki N, et al. Mortality risk in preterm and small-for-gestational-age infants in low-income and middle-income countries: a pooled country analysis. Lancet. 2013;382:417–425.
6. Lee ACC, Katz J, Blecowe H, et al. National and regional estimates of term and preterm babies born small for gestational age in 138 low-income and middle-income countries in 2010. Lancet Glob Health. 2013;1:e26–e36.
7. Kramer MS. The epidemiology of adverse pregnancy outcomes: an overview. J Nutr. 2003;133:1592S–1596S.
8. Mofenson LM. Antiretroviral therapy and adverse pregnancy outcome: the elephant in the room? J Infect Dis. 2016;213:1051–1054.
9. Bisio F, Nicco E, Calzi A, et al. Pregnancy outcomes following exposure to efavirenz-based antiretroviral therapy in the Republic of Congo. New Microbiol. 2015;38:185–192.
10. Chen JY, Ribaudo HJ, Souda S, et al. Highly active antiretroviral therapy and adverse birth outcomes among HIV-infected women in Botswana. J Infect Dis. 2012;206:1695–1705.
11. Li N, Sando MM, Spiegelman D, et al. Antiretroviral therapy in relation to birth outcomes among HIV-infected women: a cohort study. J Infect Dis. 2016;213:1057–1064.
12. Zash R, Souda S, Chen J, et al. Reassuring birth outcomes with Tenofovir/Emtricitabine/Efavirenz used for prevention of mother to child transmission of HIV in Botswana. J Acquir Immune Def Syndr. 2016;71:428–436.
13. Zash R, Jacobson D, Diseko M, et al. Comparative safety of antiretroviral treatment regimens in pregnancy. JAMA Pediatr. 2017;171:e172222.
14. Kourtis AP, Schmid CH, Jamieson DJ, et al. Use of antiretroviral therapy in pregnant HIV-infected women and the risk of premature delivery: a meta-analysis. AIDS. 2007;21:607–615.
15. Cohan D, Natureeba P, Koss CA, et al. Efficacy and safety of lopinavir/ritonavir versus efavirenz-based antiretroviral therapy in HIV-infected pregnant Ugandan women. AIDS. 2015;29:183–191.
16. Cotter AM, Garcia AG, Duthely ML, et al. Is antiretroviral therapy during pregnancy associated with an increased risk of preterm delivery, low birth weight, or stillbirth? J Infect Dis. 2006;193:1195–1201.
17. Goldenberg RL, Culhane JF, Iams JD, et al. Epidemiology and causes of preterm birth. Lancet. 2008;371:75–84.
18. Shapiro RL, Souda S, Parekh, et al. High prevalence of hypertension and placental insufficiency, but no in utero HIV transmission, among women on HAART with stillbirths in Botswana. PLoS One. 2012;7:e31580.
19. Powis K, Shapiro RL. Protease inhibitors and adverse birth outcomes: is progesterone the missing piece to the puzzle? J Infect Dis. 2015;211:4–7.
20. Papp E, Mohammadi H, Loutfy MR, et al. HIV protease inhibitor use during pregnancy is associated with decreased progesterone levels, suggesting a potential mechanism contributing to fetal growth restriction. J Infect Dis. 2015;211:10–18.
21. Fiore S, Ferrazzi E, Newell ML, et al. Protease inhibitor-associated increased risk of preterm delivery is an immunological complication of therapy. J Infect Dis. 2007;195:914–916.
22. National Statistical Office (NSO) and ICF [Malawi]. Demographic and Health Survey 2015. Zomba, Malawi and Rockville, MD: NSO and ICF; 2017.
23. Ballard JL, Novak KK, Driver M. A simplified score for assessment of fetal maturation of newly born infants. J Pediatr. 1979;95:769–774.
24. Oken E, Kleinman KP, Rich-Edwards J, et al. A nearly continuous measure of birth weight for gestational age using a United States national reference. BMC Pediatr. 2003;3:6.
25. Balcazar H, Haas J. Classification schemes of small-for-gestational age and type of intrauterine growth retardation and its implications to early neonatal mortality. Early Hum Dev. 1990;24:219–230.
26. Battai M, Frangillo EA, Sui X, et al. Use of quantile regression to investigate the longitudinal association between physical activity and body mass index. Obesity. 2014;22:E149–E156.
27. Taha TE, Dadabhai SS, Sun J, et al. Child mortality levels and trends by HIV status in Blantyre, Malawi: 1989–2009. J Acquir Immune Defic Syndr. 2012;61:226–234.
28. Lee ACC, Kozuki N, Cousens S, et al. Estimates of burden and consequences of infants born small for gestational age in low and middle income countries with INTERGROWTH-21st
standard: analysis of CHERG datasets. BMJ. 2017;358:j3677.
29. Malaba T, Newell M-L, Madlala H, et al. Methods of gestational age (GA) assessment influence the observed association between ART
exposure and preterm delivery (PTD): a prospective study in Cape Town, South Africa. IAS Conference, Paris, France, MOPDC0101 Oral Abstract, 2017