Maternal Tenofovir Disoproxil Fumarate Use During Pregnancy Is Not Associated With Adverse Perinatal Outcomes Among HIV-infected East African Women: A Prospective Study

In women who used 3-drug antiretroviral therapy (ART) during pregnancy, exposure to prenatal tenofovir disoproxil fumarate (TDF) was not associated with adverse perinatal outcomes. Pregnancies with TDF exposure were less likely to have preterm birth than pregnancies with non-TDF ART.

Abstract

Background

Tenofovir disoproxil fumarate (TDF) is commonly used in antiretroviral treatment (ART) and pre-exposure prophylaxis regimens. We evaluated the relationship between adverse perinatal outcomes and prenatal TDF use.

Methods

Longitudinal data were analyzed from human immunodeficiency virus (HIV)-infected women who became pregnant during 2 HIV prevention studies conducted among HIV-serodiscordant couples in Kenya and Uganda. Pregnancies included were singleton, not terminated by an induced abortion, and had documented 3-drug ART use. Multivariate generalized estimating equation models were used to determine the association of prenatal TDF and perinatal outcomes.

Results

The most frequent ART regimens were TDF/3TC/EFV (39%) and AZT/3TC/NVP (34%); 49% of pregnancies had prenatal TDF exposure and 6% used a protease inhibitor. Neonatal death, preterm birth, and pregnancy loss occurred in 2%, 8%, and 12% of pregnancies, respectively. No differences were observed between pregnancies with and without exposure to TDF in the frequency of pregnancy loss (adjusted prevalence rate ratio [aPRR] 1.19, P = .8) or neonatal death (aPRR 0.68, P = .6). Preterm birth occurred less frequently among pregnancies exposed to TDF (aPRR, 0.34, P = .02).

Conclusion

Maternal TDF use did not adversely affect perinatal outcomes.

Combination antiretroviral therapy (ART) containing tenofovir disoproxil and fumarate (TDF) is currently a first-line regimen for human immunodeficiency virus (HIV) treatment and Option B+ prevention of mother-to-child transmission (PMTCT) programs encouraged by the World Health Organization (WHO) [1]. Twenty-one of the 22 priority countries of the UNAIDS Global Plan for the Elimination of New HIV Infections among Children by 2015 [2] have either fully implemented PMTCT Option B+ or are in the process of scaling up nationally, and 77% of all HIV-infected pregnant women globally (approximately 1.5 million women) used ART during pregnancy in 2015 [3]. TDF and the fixed-dose combination of emtricitabine (FTC)/TDF and lamivudine (3TC)/TDF are also recommended by WHO for antiretroviral pre-exposure prophylaxis (PrEP) to prevent HIV acquisition among pregnant and breastfeeding women with high risk for HIV [4, 5]. As global elimination of mother-to-child transmission (EMTCT) efforts intensify by expanding PMTCT Option B+ and as PrEP use expands among pregnant women worldwide, the likelihood of women using TDF during pregnancy will substantially increase in settings where HIV prevalence is high.

A recent systematic review [6] that included 33 articles found no statistically significant differences between TDF use during pregnancy and comparison groups in stillbirth/pregnancy loss, preterm delivery, low birth weight, small for gestational age, birth defects, and infant or maternal mortality [7]. However, the review included few prospective studies evaluating the relationship of TDF use in pregnancy and perinatal outcomes among African cohorts, and these have mixed results [8–11]. Recently, the PROMISE (Promoting Maternal-Infant Survival Everywhere) study found higher rates of adverse perinatal outcomes (low birth weight, very preterm birth, and early infant death) among HIV-infected African mothers using TDF-based ART compared to zidovudine (AZT)-based ART; however, this was in the context of protease-inhibitor-based regimens, which may have been an effect modifier [9]. Other studies have not found similar associations between TDF-based versus non-TDF-based PMTCT regimens in African cohorts [11]. Additional data comparing perinatal outcomes between HIV-infected African women who used TDF-based and non-TDF-based ART during pregnancy may contribute to the growing safety profile of prolonged maternal prenatal TDF use.

We evaluated whether pregnancy loss, preterm birth, and neonatal death were more frequent in a cohort of Kenyan and Ugandan HIV-infected women who used TDF-containing ART during pregnancy compared to HIV-infected women who used ART during pregnancy that did not contain TDF.

METHODS

Study Population and Procedures

Longitudinal data were analyzed from women who were HIV-infected at enrollment and became pregnant during 2 studies (the Partners PrEP Study and the Partners Demonstration Project). The Partners PrEP Study was a randomized clinical trial of the safety and efficacy of daily oral PrEP for the prevention of HIV acquisition; 4758 HIV-serodiscordant couples from 9 sites in Kenya and Uganda were enrolled and followed between 2008 and 2012 [12]. The Partners Demonstration Project was an open-label implementation project evaluating the integrated delivery of PrEP and ART for HIV prevention that enrolled and followed 1013 high-risk HIV-serodiscordant couples at 4 sites in Kenya and Uganda between 2012 and 2016 [13]. At enrollment into both studies, all participants were members of a mutually disclosed HIV-serodiscordant couple, ≥18 years old, and not using PrEP or ART [13, 14].

In both studies, HIV-infected partners were followed quarterly and monitored 6-monthly for CD4 counts, WHO clinical staging, and HIV RNA levels. HIV-infected participants were referred to local clinics of their choice, including those colocated with the research clinic, to initiate treatment once eligible according to national guidelines [12, 13]. In both studies, pregnancy testing for HIV-infected women was performed when clinically indicated and women attended quarterly study visits through the duration of their pregnancy.

Definition of Exposures and Outcomes

The primary exposure was the use of any TDF-containing 3-drug ART during pregnancy. TDF use was captured as part of information on maternal ART use during pregnancy, including the type of ART regimen and date of initiation self-reported by women and verified with clinical records or pill bottles when available. ART use during pregnancy was categorized as having been initiated before pregnancy, during the first trimester, or after the first trimester. Sociodemographic characteristics and obstetrical history were also collected through self-report.

The 3 primary outcomes were pregnancy loss (any, <20 weeks, and ≥20 weeks), neonatal death within 3 days of delivery following live birth, and preterm birth (live birth <37 weeks gestation). Last menstrual period (LMP), pregnancy outcome (induced abortion, pregnancy loss, or live birth), the date the pregnancy ended, and occurrence of neonatal death were self-reported by women. Gestational age at birth or pregnancy loss was calculated by subtracting date of pregnancy end from LMP.

Statistical Analysis

Pregnancies were included in the primary analysis if they were singleton, did not terminate by induced abortion, had documentation of 3-drug ART regimen used during pregnancy and had pregnancy outcome data available. Fisher’s exact tests for proportions and Kruskal-Wallis tests for continuous measures were used to detect differences in sociodemographic and clinical characteristics between women exposed and unexposed to TDF during pregnancy. All analyses of pregnancy loss were restricted to women who initiated 3-drug ART before pregnancy or during the first trimester to ensure ART exposure preceded pregnancy loss.

Generalized estimating equation (GEE) log-binomial regression models were used to determine the association of TDF-containing 3-drug ART use during pregnancy and pregnancy loss, neonatal death, and preterm birth with use of non-TDF 3-drug ART use as the reference group. We decided a priori to adjust multivariate GEE models for study cohort and maternal age. Additionally, we assessed maternal education, number of children, marital status, time since HIV diagnosis, time since ART initiation, CD4⁺ lymphocyte count (cells/µL), maternal viral load (plasma HIV RNA [log₁₀ copies/mL]), WHO clinical stage, and protease inhibitor use as potential covariates. CD4 and viral load results were lagged forward to account for the 6-monthly testing schedule. We included covariates in the final model that changed the measure of association by >10% (crude vs adjusted). For variables that were collinear (time since ART initiation, CD4⁺ lymphocyte count, HIV RNA load, and WHO clinical stage), we included the variable with the least amount of missing data in multivariate models.

Sensitivity Analyses

We determined the distribution of adverse perinatal outcomes by specific ART regimens using descriptive statistics to account for heterogeneity among 3-drug ART combinations used within TDF-containing and non-TDF-containing regimens. We also compared the frequency of adverse perinatal outcomes between pregnancies with exposure to 3-drug ART combinations containing TDF versus AZT. Pregnancies were included in the subanalyses if the mothers prenatally used ART that did not contain protease inhibitors but contained either TDF or AZT, but not both. We performed additional sensitivity analyses by repeating our primary analyses including pregnancies that were missing birth outcome data but met all other inclusion criteria by: (1) assuming all pregnancies with missing outcome data had each respective adverse perinatal outcome; (2) assuming all pregnancies with missing outcome data did not have each respective adverse perinatal outcome; (3) using multiple imputation to predict each respective perinatal outcome among pregnancies missing outcome data [15]; and (4) using marginal structural models (MSMs) to account for potential time-dependent confounding associated with temporal changes across the data collection period in the frequency of TDF-containing ART use for PMTCT [1]. For MSMs, we computed stabilized inverse probability weights using logistic regression to predict the probability of TDF use (by study cohort, maternal age, time since maternal HIV diagnosis, HIV RNA load at first pregnancy visit, and year pregnancy occurred) as described by Cole et al [16, 17], the weights adjusted for time-dependent measures. Weights for the probability of not having birth outcome data available were computed separately from the weights for probability of TDF use; both weights had a mean 1.0, supporting correct specification. These weights were then used in a pooled logistic regression model of TDF use and each perinatal outcome.

RESULTS

Demographics and TDF Exposure

In total, 422 pregnancies (25% of the total pregnancies to HIV-infected women in the cohorts) met inclusion criteria and were included in the primary analysis (Figure 1). The median age of pregnant women was 25.4 years (interquartile range [IQR] 20.5–29.9), 21% were primigravidas, and over half of women (57%) were WHO Stage I at their first pregnancy visit. Overall, the most frequently used ART regimens were TDF/3TC/EFV (39%) and AZT/3TC/NVP (34%); no women switched regimens during pregnancy. Among the 208 (49%) pregnancies with prenatal TDF exposure, the most frequently used ART regimens were TDF/3TC/EFV (n = 167, 80%) and TDF/3TC/NVP (n = 31, 15%). Among pregnancies without TDF exposure (n = 214, 51% of all pregnancies), AZT/3TC/NVP (n = 146, 68%); AZT/3TC/EFV (n = 20, 9%), AZT/3TC/LPV/r (n = 19, 9%), and stavudine (d4T)/3TC/NVP (n = 19, 9%) were the ART regimens most frequently used during pregnancy. Overall, the frequency of protease-inhibitor use was low (6%). Compared to women who used non-TDF-containing ART during pregnancy, women who used TDF-containing ART were younger, more educated, had fewer children, more frequently initiated ART before pregnancy or during the first trimester, had lower viral loads, and used regimens that contained protease inhibitors less frequently (Table 1).

Figure 1.

Flowchart of inclusion. Abbreviations: TDF, tenofovir disoproxil fumarate; 3TC, lamivudine; EFV, efavirenz; NVP, nevirapine; AZT, zidovudine; LPV/r, lopinavir/ritonavir; d4T, stavudine.

Table 1.

Demographic and Clinical Characteristics by Prenatal Tenofovir Disoproxil and Fumarate Use Among HIV-infected Women who Used 3-Drug antiretroviral therapy During Pregnancy

Characteristic	n (%) or Median (IQR)a
	All HIV-infected womenb		P valued
	No TDF use (n = 214)	Any TDF usec (n = 208)
Demographic
Age (years)	26.6 (23.5–31.6)	24.7 (21.7–28.5)	.001*
Education completed (years)	7 (5–9)	8 (7–11)	.010*
Number of children	2 (1–3)	1 (0–2)	<.001*
Married	201 (98.1%)	187 (96.4%)	.312
Clinical
Time since first HIV diagnosis (years)	2.4 (1.4–3.9)	1.8 (1.1–2.6)	<.001*
Timing of ART initiation
Before pregnancy	55 (26.4%)	80 (39.2%)	<.001*
First trimester	27 (13.0%)	42 (20.6%)
Second or third trimester	126 (60.6%)	82 (40.2%)
CD4 (cell/µL)	526 (394–753)	739 (504–912)
Plasma HIV RNA (log10 copies/ml)	3.1 (1.3–4.3)	1.9 (1.5–4.1)	<.001*
WHO clinical stage
Stage 1	97 (50%)	120 (64%)	.002*
Stage 2	71 (37%)	61 (32%)
Stage 3	24 (12%)	7 (4%)
Stage 4	2 (1%)	0 (0%)
PI-containing maternal ART regimen	21 (10%)	2 (1%)	<.001*

Abbreviations: ART, antiretroviral therapy; IQR, interquartile range; TDF, tenofovir disoproxil fumarate; PI, protease-inhibitor. *P < .10.

^aMissing data not shown

^bAmong HIV-infected women with documented use of any 3-drug ART during pregnancy.

^cMaternal TDF use defined as documented use of TDF-containing 3-drug ART during pregnancy.

^dChi-squared test for proportions or Kruskal-Wallis test for continuous measures; Fisher’s exact test for variables with ≤5 observations.

Perinatal Outcomes

Among pregnancies in which 3-drug ART was initiated before pregnancy or during the first trimester (n = 204), pregnancy loss occurred in 24 (12%) pregnancies; 79% of these losses occurred at <20 weeks gestation. Neonatal death and preterm birth occurred in 2% and 8% of pregnancies with live births (n = 380), respectively.

Among pregnancies with exposure to 3-drug ART in the first trimester, there was no difference in the frequency of any pregnancy loss between those with TDF-containing ART exposure compared to regimens without TDF (14% versus 9%, respectively, prevalence rate ratio [PRR] = 1.63, 95% CI, 0.66–4.04, P = .2). We did not detect a difference between exposure groups in the subgroup of losses <20 weeks gestation (11% vs 7%, PRR = 1.49, 95% CI, 0.54–4.12, P = .4) nor in the subgroup of pregnancy losses ≥20 weeks gestation (2% vs 1%, PRR = 2.79, 95% CI, 0.32–24.63, P = .4), though statistical power was limited. Among pregnancies that resulted in live births, there were no differences between pregnancies with prenatal TDF-containing ART use compared to non-TDF-containing ART use in the frequency of neonatal death (1% versus 2%, PRR = 0.82, 95% CI, 0.19–3.57, P = .8) or preterm birth (6% versus 10%, PRR = 0.55, 95% CI, 0.27–1.14, P = .1). In multivariate models, we did not detect any association between prenatal TDF use and pregnancy loss or neonatal death

after adjustment for study cohort, maternal age, time since maternal HIV diagnosis, and HIV RNA viral load at first pregnancy visit (Table 2, Model 1). After adjustment, prenatal TDF use was associated with decreased likelihood of preterm birth compared to prenatal use of ART that did not contain non-TDF (aPRR = 0.34, 95% CI, 0.13–0.85, P = .02).

Table 2.

Association of Adverse Perinatal Outcomes and Prenatal Tenofovir Disoproxil Fumarate Use Among HIV-Infected Women who Used 3-Drug Antiretroviral Therapy During Pregnancy^a,b

Perinatal outcome	Univariate		Multivariatec
			Model 1e		Model 2f		Model 3g		Model 4h		Model 5i
	PRR (crude) (95% CI)	P d	Adj PRR (95% CI)	P d	Adj PRRd (95% CI)	P d	Adj PRR (95% CI)	P d	Adj PRR (95% CI)	P d	Adj OR (95% CI)	P d
Pregnancy lossj
Any	1.63 (0.66–4.04)	.3	1.19 (0.35–4.08)	.8	1.19 (0.26–4.89)	.9	1.20 (0.32–4.54)	.8	1.33 (0.33–5.30)	.7	1.05 (0.75–1.46)	.8
<20 weeks	1.49 (0.54–4.12)	.4	1.11 (0.25–4.88)	.9	1.19 (0.35–4.08)	.8	1.08 (0.24–4.89)	.9	1.15 (0.24–5.43)	.9	1.02 (0.73–1.40)	.9
≥20 weeks	2.79 (0.32–24.63)	.4	1.75 (0.41–7.82)	.4	0.95 (0.45–4.07)	.8	1.71 (0.39–7.37)	.5	2.02 (0.50–8.19)	.3	1.04 (0.95–1.13)	.4
Neonatal deathk	0.82 (0.19–3.57)	.8	0.68 (0.17–2.78)	.6	0.55 (0.17–1.77)	.3	0.68 (0.17–2.81)	.6	0.71 (0.20–2.56)	.6	1.01 (0.96–1.06)	.7
Preterm birthk, l	0.55 (0.27–1.14)	.1	0.34 (0.13–0.85)	.02	0.37 (0.15–0.89)	.03	0.33 (0.13–0.94)	.02	0.41 (0.16–1.06)	.07	0.85 (0.74–1.02)	.075

Abbreviations: CI, confidence interval; TDF, tenofovir disoproxil fumarate; PRR, prevalence rate ratio; Adj PRR, adjusted prevalence rate ratio; Adj OR, adjusted odds ratio.

^aPregnancies occurring among HIV-infected women with documented use of any 3-drug ART during pregnancy.

^bMaternal TDF use defined as documented use of TDF-containing 3-drug ART during pregnancy.

^cMultivariate models adjusted for study cohort, maternal age, time since HIV diagnosis, and HIV RNA (log₁₀ copies/mL) at first pregnancy visit.

^d P value for log-binomial generalized estimating equations (GEE) models.

^eModel 1 is a complete case analysis of 422 pregnancies that met inclusion criteria and had birth outcome information available.

^fModel 2 includes 451 pregnancies that met inclusion criteria and assumes 29 pregnancies with missing outcome information had the respective adverse perinatal outcome.

^gModel 3 includes 451 pregnancies that met inclusion criteria and assumes 29 pregnancies with missing outcome information did not have the respective adverse perinatal outcome.

^hModel 4 includes 451 pregnancies that met inclusion criteria and used multiple imputation to predict the respective adverse perinatal outcomes for 29 pregnancies missing outcome data.

ⁱModel 5 includes weighted marginal structural models for each respective adverse perinatal outcome adjusted for study cohort, maternal age, time since HIV diagnosis, and HIV RNA (log₁₀ copies/mL) at first pregnancy visit, and year pregnancy occurred; weights were truncated at the 1st and 99th percentiles.

^jAmong pregnancies in which 3-drug ART was initiated before pregnancy or during the first trimester.

^kAmong live births.

^lPreterm birth defined as gestational age <37 weeks at birth.

Sensitivity Analyses

The frequency of adverse perinatal outcomes was similar between women using TDF in combination with 3TC and EFV, 3TC and NVP, and other drugs (Table 3). In total, 371 pregnancies had exposure to either TDF- or AZT-containing 3-drug ART that did not contain both TDF and AZT, and did not contain protease inhibitors. Among those 371 pregnancies, 169 (46%) had TDF exposure and 202 (54%) had AZT exposure; 184/371 (50%) had first trimester ART exposure. There was no difference in the frequency of loss at <20 weeks between pregnancies with first trimester exposure to TDF-containing versus AZT-containing 3-drug ART (11% vs 8%, respectively, P = .6); no losses ≥20 weeks occurred among pregnancies with first trimester exposure to AZT-containing ART and 4 occurred among those with exposure to TDF-containing ART (P = .3). Among pregnancies with live births, we did not detect differences between exposure to TDF-containing or AZT-containing ART in the frequency of neonatal death (2% vs 2%, P = .9) or preterm birth (6% vs 10%, P = .2).

Table 3.

Distribution of Adverse Pregnancy Outcomes Among HIV-Infected Women who Used 3-Drug ART During Pregnancy, by Regimen

Perinatal outcome	TDF-Containing 3-Drug ART n (%)				Non-TDF-Containing 3-Drug ART n (%)
	Any	TDF+		Othera	Any	AZT+			d4T+3TC+NVP	Otherb
		3TC+EFV	3TC+NVP			3TC+NVP	3TC+EFV	3TC+LPV/r
	(n = 208)	(n = 167)	(n = 31)	(n = 10)	(n = 214)	(n = 146)	(n = 20)	(n = 19)	(n = 19)	(n = 10)
Pregnancy lossc
Anyc	17 (14%)	13 (13%)	4 (18%)	0 (0%)	7 (9%)	5 (8%)	0 (0%)	0 (0%)	1 (8%)	1 (20%)
<20 weeksc	13 (11%)	9 (10%)	4 (18%)	0 (0%)	6 (7%)	5 (8%)	0 (0%)	0 (0%)	1 (8%)	0 (0%)
≥20 weeksc	4 (2%)	4 (4%)	0 (0%)	0 (0%)	1 (1%)	0 (0%)	0 (0%)	0 (0%)	0 (0%)	1 (20%)
Neonatal deathd	3 (1%)	3 (2%)	0 (0%)	0 (0%)	4 (2%)	3 (2%)	0 (0%)	0 (0%)	1 (5%)	0 (0%)
Preterm birthd,e	10 (6%)	9 (6%)	1 (4%)	0 (0%)	20 (10%)	14 (10%)	2 (11%)	0 (0%)	2 (12%)	2 (22%)

Abbreviations: TDF, tenofovir disoproxil fumarate; 3TC, lamivudine; EFV, efavirenz; NVP, nevirapine; AZT, zidovudine; LPV/r, lopinavir/ritonavir; d4T, stavudine.

^aOther TDF-containing regimens include TDF+FTC+EFV (n = 2), AZT+TDF+3TC+EFV (n = 2), 3TC+AZT+TDF (n = 1), TDF+3TC+atazanavir (n = 1), TDF+3TC+LPV/r (n = 1), AZT+TDF+3TC+EFV (n = 1), AZT+TDF+3TC+NVP (n = 1), and TDF+NVP+emtricitabine (n = 1).

^bOther non-TDF-containing regimens include d4T+3TC+EFV (n = 4), d4T+3TC+LPV/r (n = 1), 3TC+atazanavir+abacavir (n = 1), AZT+d4T+NVP (n = 1), EFV+3TC+NVP (n = 1), AZT+NVP+EFV (n = 1), and AZT+3TC+NVP+EFV (n = 1)³. Fisher’s exact P value compares the frequency of pregnancy loss, neonatal death, and preterm birth between pregnancies with exposure to any TDF-containing 3-drug ART during pregnancy and pregnancies without any TDF-containing 3-drug ART during pregnancy.

^cAmong pregnancies in which 3-drug ART was initiated before pregnancy or during the first trimester (n = 204).

^dAmong live births (n = 380).

^ePreterm birth defined as gestational age <37 weeks at birth.

In addition to the 422 included in the primary analysis, 29 pregnancies were missing perinatal outcome data but met all other inclusion criteria; 24/29 (83%) of these pregnancies had exposure to prenatal TDF use. Results from separate multivariate GEE models that included all 451 pregnancies and accounted for missing outcomes data by assuming all 29 pregnancies did or did not have each respective adverse perinatal outcome were similar to our primary results (Table 2, Models 2 and 3). Additionally, results from multiple imputation and marginal structural models were similar to our primary results (Table 2, Models 4 and 5).

DISCUSSION

In this prospective analysis of pregnancies among HIV-infected Kenyan and Ugandan women who used 3-drug ART during pregnancy, exposure to prenatal TDF use was not associated with adverse perinatal outcomes. Compared to pregnancies with first trimester exposure to 3-drug ART regimens that did not contain TDF, we found no difference in the frequency of pregnancy loss among pregnancies with TDF exposure in the first trimester. In addition, we found no association between exposure to prenatal TDF use and neonatal death. Pregnancies with prenatal TDF exposure were less likely to have preterm birth than pregnancies with exposure to ART that did not contain TDF. Our findings support the growing evidence that prenatal TDF use is not associated with adverse perinatal outcomes, and contribute to the few prospective studies evaluating the safety of TDF use during pregnancy from African cohorts.

Our finding that TDF-containing ART use during the first trimester was not associated with pregnancy loss <20 weeks compared to non-TDF-containing ART use is novel as previous analyses have not reported on the association of prenatal TDF use and pregnancy loss at earlier than 20 weeks (excluding induced abortions) among HIV-infected women. The multicountry Development of AntiRetroviral Therapy in Africa (DART) trial found no difference in the frequency of pregnancy losses <22 weeks between HIV-infected women not on TDF-containing ART versus those on TDF-containing ART, though spontaneous and induced abortions were combined (62/137 pregnancies ending <22 weeks were induced abortions in DART) [8]. Two analyses that used data from HIV-uninfected African women enrolled in PrEP efficacy clinical trials who became pregnant while on PrEP and had in utero PrEP exposure for approximately 5 weeks found that risk of early pregnancy loss was not higher among women exposed to TDF-containing PrEP compared to placebo [18, 19].

Most pregnant women in sub-Saharan African settings present for their first antenatal care visit during the second or third trimester [20, 21], and therefore antenatal care safety studies that enroll women from antenatal care or abstract antenatal care records likely miss early pregnancy losses. Among pregnancy losses that were not induced, the frequency of pregnancy losses occurring >20 weeks (21%) in our study may be higher than observed in settings without regular pregnancy testing, though it is similar to other studies among HIV-infected African women such as the DART trial in which 26% of noninduced pregnancy losses occurred at >22 weeks [8]. Future safety studies that enroll women prior to or early in pregnancy and evaluate longer in utero TDF exposure could be helpful for assessing the association of TDF use and early pregnancy loss and other adverse neonatal outcomes. Similar to other studies among African cohorts that compared stillbirth in HIV-infected women receiving TDF ART or non-TDF ART [8, 10, 11], we found no differences in pregnancy loss at ≥20 weeks between pregnancies with and without TDF exposure.

Our finding that prenatal TDF-containing ART use among HIV-infected mothers was not associated with increased frequency of preterm birth or neonatal death compared to non-TDF-containing ART is similar to some other prospective studies in African cohorts, but not all [8, 10, 11]. In one of the largest studies to demonstrate an increase in adverse events, the PROMISE study, TDF-based ART was associated with higher rates of very preterm birth <34 weeks gestation (6.0% vs 2.6%, P = .04) and infant death within the first 14 days of life (4.4% vs 0.6%, P = .001), relative to AZT-based ART [9]. Data have suggested a possible pharmacokinetic interaction between lopinavir-ritonavir and TDF, which may explain the PROMISE results [22, 23]. In contrast to the PROMISE study, our study had infrequent protease-inhibitor use among TDF-users (33% vs 3%) [22, 23] When we excluded women who used a protease-inhibitor containing ART regimen, we did not detect differences in the frequency of preterm birth or neonatal death between pregnancies that had exposure to either TDF- or AZT-containing 3-drug ART.

Our study has limitations. We adjusted for several characteristics to account for differences between women who used TDF-containing and non-TDF-containing ART; however, because women were not randomized to ART regimens it is possible that residual confounding exists. The Partners PrEP Study and the Partners Demonstration Project did not collect obstetric history on HIV-infected women, including past adverse perinatal outcomes in previous pregnancies, and therefore we could not include obstetric history as potential covariates. The overall frequency of preterm birth in our study (8%) was lower than similar studies from Botswana (21%) [11] and South Africa (20%) [10]. Unlike previous observational studies evaluating TDF safety, all pregnant women included in our analysis were enrolled in longitudinal studies, regularly received sexually transmitted disease screening and treatment and other clinical services, and were encouraged to seek prenatal care at study visits, which may account for the lower frequency of preterm birth. We also relied on self-report of perinatal outcomes and last menstrual period to determine gestational age, which can be unreliable and may potentially lead to underreporting of preterm birth, though previous studies evaluating TDF safety have used this approach [8, 11]. Our relatively small sample size may have limited our power to detect statistical differences, though our sample size is comparable to previous studies examining the relationship of between prenatal TDF use and perinatal outcomes. We also adjusted for characteristics that differed between TDF-users and non-users that may have further decreased our statistical power in multivariate models. Larger population-based surveillance studies in African settings with adequate power to detect differences in rare outcomes like neonatal death will be important as TDF-based regimens continue to be rolled out for HIV treatment and prevention. Information on birth weight, birth length, and infant growth outcomes were not captured for HIV-infected mothers enrolled in the parent studies. Few studies have assessed the effects of prolonged prenatal TDF use on postnatal infant growth and bone health [24–28], with only 2 in African populations [8, 29], and these had mixed results. Additional longitudinal studies that evaluate longer-term postnatal outcomes of TDF use during pregnancy, including infant and child cognitive growth, are needed.

Our findings are reassuring and complement the growing body of literature indicating that TDF use during pregnancy is not associated with adverse perinatal outcomes compared to non-TDF-containing ART regimens. More specifically, our study contributes to the very limited data available on safety of TDF use during early pregnancy and did not find an association with first trimester TDF exposure and pregnancy loss at <20 weeks. Further research on longer-term effects of maternal prenatal TDF use is important given the majority of HIV-infected women are prescribed a TDF-containing PMTCT regimen. In addition, data on the safety of TDF as PrEP are needed from HIV-negative pregnant women, given PrEP scale-up in settings where fertility and HIV acquisition rates in pregnant women are high.

Notes

Authors’ contributions. J. M. B. and R. A. H. conceived of the research question; J. P. conducted the data analysis. J. P. and R. A. H. drafted the paper. N. M., C. C., and E. B. assisted in refining the research question and analysis. E. W., R. A. H., N. M., K. N., and E. B. assisted in data collection. All authors contributed to editing the text and approved the final version.

Acknowledgments. We thank the couples who participated in this study for their motivation and dedication and the referral partners, community advisory groups, institutions, and communities that supported this work.

Partners Demonstration Project Team Coordinating Center (University of Washington) and collaborating investigators (Harvard Medical School, Johns Hopkins University, Massachusetts General Hospital): Jared Baeten (protocol chair), Connie Celum (protocol co-chair), Renee Heffron (project director), Deborah Donnell (statistician), Ruanne Barnabas, Jessica Haberer, Harald Haugen, Craig Hendrix, Lara Kidoguchi, Mark Marzinke, Susan Morrison, Jennifer Morton, Norma Ware, Monique Wyatt. Project sites: Kabwohe, Uganda (Kabwohe Clinical Research Centre): Stephen Asiimwe, Edna Tindimwebwa Kampala, Uganda (Makerere University): Elly Katabira, Nulu Bulya Kisumu, Kenya (Kenya Medical Research Institute): Elizabeth Bukusi, Josephine Odoyo Thika, Kenya (Kenya Medical Research Institute, University of Washington): Nelly Rwamba Mugo, Kenneth Ngure. Data Management was provided by DF/Net Research, Inc. (Seattle, WA). PrEP medication was donated by Gilead Sciences.

Partners PrEP Study Team University of Washington Coordinating Center and Central Laboratories: Connie Celum (principal investigator, protocol co-chair), Jared M. Baeten (medical director, protocol co-chair), Deborah Donnell (protocol statistician), Robert W. Coombs, Lisa Frenkel, Craig W. Hendrix, Jairam Lingappa, M. Juliana McElrath. Study sites and site principal investigators: Eldoret, Kenya (Moi University, Indiana University): Kenneth Fife, Edwin Were; Kabwohe, Uganda (Kabwohe Clinical Research Center): Elioda Tumwesigye; Jinja, Uganda (Makerere University, University of Washington): Patrick Ndase, Elly Katabira; Kampala, Uganda (Makerere University): Elly Katabira, Allan Ronald; Kisumu, Kenya (Kenya Medical Research Institute, University of California San Francisco): Elizabeth Bukusi, Craig Cohen; Mbale, Uganda (The AIDS Support Organization, CDC-Uganda): Jonathan Wangisi, James Campbell, Jordan Tappero; Nairobi, Kenya (University of Nairobi, University of Washington): James Kiarie, Carey Farquhar, Grace John-Stewart; Kenya Medical Research Institute, Nairobi, Kenya: Nelly Rwamba Mugo; Tororo, Uganda (CDC-Uganda, The AIDS Support Organization): James Campbell, Jordan Tappero, Jonathan Wangisi.

Data management was provided by DF/Net Research, Inc. (Seattle) and site laboratory oversight was provided by Contract Laboratory Services (University of the Witwatersrand, Johannesburg, South Africa).

Financial support. This study was funded through grants from the National Institutes of Health (T32AI07140 and F32NR017125 to J. P., R00HD076679 to R. H., K24HD054314 to G. J. S.) and through assistance from the University of Washington Center for AIDS Research (NIH grant P30 AI27757) and the University of Washington’s Global Center for Integrated Health of Women Adolescents and Children. The Partners Demonstration Project was funded by the US National Institutes of Mental Health (R01 MH095507), the Bill and Melinda Gates Foundation (grants OPP47674, OPP1056051), and the US Agency for International Development (contract AID-OAA-A-12-00023). The Partners PrEP Study was funded by the Bill and Melinda Gates Foundation (grant OPP47674).

Potential conflicts of interest. All authors: No reported conflicts of interest. All authors have submitted the ICMJE Form for Disclosure of Potential Conflicts of Interest. Conflicts that the editors consider relevant to the content of the manuscript have been disclosed.