Effect of pregnancy versus postpartum maternal isoniazid preventive therapy on infant growth in HIV-exposed uninfected infants: a post-hoc analysis of the TB APPRISE trial

Background: Isoniazid preventive therapy (IPT) initiation during pregnancy was associated with increased incidence of adverse pregnancy outcomes in the TB APPRISE trial. Effects of in utero IPT exposure on infant growth are unknown. Methods: This post-hoc analysis used data from the TB APPRISE trial, a multicentre, double-blind, placebo-controlled trial, which randomised women to 28-week IPT starting in pregnancy (pregnancy-IPT) or postpartum week 12 (postpartum-IPT) in eight countries with high tuberculosis prevalence. Participants were enrolled between August 2014 and April 2016. Based on modified intent-to-treat analyses, we analysed only live-born babies who had at least one follow-up after birth and compared time to infant growth faltering between arms to 12 weeks and 48 weeks postpartum in overall and sex-stratified multivariable Cox proportional hazards regression. Factors adjusted in the final models include sex of infant, mother’s baseline BMI, age in years, ART regimen, viral load, CD4 count, education, and household food insecurity. Results: Among 898 HIV-exposed uninfected (HEU) infants, 447 (49.8%) were females. Infants in pregnancy-IPT had a 1.47-fold higher risk of becoming underweight by 12 weeks (aHR 1.47 [95% CI: 1.06, 2.03]) than infants in the postpartum-IPT; increased risk persisted to 48 weeks postpartum (aHR 1.34 [95% CI: 1.01, 1.78]). Maternal IPT timing was not associated with stunting or wasting. In sex-stratified analyses, male infants in the pregnancy-IPT arm experienced an increased risk of low birth weight (LBW) (aRR 2.04 [95% CI: 1.16, 3.68), preterm birth (aRR 1.81 [95% CI: 1.04, 3.21]) and becoming underweight by 12 weeks (aHR 2.02 [95% CI: 1.29, 3.18]) and 48 weeks (aHR 1.82 [95% CI: 1.23, 2.69]). Maternal IPT timing did not influence growth in female infants. Interpretation: Maternal IPT during pregnancy was associated with an increased risk of LBW, preterm birth, and becoming underweight among HEU infants, particularly male infants. These data add to prior TB APPRISE data, suggesting that IPT during pregnancy impacts infant growth, which could inform management, and warrants further examination of mechanisms. Funding: The TB APPRISE study Supported by the National Institutes of Health (NIH) (award numbers, UM1AI068632 [IMPAACT LOC], UM1AI068616 [IMPAACT SDMC], and UM1AI106716 [IMPAACT LC]) through the National Institute of Allergy and Infectious Diseases, with cofunding from the Eunice Kennedy Shriver National Institute of Child Health and Human Development (contract number, HHSN275201800001I) and the National Institute of Mental Health. This post-hoc analysis utilised data from the P1078 TB APPRISE trial – a randomised, double-blind, placebo-controlled, multicentre, non-inferiority study designed to evaluate the effect of pregnancy-IPT vs postpartum-IPT on maternal complications and composite adverse birth outcomes. The trial, as reported in detail previously,8 was conducted in eight countries (Botswana, Haiti, India, South Africa, Tanzania, Thailand, Uganda, and Zimbabwe) at 13 different sites with high TB prevalence (>60 cases per 100,000). Participants were randomised to receive a 28-week course of IPT (300 mg daily) either during pregnancy (pregnancy-IPT) or at postpartum week 12 (postpartum-IPT). The pregnancy-IPT arm received isoniazid daily for 28 weeks (initiated between 14- and 34-weeks gestation, immediately after enrolment), then switched to placebo until the 40th week postpartum. The postpartum-IPT arm initiated a placebo immediately after trial entry during pregnancy until the 12th week postpartum and then switched to isoniazid daily until the 40th week postpartum. All participants received vitamin B6 and a prenatal multivitamin from week 0 to week 40 postpartum. The randomisation was stratified by the gestational age at trial entry (≥14 weeks to <24 weeks or ≥24 weeks to ≤34 weeks) and was balanced at each site. As detailed in the parent paper,8 all women provided written informed consent and all local and collaborating institutional review boards approved it. An independent data and safety monitoring board reviewed it biannually. A proposal for these post-hoc analyses was approved by the IMPAACT operations team, and the manuscript was approved for publication by the IMPAACT publication team. The authors attested to the fidelity of the protocol and the accuracy of the analyses. This report conforms with CONSORT reporting guidelines. Pregnant WLWH, 14–34 weeks of gestation, weighing >35 kg, with >750 absolute neutrophil count cells/mm3, >7.5 g/dL haemoglobin, and >50,000 platelets count/mm3 were eligible. Participants were required to have liver enzymes (aspartate aminotransferase [AST], alanine aminotransferase [ALT], and total bilirubin) at or below 1.25 times the upper limit of the normal range within 30 days prior to study entry. Women with active TB, recent TB exposure, TB treatment for more than 30 days in the previous year, or peripheral neuropathy of grade 1 or higher were excluded. The original study included 956 participants, 477 randomised to pregnancy-IPT and 479 to postpartum-IPT arm. Participants were enrolled between August 2014 and April 2016. This analysis was restricted to HEU infants born to mothers participating in the RCT. Exclusion criteria for this analysis included lack of infant information (withdrawal from the study before birth or no live birth, or lack of any growth measurement), HIV infection of the infant, and twin births. Mother-infant pairs were followed up to 48 weeks postpartum. Weight and length of infants were measured at birth, 4th, 8th, 12th, 24th, 36th, 44th, and 48th weeks postpartum to the nearest 0.1 kg and 0.1 cm. The scales were calibrated regularly as per the manufacturer’s instructions. Shoes and outer layers of clothing were removed before weight measurements were taken. Infants’ lengths were measured with horizontal boards. The data collectors were trained and experienced in weight and length measurement. There was a two-week extension period for mothers who did not attend their last visit. Missing values at the scheduled last visit were replaced by measurements within two weeks after the end of the study. Low birth weight (LBW) was defined as less than 2.5 kg regardless of gestational age. Birth before completion of 37 weeks of pregnancy was regarded as preterm. Small for gestational age (SGA) was defined by weight less than the 10th percentile for gestational age using INTERGROWTH growth standards.14 Weight-for-age z-score [WAZ], weight-for-length z-score [WLZ], and length-for-age z-score [LAZ]) were defined using WHO child growth standards.15 Growth faltering was less than −2 Z-scores; with underweight defined as WAZ < –2, wasting WLZ < –2, and stunting LAZ < –2. Cofactors of growth faltering assessed in the analyses included: Infant sex and maternal characteristics at enrolment, including body mass index (BMI), age, ART regimen, viral non-suppression (viral load ≥40 copies/ml), CD4 count (cells/mm3), education, and household food insecurity. Household food insecurity was considered positive if the respondents answered yes to at least one of the following questions: did you experience a lack of resources to get food, have you gone to bed hungry in the last 30 days, and have you passed the entire day and night hungry? Means and standard deviations (SDs) were used to describe normally distributed continuous variables, medians and interquartile ranges (IQRs) to describe skewed distributions, and frequencies and percentages to describe categorical variables. Baseline maternal and infant characteristics were compared between pregnancy-IPT and postpartum-IPT randomisation groups using two-sided t-tests (Mann–Whitney U tests if assumptions were not met) for continuous variables and Pearson χ2 tests (Fisher's exact tests if assumptions were not met) for categorical variables. In the primary study, randomisation was carried out on pregnant women. The randomisation groups were compared in modified intent-to-treat analyses adjusted for predetermined potential confounding variables. For the measurement of adverse birth outcomes, the effects of pregnancy-IPT on LBW and preterm birth were examined in the primary trial publication; however, sex-stratified analyses of these outcomes were not conducted. We examined the effects of pregnancy-IPT on birth outcomes (LBW, preterm birth, and SGA) using generalised linear models with a Poisson family and a log link (to estimate relative risks) in overall and sex-stratified analyses. Multivariable generalised linear models were fitted to control potential confounders. For the measurement of growth faltering during infancy, mothers in the postpartum-IPT arm initiated IPT at 12 weeks after delivery, therefore data were censored at 12-weeks postpartum to examine the effect of pregnancy-IPT compared to no IPT during pregnancy and postpartum. In addition, to compare the longer-term effects of pregnancy-IPT on growth faltering, randomised arms were compared up to 48 weeks after birth. Growth faltering was compared between randomised groups using Cox proportional hazards regression models and generalised estimated equations (GEE). We used Kaplan–Meier survival analysis to compare, unadjusted, time to the first event of growth faltering (underweight, wasting, or stunting) to 12 weeks postpartum and 48 weeks postpartum in overall and sex-stratified analyses. Univariate and multivariable Cox proportional hazards regression models and models including interaction terms between randomisation arm and infants' sex were fit to compare the risk of experiencing the first episode of growth faltering between the randomised groups and any effect modification by infant sex. For these analyses, time zero was the randomisation date, and no failure (no growth faltering) was assumed prior to birth. Since participants were randomly assigned to either a treatment or a control group during pregnancy, whatever difference occurred between the two arms, such as gestational age at birth or low birth weight, was assumed to be attributable to the intervention. Infants lost to follow-up or who died prior to failure were censored at their last visit date. Growth data from visits following growth faltering were censored. We used multivariate Cox proportional hazards regression models to identify cofactors (maternal BMI, age, ART regimen, viral non-suppression, CD4 count (cells/mm3), education, and household food insecurity and infant sex) of growth faltering. As fewer than 5% of at-risk participants remained in the study after 60 post-randomisation weeks, the values at 60th week and after were censored. In multivariable models, we didn't adjust birth characteristics because adjusting for low birth weight, preterm birth, and/or SGA would underestimate the effect of the intervention on the outcome as they are in the causal pathway between pregnancy IPT exposure and long-term growth. Moreover, univariate and multivariable generalised estimated equations (GEE) were fit with a Poisson family and a log link (to estimate relative risk) and exchangeable correlation structure to compare risks of growth faltering (underweight, wasting, and stunting) at any time up until 12 weeks postpartum and up to 48 weeks, as well as testing for interaction by infants' sex and analyses stratified by sex. Infants who experienced growth faltering anytime were not censored in this analysis. The multivariate GEE model was used to identify cofactors of growth faltering in HEU infants. We also fitted multivariable linear regression to examine the long-term impact of IPT on growth (WAZ, LAZ, WLZ) at 48 weeks of infant age. We used R version 4.1.0 for analyses. The funder of the study had no role in study design, data collection, data analysis, data interpretation, or writing of the report. BAR, ASC, and GM had access to the dataset. ASC, SML, GJS, AG, and GM were responsible for the final decision on the submission of this manuscript for publication.