Timing of combination antiretroviral therapy (cART) initiation is not associated with stillbirth among HIV-infected pregnant women in Malawi

Abstract

Objective:

To assess the association between timing of maternal combination ART (cART) initiation and stillbirth among HIV-infected pregnant women in Malawi’s Option B+ programme.

Methods:

Cohort study of HIV-infected pregnant women delivering singleton live or stillborn babies at ≥28 weeks of gestation using routine data from maternity registers between January 1, 2012 and June 30, 2015. We defined stillbirth as death of a fetus at ≥28 weeks of gestation. We report proportions of stillbirth according to timing of maternal cART initiation (before pregnancy, 1^st or 2^nd trimester, or 3^rd trimester or labour). We used logistic regression, with robust standard errors to account for clustering of women within health facilities, to investigate the association between timing of cART initiation and stillbirth.

Results:

Of 10,558 mother-infant pairs abstracted from registers, 8,380 (79.4%) met inclusion criteria. The overall rate of stillbirth was 25 per 1,000 deliveries (95% confidence interval 22–29). We found no significant association between timing of maternal cART initiation and stillbirth. In multivariable models, older maternal age, male sex of the infant, breech vaginal delivery, delivery at <34 weeks of gestation, and experience of any maternal obstetric complication were associated with higher odds of stillbirth. Deliveries managed by a mission hospital or health centre were associated with lower odds of stillbirth.

Conclusion:

Pregnant women’s exposure to cART, regardless of time of its initiation, was not associated with increased odds of stillbirth.

INTRODUCTION

An estimated 2.6 million stillbirths, defined by WHO as a baby born with no signs of life at or after 28 weeks of gestation [1], occur every year worldwide, of which 98% are in low- and middle-income countries [2]. In sub-Saharan Africa, more than 3% of deliveries each year result in stillbirth [3]. The major causes include childbirth complications, maternal infections in pregnancy, maternal disorders (especially hypertension, obesity and diabetes), fetal growth restriction and congenital abnormalities [1,4].

Maternal antiretroviral therapy (ART) exposure may also be associated with increased risk of stillbirth [5]. By 2015, 91% of the 1.1 million women receiving antiretroviral drugs for prevention of mother-to-child transmission of HIV (PMTCT) were being offered lifelong ART as a result of the implementation of the Option B+ policy (lifelong ART for all HIV-infected pregnant and breastfeeding women regardless of CD4 cell count or clinical stage) [6]. Untreated HIV infection can increase the risk of stillbirth by about 1.7 times [7]. Exposure to ART, especially from the time of conception, may further increase the risk of stillbirth [5]. In Botswana, a study that included more than 33,000 women between 2009 and 2011 reported that 6.3% of women who conceived while receiving nevirapine-based ART had a stillbirth, vs. 4.1% of all other HIV-infected women, including those who initiated zidovudine, other ART, or no antiretroviral drugs during pregnancy, and 2.5% of HIV-uninfected women [8]. Higher risk of stillbirth has also been reported in different settings among antiretroviral-exposed women when ART was initiated prior to conception or early in pregnancy [9–13]. However, a Ugandan study in the era of Option B+ and a systematic review by Uthman et al. found no association between stillbirth and pre-conception ART or timing of ART initiation, respectively [14,15].

In this study, we used routine programme data from large health facilities in Malawi to investigate the association between timing of maternal combination ART (cART) initiation and stillbirth among HIV-infected women in the Option B+ era.

METHODS

Study setting and population

We abstracted routinely collected data from maternity registers at 20 large health facilities in central and southern Malawi. The selected facilities were among the 21 facilities that participated in our previous studies to evaluate the implementation of Option B+ in Malawi [16–22] and included 12 district hospitals, two central hospitals, three faith-based hospitals, and three large health centres. We included all women who received cART during pregnancy and delivered a singleton live birth or stillbirth at a gestational age of ≥28 weeks between January 1, 2012 and June 30, 2015. We excluded all women with missing data on age, gestational age, or timing of cART initiation. Exclusion criteria was hierarchical.

Data collection and management

In the Malawi government’s health care system, observations from labour and delivery are recorded on labour charts and summarized in the maternity register. These standard monitoring and evaluation tools capture demographic characteristics, obstetrical history, and infant outcomes. Data on HIV and ART status (for HIV-infected women) are obtained from the woman’s personal health passport, a government-issued book containing information on general history, diagnoses, treatments, antenatal consultations, and deliveries and summarised in the register. We took digital images from the paper-based maternity registers, followed by double data entry and cleaning.

Outcomes and main exposure

The main outcome was stillbirth, defined as a baby born with no sign of life at ≥28 completed weeks of gestation [1]. Stillbirth was based on gestational age at delivery, calculated by the health service providers as the difference between delivery date and the last menstrual period (LMP) self-reported by pregnant women at the first antenatal care (ANC) visit.

The main exposure was timing of maternal cART initiation (before pregnancy, 1^st or 2^nd trimester, or 3^rd trimester or labour). We defined cART as a combination of at least three antiretroviral drugs. During the study period, adult cART patients in Malawi received tenofovir/ lamivudine/ efavirenz, zidovudine/ lamivudine/ nevirapine or stavudine/ lamivudine/ nevirapine as first-line regimen, and either lopinavir or atazanavir (each boosted with small dose of ritonavir) plus tenofovir/ lamivudine or zidovudine/ lamivudine as second-line regimen [23,24].

Statistical analyses

We used Chi-square tests to test differences in proportions by timing of maternal cART initiation, and Kruskal-Wallis tests to test for differences in medians. We used logistic regression, with cluster-based robust standard errors [25] to account for clustering of women within health facilities, to model stillbirth. We calculated unadjusted and adjusted odds ratios (ORs and aORs, respectively) with 95% confidence intervals (CIs) of the associations between timing of cART initiation and stillbirth. P-values were obtained from the Wald test. We considered the following explanatory variables: infant sex, maternal complications (none, haemorrhage, obstructed/prolonged labour, other, with only the leading complication recorded in case of multiple complications), delivery setting (health facility, home, unknown), mode of delivery (spontaneous vaginal, vacuum extraction, caesarean, breech vaginal, unknown), parity (0, 1, >1), maternal age at delivery (<20, 20–34, ≥35 years), gestational age at delivery (<34, 34–36, >37 weeks), facility type (health centre, faith-based hospital, district hospital, central hospital), and location of facility (urban, rural). Explanatory variables that were significant in univariable analyses were included in the multivariable model.

All statistical tests were 2-sided, with p<0.05 considered significant. Analyses were performed using STATA version 14.1 (Stata Corporation, College Station, Texas, USA).

The Malawi National Health Sciences Research Committee and the Cantonal Ethics Committee of Bern in Switzerland granted ethical approval for the study.

RESULTS

Characteristics of women and infants by timing of cART initiation

A total of 10,558 HIV-infected mothers and their infants were recorded in the maternity registers at the 20 health facilities since the implementation of Option B+. Of these women, 643 (6.1%) were excluded because of delivery before 28 weeks or missing data on gestational age, 666 (6.3%) because they did not receive cART before or during their pregnancy, 156 (1.5%) because of delivery before January 1, 2012 or after June 30, 2015, 608 (5.8%) because of multiple gestation, and 105 (1.0%) because of missing maternal age (Figure 1).

Figure 1: Study eligibility flow chart.

^¥ Exclusion criteria was hierarchical; cART, combination antiretroviral theraphy

Of the included 8,380 HIV-infected women, 5,961 (71.1%) started cART before the current pregnancy, 1,128 (13.5%) started cART in the 1^st or 2^nd trimester of the pregnancy, and 1,291 (15.4%) started cART in the 3^rd trimester or during labour (Table 1). The median (interquartile range [IQR]) gestational age at delivery was 37 (36–38) weeks with no differences by timing of maternal cART initiation. Most women (76.4%) were between 20 and 34 years old at delivery. Most deliveries were spontaneous, vaginal (85.5%) and took place in a health facility (97.3%). Stillbirths were rare: 210 (2.5%) of the 8,380 deliveries. Of the 210 stillbirths, 108 (51.4%) were described as macerated and 102 (48.6%) as fresh.

Table 1:

Characteristics of women and infants stratified by timing of maternal combination antiretroviral therapy (cART) initiation

	Timing of cART initiation
	Before pregnancy		1st or 2nd trimester		3rd trimester or during labour		P
Number of women	5961	(71.1)	1128	(13.5)	1291	(15.4)
Maternal age (years)
<20	341	(5·7)	97	(8·6)	89	(6·9)	<0.0001
20–34	4501	(75·5)	886	(78·5)	1019	(78·9)
≥35	1119	(18·8)	145	(12·9)	183	(14·2)
Median (IQR)	29	(24–33)	27	(22–32)	28	(23–32)
Parity
0	629	(10·6)	160	(14·2)	142	(11·0)	<0.0001
1	1069	(17·9)	233	(20·7)	275	(21·3)
>1	4238	(71·1)	734	(65·1)	872	(67·5)
Missing	25	(0·4)	1	(0·1)	2	(0·2)
Median (IQR)	2	(1–4)	2	(1–3)	2	(1–3)
Gestational age (weeks)
<34	319	(5·4)	51	(4·5)	67	(5·2)	<0.0001
34–36	1682	(28·2)	385	(34·1)	515	(39·9)
≥37	3960	(66·4)	692	(61·3)	709	(54·9)
Median (IQR)	37	(36–38)	37	(36–38)	37	(36–38)
Mode of delivery
Spontaneous vaginal	5007	(84·0)	1008	(89·4)	1137	(88·1)	<0.0001
Vacuum extraction	110	(1·9)	14	(1·2)	12	(0·9)
Breech vaginal	105	(1·8)	14	(1·2)	26	(2·0)
Caesarean section	722	(12·1)	90	(8·0)	116	(9·0)
Missing	17	(0·3)	2	(0·2)
Delivery setting
Health facility	5813	(97·5)	1103	(97·8)	1239	(96·0)	<0.0001
Home	56	(0·9)	12	(1·1)	35	(2·7)
Other	84	(1·4)	13	(1·2)	14	(1·1)
Missing	8	(0·1)			3	(0·2)
Birth attendant
Trained attendant	5332	(89·4)	1065	(94·4)	1179	(91·3)	0.094
Other	114	(1·9)	25	(2·2)	38	(2·9)
Missing	515	(8·6)	38	(3·4)	74	(5·7)
Maternal obstetric complications#
None	5111	(85·7)	1007	(89·3)	1136	(88·0)	0.002
Haemorrhage	177	(3·0)	25	(2·2)	44	(3·4)
OPL	228	(3·8)	41	(3·6)	40	(3·1)
Other	414	(6·9)	49	(4·3)	63	(4·9)
Missing	31	(0·5)	6	(0·5)	8	(0·6)
Facility type
Central hospital	1233	(20·7)	155	(13·7)	237	(18·4)	<0.0001
Health centre	557	(9·3)	213	(18·9)	442	(34·2)
Mission hospital	387	(6·5)	78	(6·9)	74	(5·7)
District hospital	3784	(63·5)	682	(60·5)	538	(41·7)
Facility location
Rural	4173	(70·0)	817	(72·4)	639	(49·5)	<0.0001
Urban	1788	(30·0)	311	(27·6)	652	(50·5)
Infant sex
Male	3068	(51·5)	584	(51·8)	643	(49·8)	0.547
Female	2840	(47·6)	537	(47·6)	633	(49·0)
Unknown	53	(0·9)	7	(0·6)	15	(1·2)

Data are n (%) unless indicated otherwise. OPL, Obstructed/prolonged labour; cART, Combination antiretroviral therapy; IQR, Interquartile range.

^#Health care workers only record one leading complication

Factors associated with stillbirth

We found no significant differences in the proportion of stillbirth by timing of maternal cART initiation (Table 2). Timing of maternal cART initiation was not associated with odds of stillbirth in either the univariable or multivariable analysis. The odds of stillbirth increased with age of the woman at delivery. Women aged ≥35 years were more likely (aOR 1.49 [95% CI, 1.18–1.88]) to experience a stillbirth than women aged between 20 and 34 years. In the univariable analysis, delivery through caesarean section increased the odds of stillbirth (OR 2.51 [95% CI, 1.51–4.17]) but this association was non-significant in the multivariable analysis. Compared to spontaneous vaginal delivery, breech vaginal delivery had an almost four times higher odds (aOR 3.87 [95% CI, 1.88–8.00]) to result in a stillbirth. When compared with term delivery, a non-significant increase in the odds of a stillbirth was noted among deliveries at 34–36 weeks of gestation (aOR 1.26 [95% CI, 0.81–1.97]); the odds increased substantially in pregnancies of less than 34 weeks of gestation (aOR 5.66 [95% CI, 4.02–7.97]). Male babies had higher odds (aOR 1.21 [95% CI, 1.02–1.43]) of stillbirth than female babies. Women who had haemorrhage (aOR 5.46 [95% CI, 3.01–9.91]), obstructed or prolonged labour (aOR 4.00 [95% CI, 1.73–9.20]), or any other obstetric complication (aOR 8.52 [95% CI, 9.74–17.38]) were more likely to experience a stillbirth than those who had no complications. Deliveries managed by a mission hospital (aOR 0.63 [95% CI, 0.50–0.79]) or health centre (aOR 0.53 [95% CI, 0.30–0.95]) were less likely to result in stillbirth than those managed at a central hospital.

Table 2:

Association between timing of maternal combination antiretroviral therapy (cART) initiation and other factors with stillbirth

	Proportion with stillbirth (%)		Univariable (n = 8,380)			Multivariable (n = 8,316)
			OR	(95% CI)	P*	aOR	(95% CI)	P*
Timing of cART initiation
Before pregnancy	158/5,961	(2.7)	1.23	(0.82–1.83)	0.218	1.01	(0.68–1.49)	0.600
1st or 2nd trimester	24/1,128	(2.1)	0.98	(0.63–1.52)		0.86	(0.47–1.57)
3rd trimester or labour	28/1,291	(2.2)	1			1
Maternal age (years)
<20	7/527	(1.3)	0.55	(0.26–1.18)	<0.0001	0.50	(0.24–1.02)	0.0001
20–34	152/6,406	(2.4)	1			1
≥35	51/1,447	(3.5)	1.50	(1.19–1.90)		1.49	(1.18–1.88)
Parity
0	17/931	(1.8)	1		0.131
1	30/1,577	(1.8)	1.04	(0.54–2.00)
>1	160/5,844	(2.7)	1.51	(0.91–2.52)
Missing	3/28	(10.7)
Gestational age (weeks)
<34	55/437	(12.6)	7.57	(5.46–10.51)	<0.0001	5.66	(4.02–7.97)	<0.0001
34–36	55/2,582	(2.1)	1.15	(0.74–1.78)		1.26	(0.81–1.97)
≥37	100/5,361	(1.9)	1			1
Mode of delivery
Spontaneous vaginal	139/7,152	(1.9)	1		<0.0001	1		<0.0001
Vacuum extraction	4/136	(2.9)	1.53	(0.53–4.38)		0.92	(0.29–2.85)
Breech vaginal	20/145	(13.8)	8.07	(4.70–13.85)		3.87	(1.88–8.00)
Caesarean section	44/928	(4.7)	2.51	(1.51–4.17)		0.65	(0.31–1.37)
Missing	3/19	(15.8)
Birth attendant
Trained	190/7,576	(2.5)	1		0.252
Other	2/177	(1.1)	0.44	(0.11–1.78)
Missing	18/627	(2.9)
Maternal obstetric complications#
None	111/7,254	(1.5)	1		<0.0001	1		<0.0001
Haemorrhage	22/246	(8.9)	6.32	(3.54–11.27)		5.46	(3.01–9.91)
OPL	13/309	(4.2)	2.83	(1.46–5.48)		4.00	(1.73–9.20)
Other	62/526	(11.8)	8.60	(5.84–12.67)		9.74	(5.45–17.38)
Missing	2/45	(4.4)
Facility type
Central hospital	58/1,625	(3.8)	1		<0.0001	1		0.0006
Health centre	16/1,212	(1.3)	0.36	(0.25–0.53)		0.53	(0.30–0.95)
Mission hospital	13/539	(2.4)	0.67	(0.53–0.83)		0.63	(0.50–0.79)
District hospital	123/5,004	(2.5)	0.68	(0.56–0.83)		0.85	(0.63–1.14)
Facility location
Rural	134/5,629	(2.4)	1		0.471
Urban	76/2,751	(2.8)	117	(0.77–1.77)
Infant sex
Female	112/4,295	(2.6)	1		0.177	1		0.040
Male	94/4,010	(2.3)	1.12	(0.94–1.32)		1.21	(1.02–1.43)
Unknown	4/75	(5.3)	2.35	(0.88–6.29)		2.90	(1.07–7.88)

OR, Odds ratio; aOR, Adjusted odds ratio; CI, Confidence interval; OPL, Obstructed/prolonged labour; cART, Combination antiretroviral therapy; IQR, Interquartile range.

^#Health care workers only record one leading complication

^*P-value obtained using a Wald test

DISCUSSION

In this study of infants born to HIV-infected mothers after implementation of Option B+ in Malawi, we found a stillbirth rate of 25 per 1,000 births. This figure, though higher than the worldwide target of ≤12 stillbirths per 1000 births by 2030 set by WHO’s Every Newborn Action Plan to end preventable deaths [26], is close to the 2015 estimate of 22 stillbirths per 1,000 births (95% CI, 20–30) in the Malawian general population [27]. In settings without widespread availability of effective ART, the risk of stillbirth is generally higher [27]. In our study, stillbirths were not more common among women who initiated cART before pregnancy compared to those who started cART either during the 1^st or 2^nd trimester or during the 3^rd trimester or labour. The demographic characteristics of the study population were similar to those of pregnant women in the general population [28]. However, the percentage of women delivering in health facilities (97.3%) and the percentage of caesarean sections (11.1%) were higher than in the Malawian general population (91% and 6%, respectively) [28]. This difference may be due to several reasons. We excluded delivery data from primary health care facilities where caesarean sections do not take place. HIV-infected women may be more likely to deliver in health facilities. Furthermore, the national percentages exclude stillbirths.

Our finding that exposure to antiretroviral combinations used in this study (nucleotide/nucleoside reverse transcriptase inhibiting (N(t)RTI) and non-NRTI (NNRTI)), irrespective of timing of initiation, was not associated with increased odds of stillbirth is consistent with previous studies. A cross-sectional study in Western Uganda found that stillbirth was not more common among women who were on N(t)RTI/NNRTI-based antiretroviral combinations than in HIV-uninfected women, regardless of timing of initiation and adherence to antiretrovirals. The authors suggested that the positive effect of antiretrovirals on maternal HIV infection counterbalances their potential toxicity [14]. A study conducted in Botswana found no differences in stillbirths between pregnant women taking either tenofovir/emtricitabine/efavirenz, other triple ART regimens, or zidovudine only [29]. Despite the failure to distinguish between women who started ART before conception from those who started during pregnancy, results from a study in Malawi and Mozambique reported that a longer course of ART during pregnancy was protective against stillbirth [30].

We found a number of factors to be associated with stillbirth. The increased odds of stillbirth for older women found in our study has also been reported in systematic reviews on stillbirth [27,31]. Consistent with the study by Chi et al. [32], pregnancies lasting less than 34 weeks had an almost six times higher odds of resulting in stillbirth than term deliveries. The higher odds of stillbirth in breech vaginal deliveries was also reported in a meta-analysis that included data from clinical trials and observational studies [33]. The proportion of stillbirth was higher among deliveries managed by central hospitals, with only minor differences in proportions among deliveries managed by health centres, mission hospitals and district hospitals. In Malawi, health centres and mission hospitals offer primary healthcare and refer complicated cases to district hospitals, which in turn, depending on the degree of complexity, may refer to a central hospital. Therefore, the higher proportion of stillbirths among deliveries managed by central hospitals and lower odds of stillbirth among those managed by mission or health centre suggests a flow of delayed and complicated cases to central hospitals from surrounding health facilities. Our finding of a higher odds of stillbirth among male babies was also reported in a systematic review and meta-analysis of more than 30 million births by Mondal et al. [34]. Presence of any maternal complication during delivery was strongly correlated with stillbirth in our study, similar to findings from a systematic review by Lawn et al. [27]. Maternal co-morbidities during pregnancy may be higher among HIV-infected women on ART [35,36]. Many of these covariates are important predictors of stillbirth in the general population.

Strengths and limitations

Our study included a large sample of mother-infant pairs from a large geographical area and from diverse health facility settings, including health centres, district hospitals, mission hospitals and central hospitals. We controlled for several important factors associated with stillbirth. However, because of our reliance on routine data, we could not include some factors that may influence the association between maternal cART initiation and stillbirth, such as ART adherence, CD4 cell count or viral load. Poor ART adherence may result in drug resistance [31,32] and lower CD4 cell counts, which may further lead to adverse maternal and birth outcomes. Some studies have reported that reaching optimal ART adherence during pregnancy is challenging, particularly during the postpartum period [20,39]. However, data from the PROMOTE trial and the Kisumu Breastfeeding Study that included pregnant women regardless of clinical or immunological stage suggested high and constant levels of adherence during pregnancy and breastfeeding [40,41].

However, we had no data on other factors that may be associated with stillbirth, such as co-infections and comorbidities, distance to healthcare facility, or socio-economic status. Tuberculosis, syphilis and malaria are common among HIV-infected people and have been shown to correlate with stillbirth [2,42]. Low socio-economic status is associated with increased risk of stillbirth [43] and delayed ART initiation [44], possibly because women with low socio-economic status are less likely to make use of ANC, even if it is free.

Our study also has other limitations. First, there may have been some misclassification of stillbirth due to use of gestational age estimated from self-reported LMP. The consistency of our findings with other analyses suggests that such misclassification may be minimal. Second, we had no data on the ART regimen the women received. However, programme data report that by the end of June 2015 93% of adult ART patients were on the tenofovir/ lamivudine/ efavirenz regimen which is also recommended for Option B+, 5% on other NRTI- or NNRTI-based first-line ART regimens, and there was no use of integrase inhibitors [45,46]. Third, the use of routinely collected data may have affected data quality. Fourth, we did not have data on timing of entering ANC and hence we were not able to verify if the results were confounded by the coincidence of ANC and ART start.

Conclusion

The continued scale-up of ART programmes in most HIV-endemic countries and the introduction of universal eligibility of ART for all people living with HIV means that increasing numbers of pregnant women will be on ART throughout pregnancy. It is therefore encouraging that long exposure to cART seems not to be associated with increased odds of stillbirth. This finding is supported by the argument that ART keeps the woman healthy and prevents a number of complications that are risk factors for stillbirth. In our setting, this seems to outweigh any potential detrimental effects of antiretroviral drugs on a live birth. The strongest predictors of stillbirth were maternal obstetric complications. Analyses of routine programme data can play a relevant role in evaluating effects of ART on birth outcomes, including when new regimens are introduced on a large scale.