Preterm delivery risk in women initiating antiretroviral therapy to prevent HIV mother-to-child transmission




The aim of the study was to describe the relationship between preterm delivery (PTD; < 37 weeks of gestation) and antiretroviral therapy in a single-centre cohort of pregnant women with HIV infection.


A retrospective analysis of data for 331 women who received care in a dedicated HIV antenatal clinic between 1996 and 2010 was carried out. Data on first CD4 cell count and viral load (HIV-1 RNA copies/mL) recorded in pregnancy, class and timing of antiretroviral therapy, gestational age at delivery, and risk factors for and causes of PTD were available from a clinical database.


Overall, 13.0% of deliveries were preterm, of which 53% were severe preterm (< 34 weeks of gestation). The lowest rate of PTD was observed in women treated with zidovudine monotherapy (6.2%). Higher rates of PTD were observed in women starting combination antiretroviral therapy (cART) in pregnancy compared with women conceiving while on cART [odds ratio (OR) 2.52; 95% confidence interval (CI) 1.22–5.20; P = 0.011]. Of the women who were eligible for zidovudine monotherapy on the basis of CD4 counts and HIV viral load but who were treated with short-term cART to prevent HIV mother-to-child transmission, 28.6% delivered preterm. Women on short-term cART remained at the highest risk of PTD compared with zidovudine monotherapy in multivariate analysis (OR 5.00; 95% CI 1.49–16.79; P = 0.015).


The causes of PTD are multiple and poorly understood. The timing of initiation and type of antiretroviral therapy administered during pregnancy appear to contribute to PTD risk. Understanding this association should improve the safety of antiretroviral therapy in pregnancy without increasing the risk of transmission.


Whether exposure to antiretroviral therapy (ART) during pregnancy is associated with preterm delivery (PTD; gestational age < 37 weeks) has been a contentious issue. The association was first reported in a Swiss cohort (1996−1998) [1] and has since been supported by the analysis of observational data from a large pan-European cohort [2, 3] as well as several independent studies from European centres [4-6]. In data from North America such an association is less clear. In the Women and Infants Transmission Study, in which the overall PTD rate was 18%, initiation of ART after 25 weeks of gestation was associated with a lower rate of PTD if regimens contained the nucleoside reverse transcriptase inhibitor (NRTI) zidovudine [odds ratio (OR) 0.5; 95% confidence interval (CI) 0.3–0.8] and a higher rate of PTD if regimens were zidovudine-sparing (OR 7.9; 95% CI 1.44–44.6) [7]. In their multivariate analysis, the Pediatric Spectrum of HIV Disease study found PTD to be associated with no ART and also with protease inhibitor (PI)-based triple therapy [combination ART (cART) with at least three drugs], symptomatic HIV disease, Black race and maternal illicit drug use [8]. The authors of a meta-analysis of studies mostly conducted prior to 2000 concluded that ART was not associated with overall risk of PTD but did report a statistically significant, 1.7-fold relative risk of PTD for cART started pre-conception or in the first trimester [9]. The largest surveillance study to date (n = 33148; HIV-infected individuals, n = 9504) has demonstrated a 1.2-fold increased risk of PTD for cART started pre-conception compared with all other groups and a 1.4-fold increased risk for cART initiated in pregnancy compared with zidovudine [10]. Two large randomized controlled trials in Africa, designed to evaluate ART to prevent HIV mother-to-child transmission during pregnancy and exclusive breast feeding in women with CD4 lymphocyte counts > 200 cells/μl, also reported PTD rates. The Mma Bana study showed a 2-fold higher PTD rate with PI-based cART vs. triple NRTI (21% vs. 11%, respectively; P = 0.003) in women starting ART from 26 weeks onwards [11]. The Kesho Bora study showed no difference comparing PI-based cART vs. zidovudine monotherapy (ZDVm) (13% vs. 11%, respectively; P = 0.39); however, in this study treatment was not initiated until 34–36 weeks, and was thus less likely to impact on gestational age at delivery [12].

The question is not necessarily which data are correct but why they differ. The answer may lie in the subjects or the therapy. Not all types of ART may be associated with PTD, with a higher risk suggested for PI-based cART [4] and some data showing an association with ritonavir boosting [13]. However, biases are inevitable; for example, at some centres PI-based cART may be preferred for women with more advanced HIV infection. The timing of initiation of therapy in relation to the pregnancy may also be important. Both pre-conception cART [3, 10] and initiation of therapy during pregnancy [5] have been associated with the PTD risk. However, particularly where, or when, guidelines for initiating cART outside of pregnancy had a lower threshold than now, the contribution of advanced HIV infection per se to these risks is difficult to measure.

In the present study, we have attempted to reduce the potential biases by focusing on women who, according to the concurrent British HIV Association guidelines (CD4 cell count > 200 cells/μL until 2008 and > 350 cells/μL thereafter [14-17]; HIV viral load < 10 000 HIV-1 RNA copies/mL plasma), were eligible for treatment with ZDVm [in combination with a pre-labour Caesarean section (PLCS)] to reduce the risk of HIV mother-to-child transmission but chose to initiate a short-term course of cART as they preferred the option of a vaginal delivery.


Three hundred and thirty-one HIV-positive pregnant women were managed by a single, multidisciplinary team at St Mary's Hospital from 1996 to the end of July 2010. Anonymized linked routine clinical data pertaining to the mother, including classical risk factors for PTD, and the delivery were prospectively collected and entered into a password-protected database. The database was interrogated for all deliveries up to the end of July 2010 for ethnicity, parity, multiple pregnancies, Centers for Disease Control and Prevention (CDC) disease stage, immunovirological parameters, injecting drug use, history of assisted reproduction, type of ART, timing of ART, birth outcome and gestational age. The use of anonymized data collected during routine clinical care to audit outcomes is a recognized element of clinical governance. This was confirmed by the host institute Joint Research Office and by an independent local research ethics committee.

Data were analysed in spss version 19.0 for Macintosh (SPSS, Chicago, IL). Continuous variables were analysed using the t-test and categorical variables using the χ2 test. ORs were calculated for univariate associations with PTD. Bonferoni adjustment was made for multiple comparisons; a P-value < 0.0125 was deemed significant. Factors associated with PTD with a P-value of < 0.1 were included in a multivariate logistic regression using a backward logistic regression method.


Of the 331 women, 78% were Black African and 94% had become HIV-infected through heterosexual intercourse. The median gestational age at first antenatal visit was 13 weeks + 3 days. The median age at delivery was 32.2 years and the median parity was 1; these variables did not differ by gestational age at delivery. Only 11.5% of women gave a history of any AIDS-defining illness. At baseline (the first antenatal visit during this pregnancy), the median CD4 lymphocyte count was 360 cells/μL [interquartile range (IQR) 227–510 cells/μL] and the median viral load was 1466 (IQR 49–13 436) copies/mL plasma, and 123 women (37.2%) had a viral load < 50 HIV copies/mL plasma.

Eight women (2.4%) took no therapy prior to labour, 65 (19.6%) were treated with ZDVm, seven (2.1%) were treated with two nucleoside reverse transcriptase inhibitors (NRTIs) only and 131 (39.6%) conceived while on cART and continued it during the pregnancy. One hundred and twenty women (36.2%) initiated cART during the pregnancy either for their own health (n = 56, including two women who did so prior to first attending the antenatal clinic), in which case they continued post-partum (new continuous cART), or solely to prevent mother-to-child transmission of HIV [with either a PI or nevirapine as the third drug (short-term cART; n = 59) or with a fixed dose combination of three NRTIs only (abacavir, lamivudine and ZDV; n = 5)], in which case they discontinued therapy after delivery. The crude PTD rates are presented in Table 1.

Table 1. Preterm delivery rates according to treatment type
TreatmentnPTD%95% CIMedian values at first antenatal clinic visit
CD4 count (cells/μL)HIV RNA (copies/mL plasma)
  1. ART, antiretroviral therapy; cART, combination antiretroviral therapy (not including triple NRTI); CI, confidence interval; dual NRTI, two nucleoside reverse transcriptase inhibitors; PTD, preterm delivery; triple NRTI, three nucleoside reverse transcriptase inhibitors; ZDVm, zidovudine monotherapy.
Dual NRTI70015024689
Triple NRTI5005511041
Short-term cART591525.414.0–36.935010200
Pre-conception cART131139.94.7–15.1395< 50
New continuous cART561017.912.0–35.015023430

There were five twin pregnancies, of which four delivered preterm; two of these women were taking ZDVm, one was on short-term cART (behave CD4 count 390 cells/μL; viral load > 10 000 copies/mL) and two were on cART started pre-conception. One of these mothers conceived through intra-cytoplasmic sperm injection, the only case of assisted reproduction in the cohort. In women for whom data on smoking status were available, 13% were current smokers. There was no difference in PTD by smoking history (current smokers compared with nonsmokers: OR 1.44; 95% CI 0.32–6.57; P = 0.632). Only three women in the cohort had a history of recreational drug use – all delivered at term.

There were six (1.8%) intra-uterine deaths (IUDs) at gestational ages 27 weeks + 2 days to 40 weeks + 5 days, all in women on cART. Four were in women on new continuous cART, one in a woman on cART started pre-conception and one in a woman on short-term cART. There were two second-trimester miscarriages (both at 22 weeks), one in a woman not receiving therapy and one in a woman on new continuous cART. There were 328 live-born infants. Forty-three women delivered before the completion of the 36th week of gestation (13.0%); of these, 23 (53%) delivered prior to 34 weeks of gestation (range 26 weeks to 33 weeks + 6 days) and 11 (26%) delivered prior to 32 weeks of gestation (range 26 weeks + 4 days to 31 weeks + 4 days).

The majority of cases of PTD were accounted for by preterm spontaneous rupture of membranes (15) or labour (6) or pre-eclampsia (10). Of the remainder, early delivery was attributable to maternal health (four women), antepartum haemorrhage (two women), fetal conditions (intrauterine growth restriction and hydrops fetalis) (two women) and uncomplicated twin pregnancy (two women). In two cases, PLCS was planned for term but performed early, pre-labour, pre-rupture of membranes, as part of the individualized prevention of mother-to-child transmission (PMTCT) strategy for women with unsuppressed HIV infection.

In univariate analysis, baseline HIV viral load in this pregnancy and category of treatment but not maternal age, ethnicity, parity, disease status or first CD4 lymphocyte count recorded in pregnancy were associated with PTD. The OR for PTD with any cART exposure (pre-conception, long-term new continuous or short-term cART) in pregnancy compared with NRTI-only therapy (ZDVm, dual NRTI or triple NRTI) was 3.33 (95% CI 1.15–9.66; p 0.02).

Next, the relationship between the timing of initiation of therapy and PTD was explored. Comparing only cART initiated in pregnancy (new continuous and short-term cART) with NRTI-only therapy, the risk of PTD was increased (OR 5.07; 95% CI 1.69−15.20; P = 0.002). Initiating any cART during pregnancy (short-term or long-term) was also associated with a significantly increased risk of PTD compared with cART commenced prior to conception (OR 2.52; 95% CI 1.22–5.20; P = 0.011). Evaluating PTD risk in women who fulfilled the criteria for long-term cART (CD4 cell count < 200 cells/μL pre-2008 and < 350 cells/μL post-2008), the difference comparing women starting new continuous cART and women who had commenced cART pre-conception was not significant (OR 1.97; 95% CI 0.81−4.82; P = 0.13).

The high rates of PTD observed in women who, having CD4 T-lymphocyte counts above the recommended level for starting long-term cART, thus initiated cART only to prevent mother-to-child transmission (25.4% with short-term cART compared with 6.2% with ZDVm; OR 5.20; 95% CI 1.62–16.74; P = 0.003) suggest that the type of therapy, rather than the indication for therapy, was important. We therefore identified the subgroup of 28 women who would have been eligible for ZDVm, by concurrent British HIV Association (BHIVA) guidelines, but chose short-term cART and the option of a vaginal delivery. Eight women (28.6%) delivered preterm between 28 weeks + 2 days and 36 weeks + 3 days of gestation. Compared with ZDVm prescribed according to the same criteria (n = 53), this risk was significant in univariate analysis (OR 4.90; 95% CI 1.33–18.13; P = 0.011).

Given that the concurrent guidelines recommended starting ZDVm at 28 weeks, the timing of initiation of therapy was examined and no difference was observed between short-term cART (mean 28.6 weeks) and ZDVm (mean 29.2 weeks; p 0.49). Differences in baseline antenatal CD4 T-lymphocyte count (mean 425 vs. 466 cells/μL, respectively; P = 0.31) and baseline plasma HIV viral load (mean 3712 vs. 2528 copies/mL, respectively; P = 0.09) in the short-term cART group compared with the ZDVm group did not reach statistical significance.

Finally, the question of whether individual components of cART may be implicated in PTD was addressed. PTD occurred in 16 of 137 women treated with nonnucleoside reverse transcriptase inhibitor (NNRTI)-based cART (11.7%) and in 22 of 96 women treated with PI-based cART (22.9%; P = 0.04). Furthermore, no difference in PTD rate was seen when nevirapine-based cART was compared with PI-based cART by indication for therapy. However, of the 28 patients who were treated with short-term cART but were eligible for ZDVm, 22 received PI-based cART. All eight PTD events occurred in these 22 women (short-term PI therapy vs. ZDVm: OR 8.71; 95% CI 2.30–33.06; P < 0.0001).

In the multivariate analysis (adjusted for pregnancy baseline plasma HIV viral load, maternal age, parity, ethnicity and pregnancy baseline CD4 lymphocyte count; see Table 2), only use of short-term cART compared with ZDVm (OR 5.00; 95% CI 1.49–16.79; P = 0.015) was associated with PTD risk.

Table 2. Multivariate analysis evaluating risk factors for preterm delivery
 OR95% CIP-value
  1. Baseline refers to first antenatal clinic attendance in this pregnancy.
  2. cART, combination antiretroviral therapy; ln, natural logarithm; OR, odds ratio; ZDVm, zidovudine monotherapy.
Age (years)0.970.86−1.100.62
Black African10.700.63−183.060.11
HIV RNA at baseline (ln copies/mL)0.660.40−1.100.11
CD4 count at baseline (cells/μL)1.001.00−1.010.35
Short-term cART vs. ZDVm5.001.49−16.790.015


In the absence of a randomized controlled study powered to evaluate PTD as the primary outcome, a detailed analysis of women eligible for ZDVm but treated with short-term cART for PMTCT reduces confounding. For example, in this clinical setting short-term cART and ZDVm were both commenced at the same mean gestational age (29 weeks). This analysis confirms an association between cART and PTD in our cohort and indicates an association of PTD with PI-based regimens initiated for PMTCT only. However, the risk of PTD was not limited to this healthier group, being also seen in patients with lower CD4 counts initiating long-term therapy for maternal health (17.9%). The overall rate of PTD seen in our unselected HIV-infected cohort (13%), which is similar to the national figure (13.3%) from the National Survey of HIV in Pregnancy and Childhood [6], is higher than expected for the general population (7.7%) [18]. Of women experiencing PTD, 44% were induced for maternal or fetal complications, which is similar to data from the Agence Nationale de Recherche sur le SIDA (ANRS) French Perinatal Cohort, in whom 38% of PTDs were induced [13].

The highest risk of PTD in association with cART was observed with cART initiated during pregnancy (25%). This mirrors the finding of some groups that late use of cART is associated with a 3−7-fold increased risk of PTD [7, 19]. Conversely, other groups have demonstrated that pre-conception cART increases PTD risk [3, 10, 20, 21]. It is plausible that initiation of treatment during pregnancy induces an immune activation phenomenon with modulation of the cytokine environment at the fetal−placental unit, reducing the threshold for PTD. Early work in this area has demonstrated reduced maternal plasma anti-inflammatory interleukin (IL)-10 concentrations in women receiving ART compared with no therapy and NRTI-only treatment [22, 23]. Given that the mean gestation at initiation of short-term cART was 29 weeks, it is possible that earlier treatment, even in women with good CD4 counts and low viral load, might reduce the risk of PTD if there is a key time when cytokine changes might trigger this event or increase a predisposition.

We have considered the potential for bias in this study and find no evidence that women receiving ZDVm were at lower risk of PTD prior to commencing ART in our cohort: of the five twin pregnancies, two were managed with ZDVm plus PLCS and only one with PI-based short-term cART; of four women receiving cervical sutures, a clear indication of PTD risk, one was treated with ZDVm, two were on cART at conception and one initiated cART during the pregnancy in accordance with CD4 count and viral load criteria. Conversely, each woman is counselled regarding her treatment options and if she were a higher risk candidate for PTD it might be argued that she would be more likely to initiate a non-PI regime, thus leading to an underestimation of the risk of PTD associated with PI-based cART. Nor is there evidence that our cohort had a high risk of PTD in general: the median gestational age at booking was 13 weeks, we identified only three women with a history of recreational drug use in the cohort, all of whom delivered at term, and current tobacco use was reported by 13% of women, with no difference in rates of PTD comparing smokers and nonsmokers. PTDs associated with IUD were not included in the analysis. However, the frequency of IUD (six of 331; 1.8%) is higher than the 5.1/1000 reported in the UK for 2008, and further investigation of both IUD and late second trimester miscarriage rates in relation to HIV infection and ART is warranted.

More detailed analysis of larger cohorts is imperative to address the relationship between ART, other risk factors and clinical aspects of PTD. This is becoming increasing important with the rollout of cART in resource-limited settings where intensive care facilities for premature babies are not generally available. The difficulty of accurately diagnosing gestational age in resource-limited settings may mask this problem and surrogates such as low birth weight, stillbirth or early infant mortality rates may need to be considered.

We do not advocate any change in recommendations to initiate the treatment of pregnant women with low CD4 counts with cART, nor to avoid the use of cART to prevent mother-to-child transmission when maternal viral load is high and especially in settings where caesarean section may not be a safe option. However, the mechanisms and comparative risks of PTD and severe PTD associated with different treatment strategies and treatment timings need urgent attention, ideally within randomized trials. In summary, despite HIV treatment, rates of PTD overall remain higher than in the general population. Timing and type of treatment are driven by many factors, including knowledge of HIV status, maternal health, CD4 count and viral load and patient preference. Some studies find an association between pre-conception cART and PTD, which suggests an impact of either HIV therapy or prior or persisting immunosuppression, or both. In this study, we have examined for the first time an additional risk of PTD among women with good CD4 counts, initiating cART for the prevention of HIV mother-to-child transmission at the end of the second trimester. While women choosing ZDVm plus PLCS had a risk of PTD no greater than that of the general population, a high risk of PTD was observed in the subgroup who chose cART and a spontaneous vaginal delivery. We postulate that this risk might be a result of cytokine switching and that this merits further investigation. Factors involved in the association between PTD and pre-conception cART are likely to be different.