Background

Although new infections of human immunodeficiency virus (HIV) show a descending trend in recent years, the number of people living with HIV has been rising year by year. More than 16 million (95 % CI: 14.8 million, 17.4 million) female adults had been infected with HIV by the end of 2012 [1]. There is a risk for mother-to-child transmission of HIV. Among HIV infected women who took the highly active antiretroviral therapy (HAART), studies reported that the mother-to-child transmission (MTCT) rate ranged from 1 to 5 % [2], and it was around 10 % among women who did not [3, 4]. There is also a possibility that maternal HIV infection has severe impacts on pregnancy outcomes. It has been reported that HIV infected women are more likely to encounter adverse pregnancy outcomes, such as low birth weight (LBW) and preterm delivery (PTD) [5]. And it is suggested that LBW and PTD are important risk factors for post-neonatal mortality and morbidity and other adverse events including neurodevelopmental problems [6, 7].

Studies have provided inconsistent results for the association between maternal HIV infection and LBW/PTD. Some studies suggested that maternal HIV infection could increase the risk of LBW and PTD [8–10], but others reported no significant association between them [11–13]. Brocklehurst et al. [5] summarized the study results published between 1983 and 1996. They found the association of maternal HIV infection with adverse pregnancy outcomes such as LBW and PTD, but failed to assess the effect of antiretroviral drugs on it. There have been many new studies on the association between maternal HIV infection and adverse pregnancy outcomes reported in the past 20 years. For this reason, we conducted a meta-analysis to provide an update on the relationship between maternal HIV infection and LBW/PTD. In addition, there has been a rapid progress of medication and heath care for HIV infected women, and we assessed the influence of such a progress on the relationship.

Methods

Search strategy

We searched English databases (EMBASE, MEDLINE via PubMed and Web of Science) and Chinese databases (China National Knowledge Infrastructure, Wan Fang database, Sino Med and VIP) up to May 31, 2015 for studies reporting infant outcomes among women exposed and unexposed to HIV infection during pregnancy. Articles were searched using the following search terms: “maternal HIV infection” “HIV infected mothers/HIV infected pregnant women” combined with “pregnant outcomes/birth outcomes” “low birth weight (LBW)” “prematurity/preterm birth/preterm delivery (PTD)”. Only cohort studies were included.

Study selection

The relevant articles were manually selected based on the following criteria: (1) Types of studies: Prospective or retrospective cohort studies; unsystematic observations (case series or case reports) were excluded from all analyses. Because cohort studies demonstrate strong intensity compared with other observational studies in terms of causality. (2) Types of participants: Women with HIV infection during pregnancy. (3) Types of comparators: Women with no HIV infection during pregnancy. (4) Types of pregnancy outcomes: LBW (defined as birth weight <2500 g) and PTD (defined as gestational age <37 weeks). (5) All the articles reported or allowed the calculation of odd ratios (OR) with corresponding 95 % confidence intervals (95 % CIs). In order to standardize the results, all measures of effect were expressed as OR and 95 % CIs. (6) Published in English or Chinese, and full text was available. When the same cohort population was used in several papers, only the most recent ones were included in the current analysis, i.e., the same study data were only used once in the study.

Assessment of study quality and data extraction

We adapted the Newcastle-Ottawa Scale suggested by Zeng et al. [14] to measure study quality. Scores were awarded for each criterion met, and the maximum score was 9. Studies were graded high quality if the score was no less than 7 and low quality if the score was under 7. Abstracts and conference papers were graded 0 for quality and were included in the low quality category for sensitivity analyses.

Two researchers selected literature in accordance with the inclusion and exclusion criteria, and extracted data after independent verification. The following information was extracted from each study: subject of the literature, the author, year of publication,location of the study population, number of women in the two groups, odds ratio and its 95 % CIs for LBW and PTD associated with maternal HIV infection, exposure to maternal use of antiretroviral drugs before or during pregnancy.

Statistical analysis

Stata12.0 software was used to analyze the data. For both LBW and PTD, if no significant heterogeneity for their associations with maternal HIV exposure among the studies included was found (P > 0.1, I2 < 50 %), a fixed effect model was used to calculate ORs and their 95 % CIs. Otherwise,a random effects model was used. The sensitivity of data was analyzed by using the method which reduced the biggest weight of literature. Funnel plot and the Egger’s linear regression method were used to assess publication bias.

No ethics statement to declare.

Results

Articles included in the Meta-analysis and study select process

A total of 1148 articles were identified. After removing duplicate publications and based on our inclusion and exclusion criteria, 52 articles were included in our analysis (Table 1, Fig. 1) and of them, 43 were related to LBW and 40 were related to PTD. They were all published in English. More than 60 % of the articles were conducted in Africa (n = 24) and America (n = 11), and 8 were from Europe (n = 2) and Asia (n = 6).

Table 1 Characteristics of the included studies
Full size table
Fig. 1
figure 1

Flow chart of study selection process

Full size image

The prevalence of infants born before 37 weeks’ gestation ranged from 5.2 to 73.0 % in HIV infected women and 2.2 to 32 % in HIV uninfected women. The prevalence of LBW among HIV infected women ranged from 3.4 to 56.0 % and 2.5 to 36.9 % in HIV uninfected women. Odds ratios for the association of HIV exposure ranged from 0.65 to 6.27 with LBW and ranged from 0.42 to 3.68 with PTD.

Impact of maternal HIV infection on LBW and PTD

There was significant heterogeneity among studies for maternal HIV infection associated with LBW/PTD (I2 = 71.7 %, P < 0.05, and I2 = 51.8 %, P < 0.05 for LBW and PTD, respectively) suggesting that the summary measures need to be interpreted with caution. Pooled ORs for LBW and PTD from random effect models were presented in Figs. 2 and 3, respectively. The summary OR was 1.73 (95 % CI: 1.64, 1.82, P < 0.001) for LBW and 1.56 (95 % CI: 1.49, 1.63) for PTD, indicating that HIV infected women had approximately 2-fold risk to deliver low birth weight or preterm babies compared with uninfected ones.

Fig. 2
figure 2

Forest plot of odds ratio for low birth weight. Pooled OR = 1.73, 95 % CI: 1.64, 1.82; Heterogeneity chi-squared = 148.65 (d.f. = 42), p = 0.000, I-squared (variation in ES attributable to heterogeneity) = 71.7 %; Test of ES = 1: z = 20.06, p = 0.000

Full size image
Fig. 3
figure 3

Forest plot of odds ratio for preterm delivery. Pooled OR = 1.56, 95 % CI: 1.49, 1.63; Heterogeneity chi-squared = 80.95 (d.f. = 39), p = 0.000, I-squared (variation in ES attributable to heterogeneity) = 51.8 %; Test of ES = 1: z = 19.79, p = 0.000

Full size image

Test for publication bias

Begg’s and Egger’s Tests suggested that there were significant publication biases among the 43 studies of LBW (Begg’s Test: z = 2.15, P = 0.03; Egger’s Test: t = 3.53, P = 0.001) (Additional file 1). After adjusted by using the Trim and Filling method, the pooled OR was 1.77 (95 % CI: 1.57, 2.00), which was near to the original OR (1.73, 95 % CI: 1.64, 1.82). No significant publication bias was detected for the 40 articles of PTD (Begg’s Test: z = 0.79, P = 0.43; Egger’s Test: t = 1.69, P = 0.10) (Additional file 2).

Sensitivity analysis

After removing the article that counted the biggest weight, there was no significant change of the results. The pooled ORs were 1.98 (95 % CI: 1.76, 2.23) for LBW and 1.62 (95 % CI: 1.49, 1.77) for PTD, respectively. After stratified by location of study population, the significant associations between maternal HIV infection and LBW or PTD were still found in the studies from Africa, America and Asia, except those from Europe. Studies from Africa (OR = 2.18, 95 % CI: 1.84, 2.59), America (OR = 1.80, 95 % CI: 1.49, 2.19) and Asia (OR = 1.99, 95 %CI: 1.05, 3.74) showed significant pooled ORs for LBW, and for PTD the corresponding pooled ORs were 1.56 (95 % CI: 1.40, 1.75), 1.69 (95 % CI: 1.53, 1.88) and 1.60 (95 % CI: 1.10, 2.35), respectively. These associations were all significant in the two groups of developing and developed countries. No significant difference in the relationship was detected across study periods and their pooled ORs were all statistically significant for both LBW and PTD. But studies during 2005–2015 were at slightly higher risk for both LBW and PTD compared with the other two groups (Table 2).

Table 2 Results of subgroup analysis
Full size table

We also conducted a subgroup analysis based on the information of antiretroviral drugs (ARVs) usage. ARVs usage before or during pregnancy did not decrease the risks of both LBW and PTD. Women who took ARVs were at similar risk of delivering low birth weight infants (OR = 2.04, 95 % CI: 1.61, 2.57) compared with those did not (OR = 1.75, 95 % CI: 1.42, 2.17). For PTD, the pooled ORs were 1.77 (95 CI: 1.55, 2.02) for women who took drugs, 1.54 (95 % CI: 1.23, 1.92) for those who did not took drug and 1.55 (95 % CI: 1.32, 1.82) for those who provided no information about AVRs usage (Table 2).

Discussion

It has been previously reported that compared with unexposed children, children who encountered intrauterine HIV exposure are more vulnerable to stunting, underweight and wasting, and their birth weight, height and head circumference are generally lower than their unexposed counterparts [15]. However, it was controversial for the effect of maternal HIV infection on both LBW and PTD. Brocklehurst et al. [5] reported that the increasing risks of LBW and PTD were associated with maternal HIV infection. And they found that women in developing countries had higher risks of both LBW and PTD than those in developed countries. Our study, which included a number of newer studies in the meta-analysis also indicated that maternal HIV infection increased the risks of both LBW and PTD. We found that these associations were not affected by different study periods or antiretroviral drugs usage. In addition, we did not find marked difference in these associations between developing and developed countries.

The HIV associated LBW and PTD might be related to the damaged human immune system, especially the reduced CD4+ T cells and immunosuppression. Previous studies documented that women with CD4 cell counts <350 cells/mm3 had an increased risk of having LBW infants (RR = 1.57; 95 % CI: 1.16, 2.12) compared to women with higher CD4 cell counts [16, 17]. There is a possibility that women are immunocompromised during pregnancy [18], and if accompanied with HIV infection, disease progress might be accelerated. Simultaneously, reproductive tract infections, which are contributed to the incidence of adverse pregnancy outcomes, would occur more frequently due to the immunosuppression [19]. Some studies have reported that HIV-1 can replicate in the placenta [20], and it has also been shown that HIV-1 infection may alter the cytokine profile in the placenta [21, 22]. This may affect the function of placenta during pregnancy, and then restrict the development of fetal, which might be another incentive of LBW and PTD.

After stratified by antiretroviral drugs use before or during pregnancy, we found that HAART or other regimens of antiretroviral therapy (ART) had no obvious effect on the associations between maternal HIV infection and LBW/PTD. It is suggested that intrauterine ARVs exposure did not decrease or increase the risk of LBW or PTD in HIV infected women. And this is consistent with the findings reported by van der Merwe et al. and Townsend et al. [3, 4]. However, Papp et al. [23] suggested that Protease Inhibitor (PI)-based ART could increase the risk of adverse pregnancy outcomes mainly due to lower level of progesterone, which was significantly associated with fetal weight. Sibude et al. [24] also found that ARVs and, particularly, with the initiation of ritonavir-boosted PI therapy during pregnancy were correlated with PTD in HIV infected women. Though we did not found such effect of ARVs on the association between maternal HIV infection and LBW/PTD, no details of information for AVRs and therapy regimens may contribute to this. We found that studies during 2005–2015 were at slightly higher risk for both LBW and PTD compared with the other two groups, it might be related with the increasing use of combination ART earlier in pregnancy, and more use of PI for ART in recent years. Kourtis et al. [25] had found that use of PI in ART may increase the risk of PTD compared with ART without PI. And different initiation time of ART have different influences on the risk of PTD. Townsend et al. [26] also indicated that HARRT was associated with PTD. We have ever tried to analyze the data by the information of ART regimens, initiation time, ARVs duration, etc., however, there were not enough information to support the analysis of the data after stratification, for the information of ART were not provided or not detailed described in all articles. Thus we could not know if there any influences of different ART regimens or different initiation time on the association between maternal HIV infection and LBW/PTD. On the other hand, ARVs might be responsible for adverse pregnancy outcomes such as LBW and PTD, but its high effectiveness in the prevention for mother to child transmission outweighed its risk of LBW/PTD suggested by Santini et al. [27].

Maternal HIV infection was found to be related with both LBW and PTD in some places such as Africa and America. Moreover, no difference in the relationship was detected among the different study periods. This, to a certain degree, reflected the non-ideal treatment and health care development in reducing the risk of LBW and PTD for HIV infected women. So we look forward to finding new regimens or methods combined with other measurements like perinatal health care for pregnant women to decrease the risk of adverse pregnancy outcomes.

There are several limitations to note. Our meta-analysis dataset was not complete due to some restrictions of accessing to full texts. Another limitation is that we could not determine if maternal HIV infection had direct effects on LBW/PTD. We could not rule out the possibility that the association is secondary to HIV-associated deficiency, for example maternal nutrition deficiency, or other risk factors of adverse pregnancy, such as drug use, smoking, etc. And we could not get the association between maternal HIV infection and LBW/PTD after controlling for different clinical stages of HIV infection in this analysis because of the absence of such information. The relationship between maternal HIV infection and LBW/PTD is likely affected by the clinical stage of HIV infection. Study conducted by Coley et al. in Tanzania [28] indicated that although HIV infected asymptomatic women did not have a higher risk of having LBW infants compared with uninfected women (OR = 1.25; 95 % CI: 0.88, 1.79), symptomatic HIV-infected women who were in Stage 2 or higher according to the WHO staging system had about 2-time higher risks for low birth weight (RR = 2.29; 95 % CI:1.34, 3.92) and prematurity (RR = 1.93; 95 % CI: 1.35, 2.77) compared with HIV-uninfected ones. This is consistent with the finding of Bucceri et al. [11] and Muhangi et al. [29]. Publication bias is always a concern for systematic reviews. Publication bias in this meta-analysis was analyzed by using funnel plot and the Begg’s and Egger’s test methods. Significant publication bias was detected among studies of LBW and proper adjustment was conducted. And we found the results of this analysis had good stability by sensitivity analysis.

Conclusions

In conclusion, the findings of this study indicated that maternal HIV infection was associated with increased risks of both low birth weight and preterm delivery. However, the associations were moderate. No difference in the relationships between maternal HIV infection and adverse pregnancy outcomes was detected among different study periods. ARVs were not found to decrease risks of either low birth weight or preterm delivery associated with maternal HIV exposure.