Elective cesarean section for women living with HIV: a systematic review of risks and benefits

Supplemental Digital Content is available in the text


Introduction
Cesarean section (c-section) before labor and before rupture of membranes [elective c-section (ECS)] has been suggested as an intervention to prevent vertical transmission of HIV in high-income settings where training and resources exist to conduct c-sections safely [1]. The decision to offer ECS to women living with HIV must consider a range of potential risks as well as benefits for both the mother and the child. These risks and benefits vary depending on the underlying risk of vertical transmission of HIV during delivery, which is associated with disease stage and antiretroviral treatment (ART) use [2], as well as on the underlying risks of ECS compared with vaginal delivery for both mother and child, which is associated with the local capacity and skills to perform c-sections and treat potential complications [3]. Unfortunately, many women in low-income and middleincome countries (LMICs), in particular, lack access to high-quality obstetric services, a critical concern in the context of rising c-section rates globally [3]. Furthermore, women living with HIV may experience higher rates of some obstetric complications compared with HIV-uninfected women [4].
In 2005, Read and Newell published a Cochrane systematic review, which identified one clinical trial and five observational studies evaluating the safety of ECS versus vaginal delivery among HIV-1-infected women [5]. Taken together, these studies indicated that ECS can substantially reduce the risk of mother to child HIV transmission, whereas it also resulted in slightly higher rates of postpartum maternal morbidity, such morbidity was generally rated as minor [5]. The authors concluded that in general, the benefit of ECS outweighs the risks, but the risk-to-benefit ratio depends upon the underlying rate of vertical HIV transmission [5].
There were several limitations to the data available at the time of the Read and Newell review [5] as well as to its interpretation and applicability 12 years later. The single trial included only HIV-1-infected women taking no ART during pregnancy or taking only zidovudine. In addition to this, HIV infection, no infant outcomes were measured in any of the included studies. Furthermore, all studies were conducted in high-income countries in Europe or North America, where ECS is a relatively safe procedure. The vast majority of women living with HIV live in sub-Saharan Africa and other LMIC settings, where higher rates of morbidity and mortality may be ascribed to the c-section surgery itself. Since the single trial, published in 1999 with data collected in the mid-1990s, ART use has expanded greatly worldwide, and more effective regimens have been developed. In 2015, the WHO recommended offering immediate ART to all individuals living with HIV [2]. These actions should significantly reduce vertical HIV transmission.
Women living with HIV have the right to the most up-todate knowledge about risks and benefits of sexual and reproductive health decisions they will make, with the support of their healthcare providers [6]. To inform WHO recommendations on the sexual and reproductive health and rights of women living with HIV, we sought to update the Read and Newell review [5] to consider the current existing evidence on ECS for women living with HIV globally.

Methods
This systematic review and meta-analysis followed Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guidelines [7,8] to answer the question: does ECS in women living with HIV result in better maternal and perinatal outcomes than other modes of delivery?
Inclusion/exclusion criteria To be included in the review, an article had to present primary research comparing outcomes of ECS to other modes of delivery (e.g. non-ECS, vaginal delivery) among women living with HIV and their children; measure any of the following outcomes: morbidity and mortality among women [e.g. febrile morbidity, endometritis, hemorrhage or severe anemia, pneumonia, urinary tract infections (UTIs)], HIV infection in infants (efficacy of prevention of vertical transmission), other morbidity and mortality among infants (e.g., respiratory morbidity and skin lacerations), or breastfeeding (success or timing of initiation and continuation); and be published in a peer-reviewed journal prior to the search date of 1 October 2015. Analytic epidemiologic studies, both observational (case-control and cohort studies) and interventional (clinical trials), were included; ecological and historical-control studies were not. Mode of delivery had to be explicitly described. Studies from any geographical location including any women living with HIV of childbearing age were eligible for inclusion. Studies published in all languages were eligible for inclusion.
For this review, we defined ECS as a c-section conducted before start of labor and before rupture of membranes. However, we included any study that used the term ECS, without requiring further definition by study authors. We similarly accepted author-provided definitions for all outcomes.

Search strategy
We searched four electronic databases: PubMed, CINAHL, Embase, and the Cochrane Central Register of Controlled Trials (CENTRAL). For each online database, we used the following search strategy: (HIV OR AIDS) AND ('mode of delivery' or 'cesarean section' or 'cesarean section' or 'c-section'). We conducted secondary reference searching on all included studies and the previous Read and Newell review [5].
Titles, abstracts, citation information, and descriptor terms of citations identified through the search strategy were screened by a member of the study staff. When a citation was considered relevant or when title/abstract was deemed insufficient for inclusion/exclusion decision, the full-texts were retrieved and evaluated. Two reviewers (independently and in duplicate) assessed all full-text articles for eligibility to determine final study selection. Differences were resolved through consensus.
Data extraction and quality assessment Data were extracted by two reviewers using standardized forms. Differences in data extraction were resolved through discussion and referral to a senior study team member when necessary. The following information was gathered from each included study: (1) Study description: Study objectives; year(s); location (country/city); setting (population-based, hospital, clinic); study design; sample size; recruitment and allocation methods; follow-up periods; loss to follow-up (2) Population characteristics: Age, socioeconomic status; HIV disease stage; CD4 þ cell count; VL; ART status/ regimen; comorbidities (e.g., diabetes); obstetric characteristics (3) Intervention: Mode of delivery; method of determination (e.g., medical records, survey self-report) (4) Outcomes: Analytic approach; outcome measures and definitions (including both maternal and neonatal outcomes); comparison groups; effect sizes; confidence intervals (CIs); significance levels Authors were contacted for additional clarification if information in published articles was insufficient.
For randomized controlled trials (RCTs), risk of bias was assessed using the Cochrane Collaboration's tool [9]. This tool assesses random sequence generation (selection bias), allocation concealment (selection bias), blinding of participants and personnel (performance bias), blinding of outcome assessment (detection bias), incomplete outcome data addressed (attrition bias), and selective reporting (reporting bias). Methodological components of the studies were classified as high or low risk of bias. For observational studies using different designs, we adapted the Newcastle-Ottawa scale to consider measures of study quality [10].

Analysis
We examined results of ECS compared with both vaginal delivery and all other modes of delivery (non-ECS, forceps-assisted or vacuum-assisted delivery, etc.). Although vaginal delivery is the main comparison of interest, in observational studies, this comparison excludes women with medical indications for emergency or non-ECS, potentially biasing results. We therefore also present data for all other modes of delivery.
Where multiple studies reported the same outcome among comparable populations with adequate data, meta-analysis was conducted using random-effects models to combine odds ratios (ORs) using the program Comprehensive Meta-Analysis [11]. Heterogeneity was assessed using the I 2 statistic and interpreted according to Cochrane thresholds [12], and funnel plots were created to examine the potential for publication bias. In metaanalyses, we did not combine data from trials with data from observational studies, as results were expected to differ systematically, resulting in increased heterogeneity [12]. We attempted to identify overlapping participant data across articles by contacting study authors to avoid combining articles with overlapping data in meta-analysis.
In cases of overlap, we included only the most recent or comprehensive data in meta-analysis. We conducted stratified analyses for studies conducted in the combination antiretroviral therapy (cART) era (defined as after 1996 or cART use in country). We also conducted stratified analyses of data from women who were on cART and women who had higher CD4 þ cell counts or lower VLs (defined as CD4 þ cell count > 200 cells/ml or VL < 400 RNA copies/ml). We then further stratified for women in these categories who delivered their pregnancies at term (at or >37 weeks of gestation) (i.e. cART patients delivering at term, and women with CD4 þ cell count > 200 cells/ml or VL < 400 RNA copies/ml and delivering at term). Finally, we conducted stratified analyses of data from studies conducted in LMICs, as classified by the World Bank [13].

Results
Description of included studies Figure 1 presents a study selection flowchart. The initial database search yielded 2565 records, with two records identified through other sources; 1750 remained after removing duplicates. After the initial title/abstract review, 64 articles were retained for full-text screening. Ultimately, 36 articles met the inclusion criteria and were included in the review . Seventeen were published in 2005 or later (after the cutoff date of the previous review). Results are presented below for each of the main outcomes. Funnel plots did not indicate publication bias. Heterogeneity was not substantially significant in most meta-analyses.

Maternal health outcomes
Maternal health outcomes are reported in Table 2. In the RCT, adverse maternal health outcomes were minimal [21,42]. Postpartum fever was reported by 1.1% (2/183) of women who gave birth vaginally and 6.7% (15/225) who gave birth by ECS (P ¼ 0.002). Postpartum bleeding or intravascular coagulation disease occurred in one woman in each group. Anemia of greater than moderate severity (hemoglobin < 8 g/dl) was reported in two women who gave birth vaginally and four by ECS. No further adverse events were reported at 6-week follow-up.     One observational study in Kenya examined maternal mortality [48]. Eight deaths were reported of the 405 women delivering vaginally, five of the 74 given non-ECS, and none of the 22 given ECS [48].

Infant HIV infection
By far, the most common outcome measured was infant HIV infection ( In meta-analysis of all observational studies, ECS was also associated with a decreased odds of infant HIV infection (  [17,18,33]). Examining data from cART patients delivering at term also yielded nonsignificant results (Table 3) [17].
When focusing on data stratified by CD4 þ or VL of the mother, only two studies [16,17]  conducted in high-income countries and among women who were not on current highly effective ART regimens. Altogether, data from a single RCT and multiple observational studies indicate that ECS reduces the risk of infant HIV infection in the absence of ART. However, the association between ECS and infant HIV infection was nonsignificant in most stratified analyses of studies conducted in the cARTera and among women on cART, women with higher CD4 þ cell counts or lower VLs, and women whose deliveries were at term. Limited data on other maternal outcomes and infant health outcomes do suggest increased maternal and infant morbidity associated with ECS compared with vaginal birth, as is seen with HIV-uninfected women. However, many outcomes were relatively minor or less problematic with accurate dating of pregnancy and ECS at term.
The risk-benefit ratio of ECS likely depends upon the underlying rate of vertical HIV transmission, as well as the risks of both maternal and infant morbidities and mortality associated with ECS and other modes of delivery. For women who are on ART and virally suppressed, the risk of vertical HIV transmission is relatively low. For women in high-income countries with access to quality obstetric services, the risks associated with ECS are also relatively low. However, the risk of vertical transmission increases greatly for women in the absence of effective ART while the risks of ECS increase for women without access to high-quality obstetric services. We found only three studies from sub-Saharan Africa, and whereas one study from Kenya reported on maternal mortality, none reported on maternal or infant morbidity outcomes other than HIV infection. Future studies from sub-Saharan Africa and other LMICs would help to clarify the risks and benefits in such settings and provide useful evidence for policy-makers.
The findings from this review suggest routine ECS for women living with HIV may not be appropriate; instead, individual patients and clinicians should consider the risks and benefits for specific clients, and women's autonomy to choose their mode of delivery should be respected. This is consistent with other national guidelines and recommendations from professional groups [50][51][52][53]. US and UK guidelines, while not recommending routine ECS for all women living with HIV, do recommend that clinicians consider ECS at higher VLs. The American College of Obstetricians and Gynecologists has recommended considering ECS when VL more than 1000 [51]. UK guidelines recommend ECS with VL more than 1000 and recommend considering ECS when VL ¼ 50-999, 'taking into account the actual VL, the trajectory of the VL, length of time on treatment, adherence issues, obstetric factors and the woman's views' [53]. An examination of national guidelines across 23 European countries found that 95% 'included the recommendation that HIV-positive women on successful cARTwith a very low or undetectable VL (<1000) can have a vaginal delivery' [50]. The 2015 WHO Statement on Cesarean Section Rates emphasized the need to avoid unnecessary c-sections, especially in settings that lack the facilities and/or capacity to properly conduct safe surgery and treat surgical complications, which can extend many years beyond the current delivery and affect the health of the woman, her child, and future pregnancies [3]. However, there may be specific clinical indications, such as raised VL at delivery or known ART resistance, where ECS may be further considered, highlighting the need for individual-level consideration of risks and benefits of ECS in addition to national guidelines. When c-section is medically indicated, it should be available, accessible, and safe for all women, including women living with HIV.
It is critically important to emphasize respect for women's autonomy regarding mode of delivery. In the largest survey conducted by and for women living with HIV globally, women living with HIV reported experiencing routine lack of inclusion or choice in decision-making about their own sexual and reproductive healthcare [6]. Principles of human rights must be embedded in all healthcare policies, practices, and training, and coercion of any kind is never acceptable [54].
The issue of mode of delivery for women living with HIV is important, and no systematic review has been conducted to update the evidence in the past 12 years. Our review used a broad search strategy, double data extraction, and careful assessment of study quality. However, the findings must be seen in light of several limitations. Studies that defined ECS used a definition consistent with the one used for this review; however, the minority of studies that did not clearly specify how they defined ECS may have introduced heterogeneity into the review. Few studies were available from recent years, from women on cART, and for different subgroups. Few studies were also available from LMICs where surgical skills and health system capacity are most limited; we identified only three studies from sub-Saharan Africa and none reported maternal and infant morbidity outcomes beyond HIV transmission. All but one were observational studies with their well established and inherent limitations and bias in assessing intervention effects; in the absence of randomization, providers likely directed women to ECS or other modes of delivery based on systematically different sociodemographic characteristics, clinical presentation, or staffing capabilities. The only RCT was published in 1999. In meta-analyses, we attempted to include only nonoverlapping participant data, but the complex set of overlaps across studies made this difficult, and it is possible that duplicate data were included in some analyses. Meta-analyses also often had few studies, large CIs, and sometimes considerable statistical heterogeneity. This review points to the need for further research, particularly in low-income and middle-income countries, and particularly in sub-Saharan Africa where the majority of women living with HIV reside. However, reductions in vertical transmission due to cART mean future studies must be large (and thus expensive) to identify statistically significant differences across modes of delivery. The evidence base is therefore unlikely to be significantly strengthened in the future.
In conclusion, our findings suggest that while ECS may be protective against infant HIV infection in the absence of effective ART, this effect was not statistically significant among women on cART or who are at term and virally suppressed, and there are other risks to mothers and infants associated with ECS. Risks and benefits are likely to differ across settings. Clinicians and healthcare providers should consider the risks and benefits for individual clients, and respect women's autonomy to choose their mode of delivery.