In this large, long-running study of HIV-infected pregnant women, we have documented fluctuations in vaginal delivery and CS delivery rates over time. During 2005-2007,, a third of women delivered vaginally, half by elective CS and the remainder by emergency CS. In contrast, at the start of the HAART era, two-thirds of women delivered by elective CS. We document geographical variation in mode of delivery in the HAART era, with an increasing proportion of vaginal deliveries, mainly in the UK, Belgium and the Netherlands. In multivariable analysis of MTCT risk among MCPs with maternal HIV RNA <400 copies/mL, elective CS was associated with an 80% decreased MTCT risk. However, among women with viral loads <50 copies/mL there were only two transmissions overall.
Although clinical trials are the gold standard for clinical care, observational studies often provide the initial evidence for trial inception and design. Use of elective CS as a PMTCT intervention is a case in point: the ECS first published results showing an association between reduced MTCT risk and elective CS in 1994 [5
], with subsequent confirmation from a large meta-analysis [9
]. Our finding here, that the highest implementation of an elective CS policy had already been reached in 1999, when the mode of delivery trial was published [8
], is probably largely explained by participating clinicians changing their practices before the trial results were released as a result of the observational evidence that they helped to provide; furthermore, a number of women were concomitantly enrolled in both the trial and the ECS. The somewhat paradoxical finding of a declining elective CS rate in the years immediately following the trial publication may be partly explained by the concurrent implementation of antenatal HAART instead of mono- or dual therapy for PMTCT, when the first studies suggesting the benefit of HAART for decreasing MTCT risk were published [24
] and guidelines started to change. In the Netherlands, for instance the national guideline in 2000 only mentioned an elective CS as a rescue therapy in case of HAART failure or refusal [28
Other European studies have also documented declining elective CS rates in the HAART era. In an analysis from the French Perinatal Study involving over 5000 pregnant women receiving antenatal ART and delivering between 1997 and 2004, the elective CS rate declined from 56% in 2000 to 41% in 2004 [4
]. In the UK and Ireland National Study of HIV in Pregnancy and Childhood (NSHPC), the elective CS rate peaked in 1999 at 66%, declining to around 50% in 2006. The emergency CS rate we report here was relatively stable but high and ranged from 15-17% in the HAART era; the French Perinatal Study also reported stable emergency CS rates between 1997 and 2004, but higher at around 29% [4
]. In contrast, the NSHPC documented an increasing emergency CS rate, from 17% in 1999 to 23% in 2006 in the UK and Ireland, concurrent with updated guidelines recommending a vaginal delivery for women on HAART with viral suppression [29
]; most emergency CS deliveries were in women delivering at term, consistent with the likelihood that these women opted for a vaginal delivery but delivered by CS due to intrapartum complications [29
]. The lower emergency CS rate in our centres in Italy and Spain compared to that in Belgium, the Netherlands and the UK may be largely explained by the greater proportion of women opting for vaginal deliveries in the latter.
A prominent factor associated with likelihood of an elective CS was geographic location. In our adjusted analysis, women delivering in Belgium, the Netherlands or UK were 93% less likely to have an elective CS compared with women living in Italy or Spain by 2003-2007. Geographic differences may be explained by differences in national guidelines [13
,52] and may also reflect variation in the elective CS rate in the general population. The association between antenatal ART and mode of delivery strengthened over time: in 1998-2002, women on mono- or dual therapy were 1.6 times more likely to deliver by elective CS than women on HAART, increasing to 2.5 times by 2003-2007. Although women with a last HIV RNA viral load in pregnancy <50 copies/ml were significantly less likely to have an elective CS in the group delivering between 1998 and 2003, this was not the case in the more recent time period. This might be due to the fact that the policy to perform an elective CS was very region bound and that more CS that were intentionally prophylactic became an emergency CS because of changed guidelines with respect to the week of the planned CS (37-37+6 weeks instead of 36-36+6 weeks) [15
Prematurity is a well-defined risk factor for MTCT [2
] and some studies have suggested that infants born to HIV-infected mothers may be particularly susceptible to intrapartum acquisition of infection [32
]. In our analysis among MCPs with viral loads <400 copies/mL, infants born before 34 weeks had an eight-fold increased risk of HIV infection compared with term infants. Among premature infants, both elective CS and emergency CS were effective in reducing MTCT risk (independent of maternal CD4 count and ART), whilst in term infants, emergency CS was not associated with a significantly lower MTCT risk. Associations between prematurity and HAART use have been reported in several studies, mainly in Europe, with prematurity rates in cohorts of HIV-infected women reported up to 34% [33
]. A recent risk-benefit analysis using UK data indicated that the risk-benefit ratio associated with exclusive HAART (vs zidovudine monotherapy) was an estimated 0.59 premature infants for each infection averted [38
]. It is clear that the relationships between preterm delivery, HAART use and MTCT are complex, and the role that mode of delivery may play in these requires further research.
We found that elective CS was an effective PMTCT intervention among nearly 1000 women with viral load <400 copies/mL, with an 80% decreased risk, independent of HAART use and gestational age. This extends our previous finding, whereby elective CS was associated with a 93% decreased MTCT risk in 560 women with undetectable viral loads (around half of whom were tested with less sensitive assays than those currently used) [12
]. A decision regarding mode of delivery has to be made before labour starts on the basis of the instituted antiretroviral treatment and the last measured HIV-RNA viral load. Emergency CS can be the result of a woman with a planned elective CS starting labour earlier than the planned date or the consequence of a complication during a planned vaginal delivery. Here, emergency CS was associated with an 80% decreased risk of MTCT among infants born to women with viral loads <400 copies/mL, although only with borderline statistical significance. We also described MTCT rates by mode of delivery, reclassified as prophylactic CS and an attempted vaginal delivery to reflect intended delivery. The possibility exists that some conditions potentially favourable for MTCT like placental abruption, IUGR and infection of the lower genital tract were also included in the “started vaginally” group. Most HAART-using women here with a known HIV-RNA load in the last month of pregnancy had undetectable levels (<50 copies/mL) and virtually all had levels <500 copies/mL. In our analysis comparing MTCT rates among women delivering by prophylactic CS and those starting vaginally, we carried out a sub-analysis restricted to women with undetectable viral loads because it could be that a prophylactic CS is mainly performed in cases where there is a perceived high risk of MTCT (i.e. confounding by indication): there were no transmissions among women with viral loads <50 copies/mL starting labour vaginally at ≥34 weeks and a MTCT rate of 0.4% for those having a prophylactic CS.
Our findings suggest a protective effect of elective CS even at low maternal viral loads, but when the HIV-RNA load is <50 copies/mL we were insufficiently powered to draw any conclusions about the benefit of intended elective CS or the risk of intended vaginal delivery in this group of patients, who can achieve MTCT rates below 0.5% [1
]. The effectiveness of elective CS in PMTCT is just one of the factors requiring consideration in decision-making around mode of delivery; the potential risks of CS also need consideration as CS, particularly in HIV-infected women, may cause maternal morbidity in the short-term [20
] and in subsequent pregnancies [40
]. A further factor to consider is that delivery may not take place as planned: recent studies have shown that between 38% and 55% of women opting for a vaginal delivery have actually delivered by CS, for a variety of reasons [1
A limitation of our study is the lack of data on what the planned mode of delivery was. We thus could not address the issue of … We were not able to investigate the likely timing of transmission in infected infants ..
In conclusion, we show that implementation of obstetric interventions for PMTCT are not only influenced by evidence-based medicine but also by “opinion-based” medicine. Our data highlight the effectiveness of antenatal HAART in PMTCT, which has resulted in a very small number of infections in recent years and has contributed to a declining elective CS rate overall. The numbers needed to treat (i.e. the number of elective CS deliveries) to prevent a single transmission will be high taking into account the results of the present and other studies [1
]. Cohort collaborations and risk benefit analyses are needed to address the important question of whether elective CS has any additional benefit for women reaching the end of their pregnancy with undetectable viral loads, and to explore the role and feasibility of CS after labour and/or rupture of membranes in preventing MTCT among specific groups, including premature infants.