Background

Antiretroviral drugs reduce viral replication and can reduce mother‐to‐child transmission of HIV either by lowering plasma viral load in pregnant women or through post‐exposure prophylaxis in their newborns. In rich countries, highly active antiretroviral therapy (HAART) which usually comprises three drugs, has reduced the mother‐to‐child transmission rates to around 1‐2%, but HAART is not always available in low‐ and middle‐income countries. In these countries, various simpler and less costly antiretroviral regimens have been offered to pregnant women or to their newborn babies, or to both.

Objectives

To determine whether, and to what extent, antiretroviral regimens aimed at decreasing the risk of mother‐to‐child transmission of HIV infection achieve a clinically useful decrease in transmission risk, and what effect these interventions have on maternal and infant mortality and morbidity.

Search methods

We sought to identify all relevant studies regardless of language or publication status by searching the Cochrane HIV/AIDS Review Group Trials Register, The Cochrane Library, MEDLINE, EMBASE and AIDSearch and relevant conference abstracts. We also contacted research organizations and experts in the field for unpublished and ongoing studies. The original review search strategy was conducted in 2002 and updated in 2006 and again in 2009.

Selection criteria

Randomised controlled trials of any antiretroviral regimen aimed at decreasing the risk of mother‐to‐child transmission of HIV infection compared with placebo or no treatment, or compared with another antiretroviral regimen.

Data collection and analysis

Two authors independently selected relevant studies, extracted data and assessed trial quality. For the primary outcomes, we used survival analysis to estimate the probability of infants being infected with HIV (the observed proportion) at various specific time‐points and calculated efficacy at a specific time as the relative reduction in the proportion infected. Efficacy, at a specific time, is defined as the preventive fraction in the exposed group compared to the reference group, which is the relative reduction in the proportion infected: 1‐(Re/Rf). For those studies where efficacy and hence confidence intervals were not calculated, we calculated the approximate confidence intervals for the efficacy using recommended methods. For analysis of results that are not based on survival analyses we present the relative risk for each trial outcome based on the number randomised. No meta‐analysis was conducted as no trial assessed identical drug regimens.

Main results

Twenty‐five trials including 18,901 participants with a median trial sample size of 627 ranging from 50 to 1,844 participants were included in this update. Twenty‐two trials randomised mothers (18 pre‐natally and four in labour) and followed up their infants, and three trials randomised infants. The first trial began in April 1991 and assessed zidovudine (ZDV) versus placebo and since then, the type, dosage and duration of drugs to be compared has been modified in each subsequent trial. We present the results stratified by regimen and type of feeding.

Antiretrovirals versus placebo In breastfeeding populations, three trials found that:

ZDV given to mothers from 36 to 38 weeks gestation, during labour and for 7 days after delivery significantly reduced HIV infection at 4‐8 weeks (Efficacy 32.00%; 95% CI 1.50 to 62.50), 3 to 4 months (Efficacy 33.07%; 95% CI 5.57 to 60.57), 6 months (Efficacy 34.55%; 95% CI 9.05 to 60.05), 12 months (Efficacy 34.31%; 95% CI 9.30 to 59.32) and 18 months (Efficacy 29.74%; 95% CI 2.73 to 56.75).

ZDV given to mothers from 36 weeks gestation and during labour significantly reduced HIV infection at 4 to 8 weeks (Efficacy 43.78%; 95% CI 8.78 to 78.78) and 3 to 4 months (Efficacy 36.95%; 95% CI 2.94 to 70.96) but not at birth.

ZDV plus lamivudine (3TC) given to mothers from 36 weeks gestation, during labour and for 7 days after delivery and to babies for the first 7 days after birth (PETRA 'regimen A') significantly reduced HIV infection (Efficacy 62.75%; 95% CI 40.76 to 84.74) and a combined endpoint of HIV infection or death (Efficacy 62.75 [, ]61.00%; 95% CI 40.76 to 84.74) at 4 to 8 weeks but these effects were not sustained at 18 months.

ZDV plus 3TC given to mothers from the start of labour until 7 days after delivery and to babies for the first 7 days after birth (PETRA 'regimen B') significantly reduced HIV infection (Efficacy 41.83%; 95% CI 12.82 to 70.84) and HIV infection or death at 4 to 8 weeks (Efficacy 35.91%; 95% CI 8.41 to 63.41) but the effects were not sustained at 18 months.

ZDV plus 3TC given to mothers during labour only (PETRA 'regimen C') with no treatment to babies did not reduce the risk of HIV infection at either 4 to 8 weeks or 18 months.

In non‐breastfeeding populations, three trials found that:

ZDV given to mothers from 14 to 34 weeks gestation and during labour and to babies for the first 6 weeks after birth significantly reduced HIV infection in babies at 18 months (Efficacy 66.22%; 95% CI 33.94 to 98.50).

ZDV given to mothers from 36 weeks gestation and during labour with no treatment to babies ('Thai‐CDC regimen') significantly reduced HIV infection at 4 to 8 weeks (Efficacy 50.26%; 95% CI 13.80 to 86.72) but not at birth

ZDV given to mothers from 38 weeks gestation and during labour with no treatment to babies did not influence HIV transmission at 6 months.

Longer versus shorter regimens using the same antiretrovirals One trial in a breastfeeding population found that:

ZDV given to mothers during labour and to their babies for the first 3 days after birth compared with ZDV given to mothers from 36 weeks and during labour (similar to 'Thai‐CDC') resulted in HIV infection rates that were not significantly different at birth, 4‐8 weeks, 3 to 4 months, 6 months and 12 months.

Three trials in non‐breastfeeding populations found that:

ZDV given to mothers from 28 weeks gestation during labour and to infants for the first 3 days after birth compared with ZDV given to mothers from 35 weeks gestation through labour and to infants from birth to 6 weeks significantly reduced HIV infection rate at 6 months (Efficacy 45.35 %; 95% CI 1.39 to 89.31) but compared with the same regimen ZDV given to mothers from 28 weeks gestation through labour and to infants from birth to 6 weeks did not result in a statistically significant difference in HIV infection at 6 months. ZDV given to mothers from 35 weeks gestation during labour and to infants for the first 3 days after birth was considered ineffective for reducing transmission rates and this regimen was discontinued.

An antenatal/intrapartum course of ZDV used for a median of 76 days compared with an antenatal/intrapartum ZDV regimen used for a median 28 days with no treatment to babies in either group did not result in HIV infection rates that were significantly different at birth and at 3 to 4 months.

In a programme where mothers were routinely receiving ZDV in the third trimester of pregnancy and babies were receiving one week of ZDV therapy, a single dose of nevirapine (NVP) given to mothers in labour and to their babies soon after birth compared with a single dose of NVP given to mothers only resulted in HIV infection rates that were not significantly different at birth and 6 months. However the reduction in risk of HIV infection or death at 6 months was marginally significant (Efficacy 45.00%; 95% CI ‐4.00 to 94.00).

Antiretroviral regimens using different drugs and durations of treatment

In breastfeeding populations, three trials found that:

A single dose of NVP given to mothers at the onset of labour plus a single dose of NVP given to their babies immediately after birth ('HIVNET 012 regimen') compared with ZDV given to mothers during labour and to their babies for a week after birth resulted in lower HIV infection rates at 4‐8 weeks (Efficacy 41.00%; 95% CI 11.84 to 70.16), 3‐4 months (Efficacy 38.91%; 95% CI 11.24 to 66.58), 12 months (Efficacy 35.98 [9.25, 62.71]36.00%; 95% CI 8.56 to 63.44) and 18 months (Efficacy 39.15%; 95% CI 13.81 to 64.49). In addition, the NVP regimen significantly reduced the risk of HIV infection or death at 4‐8 weeks (Efficacy 41.74%; 95% CI 14.30 to 69.18), 3 to 4 months (Efficacy 40.00%; 95% CI 14.34 to 65.66), 12 months (Efficacy 32.17%; 95% CI 8.51 to 55.83) and 18 months (Efficacy 32.57 [9.93, 55.21]33.00%; 95% CI 9.93 to 55.21).

The 'HIVNET 012 regimen' plus ZDV given to babies for 1 week after birth compared with the 'HIVNET 012 regimen' alone did not result in a statistically significant difference in HIV infection at 4 to 8 weeks.

A single dose of NVP given to babies immediately after birth plus ZDV given to babies for 1 week after birth compared with a single dose of NVP given to babies only significantly reduced the HIV infection rate at 4 to 8 weeks (Efficacy 36.79%; 95% CI 3.57 to 70.01).

Five trials in non‐breastfeeding populations found that:

In a population in which mothers were receiving 'standard' antiretroviral for HIV infection a single dose of NVP given to mothers in labour plus a single dose of NVP given to babies immediately after birth ('HIVNET 012 regimen') compared with placebo did not result in a statistically significant difference in HIV infection rates at birth and at 4 to 8 weeks.

The 'Thai CDC regimen' compared with the 'HIVNET 012 regimen' did not result in a significant difference in HIV infection at 4 to 8 weeks.

A single dose of NVP given to babies immediately after birth compared to ZDV given to babies for the first 6 weeks after birth did not result in a significant difference in HIV infection rates at 4‐8 weeks and 3 to 4 months.

ZDV plus 3TC given to mothers in labour and for a week after delivery and to their infants for a week after birth (similar to 'PETRA regimen B') compared with NVP given to mothers in labour and immediately after delivery plus a single dose of NVP to their babies immediately after birth (similar to 'HIVNET 012 regimen') did not result in a significant difference in the HIV infection rate at 4 to 8 weeks.

An evaluation of various antiretroviral drugs given to mothers from 34 to 36 weeks and during labour with the same drugs given to their babies for 6 weeks after birth: stavudine (d4T) versus ZDV, didanosine (ddI) versus ZDV and d4T plus ddI versus ZDV did not result in statistically important differences in HIV infection rates at birth, 4 to 8 weeks, 3 to 4 months and 6 months.

TRIPLE regimens versus other

Two trials compared a regimen of three antiretrovirals given to the mother, which we refer to as TRIPLE, with other regimens.

In a breastfeeding population, a trial of TRIPLE regimen commenced at 34 weeks compared with only ZDV for the same period until labour when sdNVP was added found no babies infected with HIV at birth in either group and at 6 months post delivery there was no statistically significant difference in HIV infection between groups (Efficacy ‐84.62%, 95%CI: ‐490.35 to 321.11). The infants in the TRIPLE group did not receive any drugs while those in the ZDV group received sdNVP at birth.

In a non‐breastfeeding population, a trial compared a protease inhibitor‐based TRIPLE regimen combination of lopinavir/ritonavir, ZDV and lamivudine from 26 to 34 weeks gestation through 6 months post‐partum with a shorter regimen of ZDV from 28 to 36 weeks, then ZDV and 3TC and sdNVP at onset of labour, followed by ZDV and 3TC for one week after delivery. Infants in both groups received sdNVP within 72 hours of delivery and ZDV for one week. There was no statistically significant difference between groups in HIV infection at birth (Efficacy 18.18%, 95%CI ‐83.48 to 119.84) or at four to eight weeks (Efficacy 31.25%, 95%CI ‐29.29 to 91.79). At six months, HIV infection was higher but not statistically significantly so in the non‐TRIPLE group (Efficacy 42.35%, 95%CI ‐0.57 to 85.27). At 12 months HIV infection was statistically significantly higher in the non‐TRIPLE group (Efficacy = 42.11%, 95%CI 0.66 to 83.56). At 6 months, the HIV infection or death incidence remained higher in the non‐TRIPLE group (RR 34.13, 95%CI [‐0.29 to 68.55) and at 12 months this difference was statistically significant (RR 36.20, 95%CI 5.92 to 66.48).

TRIPLE regimen versus TRIPLE regimen

In a breastfeeding population, one trial compared two triple combination antiretroviral regimens with each other, viz abacavir, lamivudine and ZDV with lopinavir/ritonavir and ZDV and lamivudine in the mother from 26 to 34 weeks and continued for six months post‐partum. Infants in both groups received sdNVP and one month of ZDV. This trial found no significant difference in HIV infection rates at birth (Efficacy ‐189.47%; 95%CI ‐715.29 to 336.35) with incidence at six months remaining non‐significant with transmission rates being very low (< 1%).

Adverse effects

The incidence of serious or life‐threatening events was not significantly different in any of the trials included in this review.

Authors' conclusions

A regimen combining triple antiretrovirals is most effective for preventing transmission of HIV from mothers to babies. The risk of adverse events to both mother and baby appears low in the short‐term but the optimal antiretroviral combination and the optimal time to initiate this to maximise prevention efficacy without compromising the health of either mother or baby remains unclear.

Short courses of antiretroviral drugs are also effective for reducing mother‐to‐child transmission of HIV and are not associated with any safety concerns in the short‐term. ZDV given to mothers during the antenatal period, followed by ZDV+3TC intrapartum and postpartum for one week, and sd‐NVP given to infants within 72 hours of delivery and ZDV for one week, may be most effective when considering short antiretroviral courses. Where HIV‐infected women present late for delivery, post‐exposure prophylaxis with a single dose of NVP immediately after birth plus ZDV for the first 6 weeks after birth is beneficial. The long term implications of the emergence of resistant mutations following the use of these regimens, especially those containing Nevirapine, require further study.

Keywords: Female; Humans; Pregnancy; Anti-HIV Agents; Anti-HIV Agents/therapeutic use; HIV Infections; HIV Infections/prevention & control; HIV Infections/transmission; Infectious Disease Transmission, Vertical; Infectious Disease Transmission, Vertical/prevention & control; Lamivudine; Lamivudine/therapeutic use; Nevirapine; Nevirapine/therapeutic use; Pregnancy Complications, Infectious; Pregnancy Complications, Infectious/drug therapy; Randomized Controlled Trials as Topic; Reverse Transcriptase Inhibitors; Reverse Transcriptase Inhibitors/therapeutic use; Zidovudine; Zidovudine/therapeutic use

Criteria for considering studies for this review

Search methods for identification of studies

See: HIV/AIDS Collaborative Review Group search strategy.

The original review, published in 2002, used a search strategy developed for the Pregnancy and Childbirth Group as a whole. Relevant trials were identified in the Group's Specialized Register of Controlled Trials. In addition, the Cochrane Controlled Trials Register was searched with each new edition of The Cochrane Library. Conference abstracts from the International AIDS Conferences and the Conference on Retroviruses and Opportunistic Infections were also searched.

The review was updated in 2006 after refining the search strategy with the assistance of the HIV/AIDS Review Group Trials Search Co‐ordinator. We formulated a comprehensive and exhaustive search strategy in an attempt to identify all relevant studies regardless of language or publication status (published, unpublished, in press, and in progress). Full details of the Cochrane HIV/AIDS Review Group methods and the journals hand‐searched are published in the section on Collaborative Review Groups in The Cochrane Library. We used the RCT search strategy developed by The Cochrane Collaboration and detailed in the Cochrane Reviewers' Handbook in combination with terms specific to mother‐to‐child transmission. We searched the following electronic databases:

(1) MEDLINE (1966 to date) via PubMed on 17 February 2004 and updated on 31 January 2005 and again on 9 February 2006 using the strategy documented in Table 1. The searches conducted in 2005 and 2006 yielded 265 records in total of which we selected 29 for full article retrieval. (The search in 2004 yielded 276 records but no record was kept of the number of full articles retrieved for that search.)

For this update we searched MEDLINE via PubMed on 17 April 2009 using the strategy documented in Table 1 for the years 2007 to date.This search yielded 773 records of which we selected 23 records for full article retrieval. We also updated the 2006 search as the previous search had been done up until February 2006. This yielded 285 records of which we selected 8 for full article retrieval. See Figure 1 for flow diagram details.

(2) EMBASE (2000 to date) on 5 April 2004 and updated again on 31 January 2005 and 9 February 2006 using the PubMed strategy modified for EMBASE documented in Table 2.

The searches conducted in 2005 and 2006 yielded 39 records in total of which we selected 15 for full article retrieval. (The search in 2004 yielded 116 records but no record was kept of the number of full articles retrieved.)

For this 2009 update we searched EMBASE again using the strategy documented in Table 2 for the years 2006 to date. This search yielded 116 records of which we selected 7 for full article retrieval. See Figure 1.

(3) AIDSearch (1995 to date) on 31 January 2005 and again on 9 February 2006. The database includes coverage of the following conferences: ‐ International AIDS conference (1985‐2004) ‐ Conference on Retroviruses and Opportunistic Infections (1986‐2004) ‐ The British HIV Association conference (1997‐2003) ‐ International Congress on Drug Therapy in HIV infection (1994‐2002)

We limited our search to 2004 onwards, which yielded 28 records in total of which we selected eight for full article retrieval. See Table 3 for the full search strategy.

For this 2009 update, we searched NLM Gateway instead of AIDSearch as this database is no longer operational. NLM Gateway covers some relevant AIDS conferences and meetings and we used the strategy outlined in Table 4 but no meeting abstracts were found.

As we needed to identify other means of searching conference abstracts, we searched the AEGIS database (www.aegis.org). A research specialist searched all the abstract records from the following major related conferences: 1st‐5th IAS Pathogenesis (2001‐2009); 10th‐17th IAC (1994‐2008); 1st‐16th CROI (1994‐2009); US National HIV Prevention Conference ('99, '03, '05); 7th‐14th BHIVA (2001‐2008); and 8th‐9th European AIDS Society Conference (2001, 2003), using the search terms:

3TC, ABC, AZT, ZDV, d4T, TDF, FTC, NRTI, NNRTI, nucleoside, nucleotide, protease, DLV, EFV, ETR, NVP, APV, ATV, DRV, IDV, LPV, RTV, NFV, TPV, T‐20, MVC, Atripla, lamivudine, abacavir, zidovudine, stavudine, zalcitabine, didanosine, emtricitabine, epzicom, kivexa, trizivir, combivir, truvada, delavirdine, efavirenz, nevirapine, amprenavir, fosamprenavir, atazanavir, darunavir, indinavir, lopinavir, ritonavir, saquinavir, tipranavir, enfurvitide, maraviroc, raltegravir, tenofovir, breast, mother, infant, baby, pregnant, pregnancy, perinatal, postnatal, feeding, breastfeeding, vertical, mtct, pmtct.

This yielded 6099 results. Results relevant to animal research, behavioural research, etc. (e.g. simian, feline, monkey, baboon, chimpanzee, rhesus, macaque, msm, condom, behavio/ural, couples, husband, wife, boyfriend, girlfriend, violent, violence, cocaine, heroin, methamphetamine, in vitro, murine, mouse), were then excluded leaving 607 net results. No trials that had not been identified previously were identified through this process.

(4) The Cochrane Library Controlled Trials Register on 7 April 2004 and again on 31 January 2005 and 9 February 2006. The searches conducted in 2005 and 2006 yielded 12 records in total of which we selected four for full article retrieval. (The search in 2004 yielded 61 records but no record was kept of the number of full articles retrieved.) See Table 5 for search strategy.

For the 2009 update we searched CENTRAL of Issue 2 of The Cochrane Library on 11 May 2009 using the same strategy as before in Table 5 and yielded 54 records of which we identified 17 for full article retrieval.

[Table 4 contains the search strategy for NLM Gateway used in the previous update in 2005].

We also checked the reference lists of all the articles retrieved by the search strategy for relevant studies.

One of us (NS) attended the International AIDS Society Conference held in Cape Town, South Africa, from 20 to 22 July 2009 and identified additional trials first presented at the conference. Three trials were identified in this way.

Finally, we contacted research organizations and experts in the field for unpublished and ongoing studies and have included these as studies for future assessment.

Data collection and analysis

The methods detailed below refer to the process used in the 2009 update.

Description of studies

For the 2009 update, from the searches described above, we identified an additional seven relevant trials (Mashi; Thistle 2007; Chung 2005; Chung 2008; Chi 2007; Kesho Bora; Mma Bana) and one ongoing trial (BAN 2005) (also identified in the previous review) since the review was published in 2007. In total, 25 trials met our inclusion criteria. Overall the 25 trials included 18,901 participants with a median trial sample size of 627 ranging from 50 (Bhoopat 2005) to 1,844 participants (PHPT‐2). Twenty‐two trials randomised mothers (18 pre‐natally and four in labour) and followed up their infants, and three trials (Taha 2004; Gray 2005; Taha 2003) randomised infants.

Location of trials

Most trials (N=20) were conducted in a single country, all in poor regions of the world: Thailand (N=5), South Africa (N=3), Kenya (N=3), Botswana (N=2), Zimbabwe (N=2), Malawi (N=2), Cote d'Ivoire (N=1), Uganda (N=1) and Zambia (N=1). Of the five multinational trials, three were conducted across countries in Africa: the DITRAME trial was conducted in Cote d'Ivoire and in Burkina Faso; and the PETRA trial was conducted in South Africa, Tanzania and Uganda; and the Kesho Bora trial was conducted in five sites in Burkina Faso, Kenya and South Africa. Two multinational, multicentre trials were conducted in rich countries: the PACTG 076 trial in the USA and France, and the PACTG 316 in the USA, Puerto Rico, Europe, Brazil and the Bahamas.

Feeding mode

Twelve trials, all conducted in Africa, included mothers who were breastfeeding (DITRAME; HIVNET 012; PETRA; RETRO‐CI; Taha 2003; Taha 2004; Thistle 2004; Chi 2007; Chung 2005; Chung 2008; Mma Bana; Thistle 2007). Eight trials were conducted in non‐breast‐feeding populations: all of the trials conducted in Thailand (N=5) included mothers willing to formula feed (Bhoopat 2005; Limpongsanurak 2001; PHPT‐1; PHPT‐2; Thai‐CDC; ) as did the two multinational, multicentre trials (PACTG 076; PACTG 316) and the Gray 2006 trial conducted in Soweto, South Africa. Five trials included mixed feeding: in the SAINT study conducted in South Africa, 46% of the 1,319 included mothers reported breastfeeding at birth reducing to 32% at 8 weeks; in the Kiarie 2003 study conducted in Kenya, 66% of the 188 included mothers reported breastfeeding at 1 week; the mode of feeding at birth was reported to be 84% exclusive formula feeding, 15% breast milk exposure and 1% mixed feeding in the Gray 2005 trial; and in the Mashi and the Kesho Bora mode of feeding was mixed.

Dates

The first trial to assess antiretrovirals for reducing MTCT was PACTG 076, which began enrolling participants in April 1991. The trial was conducted in 59 centres in the USA and France and compared 198 mothers receiving a dose of 100 mg of ZDV five times daily from 14 to 34 weeks gestation until onset of labour with 204 mothers receiving a placebo. During labour those mothers randomised to the treatment group received a 2 mg/kg loading dose of ZDV intravenously followed by a dose of 1 mg/kg/hr until delivery and their infants received ZDV syrup at a dose of 2 mg/kg six hourly for six weeks after birth. Infants in the control group received a placebo. This trial was stopped during the first interim analysis in December 1993, as the difference in favour of ZDV was statistically significant.

Since then, the type, dosage and period of drugs to be compared have been modified in each subsequent trial. No two trials in this review compare identical interventions (see Characteristics of included studies for more detail). In addition, the diagnostic tests used to diagnose HIV infection in both mother and infants differ between trials and over time, as do the criteria for the diagnosis of HIV infection and the time‐points at which blood samples were taken (see Figure 2). Trials in the breastfeeding populations continue for a longer period than those in non‐breastfeeding in order to assess any additional benefit of antiretrovirals in reducing breast‐milk transmission.

Risk of bias in included studies

Please see Figure 3 and Figure 4 for a graphical description of the risk of bias.

Effects of interventions

We have grouped the trials according to type of comparison e.g. antiretrovirals vs placebo, and within each group stratify according to whether the study was conducted in a breastfeeding population or not. Breastfeeding is defined as greater than 90% of included women breastfeeding their infants at birth. Within strata the results are provided for each trial with the trials ordered chronologically on the date of the start of the trial rather than the date when the report was published.

For the 2009 update, no new trials were identified evaluating antiretrovirals compared with placebo or longer regimens with shorter regimens of the same antiretrovirals. We classified three of the trials identified in the 2009 update as regimens using different antiretrovirals and different durations of treatment. Three trials included evaluations of a regimen of three antiretrovirals given to the mother which we refer to as TRIPLE (also known as Highly Active Antiretroviral Treatment [HAART]) compared with other shorter regimens, and one compared two different TRIPLE regimens with each other.

ANTIRETROVIRALS VS PLACEBO (See Comparison 01)

BREASTFEEDING Three trials assessed this comparison (DITRAME; RETRO‐CI; PETRA).

DITRAME ZDV given to mothers from 36 to 38 weeks gestation, during labour and for 7 days after delivery with no treatment to infants significantly reduced HIV infection at 4‐8 weeks (Efficacy 32.00%; 95% CI 1.50 to 62.50), 3 to 4 months (Efficacy 33.07%; 95% CI 5.57 to 60.57), 6 months (Efficacy 34.55%; 95% CI 9.05 to 60.05), 12 months (Efficacy 34.31%; 95% CI 9.30 to 59.32) and 18 months (Efficacy 29.74%; 95% CI 2.73 to 56.75). The treatment did not influence risk of infant death in the first week after birth (RR 2.03; 95% CI 0.51 to 8.00), in the first 4 to 8 weeks (RR 1.77; 95% CI 0.53 to 5.97), in the first 3 to 4 months (RR 0.74; 95% CI 0.35 to 1.58), in the first 6 months (RR 0.62; 95% CI 0.35 to 1.09), in the first 12 months (RR 0.75; 95% CI 0.48 to 1.17) and in the first 18 months (RR 0.80; 95% CI 0.53 to 1.21). The study found a statistically significant reduction in the risk of premature delivery (RR 0.14; 95% CI 0.03 to 0.58 but no difference in the risk of low birthweight (RR 0.92; 95% CI 0.57 to 1.47). The difference in the risk of stillbirth was not statistically significant (RR 0.14; 95% CI 0.02 to 1.17).

Adverse events:

• Mothers: There was no significant difference in the frequency of death by 6 weeks (1 with ZDV vs 2 with placebo; p=1.0), severe anaemia at day 8 or 45 days postpartum (11 with ZDV vs 8 with placebo; p=0.47) and severe neutropenia at day 8 or 45 (1 with ZDV vs 1 with placebo; p=0.99). Both types of blood disorders were transient.

• Infants: There was no significant difference in the frequency of severe anaemia (10 with ZDV vs 12 with placebo; p=0.67) and severe neutropenia (1 with ZDV vs 4 with placebo; p=0.37); both types of disorder disappeared by 45 days.

RETRO‐CI ZDV given to mothers from 36 weeks gestation and in labour but not to babies significantly reduced HIV infection at 4 to 8 weeks (Efficacy 43.78%; 95% CI 8.78 to 78.78) and 3 to 4 months (Efficacy 36.95%; 95% CI 2.94 to 70.96) but not at birth (Efficacy ‐13.95%; 95% CI ‐163.75 to 135.85). Treatment also reduced the risk of infant death in the first week after birth (RR 0.13; 95% CI 0.02 to 0.99) and during the first 3 to 4 months after birth (RR 0.15; 95% CI 0.05 to 0.49) but not the risk of stillbirth (RR 3.50 95% CI 0.74 to 16.55).

Adverse events:

• Mothers: There was no significant difference in the frequency of postpartum death (1 with ZDV vs 5 with placebo; p=0.21), severe clinical disorder (1 reported case each of painful joints and vomiting with ZDV and 0 with placebo; p=0.25) and severe laboratory abnormality (mostly low haemoglobin) (7 with ZDV vs 9 with placebo; p=0.80).

• Infants: There was no significant difference in the frequency of congenital abnormality (2 with ZDV vs 6 with placebo; p=0.28) and severe laboratory abnormality (mostly low haemoglobin) (8 with ZDV vs 17 with placebo; p=0.09); low haemoglobin was transient.

PETRA In this study, combinations of ZDV and 3TC delivered in three different arms were compared to a placebo arm. We present the results for each of the arms compared with the placebo arm (see Table of Included Studies for full details of each arm)

PETRA a Regimen A : ZDV plus 3TC given to mothers from 36 weeks gestation through labour and continued for 7 days after delivery and to babies for the first 7 days after birth significantly reduced HIV infection at 4 to 8 weeks (Efficacy 62.75%; 95% CI 40.76 to 84.74) but not at 18 months (Efficacy 32.88%; 95% CI ‐11.75 to 77.51). The incidence of a combined endpoint of HIV infection or death was also significantly reduced (Efficacy 61.33%; 95% CI 41.34 to 81.32) at 4 to 8 weeks but not at 18 months (Efficacy 26.17%; 95% CI ‐13.62 to 65.96). There was no statistically significant difference in infant mortality in the first 4 to 8 weeks (RR 0.35; 95% CI 0.11 to 1.14) or in the first 18 months (RR 0.76; 95% CI 0.50 to 1.14) and no difference in the rates of stillbirths (RR 0.40 95% CI 0.07 to 2.15)

PETRA b Regimen B: ZDV plus 3TC given to mothers from the start of labour until 7 days after delivery and to babies for the first 7 days after birth significantly reduced HIV infection at 4 to 8 weeks (Efficacy 41.83%; 95% CI 12.82 to 70.84) but not at 18 months (Efficacy 18.47%; 95% CI ‐27.16 to 64.10). The risk of HIV infection or death was significantly reduced at 4 to 8 weeks (Efficacy 35.91%; 95% CI 8.41 to 63.41) but not at 18 months (Efficacy 7.03%; 95% CI ‐39.17 to 53.23). There was no statistically significant difference in infant mortality in the first 4 to 8 weeks (RR 0.71; 95% CI 0.28 to 1.81) or in the first 18 months (RR 1.05; 95% CI 0.73 to 1.51) and no difference in stillbirth rates (RR 1.19 95% CI 0.34 to 4.20)

PETRA c Regimen C: ZDV plus 3TC given to mothers during labour only with no treatment to babies did not significantly reduce HIV infection at either 4 to 8 weeks (Efficacy 7.19%; 95% CI ‐31.81 to 46.19) or 18 months (Efficacy 9.91%; 95% CI ‐40.99 to 60.81). There was also no significant difference in the combined outcome of HIV infection or death at 4 to 8 weeks (Efficacy 3.31%; 95% CI ‐31.69 to 38.31) and at 18 months (Efficacy 1.56%; 95% CI ‐44.60 to 47.72). Infant mortality was similar in the comparison groups in the first 4 to 8 weeks (RR 0.98; 95% CI 0.41 to 2.34) and in the first 18 months (RR 0.96; 95% CI 0.66 to 1.39) and no significant difference in the risk of stillbirth (RR 0.80 95% CI 0.20 to 3.18)

Adverse events:

• Mothers: There was no significant difference in the frequency of grade 3 and 4 laboratory events (with respect to haemoglobin, leucocytes, lymphocytes, thrombocytes, creatinine, or transaminase levels) before week 6 (32 with regimen A, 22 with regimen B, 26 with regimen C, and 29 with placebo; p=0.47).

• Infants: There was no significant difference in the frequency of grade 3 and 4 laboratory events (20 with regimen A, 18 with regimen B, 24 with regimen C, and 28 with placebo; p=96), neurological events up to 18 months (9 with regimen A, 13 with regimen B, 15 with regimen C, and 10 with placebo; p=0.61) and congenital abnormalities 28 with regimen A, 27 with regimen B, 24 with regimen C, and 28 with placebo; p=0.89)

NON BREASTFEEDING

Three trials assessed this comparison (PACTG 076; Limpongsanurak 2001; Thai‐CDC)

PACTG 076 ZDV administered to mothers from 14 to 34 weeks gestation and continued through labour and to babies for the first 6 weeks after birth significantly reduced HIV infection in babies at 18 months (Efficacy 66.22%; 95% CI 33.94 to 98.50). The treatment did not significantly reduce the risk of infant death in the first 18 months (RR 1.33; 95% CI 0.30 to 5.87) or the risks of premature delivery (RR 1.23; 95 CI 0.60 to 2.49) and low birth weight (RR 0.75; 95% CI 0.48 to 1.19). No significant difference in stillbirth rates was found (RR 0.33; 95% CI 0.01 to 8.11)

Adverse events:

• Mothers: There were no deaths in either group. The rates of severe haematological toxicity (18 with ZDV and 16 with placebo; p=0.82) and severe chemistry toxicity (7 with ZDV and 2 with placebo; p=0.09) were not significantly different.

• Infants: The frequency of anaemia (Hb<9.0g/dL) in the first 6 weeks after birth was higher in the ZDV group (44 with ZDV vs 24 with placebo; p=0.001); of these 4 infants in each group had a Hb <7.0 g/dL. The anaemia resolved by 12 weeks.

Limpongsanurak 2001 ZDV given to mothers from 38 weeks gestation and in labour but not to their babies did not influence HIV transmission at 6 months (Efficacy 8.59%; 95% CI ‐26.63 to 43.81). There was also no significant difference in the risk of stillbirths between the two groups (RR 3.07; 95% CI 0.13 to 74.28).

Adverse events:

• Not reported

Thai‐CDC ZDV given to mothers from 36 weeks gestation and in labour with no treatment to babies significantly reduced HIV infection at 4 to 8 weeks (Efficacy 50.26%; 95% CI 13.80 to 86.72) but not at birth (Efficacy 29.71%, 95% CI ‐43.48 to 102.90). There was no significant reduction in the risk of stillbirth (RR 3.02; 95% CI 0.12 to 73.57), low birth weight (RR 0.61; 95% CI 0.30 to 1.27) or infant death in the first 4‐8 weeks after birth (RR 0.50; 95% CI 0.05 to 5.50).

Adverse events:

• Mothers: There were no deaths in either group. There was no significant difference in the frequency of severe postpartum anaemia (haematocrit <24%) (13 with ZDV vs 10 with placebo; p not reported).

• Infants: There was no significant difference in the risk of grade 3 haematological effects: haematocrit<36% at birth (2 with ZDV and 1 with placebo) or <21% at 2 months (0 with ZDV and 1 with placebo; p values not reported). The number of congenital abnormalities was similar (4 with ZDV vs 3 with placebo; p not reported). During 18 months follow up no statistically significant differences were observed in the two groups in terms of growth, haematological, immunological or clinical events.

LONGER VERSUS SHORTER REGIMENS USING THE SAME ANTIRETROVIRALS (See Comparison 02) BREASTFEEDING

One trial assessed this comparison (Thistle 2004)

Thistle 2004 An 'ultrashort' ZDV regimen (administered to mothers in labour only and to their babies for the first 3 days after birth) was compared to a regimen similar to 'Thai‐CDC' (ZDV to mothers from 36 weeks and through labour with no treatment given to babies). There was no significant difference in HIV infection rates at birth (Efficacy ‐15.10%, 95% CI ‐152.08 to 121.88), 4‐8 weeks (Efficacy 16.72% 95% CI ‐42.43 to 75.87), 3 to 4 months (Efficacy 13.95%, 95% CI ‐38.50 to 66.40), 6 months (Efficacy 8.28%, 95% CI ‐41.76 to 58.32) and 12 months (Efficacy 8.62%, 95% CI ‐33.93 to 51.17). Furthermore, the risk of infant deaths was not statistically different during the following time periods after delivery: 4 to 8 weeks (RR 1.00; 95% CI 0.21 to 4.85), 3 to 4 months (RR 1.75; 95% CI 0.53 to 5.81), 6 months (RR 2.00; 95% CI 0.71 to 5.66) and 12 months (RR 2.00; 95% CI 0.71 to 5.66). Rates of prematurity were not significantly different (RR 1.75, 95% CI 0.53 to 5.81) and no stillbirths occurred in either group.

Adverse events:

• Mothers: The frequency of deaths was not significantly different (during childbirth 1 in each group, at two months postpartum 1 with 'ultrashort regimen and 0 with 'Thai‐CDC' regimen, at six months postpartum 0 with 'ultrashort' regimen and 3 with 'Thai‐CDC' regimen and at one year postpartum 1 with 'ultrashort' regimen and 0 with 'Thai‐CDC' regimen). The trialists report that there were no significant adverse effects due to medication.

• Infants: not reported.

NON BREASTFEEDING

Three trials assessed this comparison ( PHPT‐1; PHPT‐2; Bhoopat 2005)

PHPT‐1 The study evaluated four regimens: 1) 'long‐long' ‐ ZDV given to mothers from 28 weeks gestation through labour and to infants from birth to 6 weeks; 2) 'long‐short' ‐ ZDV given to mothers from 28 weeks gestation through labour and to infants for the first 3 days after birth; 'short‐long' ‐ ZDV given to mothers from 35 weeks gestation through labour and to infants from birth to 6 weeks; and 4) 'short‐short' ‐ ZDV given to mothers from 35 weeks gestation through labour and to infants for the first 3 days after birth.

PHPT‐1 a: 'long‐long' regimen vs. 'short‐long' regimen There was no significant difference in HIV infection rates in infants at 6 months (Efficacy 24.42%; 95% CI ‐20.25 to 69.09) and in the risk of HIV infection or death at 6 months (Efficacy 14.61%; 95% CI ‐30.80 to 60.02). There was also no significant difference in infant deaths during the first 6 months (RR 0.82; 95% CI 0.24 to 2.82), stillbirth rate (RR 0.55; 95% CI 0.23 to 1.33) and the risk of premature delivery (RR 2.01; 95% CI 0.94 to 4.31). There was, however, a higher risk of low birth weight with the long‐long regimen (RR 1.65; 95% CI 1.04 to 2.60)

PHPT‐1 b: 'long‐short' regimen vs. 'short‐long' regimen The long‐short regimen significantly reduced HIV infection rate at 6 months (Efficacy 45.33%; 95% CI 1.39 to 89.31). While the risk of HIV infection or death at 6 months was also reduced this was not statistically significant (Efficacy 37.08%; 95% CI ‐6.88 to 81.04). There was no significant difference in the risk of infant mortality during the first 6 months (RR 1.38; 95% CI 0.44 to 4.31), premature birth (RR 1.86; 95% CI 0.84 to 4.12), low birth weight (RR 1.50; 95% CI 0.92 to 2.43) and stillbirth (RR 0.33; 95% CI 0.11 to 1.01).

As the short‐short regimen seemed not to reduce transmission of HIV, it was discontinued at the first interim analysis.

Adverse events:

• Mothers: The rate of serious adverse events in the groups was not significantly different (deaths: 3 with 'long maternal regimen' and 8 with 'short maternal regimen'; severe anaemia: 7 with 'long maternal regimen' and 4 with 'short maternal regimen'; neutropenia: 0 with 'long maternal regimen' and 1 with 'short maternal regimen'. All maternal deaths occurred post‐partum with five women dying from pneumonia, two from sepsis, one from cryptococcal meningitis, one from an AIDS‐defining even and two from suicide. All cases of anaemia and neutropenia resolved spontaneously after treatment ended.

• Infants: The rate of serious adverse events in the groups was not significantly different (severe anaemia: 4 with long‐long, 0 with long‐short, 1 with short‐long, 4 with short‐short courses; neutropenia or leukopenia: 7 with long‐long, 3 with long‐short, 5 with short‐long, 2 with short‐short courses; congenital abnormalities: 7 with long‐long, 7 with long‐short, 6 with short‐long, 1 with short‐short courses.

PHPT‐2 In a programme where mothers were routinely receiving ZDV in the third trimester of pregnancy and babies were receiving one week of ZDV therapy, a single dose of NVP given to mothers in labour and to their babies soon after birth ('NVP‐NVP' arm) was compared with a single dose of NVP given to mothers only ('NVP‐placebo' arm). HIV infection rates at birth (Efficacy 37.50%; 95% CI ‐40.94 to 115.94) and at 6 months (Efficacy 28.57%; 95% CI ‐26.09 to 83.23) were not significantly different in the two groups. The difference in the risk of HIV infection or death at 6 months was not statistically significant (Efficacy 45.00%; 95% CI ‐4.00 to 94.00). Infant death rates in the first 6 months was significantly reduced in the NVP‐NVP arm (RR 0.20; 95% CI 0.04 to 0.91) but differences in stillbirth rates (RR 0.25; 95% CI 0.05 to 1.17) and low birth weight (RR 0.85; 95% CI 0.60 to 1.19) were not statistically significant.

Adverse events Mothers: Rates of serious adverse effects were reported as being similar in the comparison groups; 59% of adverse events were attributed to pregnancy, 26% to infections including HIV, 7% possibly to zidovudine (anaemia), and 7% possibly to nevirapine and other conditions. Infants: Rates of serious adverse events were reported as being similar in the comparison groups and were attributed as follows: 11% to neonatal and obstetric conditions, 6% to congenital abnormalities, 72% to infections including HIV, 2% possibly to zidovudine (anaemia), and 9% to other causes, of which 1% involved neonatal icterus possibly due to nevirapine use.

Bhoopat 2005 This study compared a long course ZDV regimen (given to women 62 to 92 days before labour and continued through labour; median 76 days) with a short course ZDV regimen (given to women from 14 to 35 days before labour and continued through labour, median 28 days). Babies did not receive antiretrovirals in either arm. HIV infection rates were not significantly different at birth (no infected babies with the 'long course' regimen vs. 1 infected baby with the 'short course regimen; Efficacy 100.00%; 95% CI ‐294.36 to 494.36) and 3 to 4 months (no infected babies with the 'long course' regimen vs 4 infected babies with the 'short course' regimen'; Efficacy 100.00%; 95% CI ‐16.50 to 216.50).

Adverse events:

• Not reported

REGIMENS USING DIFFERENT DRUGS AND DURATIONS OF TREATMENT (See Comparison 03)

This section describes those trials which compared regimens comprising different drugs either in a mono‐ or dual therapy manner and for varying periods of time. Trials which evaluate antiretrovirals comprising triple therapy are described in the following sections and Comparisons 04 and 05.

BREASTFEEDING

Six trials assessed this comparison (HIVNET 012; Taha 2003; Taha 2004; Thistle 2007; Chung 2005; Chi 2007). Three of these trials were identified in the most recent 2009/10 update (Thistle 2007; Chung 2005; Chi 2007).

HIVNET 012 A single dose of NVP given to mothers at the onset of labour plus a single dose of NVP given to their babies immediately after birth compared with ZDV given to mothers during labour and to their babies for a week after birth was associated with significantly lower HIV infection rates at 4‐8 weeks (Efficacy 41.00%; 95% CI 11.84 to 70.16), 3‐4 months (Efficacy 38.91%; 95% CI 11.24 to 66.58), 12 months (Efficacy 35.98%; 95% CI 9.25 to 62.71) and 18 months (Efficacy 39.15%; 95% CI 13.81 to 64.49). In addition, the NVP regimen significantly reduced the risk of HIV infection or death at 4‐8 weeks (Efficacy 41.74%; 95% CI 14.30 to 69.18), 3 to 4 months (Efficacy 40.00%; 95% CI 14.34 to 65.66), 12 months (Efficacy 32.17%; 95% CI 8.51 to 55.83) and 18 months (Efficacy 32.57%; 95% CI 9.93 to 55.21). There was no significant difference between the two groups for infant mortality in the first week after birth (RR 2.50; 95% CI 0.49 to 12.79), in the first 4 to 8 weeks (RR 2.50; 95% CI 0.79 to 7.89) and in the first 18 months (RR 1.24; 95% CI 0.81 to 1.89). There was also no significant difference in the risk of stillbirth (RR 2.00; 95% CI 0.18 to 21.94) or low birth weight (RR 0.67; 95% CI 0.35 to 1.29).

Adverse events:

Serious adverse events were defined as fatal or life threatening, permanently disabling, requiring inpatient admission, a congenital anomaly, cancer or overdose, or otherwise judged to be serious by onsite clinician, or death.

• Mothers: Differences in the rate of serious maternal adverse events in the first 8 weeks were not statistically significant (15 with NVP vs 11 with ZDV; p=0.443). Deaths at 8 weeks were 0 in the NVP group and 3 in the ZDV group; p=0.081.

• Infants: Rates of serious adverse events were not significantly different: at 8 weeks 29 with NVP vs 35 with ZDV; p=0.348 and at 18 months 109 with NVP vs 97 with ZDV; p=0.476. There were no significant differences in the rates of grade 3 and 4 laboratory toxic effects or grade 3 and 4 abnormalities in the alanine aminotransferase.

Taha 2003 A regimen that combined a single dose of NVP given to babies immediately after birth with ZDV given to babies for the first week after birth compared to a regimen consisting of a single dose of NVP to babies only significantly reduced HIV infection at 4 to 8 weeks (Efficacy 36.79%; 95% CI 3.57 to 70.01). The study did not find a statistically significant difference in the risk of infant death in the first 4 to 8 weeks (RR 1.26; 95% CI 0.60 to 2.67).

Adverse events The trialists reported safety analyses using data from all babies, including those excluded from primary analyses because maternal HIV infection was not confirmed. The rate of grade 3 to 4 adverse events did not differ between groups with 31(5·6%) events in the NVP‐only group and 44 (7·8%) events in the NVP + ZDV group (p=0·16). Fewer than 1% of these events in the whole study population were judged to be related to the interventions. Severe adverse events were reported as mainly related to infections or fever.

Taha 2004 This study compared the HIVNET regimen with the addition of ZDV given to babies for one week with the HIVNET regimen alone and found no significant difference in HIV infection at 4 to 8 weeks (Efficacy ‐15.60%; 95% CI ‐50.21 to 19.01) or in the risk of infant death in the first 4 to 8 weeks (RR 1.74; 95% CI 0.51 to 5.91).

Adverse events:

• Mothers: not reported

• Infants: Grade 3 and 4 adverse events were similar between treatment groups 22 (4.9%) with NVP versus 24 (5.4%) with NVP plus ZDV. Severe adverse events were considered to be mainly due to infections; of the 46 grade 3 or 4 adverse events, 25 cases (54%) were coded as pneumonia, diarrhoea, or malnutrition possibly related to infections. Severe haematological changes were observed in 15 (4.39%) of 342 of the NVP plus ZDV group and 7 (2.13%) of 329 of the NVP only group (p= 0.13).

Thistle 2007

This Zimbabwean trial compared the ultra‐short ZDV regimen given to mothers at the onset of labour and 3 hourly during labour, combined with a single 200mg dose of NVP during labour; infants received ZDV four times daily for 72 hours post delivery and a single dose of NVP. Mothers in the placebo group received a placebo rather than ZDV and the single dose of NVP in labour; their infants received placebo for ZDV and the single dose of NVP within 72 hours of delivery. There was no difference in HIV infection rates at birth between the groups (Efficacy ‐5.45%; 95%CI ‐79.95 to 69.05) and at four to six weeks, HIV infection efficacy remained statistically non‐significant with an efficacy of 9.21% (95%CI ‐27.56 to 45.8). For HIV infection or death at four to six weeks, there was no statistically significant difference between the groups (Efficacy 3.97%,95%CI ‐26.14 to 34.08). Infant deaths during the first four to six weeks was similar in both groups (RR = 1.08; 95% CI 0.61 to 1.92).

Adverse events

Mothers:There were 9 maternal deaths in the ultra‐short ZDV regimen and 13 in the NVP group. Infants: No cases of severe drug toxicity were observed in the infants in both groups.

Chung 2005

In this trial the so‐called Thai‐CDC regimen of ZDV given to mothers from 34 weeks until onset of labour and during labour was compared with the HIVNET 012 regimen of NVP given to mothers at onset of labour and to infants within 72 hours of delivery. There was no statistically significant difference between HIV infection at 2 weeks in infants (Efficacy ‐100%, 95%CI ‐536.39 to 336.39). At 4 to 8 weeks HIV infection in infants of the mothers receiving the NVP regimen was statistically significantly lower than those in the ZDV group (Efficacy ‐345.59%; 95%CI ‐630.00 to ‐61.18). The risk of infant death at 6 weeks was lower in the NVP group compared with the ZDV group (RR 5; 95%CI 0.25 to 99.95) but not statistically significant. As there were no events in the NVP group the RR must be treated with caution. The RR for infants weighing less than 2.5kg in the ZDV group was half that in the NVP group (RR 0.50; 95%CI 0.10 to 2.53), which was not statistically significant.

Adverse events:

• There were no adverse events or side effects reported in either intervention group in the mothers or babies.

Chi 2007

In this Zambian trial all mothers received the standard of care at the time, which was ZDV 300mg twice daily from 32 weeks until labour, sdNVP 200mg at onset of labour and all infants received the standard of care of NVP 2mg/kg once within 72 hours of delivery and ZDV 4mg/kg orally twice daily for 7 days post delivery. Mothers were randomised to either receive the addition of Tenofovir 300mg and Emtricitabine 200mg (co‐formulated as Truvada) orally in labour in addition to standard of care or only standard of care. The primary aim of this trial was to evaluate maternal resistance to non‐nucleoside reverse transcriptase inhibitor drugs at 6 weeks post‐partum.

HIV infection at birth was not statistically different between groups (Efficacy 20.40%, 95%CI ‐66.86 to 107.66) or at four to eight weeks (Efficacy 28.93%, 95%CI ‐40.84 to 98.70). The risk of death in the first four to eight weeks after birth was not statistically different between the groups (RR 1.26; 95%CI 0.34 to 4.61), nor was the risk of stillbirth (RR 1.99, 95%CI 0.18 to 21.77).

Adverse events:

• Mothers: The primary aim of this trial was to evaluate maternal resistance to non‐nucleoside reverse transcriptase inhibitor drugs at 6 weeks post‐partum. Women assigned to the intervention were reported to be statistically significantly less likely to develop resistance to NNTRI at 2 weeks (RR 0.27; 95%CI 0.11 to 0.66) and 6 weeks (RR 0.47; 95%CI 0.29 to 0.76) post childbirth.

• Seven of 198 mothers receiving the intervention and nine of 199 in the control group experience serious adverse events, four each group had postpartum anaemia. There was no statistically significant difference between the groups.

• Infants: Ten percent (20/198) of infants in the intervention group and 12% (23/199) in the control group experienced serious adverse events, including septicaemia (22) and pneumonia (8). There was no statistically significant difference between the groups and the authors report that none of the events were judged to be due to the study intervention.

NON‐BREASTFEEDING

Five trials assessed this comparison (PACTG 316; SAINT; Gray 2006; Kiarie 2003; Gray 2005) with one additional trial (Mashi) identified in the 2009/10 update.

PACTG 316 In a population in which mothers were receiving 'standard' antiretrovirals for HIV infection a single dose of NVP given mothers in labour plus a single dose of NVP given to babies immediately after birth (HIVNET 012 regimen) compared with placebo did not result in a statistically significant difference in HIV infection rates at birth (Efficacy 2.22%, 95% CI ‐140.39 to 144.83) and 4 to 8 weeks (Efficacy 12.50%, 95% CI ‐82.89 to 107.89) or in deaths at 4 to 8 weeks (RR 0.60, 95% CI 0.14 to 2.50). There was also no significant difference in the risks of stillbirth (RR 2.99; 95% CI 0.12 to 73.33), low birth weight (RR 1.14; 95% CI 0.85 to 1.53) or prematurity (RR 1.05, 95% CI 0.83 to 1.32).

Adverse events:

• Mothers: The risks of severe adverse events did not differ significantly (deaths after delivery: 2 with NVP and 1 with placebo; grade 3 and 4 toxicity ‐ rash 1 with NVP and 1 with placebo; non‐rash toxicity 40 with NVP and 38 with placebo; and hepatic toxicity 5 with NVP and 5 with placebo )

• Infants: The rates of severe adverse events did not differ significantly (grade 3 and 4 toxicity ‐ rash 1 with NVP and 3 with placebo; non‐rash toxicity 235 with NVP and 195 with placebo; and hepatic toxicity 1 with NVP and 2 with placebo

SAINT A regimen of ZDV plus 3TC given to mothers in labour and for a week after delivery and to their infants for a week after birth (similar to PETRA arm b) was compared with a regimen of NVP given to mothers in labour and immediately after delivery plus a single dose of NVP to their babies immediately after birth (similar to HIVNET 012). There was no significant difference in HIV infection rates at 4 to 8 weeks (Efficacy 24.39%; 95% CI ‐4.15 to 52.93), the risk of low birth weight (RR 1.41; 95% CI 0.45 to 4.42) and infant death rates at 4 to 8 weeks 1.01 (0.54 to 1.89)

Adverse events:

• Mothers: The risk of serious adverse events was reported as similar in the two groups: There were 5 deaths (0.8%) with NVP vs 4 (0.6%) deaths with ZDV plus 3TC. No hepatic or haematological adverse events were reported for either group.

• Infants: The risks of serious adverse events (clinical or laboratory abnormalities) were not significantly different: 9.0% in the NVP group and 10.4% in the ZDV plus 3TC group. The most frequent serious adverse events were respiratory system disorders (including asphyxia, respiratory distress syndrome, aspiration, and dyspnoea) (NVP: 4.1%; ZDV plus 3TC: 4.2%) and infections (NVP: 2.6%; ZDV plus 3TC: 3.2%).

Gray 2006 This study compared various antiretroviral drugs given to mothers from 34 to 36 weeks and through labour as well as the same drugs given to their babies for 6 weeks

Gray 2006 a: d4T versus ZDV There was no significant difference in the risks of HIV infection at birth (Efficacy 26.65%; 95% CI ‐122.44 to 175.74), 4 to 8 weeks (Efficacy ‐120.05%; 95% CI ‐301.07 to 60.97), 3 to 4 months (Efficacy ‐144.51%; 95% CI ‐329.90 to 40.88) and 6 months (Efficacy ‐116.07%; 95% CI ‐280.57 to 48.43). There was also no significant difference in the risk of infant mortality in the first 6 months (RR 2.97; 95% CI 0.83 to 10.61). Gray 2006 b: ddI versus ZDV HIV infection between the two groups did not differ significantly at birth (Efficacy 52.66%; 95% CI ‐88.01 to 193.33), 4 to 8 weeks (Efficacy ‐42.02%; 95% CI ‐205.79 to 121.75), 3 to 4 months (Efficacy ‐113.03%; 95% CI ‐290.84 to 64.78) and 6 months (Efficacy ‐89.29%; 95% CI ‐247.73 to 69.15). There was no significant difference in the risk of infant mortality in the first 6 months (RR 1.94; 95% CI 0.50 to 7.52).

Gray 2006 c: d4T plus ddI versus ZDV No significant difference in the risks of HIV infection were found at birth (Efficacy 49.43%; 95% CI ‐94.88 to 193.74), 4 to 8 weeks (Efficacy 24.15%; 95% CI ‐127.10 to 175.40), 3 to 4 months (Efficacy ‐1.14%; 95% CI ‐158.84 to 156.56) and 6 months (Efficacy 17.86%; 95% CI ‐118.79 to 154.51). There was also no significant difference in the risk of infant mortality in the first 6 months (RR 0.66; 95% CI 0.11 to 3.86).

Adverse events:

• Mothers: Rates of serious adverse events were low and did not differ significantly between the groups. There were two deaths (1 woman on ZDV died of haemorrhage after an extrauterine pregnancy and 1 woman on d4T plus ddI due to left ventricular dysfunction; neither death was judged to be due to treatment).

• Infants: Grade 3 and 4 adverse events were more frequent with d4T plus ddI (18%) and ZDV (14%) than with either d4T (9%) or ddI (6%). Death, gastroenteritis and pneumonia were the most common serious adverse events reported and most were considered not to be related to treatment.

Kiarie 2003 A 'Thai CDC' regimen compared with 'HIVNET 012' regimen found no significant difference in HIV infection at 4 to 8 weeks (Efficacy 59.09%; 95% CI ‐3.90 to 122.08). There was also no difference in the risk of infant mortality in the first 4 to 8 weeks (RR 0.99; 95% CI 0.21 to 4.72), stillbirth (RR 1.48; 95% CI 0.25 to 8.58), prematurity (RR 1.97; 95% CI 0.37 to 10.42) and low birth weight (RR 1.97; 95% CI 0.18 to 21.24).

Adverse events:

• Not reported.

Gray 2005 A single dose of NVP given to babies immediately after birth compared to ZDV given to babies for the first 6 weeks after birth did not result in a statistically significant difference in HIV infection rates at 4‐8 weeks (Efficacy 35.37%; 95% CI ‐10.00 to 80.74) and 3 to 4 months (Efficacy 39.69%; 95% CI ‐1.08 to 80.46). Infant mortality in the first 4 to 8 weeks was also similar in the two groups (RR 1.15; 95% CI 0.52 to 2.54)

Adverse events:

• There was no significant difference in the overall risk of serious adverse events in infants between ZDV and NVP (118 with ZDV versus 94 with NVP); none were thought to be due to study medication and most were due to infections such as pneumonia (22 with ZDV versus 18 with NVP) and gastroenteritis (20 with ZDV versus 15 with NVP). Other serious events included birth‐related conditions (14 with ZDV versus 6 with NVP), physiological jaundice (10 with ZDV versus 5 with NVP) and neonatal septicaemia (7 with ZDV versus 13 with NVP). Of these serious adverse events, 17 (18.1%) in the NVP arm and 36 (30.5%) in the ZDV arm were related to HIV.

Mashi

This trial was conducted in Botswana and compared sdNVP with placebo in mothers receiving ZDV from 34 weeks and in labour and infants receiving ZDV twice daily for one month post delivery. Initially the trial aimed to compare sdNVP with placebo in both mothers and infants but this was terminated in August 2002 after the results from the PHPT‐2 trial showed the superiority of sdNVP in infants. The trial continued to evaluate sdNVP versus placebo in mothers only. Infants in both groups then received sdNVP.

For HIV infection at birth, there were no differences between the groups (Efficacy ‐64.38%, 95%CI ‐183.72 to 54.96). At one month post‐partum HIV infection rates were not statistically significantly different between the groups (Efficacy ‐16.22%; 95%CI ‐102.24 to 69.80). Infant deaths in the first month were not statistically different between groups (RR 0.54, 95%CI 0.22 to 1.35). Stillbirth rates were similar in both groups (RR 1.50, 95%CI 0.43 to 5.28) and there was no difference in the incidence of premature delivery (RR = 1.00, 95%CI 0.40 to 2.50).

In October 2002, TRIPLE became available in Botswana and so all participating women with CD4 < 200 cells/μL or with an AIDS‐defining illness were offered TRIPLE and did not receive sdNVP or placebo at onset of labour.

Adverse events:

• Mothers: Twenty women experienced life‐threatening or serious events between study entry and three months after delivery. Of these, the report states that one event of anaemia was possibly related to sdNVP and two events of rash and abnormal liver function tests were possibly related to sdNVP or placebo as both of these were in the placebo arm.

• Infants: Twelve stopped ZDV before one month due to toxicity (mainly due to anaemia or neutropaenia); there were five serious or life‐threatening events possibly related to sdNVP. Four were serious rashes and one was raised bilirubin levels and jaundice. Two were hospitalised but did not die due to adverse events.

TRIPLE REGIMENS VERSUS OTHER (See Comparison 04)

Two trials evaluated TRIPLE antiretroviral regimens compared with previously used regimens which did not include triple antiretrovirals (Chung 2008; Kesho Bora).

BREASTFEEDING

Chung 2008

This single‐centre Kenyan trial compared TRIPLE (ZDV 300mg twice daily, Lamivudine 150mg twice daily, NVP 200mg twice daily from 34 weeks until 6 months post delivery) with mothers receiving ZDV 300mg twice daily from 34 weeks until labour and sdNVP 200mg at onset of labour. Infants in the TRIPLE group received no additional drugs while those in the ZDV/sdNVP group received sdNVP 2mg/kg within 72 hours of delivery. No babies were infected with HIV at birth in either group; at 6 months post delivery there was no statistically significant difference in HIV infection between groups (Efficacy ‐84.62%, 95%CI: ‐490.35 to 321.11). There was no statistically significant difference for the outcome of HIV infection or death at two weeks (Efficacy ‐84.62%, 95%CI: ‐490.35 to 321.11), or at 12 months (Efficacy 35.90%; 95%CI: ‐118.33 to 190.13). There were two stillbirths in the TRIPLE group and one in the ZDV/sdNVP group (RR 1.87, 95%CI 0.18 to 19.47). There were no infant deaths in the TRIPLE arm and one at 2 weeks from sepsis (RR 0.31, 95%CI 0.01 to 7.30). There were five infants weighing less than 2.5kg in the TRIPLE group and none in the ZDV/sdNVP group (RR 10.29, 95%CI 0.59 to 177.97) but this was not statistically significant.

Adverse events:

• Mothers: Two adverse events were reported in women receiving TRIPLE. These were not graded in the report but one was a diffuse rash pre‐delivery ascribed to NVP and the other was anaemia arising two months post‐partum resolving after blood transfusion.

• Infants: None reported.

MIXED FEEDING

Kesho Bora

This trial was conducted in five sites in three countries. TRIPLE (Lopinivar 200mg /ritonavir 50mg and ZDV 300mg and lamivudine 150mg from 26 to 34 weeks gestation through 6 months post‐partum) given to the mother was compared with ZDV given to the mother from 28 to 36 weeks, ZDV and lamivudine and sdNVP given at onset of labour and ZDV and lamivudine for one week after delivery. Infants in both groups received sdNVP within 72 hours of delivery and ZDV for one week.

There was no statistically significant difference between groups in HIV infection at birth (Efficacy 18.18%, 95%CI ‐83.48 to 119.84) or at four to eight weeks (Efficacy 31.25%, 95%CI ‐29.29 to 91.79). At six months, HIV infection was higher but not statistically significantly so in the non‐TRIPLE group (Efficacy 42.35%, 95%CI ‐0.57 to 85.27). At 12 months HIV infection was statistically significantly higher in the non‐TRIPLE group (Efficacy = 42.11%, 95%CI 0.66 to 83.56). The incidence of HIV infection or death at four to eight weeks was higher in the non‐TRIPLE group (RR 20.00, 95%CI ‐34.13 to 74.13) but this was not statistically significant. At 6 months, the HIV infection or death incidence remained higher in the non‐TRIPLE group (RR 34.13, 95%CI [‐0.29 to 68.55) and at 12 months this difference was statistically significant (RR 36.20, 95%CI 5.92 to 66.48).

The number of infants dying in the first week after birth was not statistically significantly different between groups (RR 2.00, 95%CI 0.37 to 10.88), nor in the first 4 to 8 weeks (RR 0.86, 95%CI 0.29 to 2.53), or first 6 months (RR 0.78, 95%CI 0.43 to 1.43). In the first 12 months, more babies died in the non‐TRIPLE group although this difference was not statistically significant (RR 0.65, 95%CI 0.40 to 1.07).

There was no statistically significant difference between groups for prematurity as defined by the authors (RR 1.44, 95%CI 0.93 to 2.24) or in those weighing less than 2.5kg (RR 1.25, 95%CI 0.86 to 1.82). The risks of stillbirth were the same between groups (RR 1.00, 95%CI 0.25 to 3.95).

Adverse events:

• Mothers: There were four maternal deaths in both groups. Other adverse events are reported as percentages only which we calculated backwards to determine numbers. In the TRIPLE group, 11 (2.7%) of mothers developed anaemia at delivery with 7 (1.8%) in the non‐TRIPLE group. In the TRIPLE group. one mother developed Grade 3 neutropaenia and two mothers had ALT level raised five times above the upper limit of normal at delivery . Five deaths were attributed to serious adverse events in both groups.

• Infants: Adverse events are reported as percentages only. In the TRIPLE group 14.6% of infants developed a grade 3 anaemia and 12% in the non‐TRIPLE group. In the TRIPLE group, 7.8% of infants developed a grade 3 or 4 neutropaenia and 8.4% in the non‐TRIPLE group. Fatalities attributed to serious adverse events were 6.7% in the TRIPLE group and 10.2% in the non‐TRIPLE group.

TRIPLE ANTIRETROVIRAL REGIMENS VERSUS TRIPLE ANTIRETROVIRAL REGIMENS (See Comparison 05)

To date, one trial (Mma Bana) has evaluated the superiority of the drug composition of two different types of triple antiretroviral regimens given to mothers.

BREASTFEEDING

Mma Bana

This trial was conducted in the same centres in Botswana as the Mashi trial. Mothers either received Abacavir 300mg, lamivudine 150mg, and ZDV 300mg (co‐formulated as Trizivir (TZV)) one tablet taken twice daily from 26 to 34 weeks gestation or Lopinivar 400mg /ritonavir 100mg tablet and ZDV 300mg and Lamivudine 150mg (co‐formulated as Combivir (CBV)) taken orally twice daily from 26 to 34 weeks gestation through 6 months post‐partum. Infants in both groups received sdNVP within 72 hours of delivery and one month of ZDV following birth.

There was no significant difference in HIV infection rates at birth (Efficacy ‐189.47%; 95%CI ‐715.29 to 336.35) with three babies infected in the TZV group and one in the CBV group. At three months the incidence in infant HIV was still not significantly different (5 versus 1) between groups (Efficacy ‐390.88%; 95%CI ‐981.63 to 199.87). The differences between groups in the number of infants dying in the first eight days after birth (RR 1.96; 95%CI 0.18 to 21.53) and in the first six months after birth (RR 0.98; 95%CI 0.35 to 2.76) were not significantly different. Incidence of prematurity defined as gestational age < 37 weeks was higher in the CBV group (RR 0.66; 95%CI 0.47 to 0.95) but not statistically significant. Low birthweight less than 2.5kg was also higher in the CBV group but was not statistically significant (RR 0.79; 95%CI 0.53 to 1.19). Stillbirth rates did not differ statistically significantly between the groups (RR 1.54; 95%CI 0.51 to 4.66).

Adverse events:

• Mothers: Six percent (17) of mothers in the TZV group and six percent (16) in the CBV group had a new grade 3+ diagnosis after initiating antiretrovirals. Of these, four mothers in the TZV group and five in the CBV group experienced eclampsia or pre‐eclampsia with two and one intra‐uterine death following respectively. The site investigators judged the death in the CBV group 'probably not' due to the study drug but that the deaths in the TZV group were 'possibly' related to the study drug.

• Infants: There were 28 (10%) diagnoses of grade 3 or more severity in the TZV group and 17 (6%) in the CBV group.

Summary of main results

This update of the review shows that a combination of antiretroviral drugs is the most effective antiretroviral regimen for decreasing mother‐to‐child transmission of HIV. It builds on the findings from our previous review that antiretroviral treatment administered in the perinatal period lowers the risk of mother‐to‐child transmission of HIV. This information is relevant to the care of pregnant women who are HIV‐infected in both resource‐rich and resource‐constrained settings. The choice of which combination of triple antiretrovirals, and the optimal time to initiate antiretrovirals in pregnancy, remains unclear. None of the trials in this review follow‐up the infants for longer than 24 months and the long‐term adverse effects of in‐utero or early neonatal exposure to antiretrovirals is not known. Importantly, data on the emergence and duration of resistance is currently limited.

Summary of main results from this update

In this update, the most noteworthy addition to the evidence base is the inclusion of results from trials which compare antiretroviral regimens comprising three drugs either with other regimens (Kesho Bora;Chung 2008) or with another triple antiretroviral combination as in the case of the Mma Bana trial. In the Kesho Bora trial TRIPLE ART including a protease‐based inhibitor was given from 26 to 34 weeks gestation through six months post‐partum and compared with a ZDV‐based regimen starting later at 28 to 36 weeks with no post‐partum component. Infants in the TRIPLE antiretrovirals group were significantly less likely to have the virus at 12 months. The trial population was classified as mixed feeding and the results are likely to indicate the importance of mothers continuing with antiretrovirals should they choose to breastfeed. The Chung 2008 trial was conducted in a breastfeeding population, with a TRIPLE antiretroviral regimen commenced at 34 weeks compared with only ZDV for the same period until labour when sdNVP was added found no babies were infected with HIV at birth in either group and at 6 months post delivery there was no statistically significant difference in HIV infection between groups. Although the lack of events rates does not allow inferences regarding superiority of the TRIPLE regimen over the ZDV regimen, it does confirm the efficacy of providing antiretrovirals, even if monotherapy ZDV, to mothers earlier in pregnancy. In Mma Bana transmission rates are again gratifyingly low both at birth and later at six months, suggesting that either choice of triple combination antiretroviral is effective. Our results concur with those of the Sturt 2010 Cochrane review which evaluated the efficacy of treatment for HIV‐infected pregnant women and considered transmission as a secondary outcome. Adverse event rates both for mothers and babies were low in all three trials with the exception of a significant difference seen in stillbirths in the Kesho Bora trial.

The results from the additional four trials identified in this update contributed to the evidence base for the category of antiretroviral regimens using different drugs and durations of treatment. None of the trial results provided significantly different results to those previously documented.

Summary of main results from previous versions of the review

ZDV has been the most extensively studied drug and forms a component of treatment in all 25 trials included in this review. In a non‐breastfeeding population a long course of ZDV monotherapy used in the antepartum and intrapartum periods and postnatally in babies led to an impressive 66% reduction in HIV infection risk in babies at 18 months (PACTG 076). Important benefits were also found with shorter courses of ZDV monotherapy, although these appear to be less pronounced, with effects on HIV infection rates at 4 to 8 weeks ranging from halving to a one third reduction compared with placebo. Among the shorter course regimens, a combination of ZDV and 3TC commenced at 36 weeks and continued through labour and up to a week postpartum and given to babies for the first week after birth appears to be highly worthwhile. Compared with placebo the efficacy of this regimen in reducing HIV infection risk at 4 to 8 weeks was as much as 63% (PETRA).

We found no trials that have evaluated the effects on mother‐to‐child transmission of NVP monotherapy against placebo. In a breastfeeding population a single dose of nevirapine given to mothers at the onset of labour and to their babies soon after birth compared with a suboptimal course of zidovudine given to mothers only during labour and to their babies for the first week after birth reduced the risk of mother‐to‐child transmission at 4 to 8 weeks by about 40% with the benefit persisting at 18 months (HIVNET 012). A regimen comparable to HIVNET 012 but with mothers receiving an additional dose of NVP postpartum resulted in similar HIV infection rates at 4 to 8 weeks as a regimen consisting of ZDV plus 3TC given to mothers during labour and for one week postpartum and to babies for one week after birth (PETRA regimen B) (SAINT). One small trial comparing the HIVNET 012 and Thai CDC regimens failed to demonstrate a significant difference on HIV transmission. (Kiarie 2003). Furthermore, in a population where mothers were routinely receiving ZDV in the last trimester of pregnancy and their babies were receiving ZDV in the first week after birth, the HIVNET regimen compared with a single dose of NVP given only to mothers did not result in a significant reduction in HIV transmission at 6 months (PHPT‐2). The trial did, however, find a reduction in HIV infection and death at 6 months which was marginally significant as well as a significantly lower risk of infant deaths in the first 6 months with the HIVNET 012 regimen (PHPT‐2) The addition of ZDV treatment to babies in the first week after delivery in addition to the HIVNET 012 regimen had no significant effect on HIV transmission compared with an HIVNET regimen alone (Taha 2004).

Length and timing of antiretroviral treatment

For TRIPLE antiretroviral regimens, there are no trials evaluating provision of antiretrovirals before 26 weeks gestation. However, the World Health Organisation has now suggested that ARV can be started as early as 14 weeks of pregnancy once the period of major teratogenicity is passed (WHO 2009) because of the paucity of evidence to suggest any harmful effects after 14 weeks. This evidence of lack of harm, however, is limited to the immediate effect on the fetus and the child up until the end of follow‐up, which is often short. The experience with other drugs in pregnancy, particularly diethylstilboestrol is that the toxic effects of in‐utero exposure (vaginal cancers in the female offspring and sub‐fertility in the male offspring) only became apparent when the offspring were young adults (Giusti 1995). Although this exposure was in the first trimester of pregnancy, there are more recent examples where erythromycin given to women in threatened preterm labour lead to an increase in the risk of cerebral palsy at age 7 (Kenyon 2008) and dexamethasone given to preterm neonates resulted in a excess of cerebral palsy with only limited beneficial effects in the neonatal period ([3]. These examples illustrate that long term effects of drug exposures in pregnancy may not be realised until many years, and perhaps decades, after exposure. These later effects are unlikely to prevent the use of antiretrovirals to prevent mother‐to‐child transmission, but as more regimens of similar efficacy become available, and as the majority of exposed children will remain free of HIV and survive into adulthood, the relative safety (both short term and long term) of different drugs will be important in guiding treatment choices. The importance of long‐term follow‐up of these cohorts of exposed children is therefore essential.

With ZDV‐based regimens, the length of treatment seems to influence transmission rates with a longer antenatal component being the most beneficial. In placebo‐controlled trials ZDV given to mothers from 36 weeks gestation and during labour (Thai‐CDC; RETRO‐CI) was more effective than ZDV given from 36 to 38 weeks, through labour and continued for 7 days postpartum (DITRAME), regardless of whether babies were breast fed. Similarly, in the PETRA study a combination of ZDV and 3TC given to mothers from 36 weeks gestation, during labour and continued for a week postpartum as well as to their babies for a week (regimen A) resulted in a greater reduction in HIV infection rate at 4 to 8 weeks compared with placebo than the same treatment but without an antenatal component (regimen B), a third arm using the combination treatment during labour only (regimen C) was found to be no better than placebo. Furthermore, a direct head‐to‐head comparison of ZDV given to mothers from 28 weeks gestation and during labour and to babies for 3 days after delivery ('long‐short' regimen) resulted in a lower risk of HIV infection or HIV infection or death at 6 months than ZDV given to mothers from 35 weeks and during labour and to infants for 6 weeks after delivery ('short‐long' regimen) (PHPT‐1). On the other hand, two smaller trials comparing longer and shorter zidovudine regimens found no difference between their effects on HIV infection rates in babies (Thistle 2004; Bhoopat 2005).

Antiretroviral treatment in the antenatal and intrapartum periods is not an option for HIV infected women who present late for delivery. In these cases, post‐exposure prophylaxis limited to babies can be valuable. In one trial, a single dose of NVP immediately after birth plus ZDV for the first 6 weeks after birth compared with a single dose of NVP only reduced infection rates by 37% at 4 to 8 weeks (Taha 2003). A further trial compared a single dose of NVP immediately after birth with ZDV given for the first 6 weeks and did not demonstrate a statistically significant difference in the rate of infection (Gray 2005).

Influence of breastfeeding

Breast milk is a source of HIV and has been shown to increase the risk of mother‐to‐child transmission of the virus. It is not clear to what extent breastfeeding influences the efficacy of antiretroviral treatment as only a few trials in breastfeeding populations (all of them in Africa) report HIV transmission rates beyond 4 to 8 weeks and the results of these studies are conflicting (DITRAME, PETRA; Thistle 2004; HIVNET 012). In the DITRAME trial the initial reduction in perinatal transmission was maintained at 18 months; this was also the case in the HIVNET 012 trial. On the other hand, in the PETRA study early reductions in HIV infection rates as well as HIV infection or death at 4 to 8 weeks declined and became statistically non‐significant by 18 months. One trial did not find a difference in HIV infection rates at any stage of follow up (Thistle 2004)

It should be noted that reliable information on the extent or type of breastfeeding (mixed versus exclusive) across studies was not available to us; neither were we able to take into account other factors influencing transmission risk, such as maternal viral load. A 2005 individual patient data meta‐analysis compared the efficacy of various antiretroviral regimens in reducing mother‐to‐child transmission risk at 6 weeks in African breastfeeding populations using multivariate methods to adjust for maternal CD4 count, breastfeeding and birthweight in (Leroy 2005). The authors concluded that the longest course of ZDV and 3TC (PETRA Regimen A) was more effective than shorter courses of the same drug combination and more effective than any other antiretroviral monotherapy regimen. These results concord with the findings of our review. Of note, is that the Mma Bana trial provided TRIPLE antiretrovirals to breastfeeding mothers and the low transmission rates in both the protease‐inhibitor‐based and the nucleoside reverse transcriptase inhibitor arms confirm that either TRIPLE antiretroviral combination was effective in maintaining maternal virologic suppression throughout breastfeeding.

Adverse effects

None of the trials in our review found statistically significant differences in the rates of serious or life threatening events in either mothers or babies between intervention groups. Mild transient anaemia seems to occur and is most frequently seen in mothers and babies exposed to the long course PACTG076 regimen. No cases of lactic acidosis or hepatic steatosis were found in the trial evaluating d4T and ddI, and nevirapine did not significantly increase the incidence of severe hepatoxicity or rash. Furthermore, antiretrovirals did not increase the likelihood of congenital defects, abnormal growth patterns or the occurrence of childhood cancers, although follow‐up was not sufficiently long in most of the trials to assess these later outcomes.

The emergence of resistant mutations following the use of antiretroviral regimens in the prevention of vertical transmission is, however, a cause for concern as this may potentially compromise future treatment of infected mothers and babies including future efforts to reduce mother‐to‐child transmission (Eshleman 2005a; Eshleman 2005b; Johnson 2005; Eshleman 2005c). A single dose of NVP was associated with an increase in resistant mutations in a subsample of mother‐infant pairs in the HIVNET 012 trial (Eshleman 2001). It has also been shown that a combination of zidovudine and 3TC does not prevent the emergence of 3TC resistance (Mandelbrot, 2001). These resistant mutations disappear with time and their clinical importance is therefore by no means clear, yet the finding that viral response to NVP treatment may be reduced in women previously receiving a regimen comprising a single dose NVP plus a short course of ZDV as prophylaxis is worrying (Jourdain 2004).

In this update, the completeness of the evidence is again limited by the lack of information pertaining to the emergence and duration of resistance. In a trial of ritonavir‐boosted lopinavir compared with NVP‐based therapy for treating HIV, 241 HIV‐infected mothers who had previously been exposed to single dose NVP six months or more before randomisation, were followed up for two years (Lockman 2010). In the group receiving NVP, women with resistance mutations to NVP detectable at treatment initiation had a 73% failure rates compared with 19% of those without resistance mutations. Findings were similar in a trial of 164 HIV‐infected children (Palumbo 2010) . GIven the high costs of antiretrovirals and the fact that in resource‐constrained settings many women present late to antenatal care, NVP will continue to play an important role in preventing mother‐to‐child transmission (Lallemant 2010). This will continue to have implications for clinicians determining which treatment to provide to HIV‐infected women and children. Lallemant and Jourdain question whether there is a role for resistance screening to identify individuals who would require non‐NVP based treatment (Lallemant 2010) but the costs of this are likely to limit its usefulness.

Overall completeness and applicability of evidence

This 2010 update added evidence from seven new trials to that of the previous 18 trials reported in the update in 2006. None of the new trials compared the same interventions with each other or comprised identical comparisons to those comparisons reported in the previous update.

Four of the new trials contributed data to the comparisons between different antiretroviral regimens of different durations first described in the previous review (Chi 2007; Thistle 2007; Chung 2005; Mashi). Three trials evaluated interventions in breast‐feeding populations and one in a non‐breastfeeding population. None of the new included data changed our previous results or interpretations significantly.

Two new trials contributed data to a new comparison between TRIPLE and other regimens (Chung 2008; Kesho Bora), both conducted in breast‐feeding populations. The remaining new trial, Mma Bana, provided data on comparing two types of TRIPLE regimens aimed at treating both the mother and preventing transmission to the infant and was conducted in a mixed‐feeding population.

The addition of these seven trials to previous evidence, raised a number of questions including the most effective combination of TRIPLE antiretrovirals and the optimal time to initiate TRIPLE antiretrovirals during pregnancy for both the mother and her baby. Future trials may provide clearer guidance but much of this information is likely to be derived from cohort studies, particularly in the light of the World Health Organization guidelines to commence prevention antiretrovirals at 14 weeks gestation (WHO 2009).

As almost all trials were conducted in African countries in populations of women representative of the general population, the level of applicability of the data from these trials can be viewed as high. However, the decision regarding which combination of antiretrovirals to use will likely be determined by resource availability, setting and costs and it may not always be possible to implement optimal recommendations from guidelines. In such situations results from our previous versions of this review remain pertinent.

Quality of the evidence

Heterogeneity of studies Comparison of the results of the trials included in this review is a challenging task given the differences in study populations, trial methodology and treatment regimens. In addition, the methods for assessing HIV transmission varies across studies; both the criteria for establishing HIV infection and the sensitivities of tests have changed over time (Hill 1999; Bartlett 2009). As a result of these differences, any indirect comparisons between regimens across trials must be treated with considerable caution. To reduce the amount of clinical heterogeneity between trials, we stratified our analysis according to intervention and choice of feeding. Despite this stratification, we were unable to statistically combine the results of the trials using meta‐analytic techniques. Given that many of the individual trials were small to moderate in size and that meta‐analysis was not possible we cannot rule out the play of chance as an explanation of the findings reported, neither can we rule out the possibility that important but infrequent adverse effects were missed. However, the increasing consistency in reduction of transmission rates using combined antiretrovirals over time provides good evidence in our opinion, to provide triple combination antiretrovirals to all HIV‐infected pregnant mothers.

Determination of infection status Determining exactly when transmission occurs is complex given that the age of infection is known only to be between the ages at the last negative and first positive tests (Ghent 2001). The risk of paediatric infection over time also depends on whether or not a mother continues to breastfeed, but few trials collect and provide accurate information on the exact age of complete weaning (Ghent 2001). In some trials, it has been noted that infants may become infected well after the reported time of weaning (Ghent 2001). The Ghent Working Group, comprising methodologists and trialists, has developed a consensus approach to the statistical analysis applicable across all MTCT trials to facilitate reviewing and meta‐analysing such trial data. They stress the importance of good quality information about age at weaning to estimate transmission probabilities in breastfeeding populations in future trials. They also provide a framework for statistical analysis using interval‐censored data similar to what we employed in this review. It is important to note that interval‐censored data analysis as recommended by the Ghent group does not estimate the 'time to infection' but 'time to detection' of HIV, which is always dependent on the testing schedule. Although we made this observation in our previous update, none of the included trials used the recommended standardised testing schedules which would allow for better comparisons to be made between trials.

Multiple births Because inclusion of all children from multiple births in an analysis violates the requirement of most statistical methods that observations be independent of one another (Ghent 2001), trials have to account for this and we described the different methods each trial used to do this in our review (see Table of Included Studies). However, the Ghent group recommend that all future trials use the first born baby of each multiple birth in studies of MTCT transmission. As observational studies have shown that the first born of twins has a higher risk of infection (Ghent 2001), using this method of analysis would provide a more conservative estimate of efficacy but would ensure consistency between study comparison groups and between trials.

Short‐term and long‐term efficacy HIV infection in specimens collected between the first day of the neonatal period and day 60 can be used to effectively assess short‐term efficacy (Alioum 2003). Long‐term efficacy estimates of HIV infection are more problematic. Estimates may be biased if only one test is conducted between 15 and 24 months of age because loss to follow‐up and deaths are unlikely to be evenly distributed between trial arms (Alioum 2003). It has therefore been suggested that HIV‐free survival provides a less biased estimate of long‐term efficacy (Ghent 2001). The difficulties of providing such long‐term follow‐up beyond trial termination may not always be feasible where public health resources are limited and such ideal information may rarely be available.

Potential biases in the review process

As recommended for all Cochrane reviews we conducted an extensive and comprehensive search of multiple electronic bibliographic databases and searched conference abstracts and conducted hand‐searching of relevant medical journals to identify eligible trials for inclusion. Where possible, we also engaged directly with trialists to obtain missing data or for clarification where this was required. Completed but as yet unpublished trial data, such as that of the trial reported in Characteristics of ongoing studies will be included in the following update of this review. Our ongoing interaction with investigators of trials and our regular communication with global policymakers at WHO, ensured we were aware of all trial activity in this field. Furthermore, given the high profile nature of the intervention, it is unlikely that our search strategy failed to detect existing current trial evidence.

Potential bias in the conduct of our review was also minimised by having two independent researchers extracting data and assessing the methodological quality of each study. This detailed process allows for a thorough assessment of trial conduct and an exploration of the possible biases that may be present in each trial. In this update we re‐evaluated the possible risk of bias in four domains for not only the new trials, but also for all of the previously included 18 trials, ensuring that newer methods were incorporated into the overall assessment. For many of the trials reported in earlier years, quality of reporting of these domains is poor and, as such, we were compelled to judge many of the risks as unclear. Empirical evidence has found that poor quality of reporting is not always associated with actual poor conduct and our judgments must be viewed in this light.

The review is of aggregate trial data. Published aggregate data from trials can be variable in the completeness of outcome reporting and in definitions of both the intervention and outcomes used between trials. This is particularly relevant where survival analysis has been used for outcome reporting and individuals have been censored at different endpoints during a trial, as occurred in many of the included trials. Ideally we would have obtained the individual patient data (IPD) from each trial allowing us to calculate efficacy measures using identical methods across each trial. Differences in analytical methodology does and can make a difference. For instance our results for transmission rates from those trials included in both this review and the Sturt 2010 Cochrane review differ slightly. This is because the Sturt 2010 review calculated relative risks from the presented data rather than incorporating the reported survival curves into the analysis as was done in our review. While the latter is arguably desirable given that the time‐to‐event data is accounted for, the differences in methods to calculate the curves across trials means that greater variation may have been introduced into the final calculations. Obtaining IPD would eliminate this potential source of bias.

Implications for practice.

Implications for research.

Comparison 1. Antiretrovirals versus Placebo.

Comparison 2. Longer vs shorter regimens of the same antiretrovirals.

Comparison 3. Regimens of different antiretrovirals and durations of treatment.

Comparison 4. TRIPLE regimens versus other.

Comparison 5. TRIPLE regimen vs TRIPLE regimen.

Characteristics of included studies [ordered by study ID]

Characteristics of excluded studies [ordered by study ID]

Characteristics of ongoing studies [ordered by study ID]

Internal sources

• Biostatistics Unit, Medical Research Council, South Africa.

• South African Cochrane Centre, South Africa.

External sources

• World Health Organization, Switzerland.

  #200123310. Systematic reviews and development of GRADE profiles, based on the 2008 WHO GRC guidelines, for the "WHO Guidelines on Antiretroviral Drugs for Treating Pregnant Women Living with HIV and Preventing HIV Infection in Infants ‐ 2009 revision"

References to studies included in this review