Factors associated with mother-to-child transmission of the human immunodeficiency virus in Pernambuco, Brazil, 2000–2009


Corresponding Author Pedro Alves da Cruz Gouveia, Programa de Pós-Graduação em Medicina Tropical, Bl. A – Térreo do HC/UFPE, Cidade Universitária, Av. Prof. Moraes Rego, s/n. 50670-901 Recife/PE, Brazil. E-mail: pedroalves@doctor.com



To identify risk factors associated with mother-to-child transmission of HIV in the Brazilian state of Pernambuco.


Retrospective cohort study with 1200 HIV-exposed children born in Pernambuco, registered up to the age of 2 months in a public programme to prevent vertical transmission. Univariate and multivariate logistic regression analyses were conducted for maternal and peripartum characteristics and prophylactic interventions, to identify risk factors for mother-to-child transmission of HIV.


The transmission rate was 9.16% (95% CI: 7.4–10.9). The following risk factors were independently associated with transmission: non-use of antiretroviral during pregnancy (OR: 7.8; 95% CI: 4.1–15); vaginal delivery (OR: 2.02; 95% CI: 1.2–3.4); prematurity (OR: 2.5; 95% CI: 1.3–4.7); and breastfeeding (OR: 2.6; 95% CI: 1.4–4.6).


This mother-to-child transmission rate is unacceptably high, as prophylactic interventions such as antiretroviral therapy and infant feeding formula are free of charge. Absence of antiretroviral therapy during pregnancy was the main risk factor. Therefore, early identification of exposed mothers and initiating prophylactic interventions are the main challenges for controlling transmission.



Identifier les facteurs de risque associés à la transmission mère-enfant du VIH dans l’état brésilien de Pernambuco.


Etude rétrospective de cohorte de 1200 enfants exposés au VIH, nés à Pernambuco et enregistrés jusqu’à l’âge de deux mois dans un programme public pour la prévention de la transmission verticale. Des analyses de régression logistique univariée et multivariée ont été effectuées pour les caractéristiques maternelles et péri-partum et les interventions prophylactiques, afin d'identifier les facteurs de risque pour la transmission mère-enfant du VIH.


Le taux de transmission était de 9,16% (IC95%: 07,04 à 10,09). Les facteurs de risque suivants étaient indépendamment associés à la transmission: la non-utilisation des antirétroviraux pendant la grossesse (OR: 7,8; IC95%: 4,1 à 15), l'accouchement par voie vaginale (OR: 2,02; IC95%: 1,2–3,4), la prématurité (OR: 2,5; IC95%: 1,3 à 4,7), l'allaitement maternel (OR: 2,6; IC95%: 1,4 à 4,6).


Ce taux de transmission mère-enfant est trop élevé, car les interventions prophylactiques telles que la thérapie antirétrovirale et l'alimentation du nourrisson au biberon sont gratuits. L'absence de traitement antirétroviral pendant la grossesse était le principal facteur de risque. Par conséquent l'identification précoce des mères exposées et l'initiation des interventions prophylactiques sont les principaux défis pour contrôler la transmission.



Identificar los factores de riesgo asociados con la transmisión madre-hijo del VIH en el estado Brasilero de Pernambuco.


Estudio retrospectivo de cohortes con 1,200 niños expuestos al VIH y nacidos en Pernambuco, registrados hasta una edad de dos meses dentro de un programa público para prevenir la transmisión vertical. Se realizaron análisis de regresión logística univariada y multivarida para las características maternas y del periparto e intervenciones profilácticas, con el fin de identificar los factores de riesgo asociados a la transmisión madre-hijo del VIH.


La tasa de transmisión era del 9.16% (IC 95%: 7.4–10.9). Los siguientes factores de riesgo estaban independientemente asociados con la transmisión: el no utilizar antirretrovirales durante el embarazo (OR: 7.8; IC 95%: 4.1–15); el parto vaginal (OR: 2.02; IC 95%: 1.2–3.4); la prematuridad (OR: 2.5; IC 95%: 1.3–4.7); y el amamantar (OR: 2.6; IC 95%: 1.4–4.6).


Esta tasa de transmisión madre-hijo del VIH es inaceptablemente alta, puesto que intervenciones profilácticas tales como la terapia antirretroviral o la leche de fórmula para alimentar al recién nacido son gratuitas. La ausencia de terapia antirretroviral durante el embarazo es el principal factor de riesgo. Por lo tanto la identificación temprana de las madres expuestas y el inicio de las intervenciones de profilaxis son los principales retos para controlar la transmisión.


Brazil was the first developing country to introduce a national programme for the prevention of mother-to-child transmission of HIV (MTCT) (Veloso et al. 2010). The protocol provides HIV testing during pregnancy and delivery, highly effective antiretroviral treatment (HAART) at an early stage, between the 14th and 28th week of gestation, Zidovudine (ZDV) during labour, ZDV for infants up to the sixth week of life and infant feeding formula (Brazilian Ministry of Health 2009, 2010), as in Europe and the United States These prophylactic interventions have led to the gradual reduction in reported HIV cases in children throughout the country after 1997 (Brito et al. 2006; Ramos et al. 2011).

In Brazil, initial studies showed a MTCT rate of 16% (Tess et al. 1998), but there has been a significant decrease over recent years. In a national multicentre study covering 2924 children (Succi 2007), the MTCT rate was 8.6% in 2000 and 7.1% in 2001, with some regional differences: the north and north-east regions had a lower proportion of antiretroviral use during pregnancy and also higher rates of transmission. In recent studies undertaken in the south and south-east regions, the wealthiest parts of Brazil have demonstrated lower MTCT rates of between 2.4% and 4.9% (João et al. 2003; Kakehasi et al. 2008; Tornatore et al. 2010; Matida et al. 2011).

Despite this progress, prevention has not yet been universally applied on a nationwide basis (Succi 2007; Kerr et al. 2011). Effective prophylactic intervention has not yet been fully achieved in Brazil, owing to the failure to provide prenatal care (Nishimoto et al. 2005), a lack of access to early HIV testing during pregnancy (Ramos et al. 2009), the diagnosis of maternal HIV during delivery or postnatal period (Succi 2007) and the inability to provide antiretroviral therapy for all mothers and children (João et al. 2003; Kakehasi et al. 2008; Veloso et al. 2010).

Another way to address the issue is the detection of the disease in children. The incidence of AIDS in children under 5 years dropped from 5.3 to 3.5 per 100 000 inhabitants between 2000 and 2009 in Brazil. In the north-east, however, it rose over the same period from 1.4 to 2.6. The State of Pernambuco has followed the trend of the north-east, with an increase from 2.4 to 3.4 (Brazilian Ministry of Health 2011). This increase in reported cases of AIDS in children under 5 years in the north-east reflects an inability to reduce and control MTCT.

In Brazil, there is a disparity between the existence of internationally recognised protocols to reduce MTCT and the inability to fully apply them in certain parts of the country, especially in the north-east, where coverage of health services is poor (Brazilian Ministry of Health 2010). Poverty, especially in a heterogeneous country such as Brazil, seems to be a strong determinant of MTCT, as it is associated with inadequate care during pregnancy (Barcellos et al. 2009). These situations represent an incomplete process of epidemiological and demographic transition, with the coexistence of old and new clinical and epidemiological challenges (Ramos et al. 2011). Our aim was to identify risk factors associated with MTCT in the State of Pernambuco, in the north-east, which is one of the most underprivileged regions in the country.


Study design

A retrospective cohort study was carried out at the Instituto de Medicina Integral Professor Fernando Figueira (IMIP), the referral centre in the State of Pernambuco for paediatric HIV, from January 2000, when the compulsory registration of pregnant women with HIV became law in Brazil, to December 2009. IMIP is a teaching and research hospital that exclusively treats patients from the Brazilian public healthcare system. The study included children born in Pernambuco and registered before 2 months of age to the MTCT programme at IMIP. All children were followed up for at least 18 months, until June of 2011. Children were excluded from the study if their mothers did not have a confirmed diagnosis of HIV, if they were symptomatic and were referred to the programme with suspected HIV/AIDS or if they were second-born twins. Children whose mothers had a positive rapid HIV test at delivery and who subsequently tested HIV negative were excluded as they had not actually been exposed to MTCT. Children referred with suspected HIV/AIDS were also excluded to avoid a selection bias that could artificially increase transmission rate, as IMIP is the main referral centre in the state for paediatric AIDS. Finally, second live-born twins were also excluded from the study because they share the same epidemiological characteristics as their siblings.

HIV-1 infection was defined using a modified algorithm from the Brazilian Ministry of Health, on children who presented a detectable viral load in two samples after 1 month of life, a persistent positive antibody test for HIV-1 after 18 months or the presence of any class B or C diagnosis according to Centers for Disease Control and Prevention paediatric HIV-1/AIDS case definition, in the absence of laboratory confirmation. Children were defined as uninfected if they provided two samples with an undetectable viral load after the first month of life or a negative serological test after 12 months. If a child had been breastfed, the algorithm began 2 months after the cessation of breastfeeding, as a way of minimising the occurrence of negative results before interrupting exposure to maternal milk. If a child had not completed any of the above steps and the HIV antibody test was not available after 12 months, it was considered that the child had not completed follow-up and was therefore excluded from the MTCT rate calculation and the risk factor analysis. Analysis was carried out after comparing the characteristics of the children studied with those who were not followed up to identify any potential selection bias.

Data collection

The study received approval from the IMIP Research Ethics Committee. Information was initially obtained from the records database at IMIP, where the compulsory notification of MTCT is registered, and was complemented by notes from medical charts collected by the researcher from the hospital archives. Maternal data included age, educational level, place of residence, illicit drug use, moment of HIV diagnosis and prenatal care. Peripartum data included mode of delivery, duration of ruptured membranes, gestational age, birthweight, date of childbirth and breastfeeding. Antiretroviral data comprised time of initiation, therapeutic regimen used by the pregnant woman, use of antiretroviral during the antenatal period, delivery and the newborn. Data were inserted in an Excel spreadsheet and systematically checked for internal consistency. The final database was converted and analysed using Stata-10.0 (STATA, College Station, TX, USA).

Statistical analysis

Univariate analysis was performed for maternal and peripartum characteristics and prophylactic interventions to identify factors associated with MTCT. For each variable, only those individuals for whom information was available were considered. To measure the association between risk factors and MTCT, the odds ratios (OR) were calculated with a 95% confidence interval (CI) and chi-squared tests or Fischer's exact test were performed.

A multivariate logistic regression analysis was performed with the inclusion of independent variables with P < 0.2 in the univariate analysis (Katz 1999). In the multivariate model, the variable ‘moment of initiating antiretroviral therapy during pregnancy’ was deleted because it was strongly correlated with the variable ‘antiretroviral use during the antenatal period’. To identify the model with the best adjustment and biological plausibility, the stepwise forward method was used for blocks of risk factors. Variables were analysed in three blocks: maternal factors (place of residence, educational level, time of maternal HIV diagnosis and prenatal care), peripartum factors (mode of delivery, duration of ruptured membranes, gestational age, birthweight and breastfeeding) and prophylactic interventions (antiretroviral use during antenatal period, ZDV administered at delivery and to the newborn). As information was unavailable for some individuals in one or more variables, 138 cases (13.6%) were not included in the final multivariate analysis. The variables that remained in the model for each block with P < 0.05 were analysed using a general model of multiple logistic regression, and all findings had a significance level of 5%.


A total of 1297 children met the inclusion criteria for the study, although 97 were excluded: 35 with a false-positive maternal diagnosis, 54 in the postnatal period referred with symptoms suggestive of HIV/AIDS and eight twins (Figure 1). The retrospective cohort comprised 1200 children, of whom 185 had not been fully followed up to investigate the infection (an incomplete follow-up rate of 15.4%). Of the 1015 children who completed the follow-up, 93 were infected, corresponding to a MTCT rate of 9.16% (95% CI: 7.39–10.93).

Figure 1.

Study Flow Chart.

Table 1 compares the characteristics of the 1015 mother–child pairs who completed follow-up with the 185 pairs who were lost to follow-up. The two groups were similar with regard to the main risk factors associated with HIV transmission, although the mothers who did not complete follow-up were mostly from the interior of Pernambuco, had started prophylaxis during delivery and had a higher prevalence of illicit drug use.

Table 1. Characteristics relating to mother, peripartum and prophylactic interventions by follow-up of children exposed to mother-to-child transmission of HIV in Pernambuco, Brazil, 2000–2009
 Complete follow-up (n = 1015)Lost follow-up (n = 185)χ2P-value
N % N %
  1. ARV, antiretroviral; DORM, duration of ruptured membranes; HAART, highly effective antiretroviral therapy; ZDV, zidovudine.

  2. a

    Data collected for 696 women.

  3. b

    Data collected for 678 women.

Maternal age
<20 years11278.903021.103.790.052
≥20 years88085.2015314.80
Unknown23 2 
Educational level of mother
High school education31087.104512.90
Higher education2890.3039.70
Unknown50 4 
Place of residence
Metropolitan region37886.16113.9
Illicit drugs use
Unknown104 13 
Time of maternal HIV diagnosis
Before pregnancy37084.17015.96.950.074
During pregnancy39886.36313.7
At delivery14078.73821.3
After delivery7788.501011.50
Unknown30 4 
Prenatal care
Unknown100 25 
Mode of delivery
Unknown2 1 
<1 h60384.710915.30.310.577
≥1 h27786.04514.0
Unknown135 31 
<2500 g17489.22110.82.570.109
≥2500 g75284.813515.2
Unknown89 29 
Gestational age
<37 weeks11385.02015.00.060.798
≥37 weeks77985.812914.2
Unknown123 36 
Unknown6 2 
Child′s year of birth
Moment of initiation of ARV prophylactic
During prenatal67586.510513.514.610.002
At delivery13175.34324.7
In the newborn9982.52117.5
Not used10586.801613.20
Unknown5 0 
ARV regimen used in the pregnant womana
Dual-drug therapy1688.9211.1
Single ZDV3480.9819.1
Moment of initiation of ARV during pregnancyb
1st trimester22689.32710.70.190.909
2nd trimester23188.23111.8
3rd trimester14488.31911.7
ZDV administered at delivery
Unknown58 9 
ZDV administered to the newborn
Unknown5 1 

Characterisation of the sample who completed follow-up

The mean age of mothers was 26 (95% CI: 25.65–26.35), ranging from 15 to 45 years. Most women had only basic education (65.0%), were from the capital city and its metropolitan region (76.7%) and had received prenatal care (92.3%). Among the 911 women for whom information was available regarding illicit drug use, 8.6% reported current or prior use. Most mothers received the HIV diagnosis before delivery: 40.4% during pregnancy and 37.6% before pregnancy. Maternal diagnosis occurred during delivery in 14.2% of cases and after childbirth in 7.8%. Antiretroviral was started during pregnancy in 66.8% of cases, at delivery in 13.0%, with the newborn in 9.8%, and in 10.4% of cases no treatment was started. Among the 601 mothers on antiretroviral therapy, 76.0% initiated before the second trimester, and HAART regimens were used in 91.8%.

Most deliveries were caesarean (68.4%) and, among 880 cases with a record of duration of ruptured membranes, duration was <1 h in 68.5%. Intravenous ZDV was administered during delivery in 75.0% of 957 cases who presented this information. A minority of children was born preterm at <37 weeks (12.7% of 892 registers) and with low birthweight, characterised as <2500 g (18.8% of 926 registered). Maternal breastfeeding occurred in 10.1% of cases and 87.5% of the children were receiving ZDV prophylaxis.

Factors associated with MTCT

Table 2 compares infected and uninfected children, illustrating the association with risk factors. After the data were subjected to univariate analysis, the variables with no statistically significant association with the outcome were removed, and a multivariate analysis was performed. The variables that remained in the multivariate logistic regression model with P < 0.05 for each block of risk factors (maternal factors: time of maternal HIV diagnosis and prenatal care; peripartum factors: mode of delivery, gestational age and breastfeeding; and prophylactic interventions: antiretroviral use during antenatal period and ZDV administered to the newborn), were analysed using a general model of multiple logistic regression. The following factors remained independently associated with MTCT: the non-use of antiretrovirals during pregnancy (OR 7.84; 95% CI 4.1–15); vaginal delivery (OR 2.02; 95% CI 1.19–3.43); preterm birth with gestational age <37 weeks (OR 2.49; 95% CI 1.31–4.73); and breastfeeding (OR 2.58; 95% CI 1.45–4.58), as described in Table 3.

Table 2. Univariate analysis of risk factors relating to mother, peripartum and prophylactic interventions associated with mother-to-child transmission of HIV in Pernambuco, Brazil, 2000–2009
 Infected child (n = 93)Uninfected child (n = 922)Odds-ratioχ2P-value
N % N %CI 95%
  1. ARV, antiretroviral; DORM, duration of ruptured membranes; ZDV, zidovudine.

≥20 years819.279990.810.0090.924
<20 years108.910291.10.97 (0.49–1.93)
Unknown2 21  
Educational level
>8 years of study247.131492.912.790.096
≤8 years of study6510.456289.61.51 (0.93–2.47)
Unknown4 46  
Place of residence
Capital or metropolitan region658.3071491.7012.70.101
Interior2811.9020888.101.48 (0.92–2.36)
Illicit drugs use
Yes810.37089.71.09 (0.51–2.35)
Unknown6 98  
Time of maternal HIV diagnosis
Before/during pregnancy364.773295.3189.12<0.001
At/after delivery5625.816174.27.07 (4.50–11.12)
Unknown1 29  
Prenatal care
No1927.15172.94.26 (2.38–7.62)
Unknown6 94  
Mode of delivery
Vaginal6018.726081.34.62 (2.95–7.23)
Unknown0 2  
<1 h335.557094.5127.27<0.001
≥1 h4516.223283.83.35 (2.08–5.38)
Unknown15 120  
≥2500 g678.968591.114.70.032
<2500 g2514.414985.61.72 (1.05–2.81)
Unknown1 88  
Gestational age
≥37 weeks688.771191.317.330.008
<37 weeks1916.89483.22.11 (1.22–3.67)
Unknown6 117  
Yes3332.36967.77.00 (4.28–11.46)
Unknown2 4  
Child′s year of birth
2000–20044010.035990.01.18 (0.77–1.82)
Moment of initiation of ARV during pregnancy
Up to 2nd trimester92.142897.91106.29



3rd trimester onwards53.115996.91.49 (0.49–4.53)
No treatment7823.225876.814.38 (7.09–29.16)
Unknown1 77  
ARV use during antenatal period
No7723.125676.914.03 (7.80–25.25)
Unknown2 13  
ZDV administered at delivery
No5221.818778.25.77 (3.63–9.19)
Unknown8 50  
ZDV administered to the newborn
No3628.69071.45.91 (3.69–9.48)
Unknown1 4  
Table 3. Multivariate analysis of risk factors associated with mother-to-child transmission of HIV in Pernambuco, Brazil, 2000–2009
 Adjusted odds ratio95% confidence intervalP-value
  1. ARV, antiretroviral.

Preterm birth (gestational age <37 weeks)2.491.31–4.730.005
Vaginal delivery2.021.19–3.430.009
Non-use of ARV during pregnancy7.844.10–15.01<0.001


The MTCT rate in the state of Pernambuco was 9.16% (95% CI: 7.39–10.93), an average rate compared with that in a Brazilian multicentre study for the north-east region of 7.9% in 2000 and 14.3% in 2001 (Succi 2007), and very similar to a rate of 9.9% recorded in Manaus in the northern region, between 2007 and 2009 (Soeiro et al. 2011). It should be emphasised that the MTCT rate encountered is high, even with the exclusion of those children referred on suspicion of HIV infection. There would appear to be no reduction in HIV transmission in the north-east, as the prevalence of HIV in children younger than 2 years is similar to that in children aged 3–4 years and 5–12 years (Kerr et al. 2011). In the south-east, however, the wealthiest region of the country, there has been a tendency for MTCT to decrease: the rate was 2.7% in Sao Paulo in 2006 (Matida et al. 2011); and in Minas Gerais, it fell from 20% in 1998 to 3% in 2005 (Kakehasi et al. 2008). One study in Rio Grande do Sul between 2003 and 2007 encountered a transmission rate of 4.8%, and the decline over the years was attributed to the successful control of factors, which cause peripartum and postnatal transmission (Tornatore et al. 2010). This wide variation in transmission rates in different regions of the country translates major social, operational and economic inequalities, and the regional differences of coverage and quality of health services in terms of prenatal care and the adoption of measures to prevent MTCT (Nemes et al. 2009; Ramos et al. 2011). For example, children monitored in the north-eastern region were those who were proportionately more likely to be breastfed (Succi 2007). Various studies have pointed to maternal diagnosis at delivery or postnatal as an independent risk factor for MTCT (Succi 2007; Birkhead et al. 2010) and reflect the quality of prenatal health services. The estimated coverage for HIV testing in 2006 during pregnancy in Brazil ranges from 40.6% in the north-east to 85.8% in the south (Szwarcwald et al. 2008). Although 92% of mothers in our sample had received prenatal care, more than 20% were only diagnosed at delivery or later. Our investigation is limited, as the quality of prenatal care provided, such as the number of consultations and tests carried out, was not evaluated.

In 15% of the children in the present cohort, it was not possible to determine HIV transmission, and this loss of follow-up could have led to a selection bias. However, if there was a bias, it appears to have had little impact on the results, because a comparison of children with unknown and known infectious status showed similar main risk factors associated with MTCT. Children with an unknown infectious status were mostly from the interior of Pernambuco and linked to a higher prevalence of maternal illicit drug use. The reasons for missing follow-up appointments, namely transport difficulties, distance from health facility, prior use of illicit drugs as a non-adherence marker for the PMCT programme, have all been described (Cohn et al. 2008; Ong'ech et al. 2012).

Antiretroviral therapy has proven to be the prophylactic MTCT intervention with the greatest impact of reducing the number of AIDS cases in infancy (Amaral et al. 2007; Tornatore et al. 2010), because it reduces HIV RNA to undetectable levels in mothers. In the present study, non-use of antiretroviral therapy during pregnancy remained in the final model and was the main risk factor for transmission with an OR = 7.8 (95% CI 4–15), similar to that found in other Brazilian studies (João et al. 2003; Fernandes et al. 2010). Some studies suggest that HAART during pregnancy is superior to ZDV therapy alone in reducing MTCT (Cooper et al. 2002; Peters et al. 2008). Although only two-thirds of pregnant women from the sample had received antiretroviral therapy, more than 91% had received the HAART regimen. Thorne et al. (2005) showed that, with regard to the increased use of antenatal HAART, there was a significant trend over time towards a higher prevalence of undetectable HIV RNA levels, from 29% to 50% with P < 0.001.

Preterm birth, in relation to exposure of the infant to HIV, is a multifactor event and may be either a consequence of the mother's disease or a cause of the child's infection. The association between preterm birth and AIDS in the mother has been reported in several studies. Advanced maternal HIV disease during pregnancy has higher rates of adverse outcomes, including preterm birth (Tuomala et al. 2002; Marazzi et al. 2011). A statistically significant association has also been demonstrated between preterm birth and MTCT in international studies (John et al. 2001; Tournoud et al. 2008). Charurat et al. (2009) suggest that the increased susceptibility of preterm children with HIV infection results from the immaturity of the mucous barriers and the immune system. A study in the city of Santos, Sao Paulo, found an association between preterm birth and MTCT (Nishimoto et al. 2005); however, the present study is the first in Brazil to find such association using a final logistic regression model (OR 2.49; 95% CI: 1.31–4.73). Nevertheless, these conclusions are limited because it was not possible to control potential confounding factors, as the stage of the disease in the mother was not evaluated in our model. Although preterm birth increases the likelihood of HIV transmission, more than half of obstetricians in one state capital in the north-east region were unaware of this (Farias et al. 2008), which may reduce the chances of minimising this risk factor.

Vaginal delivery had a twofold greater risk of MTCT in the study population, a finding consistent with Brazilian data: a national multicentre study showed a fairly similar twofold increase in transmission by this mode of delivery (Succi 2007), and in Belo Horizonte, the OR was 3.3 (95% CI 1.8–5.9) for vaginal delivery (Kakehasi et al. 2008). Although the present study did not differentiate between elective and emergency caesarean section, vaginal delivery increases the risk of transmission when compared with any kind of caesarean procedure. Studies in state capitals of the Brazilian north-east show that the main obstacles for prophylactic interventions during delivery are a lack of antenatal information, unawareness of HIV status among pregnant women and the unavailability of an opportune rapid test (Farias et al. 2008; Ramos et al. 2009). Obstetric practices are still therefore a determining factor in MTCT control in the north-east.

Children who were breastfed in this cohort had a 2.6-fold greater risk of HIV infection, even when the effect of breastfeeding was adjusted for other variables. Other authors have also found this association in Brazilian studies (Succi 2007; Kakehasi et al. 2008; Matida et al. 2011). Coutsoudis et al. (2004) in a meta-analysis of nine studies of populations where breastfeeding is common suggests that late postnatal transmission represents at least 24% of MTCT, with a continued risk of transmission throughout the breastfeeding period reaching a cumulative probability of 9.3% in 18 months. Although we did not measure the duration of breastfeeding, any reported exposure to breast milk in itself constituted a risk factor for transmission, as reported by other authors (John et al. 2001; Martino et al. 2002). It should be noted that in a country where the public health system provides infant feeding formula free of charge, about 10% of children were breastfed. This indicates that a late diagnosis of the mother leads to a lost opportunity of advising against breastfeeding.

This study has the limitation of having been conducted at a single centre, and the study population could not be representative of children exposed to MTCT throughout the state of Pernambuco. However, this may be offset, as the chosen centre, IMIP, receives the greatest regional demand for HIV-exposed children. During the study period, IMIP attended around 60% of all infected children <5 years in the state. Another limitation was that in 12% of the mother–child pairs, there was a lack of information regarding the duration of ruptured membranes and gestational age, both of which are variables that remained in the final multivariate model. The analysis was repeated, taking into account cases where the values for these two variables were unknown, and there was no change in the levels of association. It is therefore likely that 12% of pairs with missing information regarding these two variables demonstrated a distribution of exposure levels similar to those observed in pairs for whom this information was available.

The MTCT rate observed was relatively high, which is unacceptable considering that the Brazilian Ministry of Health provides free prophylactic interventions, including antiretroviral and infant feeding formula. While some regions of the country have been successful in controlling factors that cause peripartum and postnatal transmission, breastfeeding and the mode of delivery are still determining factors for HIV transmission in Pernambuco. Special attention should be given to prenatal care as a way of providing the early identification of exposed mothers, therefore beginning prophylactic interventions in a timely manner, as the absence of antiretroviral therapy during pregnancy was the main risk factor for MTCT. Therefore, it may be necessary to institute additional public health measures to reduce inequality in MTCT rates throughout the country. To establish the active surveillance of pregnant women at risk and for the health system to perform rapid HIV diagnostic tests, it is of utmost importance to improve the coverage of prenatal care in poor regions, such as the north-east. A further challenge would be to expand the number of high-risk maternity hospitals in the interior, which are qualified to treat pregnant HIV women. Maternity units need to ensure that these women receive preferential care and safe obstetric practices. It is clear that only greater investment in maternal and child health care in the north and north-east will be able to bring about homogeneous reduction in MTCT in Brazil.