Primary HIV-1 infection in African children infected through breastfeeding

Objectives: To describe acute retroviral syndrome and associated primary viraemia in African children infected with HIV-1 through breastfeeding.

Design: Matched case-control study performed retrospectively within the ANRS 049 DITRAME project conducted in 1995-1998 in Abidjan, Côte d'Ivoire.

Methods: Cases were children infected by HIV-1 postnatally through breastfeeding. All were HIV-1 negative by DNA PCR at least 45 days of age, but positive on a subsequent sample. This period was considered as surrounding the estimated date of postnatal contamination. Signs/symptoms occurring within this period were recorded in cases and compared with those occuring during the same time period in uninfected breastfed children (controls). For cases, plasma specimens were tested for HIV-1 plasma RNA using the branched DNA assay.

Results: Of 22 infants infected postnatally (median age at first positive sample, 185 days; range, 87-373 days), 21 (95.5%) exhibited at least one clinical sign, compared with only 27 of the 44 (61.4%) uninfected children (P = 0.003). Three independent factors were associated with primary HIV-1 infection: mononucleosis-like syndrome [odds ratio (OR), 8.3; 95% confidence interval (CI), 1.4-47.8], dermatitis (OR, 6.0; CI, 1.1-31.9), and generalized lymphadenopathy (OR, 26.5; CI, 2.0-348.4). Among cases, initial median plasma HIV-1 RNA viral load was 5.92 log₁₀ copies/ml; this declined to 4.96 log₁₀ 12 months after the first positive viral load.

Conclusions: These results may be useful for the recognition of early paediatric cases of postnatal transmission in Africa and could enable targeting of those who should benefit from HIV RNA or DNA testing for primary HIV-1 infection and their subsequent care.

Introduction

In HIV-1 infection, an initial acute retroviral syndrome (ARS), when present, is characterized by various clinical symptoms, including a mononucleosis-like illness (MLI) occurring around the time of seroconversion [^1,2]. The rate of progression to AIDS is already determined during the first year of infection [³] and is higher among HIV seroconverters with ARS than among those with asymptomatic seroconversion [⁴].

Many of the studies related to primary HIV-1 infection (PHI) have been performed in developed countries, especially among adults (i.e., homosexual men and injecting drug users). By contrast, little is known about this syndrome in children. Indeed, because of the timing of infection (in utero or perinatally) in industrialized countries [⁵], it is difficult to assess the clinical picture of an acute illness associated with HIV seroconversion. In sub-Saharan Africa, where more than 90% of the HIV-infected children are living, the timing of mother-to-child transmission (MTCT) of HIV-1 is substantially different: about one-third of all transmission occurs as a result of postnatal transmission (PT) through breastfeeding [⁶]. Improved understanding of the clinical course of PHI through PT is important, especially in the context of interventions aimed to reduce MTCT, as recently developed and evaluated among breastfed populations in developing countries [^7-9].

We compared clinical features in children infected postnatally by breastfeeding and in uninfected children, born to HIV-seropositive mothers recruited in the ANRS 049 DITRAME project on the prevention of MTCT performed in 1995-1998 in Côte d'Ivoire. The main goals of this study were to estimate the frequency of ARS associated with paediatric HIV-1 infection, and to describe its symptomatology and associated viral parameters.

Methods

Subjects

The present study was conducted retrospectively within two clinical trials and in an open-label zidovudine (ZDV) cohort performed between September 1995 and November 1998 as follows. (i) The DITRAME ANRS 049a trial was a multi-centre, randomized, double-blind, placebo-controlled trial of ZDV for the prevention of MTCT of HIV-1 in Côte d'Ivoire and Burkina-Faso [⁷]. Women (n = 250) randomized at 36-38 weeks' gestation received in the peripartum period ZDV or a matching placebo. (ii) The DITRAME ANRS 049b trial performed in 40 women living in Côte d'Ivoire was a randomized, double-blinded, placebo-controlled trial using benzalkonium chloride vaginal cleansing [¹⁰]. (iii) The open-label ZDV cohort followed the DITRAME ANRS 049a trial. Women (n = 137) received the same ZDV regimen as described for the 049a trial [¹¹].

No ZDV treatment was given to the neonates, and all liveborn infants (n = 418) except 14 were breastfed by their mothers. Mean breastfeeding duration was 210 days (range, 0-540 days). Blood samples and clinical information on children were collected at scheduled visits carried out at day 1 or 8, then at day 45, 90, and every 3 months until 18 months of age. At each visit, physical examination was conducted by two trained and dedicated clinicians. Clinical disease was classified according to the 1994 Centers for Disease Control classification system for HIV infection [¹²]. Paediatric HIV-1 infection was diagnosed using a qualitative home-made HIV-1 DNA PCR. All samples collected at day 180, or earlier were systematically processed by PCR. If positive, timing of infection was retrospectively assessed from all previous available samples. If negative, no additional virologic analysis was performed, except if HIV serology was positive at 18 months of age. PCR tests were then performed on all samples collected quarterly after 6 months.

Definitions and clinical symptoms collection

The design of the present study, which used data from Côte d'Ivoire only, was a matched case-control study nested in three cohorts. All infants infected by HIV postnatally via breastfeeding from the three cohorts were considered as cases and included in this study. PT was considered in a child with a negative HIV-1 DNA PCR from a sample obtained at least at day 45 who later became infected as defined above. All children infected in utero (positive HIV-1 DNA PCR at birth) or during the peripartum period (positive HIV-1 DNA PCR within 45 days of life) were excluded from this analysis. Controls were children uninfected at month 6 by HIV-1 DNA PCR, breastfed, HIV antibody negative at month 18, and matched 1 : 2 to cases by birth date and trial period. For infected children, the estimated date of contamination was calculated as the midpoint between the last negative and the first positive samples obtained by HIV-1 DNA PCR.

In infected children, clinical signs and symptoms occurring within this 'contamination window', i.e., between the last negative HIV-1 DNA PCR test and the first positive one, had been recorded and were retrieved from the data base. They were then compared to those observed and collected among uninfected children followed-up during the same time period of their life. Thus, for a given case infected for instance between 3 (last negative HIV-1 DNA PCR) and 6 months (first positive HIV-1 DNA PCR), two controls were clinically studied from 3 to 6 months of age as well. Further, the visits scheduled just before (pre-infection period) and immediately after (post-infection period) the contamination window were compared between infected cases and uninfected controls to see if the symptoms observed during the contamination window were really acute events.

An acute primary syndrome was defined on the basis of the following conditions: MLI (including fever, sore throat), plain fever, diarrhoea, non-specific dermatitis, ear or eye infections, generalized lymphadenopathy (lymph nodes ≥ 1 cm at two or more extra-inguinal sites), and oral candidiasis.

HIV-1 DNA and RNA measurements

HIV-1 DNA PCR on peripheral blood mononuclear cells was carried out as described previously [⁷]. Quantification of HIV-1 RNA in plasma (viral load, VL) was assessed by branched DNA assay (Quantiplex HIV-1 RNA 3.0, Chiron Diagnostics Corporation, East Walpole, Massachusetts, USA) on the last negative and on all available positive samples. This HIV-1 RNA tool has given concordant results with HIV-1 DNA PCR in this population [¹³]. Its sensitivity threshold was 250 HIV RNA copies/ml using 0.2 ml of plasma. The specimens with results above the quantification upper limit (500 000 copies/ml) were retested after 1 : 10 dilution.

Statistical analysis

Sensitivity of each clinical sign was calculated as the number of HIV-infected children exhibiting this sign divided by the sum of HIV-infected children exhibiting or lacking this sign. Its specificity was calculated as the number of uninfected children lacking this sign divided by the sum of uninfected children lacking or exhibiting this sign. If the same clinical sign was identified at several clinical visits, it was taken into account only once. Positive predictive value was calculated as (sensitivity × transmission rate)/([sensitivity × transmission rate] + [1 - specificity) × (1 - transmission rate)]). Negative predictive value was calculated as (specificity × [1 - transmission rate])/([specificity × (1 - transmission rate)] + (1 - sensitivity) × transmission rate]). In this work, we have considered two distinct MTCT cumulative incidence rates: 18% for children belonging to the ZDV group, and 27.5% for the placebo group, estimated at 6 months of life as published previously [⁷]. Univariate analysis was performed by using the χ² test, Fisher's exact test, or MacNemar test when appropriate. Comparison of means was performed by analysis of variance (F test). A conditional (because of the matched design of the analysis) logistic regression was performed to select independent symptoms associated with PHI. Only variables with a P value < 0.25 were included in the multivariate model. All statistical analyses were carried out with Stata software version 6.0 (Stata Corporation, Texas, USA). To analyse the dynamic pattern of HIV-1 RNA VL among infected children, late undetectable HIV-1 RNA values were all considered as the first time measurement point although obtained either at day 45 or month 3, or 6, or 9.

Results

Characteristics of the children

Of 418 children born to 427 seropositive mothers, 85 (20.3%) were diagnosed as HIV-1-infected, including 22 (25.9%) (12 boys and 10 girls) postnatally HIV-infected infants as follows: 12 from the ANRS 049a trial, three from the ANRS 049b trial, and seven from the ZDV open cohort. These 22 children (cases) were all infected through breastfeeding and matched to 44 uninfected controls. Among the cases, seven were contaminated between day 45 and 90, six between day 90 and 180, five between months 6 and 9, and one between months 9 and 12. The three remaining children were contaminated in a larger window period using the samples available, one between day 45 and month 6, one between months 3 and 9, and one between months 6 and 12. The median age at the first positive sample was 185 days (range, 87-373 days). The median interval between the last negative and the first positive HIV-1 DNA PCR was 91.5 days (range, 29-195 days).

Clinical signs of primary HIV-1 infection

Overall, 21 cases (95.5%) exhibited at least one clinical sign around the estimated date of contamination, but only 27 (61.4%) uninfected children of the same age (P = 0.003). The mean number of symptoms during this contamination window was 2.4 in cases (52 clinical episodes for 22 children) and 1.1 (50 episodes for 44 infants) in controls (P < 0.001). For infected infants, the median timing of occurrence of these clinical events was 177 days (range, 45-336 days). As shown in Table 1, MLI was the most sensitive of these symptoms for the diagnosis of PHI (72.7%) with a negative predictive value of 85.1% or 90.8% according to MTCT rate. The most specific symptom for PHI was generalized lymphadenopathy (95.5%). The combination of MLI and generalized lymphadenopathy was observed in six (27.3%) cases versus one (2.3%) control (P = 0.007). Dermatitis and oral candidiasis were also noted more frequently among infected children compared with uninfected children. No significant association could be found between PHI and other less common symptoms including diarrhoea, plain fever, otitis, and conjunctivitis.

Table 1

The proportions of cases with at least one clinical sign or symptom recorded during the pre- (7 of 22, 31.8%) and the post- (13 of 22, 59.1%) infection periods did not differ significantly from those documented in uninfected children (17 of 44, 38.6%, P = 0.59; and 26 of 44, 59.1%, P = 1.0, respectively). For each clinical sign, no significant difference was observed between infected and uninfected infants during these two periods (data not shown). Further, the proportions of cases with at least one clinical sign or symptom in the pre- (31.8%) and the post- (59.1%) infection periods were significantly lower than in the window period (95.5%) (McNemar test, P < 0.001 and P = 0.01, respectively). The proportion of controls with at least one clinical sign in the pre-infection period (38.6%) was also significantly lower than in the window period (61.4%) (McNemar test, P = 0.03) whereas this proportion in the post-infection period (59.1%) was not (McNemar test, P = 0.79).

In the final multiple logistic regression model, three independent clinical factors remained associated with PHI: MLI [odds ratio (OR), 8.30; 95% confidence interval (CI), 1.44-47.81], dermatitis (OR, 5.99; CI, 1.13-31.90), and generalized lymphadenopathy (OR 26.53; CI, 2.02-348.37; Table 2). The occurrence of oral candidiasis was no longer associated with paediatric PHI.

Table 2

HIV-1 RNA dynamics

As shown in Fig. 1, whereas HIV-1 RNA VL was undetectable in the 18 available samples found negative by HIV-1 DNA PCR, a peak in viral replication was observed in the 20 available initial samples also found positive by HIV-1 DNA PCR, with a median HIV-1 RNA VL of 5.92 log₁₀ copies/ml. Following this peak, a decrease in HIV-1 RNA levels was observed until 12 months after the first positive HIV-1 RNA VL, reaching a median of 4.96 log₁₀ copies/ml. When considering the 16 children with at least three HIV-1 RNA measurements (one initial undetectable and two positive thereafter), this trend (peak and then decline of VL) was observed at the individual level for 11 children whereas VL increased or remained stable for the five remaining infants (data not shown). No correlation was found between initial HIV-1 RNA levels and the clinical severity of PHI, when comparing infected children having developed at least two significant clinical signs during primary infection with those having exhibited one sign or none. No HIV-1 RNA difference was observed between children infected before 6 months of life and those infected later.

Fig. 1

Discussion

To our knowledge, our study is the first report on the occurrence of an ARS in African children infected through breastfeeding, identifying three independent clinical signs (MLI, dermatitis, and generalized lymphadenopathy) associated with paediatric PHI. These three signs are non-specific, do not allow an easily targeted screening for PHI, but their combination (for instance, MLI and lymphadenopathy) seems to reach better predictive values. Even in industrialized countries, the optimal approach for diagnosing PHI has not been defined yet [^1,14]. However, the signs/symptoms we identified appeared consistent with those described previously among African populations. For instance, Colebunders et al. [¹⁵] have compared in Kinshasa (Zaire) the clinical consequences of receiving HIV-1 seropositive and seronegative blood in HIV-1 seronegative recipients, mostly children. In this study, fatigue, diarrhoea, fever, cough, pruritis, oral candidiasis, and polyadenopathy were observed (within the first 3 months of transfusion) more often among seroconverters than controls. Lavreys et al. [¹⁶] have studied the occurrence of clinical manifestations in a prospective cohort of 103 seroconverter female sex workers in Mombasa, Kenya. Eleven signs (including fever, skin rash, inguinal and extrainguinal lymphadenopathies, and vaginal candidiasis) were significantly associated with PHI illness. Morgan et al. have found that five out of 27 recent HIV-1 seroconverters (18.5%) in a population-based cohort in rural Uganda and four out of 22 HIV-negative controls (18.2%) reported a flu-like illness in the 12 months preceding seroconversion. These authors concluded that there is no recognizable seroconversion illness in Africans infected with HIV-1 subtype A or D [¹⁷], given the background level of other illnesses. Even if our study also revealed an important overlap between signs associated with PHI and other paediatric common health problems occurring in Ivory Coast (such as malaria and measles), we were able to identify meaningful signs/symptoms associated with paediatric PHI. The fact that we studied only children infected postnatally after day 45 of life, thus eliminating the interference of neonatal infections and focusing on the symptomatology during a short time period (between the last negative and the first positive HIV-1 DNA PCR test), has probably limited this background effect.

Our study also suggests that a viral peak occurred soon after primary infection in infants infected by breastfeeding. Its median value (close to 1 × 10⁶ copies/ml) seems to be similar to HIV-1 RNA levels obtained by Biggar et al. in Malawi among six children infected peripartum or by early breastfeeding and closely followed through the first month of life [¹⁸]. The branched DNA assay is very sensitive for PHI [¹⁹], even in Côte d'Ivoire where the large majority of HIV-1 subtypes belong to the CRF02 recombinant form AG [²⁰]. Finally, as described in developed countries [²¹] and in Africa [^18,22], infected infants have higher initial HIV-1 RNA VL compared with adults. Additionally, in contrast with the adult pattern, after the first birthday, HIV RNA copy number slowly decreases to reach a plateau (i.e., the virologic set point) over the next few years of life.

Our survey has several limitations such as the lack of information on both the duration of the clinical signs associated with PHI and the rate of progression to AIDS according to the occurrence of ARS. From a biological point of view, concomitant CD4 cell count measurements are missing. Further, the HIV-1 RNA decrease over time may reflect a selection bias considering the declining number of samples tested during follow-up. Indeed, specimens from healthier children may have been more often available for RNA testing over time. However, our retrospective survey provides meaningful information on clinical and virological features associated with PHI in African children infected by breastfeeding. This emphasizes the urgent need to implement alternatives to predominant breastfeeding in order to prevent the high risk of HIV-1 PT observed in African cohorts. Our findings may also be useful to identify early cases of PT and encourage adequate HIV screening in corresponding mothers and families. Finally, as combination antiretroviral therapy during PHI reduces progression to AIDS [²³], and given the current context of increasing access of antiretroviral drugs in Africa, our study may also be useful to propose effective antiretroviral treatment to control the early viral replication observed in this frequent form of paediatric HIV infection.

Acknowledgments

The authors acknowledge the contribution from the National AIDS authorities in Côte d'Ivoire and women and children enrolled in the project. We are grateful to Chiron Corporation for technical support to this study, and especially to F. Huisse, P. Simpson and A. Kéchichian.

Sponsorship: Supported by the French Agence Nationale de Recherches sur le SIDA (ANRS, Paris, France) and the French Ministry of Foreign Affairs, within the Coordinated Action AC12.