Determinants of Mother-to-Infant Human Immunodeficiency Virus 1 Transmission Before and After the Introduction of Zidovudine Prophylaxis FREE

AT THE END of 1994 the randomized controlled Pediatric AIDS [aquired immunodeficiency syndrome] Clinical Trial Group (PACTG) protocol 076 clearly demonstrated that a 3-part regimen of zidovudine treatment during pregnancy, labor, and the neonatal period decreases the mother-to-infant transmission of human immunodeficiency virus 1 (HIV-1) by about 70%.¹ Subsequent observational studies have confirmed the effectiveness of this practice.^2- 3 In the late 1990s, randomized controlled trials demonstrated the protective effects of elective cesarean delivery⁴ and exclusive formula feeding,⁵ each reducing transmission risk by approximately half. A further reduction in transmission, up to 87%, for these 3 interventions combined has been reported by cohort studies^6- 7 and a recent population survey.⁸ Current American treatment guidelines for HIV-1–infected pregnant women recommend the use of potent antiretroviral treatment (ART), in addition to zidovudine prophylaxis,⁹ while European recommendations are less aggressive owing to concerns about effects on fetal growth.¹⁰ Only recently have data from larger cohorts been published on the effects of combined ART on mother-to-infant transmission and infant outcome.² Findings suggest that combined ART may further decrease mother-to-infant transmission.

Data from observational studies offer the opportunity of providing important measures of the effectiveness of treatments at the population level that clinical trials cannot.¹¹ Thus, it is crucial to monitor trends in prophylactic practices, therapeutic intervention usage, and mother-to-infant transmission rates using population-based surveillance. Recent population-based surveys in the United States¹² and Great Britain⁸ have found significant declines in mother-to-infant transmission rates associated with increased use of interventions. Herein we use data from the Italian Register for HIV Infection in Children to evaluate time trends in prophylactic interventions and the determinants of mother-to-infant transmission both before and after the introduction of the PACTG 076 protocol, and in treated and untreated mother-infant pairs after 1995.

Instituted in 1985, The Italian Register for HIV Infection in Children is a nationwide multicenter study of children perinatally exposed to HIV-1 infection. The register's data come from a network of 106 voluntarily participating pediatric clinical centers located throughout Italy that aim to enroll all infants born to mothers with documented HIV-1 infection during pregnancy or at the moment of delivery and is representative of the overall population of exposed infants in Italy.^13- 14 All children are enrolled, whether they are followed up prospectively or retrospectively from birth, but only those followed up prospectively are included in studies on risk factors or the course of infection.¹²

Data regarding mother-infant pairs are collected as previously described.^13- 15 In particular, baseline information is collected on the infant's demographics, age at first observation, mother's risk factors for HIV-1 infection, mother's place of origin, mother's clinical condition at delivery (defined according to the Centers for Disease Control and Prevention [CDC] classification system for HIV-1–infected adults¹⁶), mode of delivery, gestational age, type of feeding, and whether the woman had other children after her HIV-1 infection had been diagnosed. Starting in 1989, information on ARTs during pregnancy was collected as treated or not treated; from 1993 onward, data on type of therapy and gestational age at the beginning and end of therapy were also included. Before the PACTG 076 protocol, zidovudine had been administered to pregnant women in selected centers in our country¹³; thereafter, the PACTG 076 protocol prophylactic regimen was introduced, though not universally or simultaneously.

Both baseline and follow-up visits record information on infant infection status, HIV-1 antibodies (assessed by enzyme-linked immunosorbent assay and Western blot test), and viral markers (proviral DNA, and/or virus culture, and/or free and complexed p24 antigenemia). Following the CDC recommendations for HIV-1 infection in children,¹⁷ infection is diagnosed by the persistence of HIV-1 antibodies after 18 months, or before 18 months, by detection of viral markers on at least 2 occasions. Infants whose infection status is indeterminate are followed up at least until infection status can be ascertained. The register has no information available concerning maternal viral load at the time of delivery, concomitant infections, and treatments prior to pregnancy. The maternal immunologic condition at the time of delivery is known in a few cases.

Children were classified as infected, uninfected, or indeterminate. Indeterminate children were either lost to follow-up or died prior to HIV-1 diagnosis, or had indeterminate HIV-1 status (because they were either younger than 18 months and had not seroreverted or detection of viral markers was only performed once) when the study was closed. Negative proviral DNA or viral cultures carried out at birth are not used to assign infection status. We considered an incomplete PACTG 076 protocol to be anything less than all 3-part regimens,¹⁸ based on the similarity of results for having 1- or 2-part regimens and so as to increase the statistical power. Women who received zidovudine therapy prior to the publication of the PACTG 076 protocol results were classified as having an incomplete PACTG 076 protocol regimen. We defined combined ART in mothers as both double (2 nucleoside reverse transcriptase inhibitors [NRTIs]) and triple (NRTI and 1 protease inhibitor, or 3 NRTIs, or 2 NRTIs and 1 nonnucleoside reverse transcriptase inhibitor) combined ART.

Analyses were performed on data reported up to December 31, 2000, on children born to an HIV-1–positive mother between June 1, 1985, and December 31, 1999, and whose follow-up started within the first month of life. Trends in vertical transmission rates, use of ART, and elective cesarean delivery from June 1, 1985, to December 31, 1999, were estimated; 1985 and 1986 were combined owing to low numbers and considered as the reference group.

Logistic regression random effects models were used to estimate crude and adjusted odds ratios for several factors potentially influencing vertical transmission for 2 periods—1985 through 1995 (January 1, 1985, through December 31, 1995) and 1996 through 1999 (January 1, 1996, through December 31, 1999)— and for the estimate crude and adjusted odds ratios occurring between treated and untreated children after 1995, with clinical center considered as a random effect. Children with the same mother and twins were treated independently since excluding them from the analysis did not change the results. The periods were chosen based on the timing of the gradual introduction of zidovudine prophylaxis in Italy following the publication of the results of the PACTG 076 protocol at the end of 1994.¹³ The factors considered were maternal clinical condition at time of delivery based on the CDC classification (ie, asymptomatic, HIV-1–related symptoms but not clinical AIDS, clinical AIDS, unknown¹⁶); exposure category to HIV-1 of the mother (ie, injection drug use, sexual exposure, transfusion, other/unknown); mother's area of origin in an HIV-1 endemic area or not; gestational age (ie, ≤32, 33-36, 37-39, ≥40 weeks, or unknown); infant sex; mode of delivery (ie, elective cesarean section, emergency cesarean delivery, cesarean delivery type unknown, vaginal delivery, or unknown); type of infant feeding (ever breastfeeding, exclusively formula feeding, or unknown); and type of prophylactic ART received in the perinatal period (ie, none, incomplete PACTG 076 protocol, complete PACTG 076 protocol, or maternal combined ART). In the "Results" section below some categories of variables are grouped based on similarities in the estimated coefficients in the preliminary multivariate analyses. A CD4-positive, T-lymphocyte count at the time of delivery was among the factors initially considered. However, many of the mothers had missing data and this variable was strongly associated with maternal clinical condition (ie, the lower the CD4-positive, T-lymphocyte count, the more advanced the CDC clinical category). Thus, this variable was excluded from the final analyses. To assess the bias due to incomplete follow-up of indeterminate children, analyses were done first excluding, then including such children, considering those who died prior to definitive diagnosis as infected and all others as uninfected.

By the end of December 1999, 3770 infants born to HIV-1–positive women had been reported to the register within 1 month of birth. Characteristics of mothers and children in the 1985-1995 and 1996-1999 periods are reported in Table 1. The proportion of mothers reporting sexual contact as an HIV-1 risk factor increased from 23.3% to 34.8% over the 2 study periods. Injection drug use remained the most common risk factor for HIV-1 infection in mothers but decreased from 71.0% in the 1985-1995 period to 47.5% in the 1996-1999 one. The number of HIV-1–positive women giving birth increased in 1998 (295 births) and 1999 (332 births), after slightly declining for several years from 305 births in 1990 to 238 in 1997.

Antiretroviral therapy usage increased from 8.1% of mother-infant pairs in the 1985-1995 time to 78.7% in the 1996-1999 time. Table 2 shows types of antiretroviral regimens administered to mothers during pregnancy. Median duration of treatments was 5 months (range, 1-9 months). During the 1985-1995 period, the median duration was 3 months (range, 1-8 months), whereas during the 1996-1999 period, the median duration was 5 months (range, 1-9 months). The difference between duration in the 2 different periods was statistically significant (P = .001). The most frequently used triple-combination therapy in the 1996-1999 period was zidovudine, lamivudine, and indinavir (n = 27). By 1999, 89.9% of 332 mother-infant pairs received ART: 55.1% received combined ART, 22.9% complete PACTG 076 protocol, and 11.1% an incomplete PACTG 076 protocol regimen. Intrapartum and neonatal prophylaxis consisted of zidovudine therapy in all but the few cases where nevirapine therapy was used. Of the 289 women receiving combined ART, 66 (22.8%) received only intrapartum zidovudine, 5 infants (1.7%) received only oral zidovudine, 136 mother-infant pairs (47.0%) received both treatments, and 82 (28.4%) received none. Zidovudine therapy was part of the antepartum regimen for 232 women (80.3%). Of the 57 women who did not have zidovudine as part of their combination regimen, 42 (73.7%) had 1 or both of the other 2-part regimens of the PACTG 076 protocol.

The proportion of elective cesarean deliveries increased from 16.6% in the 1985-1995 period to 64.2% in the 1996-1999 one. Rates remained stable at around 13% until 1993, rising from 19.6% in that year to 81.3% in 1999. Breastfeeding decreased from 3.4% to 1.4% over the 2 study periods, with only 16 women breastfeeding in the 1996-1999 period.

The percentage of indeterminate infants was stable at under 10% per year until 1998 and 1999 when the proportions were 13.2% and 18.7%, respectively. The proportion of indeterminate children who died prior to diagnosis was 12.0% (25/209) in the 1985-1995 period and 3.8% (5/130) in the 1996-1999 period (P = .01).

Excluding indeterminate children, the vertical transmission rate was 18.5% (95% confidence interval [95% CI], 14.0%-23.6%) in the 1985-1986 period and remained stable until 1996 when incidence began to steeply drop, reaching 4.4% (95% CI, 2.3%-7.6%) in 1999. When indeterminate children were included, rates in the 1985-1986 period were 17.7% (95% CI, 13.5%-22.6%), falling to 3.9% (95% CI, 2.1%-6.6%) in 1999.

Risk factors for mother-to-infant transmission in the 1985-1995 period and the 1996-1999 one for the 3431 children with known infection status are presented in Table 3. In the 1985-1995 period, after adjusting for all other factors, CDC clinical categories B or C increased the odds of transmission by 53% (P = .004), elective cesarean delivery reduced transmission by 46% (P = .001).

In the 1996-1999 period, in multivariate analysis, risk of transmission was reduced by 76% in mother-infant pairs receiving an incomplete PACTG 076 protocol regimen, by 88% in those receiving all 3-part regimens, and by 93% for those in which the mother received combination ART (P<.001 for all categories) compared with no treatment. The 50% reduction in transmission attributable to elective cesarean delivery was of borderline statistical significance (P = .08). While breastfeeding was not associated with infection in the period prior to 1996, in the 1996-1999 period, children who were breastfed were 10 times more likely to be infected than those who were bottle-fed (P = .001). The mother-to-infant transmission rate in breastfeeding women was 21.7% in the 1985-1995 period and 53.8% in the 1996-1999 one.

The separate and combined effects of type of feeding, mode of delivery, and ART in the periods 1985-1995 and 1996-1999 and overall are presented in Table 4. For breastfeeding mothers without any prophylactic interventions, the transmission rate was 23.4% in the earlier period and 71.4% in the later, but the 95% CI in the second period was very large (29.9%-96.3%) owing to the few women who breastfed. Transmission rates were similar in both periods for nontreated, nonbreastfeeding mother-infant pairs without elective cesarean delivery (18.5% vs 24.3%) and with elective cesarean delivery (10.0% vs 8.6%). For nonbreastfeeding mother-infant pairs without elective cesarean delivery and receiving any type of therapy, transmission rates were higher in the earlier (15.3%) than in the later (4.4%) period. After 1995, the transmission rate in nonbreastfeeding mother-infant pairs receiving any type of therapy with elective cesarean delivery was 2.4%.

In the 1996-1999 period the nonbreastfeeding pairs with elective cesarean delivery had a transmission rate of 4.2% (7/167) with an incomplete PACTG 076 protocol regimen, 2.6% (5/195) with a complete PACTG 076 protocol regimen, and 0.52% (1/192) when the mother received combined ART. In the 289 mothers who took some form of combined ART during pregnancy, zidovudine therapy was not prescribed in 59 (20.4%) (though 3 of these women received nevirapine therapy). One child was infected in the no-zidovudine group (1.9%) and 3 (1.3%) in the zidovudine group.

Since 1996, the use of interventions directed to prevent mother-to-infant HIV-1 transmission and changes in treatment of adult HIV-1 infection, including potent combined therapies, has greatly increased in practices in Italy, with concomitant reductions in transmission. The transmission rate was more than 27% in the absence of any intervention. Antiretroviral treatment in mothers for their own health or prophylactic interventions, including zidovudine treatment in mother-infant pairs, elective cesarean delivery, and combined with formula feeding, reduced the risk by more than 90%.

Even where use of all 3 interventions was reported, more than 2% of infants became infected. This point should be adequately discussed when counseling an infected woman who is considering pregnancy. Strategies should be designed to further reduce the risk, possibly acting on transmission that occurs early in pregnancy and that may not be prevented by current interventions.¹³ The benefits of new approaches can only occur if infected women are diagnosed before or very early in pregnancy. On the other hand, mother-infant pairs not receiving ART accounted for almost two thirds of the transmission in the 1996-1999 period. This mainly reflects a failure to identify HIV-1 infection until or even after delivery, but it also reflects the refusal of some women to take antiretroviral drugs owing to safety or other concerns. As seen elsewhere,¹⁹ the proportion of women infected by sexual contact has increased in recent years in Italy. The reduced awareness of at-risk behavior in women infected via the sexual route²⁰ supports recommendations for routine testing of all pregnant women.

We found a clear hierarchy of effectiveness of interventions in both study periods, with transmission decreasing as interventions were combined. Transmission rates were similar for each combination in both periods except when therapy was included. For prophylaxis with or without elective cesarean delivery, transmission rates were higher in the earlier period, reflecting the lack of a clear prophylaxis protocol whereas elective cesarean delivery was likely to be carried out correctly in both periods. Antiretroviral prophylaxis seems to be the most effective intervention, but our findings suggest that the best prophylactic strategy is a combined intervention. In fact, although the effect of cesarean delivery was not statistically significant after 1996, we observed the largest reduction in transmission in the group where all prophylactic interventions were used.

We found the complete PACTG 076 protocol regimen to be more effective in preventing mother-to-infant HIV-1 transmission than an incomplete regimen. However, an incomplete regimen alone exerts some protective effect. The diffuse use of combined ART in HIV-1–infected women for clinical indications seems to contribute to the reduced transmission rate and there was an indication that combined ART plus the other part of the PACTG 076 protocol regimen may be more efficacious than that with zidovudine monotherapy. The risk of transmission in mothers receiving combined ART agrees with those reported in other studies.^2- 3 Guidelines for HIV-1–infected pregnant women recommend that zidovudine prophylaxis should be associated with combined ART⁹ since, at present, only zidovudine and nevirapine therapies have been demonstrated to effectively prevent mother-to-infant transmission.^1,21 In our study, 95% of women receiving combined ART had zidovudine included in their regimen and/or as part of the other 2 parts of the PACTG 076 protocol. Whether perinatal HIV-1 transmission may be equally preventable by other combined antiretroviral regimens independent of whether zidovudine is included or not remains to be seen. Confirmation of this observation would ideally come from clinical trials. Questions remain regarding the long-term safety for the child of combinations of antiretroviral drugs in pregnancy that will only be answered by the follow-up of large numbers of indeterminate children.

By 1999, four fifths of the infants reported to the register were delivered by elective cesarean delivery. The estimated 50% reduction in transmission attributable to elective cesarean delivery after adjusting for zidovudine prophylaxis is consistent with the findings of other studies^3,6,8 where the adjusted odds ratio ranged from 0.39 to 0.43, even though, similar to the report from Great Britain,⁸ this result was of borderline statistical significance in the 1996-1999 period. In treated mother-infant pairs this effect was reduced to about 30% and was not significant. However, the strong correlation between antiretroviral prophylactic therapy and mode of delivery could have partly masked the effectiveness of elective cesarean delivery. Supporting this interpretation, the transmission rate was 2.4% in nonbreastfeeding mother-infant pairs receiving prophylactic ART and having an elective cesarean delivery, compared with a rate of 4.4% with no breastfeeding and prophylaxis alone, consistent with reports of 2% and 4%, respectively, in the European Collaborative Study.³

The importance of avoiding breastfeeding was confirmed. Breastfeeding was rare in both study periods and was a significant risk factor for mother-to-infant HIV-1 transmission only in the 1996 through 1999 period when prophylactic interventions had come into widespread use. Based on our results, the population attributable risk (or etiologic fraction) of breastfeeding is approximately 10% even when the prevalence of breastfeeding is less than 2%.

While the Italian Register for HIV Infection in Children is population based and, therefore, able to reliably monitor intervention usage and its influence on HIV-1 mother-to child transmission, some biases may be present. The most important potential bias is related to the 339 children (9.0% of the total) who were still indeterminate at the time of the analysis. To evaluate this bias we repeated the analyses, including indeterminate children, assuming that deceased children were infected and those lost to follow-up were uninfected. Uninfected children, being asymptomatic, may be more likely to be lost to follow-up before final definition of infection status. Conversely, infants who died before a definitive diagnosis are likely to have been HIV-1 infected. The fact that the results of the analyses including and excluding those of indeterminate status are consistent suggests that the bias due to indeterminate children should not be very large.

These population-based data clearly demonstrate the effectiveness of prophylactic interventions in reducing mother-to-infant transmission in Italy, confirming the results of randomized trials and cohort studies. Given that transmission rates are very low, more women receive ART before pregnancy, and treatment regimens continue to diversify, it may be increasingly difficult to assess the importance of other ARTs and combined approaches with clinical trials. Observational studies, including population-based surveillance, may provide the most practical and reliable method to monitor intervention usage and HIV-1 mother-to-infant transmission rates.

Accepted for publication May 15, 2002.

Writing Committee

Department of Pediatrics, University of Florence, Florence: Maurizio de Martino, MD; Luisa Galli, MD. Department of Pediatrics, University of Turin, Turin: Pier-Angelo Tovo, MD; Clara Gabiano, MD. AIDS and Sexually Transmitted Disease Unit, Istituto Superiore di Sanità, Rome: Patrizio Pezzotti, Dstat; Theresa M. Wagner, MPH; Giovanni Rezza, MD.

Participants of the Italian Register for HIV Infection in Children

Ancona: P. Osimani, MD. Bari: D. De Mattia, MD; C. Di Bari, MD. Bergamo: M. Ruggeri, MD. Bologna: F. Baldi, MD; M. Ciccia, MD; M. Lanari, MD; M. Masi, MD; V. Venturi, MD. Bolzano: L. Battisti, MD. Brescia: M. Duse, MD. Brindisi: P. G Chiriacò, MD. Cagliari: R. Cavallini, MD; C. Dessì, MD; C. Pintor, MD. Catanzaro: E. Anastasio, MD. Chieti: G. Sabatino, MD. Como: M. Sticca, MD. Cuneo: G. Pomero, MD. Ferrara: T. Bezzi, MD. Florence: E. Chiappini, MD; M. De Luca, MD; P. Gervaso, MD. Forlì: M. T. Cecchi, MD. Genoa: D. Bassetti, MD; C. Gotta, MD; R. Rosso, MD; A. Timitilli, MD. Imperia: U. Tondo, MD. Mantova: P. Mussini, MD. Milan: D. Bricalli, MD; A. Bucceri, MD; G. Ferraris, MD; M. Giovannini, MD; F. Mosca, MD; R. Lipreri, MD; A. Plebani, MD; E. Riva, MD; S. Riva, MD; A Viganò, MD; G. V. Zuccotti, MD. Modena: M. Cellini, MD. Naples: W. Buffolano, MD; A. Guarino, MD; L. Tarallo, MD. Padua: R. D'Elia, MD; C. Giaquinto, MD; O. Rampon, MD. Palermo: E. R. Dalle Nogare, MD; A. Romano, MD. Pavia: D. Caselli, MD; A. Maccabruni, MD. Pisa: R. Consolini, MD. Parma: G. Benaglia, MD. Reggio Emilia: C. Magnani, MD. Roma: G. Anzidei, MD; A. M. Casadei Pistilli, MD; G. Castelli Gattinara, MD; S. Catania, MD; C. Facente, MD; P. Falconieri, MD; C. Fundarò, MD; O. Genovese, MD; C. Rendeli, MD. Sassari: S. Bionda, MD. Taranto: L. Cristiano, MD. Turin: S. Garetto, MD; C. Riva, MD; E. Palomba, MD. Trapani: V. Portelli, MD. Trento: A. Mazza, MD. Trieste: C. Salvatore, MD. Varese: A. Pellegatta, MD. Verona: M. Molesini, MD.

Trials demonstrated that exclusive formula feeding or elective cesarean delivery or a 3-part regimen of zidovudine therapy during pregnancy, labor, and the neonatal period, significantly decreases the mother-to-infant transmission of HIV-1. Observational studies offer the opportunity of providing important measures of the effectiveness of each or combined prophylactic interventions at the population level.

A clear hierarchy of effectiveness of interventions (antiretroviral prophylaxis, exclusive formula feeding, and elective cesarean delivery) was found. Antiretroviral prophylaxis (and, in particular, combined antiretroviral prophylaxis) was the most effective intervention, but transmission decreased as interventions were combined. The effectiveness of prophylactic interventions in reducing mother-to-infant transmission is clearly demonstrated at a population level, confirming the results of randomized trials and cohort studies.

Corresponding author and reprints: Maurizio de Martino, MD, Department of Pediatrics, University of Florence, via Luca Giordano 13, I-50132 Florence, Italy (e-mail: maurizio.demartino@unifi.it).