Maternal HIV Infection and Vertical Transmission of Pathogenic Bacteria
BACKGROUND: HIV-exposed newborns may be at higher risk of sepsis because of immune system aberrations, impaired maternal antibody transfer and altered exposure to pathogenic bacteria.
METHODS: We performed a secondary analysis of a study (clinicaltrials.gov, number NCT00136370) conducted between April 2004 and October 2007 in South Africa. We used propensity score matching to evaluate the association between maternal HIV infection and (1) vaginal colonization with bacterial pathogens; (2) vertical transmission of pathogens to the newborn; and (3) sepsis within 3 days of birth (EOS) or between 4–28 days of life (LOS).
RESULTS: Colonization with group B Streptococcus (17% vs 23%, P = .0002), Escherichia coli (47% vs 45%, P = .374), and Klebsiella pneumoniae (7% vs 10%, P = .008) differed modestly between HIV-infected and uninfected women, as did vertical transmission rates. Maternal HIV infection was not associated with increased risk of neonatal EOS or LOS, although culture-confirmed EOS was >3 times higher among HIV-exposed infants (P = .05). When compared with HIV-unexposed, neonates, HIV-exposed, uninfected neonates (HEU) had a lower risk of EOS (20.6 vs 33.7 per 1000 births; P = .046) and similar rate of LOS (5.8 vs 4.1; P = .563). HIV-infected newborns had a higher risk than HEU of EOS (134 vs 21.5; P < .0001) and LOS (26.8 vs 5.6; P = .042).
CONCLUSIONS: Maternal HIV infection was not associated with increased risk of maternal bacterial colonization, vertical transmission, EOS, or LOS. HIV-infected neonates, however, were at increased risk of EOS and LOS.
- EOS —
- early-onset sepsis
- GBS —
- group B Streptococcus
- HEU —
- HIV-exposed uninfected
- HUU —
- HIV-unexposed uninfected
- LOS —
- late-onset sepsis
- PCR —
- polymerase chain reaction
- PoPS —
- Prevention of Perinatal Sepsis
- ROM —
- rupture of membranes
- UTI —
- urinary tract infection
What’s Known on This Subject:
Neonatal sepsis is an important cause of under-5 childhood mortality. Infants born to HIV-infected mothers are at increased risk of morbidity and mortality, even if not having acquired HIV. This association needs further study during the neonatal period.
What This Study Adds:
Maternal HIV infection was associated with increased vaginal colonization by Escherichia ecoli but not group B Streptococcus. Neonates born to HIV-infected mothers were only at increased risk of sepsis if they had acquired HIV-infection, but not if HIV-uninfected.
Neonatal sepsis is a leading (26%–49%) cause of neonatal mortality in developing countries and contributes disproportionately to mortality among children <5 years of age.1,2 A recent systematic review reported a paucity of data on the incidence of neonatal sepsis in sub-Saharan Africa.3 Estimates of culture-confirmed neonatal sepsis ranged between 5.46 to 21 per 1000 live births.4–6 Published studies rarely distinguished between sepsis occurring during the early neonatal period (first 3 or 7 days of life), which likely reflects bacteria acquired directly from the mother’s genital tract, and sepsis in older neonates which may be acquired by vertical transmission or through environmental exposures.
There is also a lack of data on the impact of maternal HIV exposure on clinical or pathogen-specific burden of neonatal sepsis.3 A maternal immune system compromised by HIV infection may result in reduced transplacental transfer of antibodies to the fetus in utero,7,8 possibly increasing neonatal sepsis susceptibility. Although HIV infection increases bacterial vaginosis9 and the diversity of vaginal microbiota,10 it has not been associated with any difference in overall prevalence of group B Streptococcus (GBS) vaginal colonization compared with HIV-uninfected women.11
High maternal HIV prevalence rates (5%–26%) in sub-Saharan Africa have led to a reversal of decades of improvement in child survival.12 Morbidity and mortality in HIV-infected children is significantly higher than in HIV-unexposed children.13,14 In addition, HIV-exposed uninfected (HEU) infants have a higher than expected risk of morbidity and mortality compared with HIV-unexposed uninfected (HUU) infants; an effect related to maternal immunologic status and early weaning among HIV-exposed infants.14–17
We performed a secondary analysis of a South African cohort of >8000 mother-baby pairs to assess the impact of maternal HIV infection on (1) the prevalence of maternal vaginal colonization with pathogens associated with neonatal sepsis; (2) vertical transmission of bacterial pathogens to the newborn; and (3) sepsis rates during the early and late neonatal periods.
This analysis reports on secondary objectives of a cohort of women and their newborns enrolled in the Prevention of Perinatal Sepsis (PoPS) trial.18 PoPS was a randomized, placebo-controlled trial to determine the efficacy of intrapartum chlorhexidine maternal vaginal washes and newborn skin wipes for reducing early-onset neonatal sepsis and vertical acquisition of pathogenic bacteria by newborns. The study was conducted between April 2004 and October 2007 and found no effect of interventional wipes against primary end points of overall early-onset sepsis (EOS) and vertical transmission of colonizing pathogenic bacteria.18
Pregnant women attending an antenatal clinic or presenting in labor at Chris Hani-Baragwanath Hospital, a secondary-tertiary level of care hospital, in Soweto, South Africa, were enrolled into the study. The prevalence of HIV in antenatal clinic attendees in the study population during the trial period was 29.9%.19 The study staff had no input into clinical management of study participants, including management of HIV infection. The standard of care for prevention of mother-to-child transmission of HIV during the trial period included a single dose of nevirapine (sd-NVP) to both the HIV-infected mother at onset of labor and the newborn. Before triple antiretroviral treatment of HIV-infected individuals became routinely available in 2007, CD4+ and HIV viral load testing by attending physicians was limited. From 2007, pregnant women with WHO Stage IV disease or CD4+ count <200 cells/mm3 were offered treatment that included stavudine, lamivudine, and nevirapine from at least 34 weeks gestation.20 All mothers and newborns are provided with free public health care in South Africa. In total, 8011 mothers and their 8129 neonates were enrolled in the trial. HIV polymerase chain reaction (PCR) testing of HIV-exposed infants is offered routinely at 4 to 6 weeks of age.
Details of study procedures have been described.18 In brief, swabs of the lower vagina were obtained from maternal participants from May 2005 onward before administering the first study interventional wipe. A surface swab of the umbilicus, nares, and outer ear for culture was collected from neonates after delivery but before bath and study wipe. Swabs were cultured for Streptococcus agalactiae (GBS), Escherichia coli, and Klebsiella pneumoniae by using standard methods.18
Study doctors conducted active surveillance of neonatal and pediatric wards to identify hospitalized study neonates and abstracted information related to neonatal sepsis end points from medical records. Sterile site cultures were collected at the discretion of attending physicians before initiation of antibiotic treatment and processed per standard practice at the hospital microbiology laboratory. Blood was sampled for bacterial growth by using the BacT/Alert microbial system (Organon Teknika, Durham, NC). Routine methods were used for culture and identification of invasive pathogens from other sterile sites. Active laboratory-based surveillance was conducted to confirm that all sterile site cultures obtained from study neonates were captured. Maternal CD4+ T-lymphocyte count results were documented when available as part of standard of care.
End Point Definitions
Maternal vaginal colonization was defined as isolation of GBS, E coli, or K pneumoniae from vaginal swabs collected during labor before the first interventional wipe. Vertical transmission of bacteria was defined as neonatal surface colonization at birth with the same bacterium isolated from the maternal vagina. Early-onset sepsis (EOS) was defined as sepsis occurring within the first 3 days of life and late-onset sepsis (LOS) as that from the fourth to 28th day of life (Table 1). Culture-confirmed episodes of sepsis not fulfilling clinical sepsis criteria and records of all stillborn infants and neonates dying within 2 hours of birth were reviewed by 3 neonatologists to determine if the case represented a sepsis episode. Neonatal and maternal HIV infection status was based on documented HIV testing results; those with unknown status were excluded from analyses involving HIV infection status.
To specifically evaluate maternal HIV infection as a neonatal sepsis-related risk factor, we used propensity score matching to reduce bias with respect to important covariates for each defined end point.21,22 Covariates considered for impact on end points included: (1) for maternal vaginal colonization: maternal age (> or ≤25 years), rupture of membranes (ROM) before swab collection, and history of antibiotics in week before delivery; (2) for vertical transmission: prolonged ROM, antibiotic use in week preceding labor and intrapartum, number of per vaginal examinations; and (3) for early- and late-onset neonatal sepsis: mode of delivery, gestational age, prolonged ROM, maternal fever ≥38.0°C, intrapartum antibiotic use, known maternal GBS colonization and urinary tract infections (UTIs). We chose propensity score matching because key covariates differed between HIV-infected and HIV-uninfected women, and standard modeling approaches provide estimates even in the absence of an appropriate comparison population.
We used multivariable logistic regression to estimate the propensity score for each mother (or infant). The propensity score is the conditional probability of the mother being HIV-infected given the variables in the model. We matched each HIV-infected mother to an HIV-uninfected mother with the closest propensity score. For each propensity score matched analysis, balance for each covariate included in the model was evaluated for HIV-infected and HIV-uninfected mothers. The propensity score model that achieved the best balance was used. The model for maternal colonization was limited to mothers with known HIV result and colonization swab with a result. The model for vertical transmission was limited to mothers with a known HIV result, vaginal delivery, a colonization swab with a positive result (ie, for GBS, E coli, or K pneumoniae) and a newborn with a microbiologic swab result.
For the EOS and LOS sepsis end points, we considered 2 additional comparison groups: HUU versus HEU neonates and HEU versus HIV-infected neonates. We built propensity score models as above, with the appropriate outcome variable for the comparison, balancing on the variables listed above for EOS and LOS. CD4+ cell count results were available for only 34.7% (725/2090) of HIV-infected women, thus maternal CD4+ cell count was not included as a covariate.
We also conducted unadjusted analyses on the total cohort to understand the influence of covariates associated with HIV exposure or infection status on the endpoints evaluated.
χ2 tests were used to compare proportions. Relative risks and 95% confidence intervals were used to assess HIV as a risk factor. Denominator for incidence was per 1000 births. Analyses were conducted by using SAS version 9·1.
The study was approved by the Human Research Ethics Committee of the University of the Witwatersrand, South Africa and the Institutional Review Board of the Centers for Disease Control and Prevention, USA. Written, informed consent was obtained from the mother before any study procedure was undertaken. The trial was registered at ClinicalTrials.gov (NCT00136370).
HIV infection status was available for 98.5% (7894/8011) of mothers, 26.5% (2090/ 7894) of whom were HIV infected. Among HIV-infected women with CD4+ results available (725/2090), 21.8% (158/725) had CD4+ counts <200 cells/mm3, 27.4% (199/725) between 200 and 350 cells/mm3, and 50.8% (368/725) >350 cells/mm3. The majority of CD4+ results available (467/725, 64.4%) were from women enrolled between June 2006 and October 2007. Antiretroviral treatment to prevent mother-to-child transmission of HIV was provided to 92.3% (1929/2090) of HIV-infected mothers, including 97.6% (1882/1929) who received sd-NVP and 2.4% (46/1929) who received triple antiretroviral therapy, which was initiated a mean of 151 (range 1 to 1095) days before delivery. Overall, 2130 newborns were born to 2090 HIV-infected mothers.
Significant differences in baseline demographic and clinical characteristics of HIV-infected and -uninfected mothers included HIV-infected women being older (P < .0001) and more likely to have UTIs (13% vs 10%), hemoglobin <10 mg/dL (13% vs 7%), tuberculosis (1.5% vs 0.2%), receive antibiotics before labor onset (30% vs 21%) and during labor (12% vs 10% for women with vaginal delivery); P < .01 for all observations (Table 2). HIV-infected com-pared with -uninfected women had higher frequencies of fever during labor (0.7% vs 0.05%); nonelective caesarian sections (26% vs 23%); and premature deliveries (5.9% vs 3.1%). HIV-exposed newborns had lower birth weights than HUU neonates (Table 2).
Infant HIV PCR testing was undertaken at a median of 42 days of age (1–347 days) in 64.2% (1367/2130) of HIV-exposed infants, of whom 8.2% (112/1367) were HIV infected. Chlorhexidine wipes did not affect vertical HIV transmission rate (65/713 [9.1%] vs 47/654 [7.2%] PCR positive in interventional versus control arms, P = .19). Differences in baseline demographics among HIV-exposed neonates with PCR testing results included that HIV-infected compared with HEU neonates had a lower median gestational age (38 vs 39 weeks; P = .005), lower birth weight (median 2995 vs 3100 g; P = .001) and higher frequency of exposure to meconium-stained liquor (21% vs 11%; P = .002) during labor.
Vaginal Colonization and Vertical Transmission of Genital-Tract Bacteria
The total colonization cohort included 5099 (3752 HIV-uninfected, 1347 HIV-infected) women with known HIV status. The matched subset analysis for assessment of vaginal colonization included 1346 HIV-infected and 1346 HIV-uninfected participants. HIV-infected women were less likely to be colonized than HIV-uninfected women with GBS (17% vs 23%; P = .0002), or K pneumoniae (7% vs 10%; P = .008) in the matched subset analysis. In the total cohort, prevalence of GBS colonization was similar to that observed in the matched analysis, whereas higher rates of E coli colonization (47% v. 43%, P = .039) and no difference in K pneumoniae colonization were observed in HIV-infected compared with HIV-uninfected women (Table 3).
There was no difference in the rate of vertical acquisition of GBS or K pneumoniae in HIV-exposed compared with HUU newborns. The rate of vertical acquisition of E coli was, however, greater in HIV-exposed than HUU neonates in the matched-subset populations (60% vs 52%, respectively; P = .015) as well as in the total cohort (60% vs 53%; P = .0066; Table 4)
Maternal HIV-Infection Status and Sepsis Within 3 Days of Age (EOS)
Two hundred and ninety (3.6%) of the 8129 infants were hospitalized for sepsis within the first 3 days of life; 29 (0.36%) with culture-confirmed sepsis and 258 (3.2%) with clinical sepsis. The incidence of EOS did not differ significantly between HIV-exposed and HUU neonates in the matched subset analysis or the total cohort. The incidence of culture-confirmed EOS was, however, 3.3-fold greater among HIV-exposed compared with HUU in the matched analysis (P = .05), and 1.67-fold increased in the total-cohort analysis (P = .167; Table 5). In the total cohort, among HIV-exposed neonates born to mothers with CD4+ results, the incidence of EOS (per 1000 births) was inversely associated with the immunologic status of the mother (CD4+ cells/mm3 <200, 75.9; 200–350, 40.2; >350, 19.0; P = .0065, trend is linear).
Among HIV-exposed newborns tested by HIV PCR, there was a higher incidence of clinically diagnosed EOS in HIV-infected (102) compared with HEU newborns (30.6) in the matched subset analysis (P = .033), with a similar increase in the total-cohort analysis (P < .0001; Table 6). In the matched-subset analysis, HEU newborns had a lower incidence (20.6/1000 births) of EOS than HUU newborns (33.7; P = .045); with a similar difference observed in the total-cohort analysis (P = .004; Table 7). There was no difference in incidence of culture-confirmed EOS between HEU and HUU in the matched-subset or total-cohort analysis.
Maternal HIV-Infection Status and Neonatal Sepsis Between Days 4 and 28 (LOS)
The incidence of LOS was similar between HIV-unexposed and HIV-exposed neonates in the matched analysis and total cohort (Table 5). Among HIV-exposed neonates, the incidence of LOS was greater in HIV-infected (26.8) than HEU (5.6; P = .042) neonates in the total cohort, with a similar trend observed in the matched analysis (30.6 vs 10.2; P = .62; Table 6). The incidence of LOS did not differ between HEU and HUU in the matched or total-cohort analysis, Table 7.
Among neonates with maternal CD4+ results available, the incidence of LOS was inversely associated with maternal immunologic status (CD4+ <200 cells/mm3, 19.0; 200–350 cells/mm3, 10.0; and >350 cells/mm3, 8.1; P = .55).
To our knowledge, this is the first study to have reported on the impact of maternal HIV infection on vertical transmission of bacteria to newborns, and on the relative incidence of EOS and LOS between HIV-exposed and HIV-unexposed neonates. Our data come from a setting of high maternal HIV infection (29%), high usage of nevirapine to prevent mother-to-child HIV transmission, and extremely low usage of triple antiretroviral therapy among mothers. Whereas it has been hypothesized that HEU neonates are at higher risk of developing sepsis due to impaired maternal transfer of antibody,7 data from our study do not corroborate such speculation. On the contrary, EOS rates were marginally lower in HEU than HUU in the total cohort and matched subset analysis; and no difference was observed for LOS rates between these groups.
The absence of an increased risk of GBS colonization among HIV-infected women in our study is supported by another recently published study from Malawi, which found no evidence of increased GBS colonization associated with HIV, except in women with CD4+ counts >500mm3.11 Overall, our observations that maternal colonization and vertical transmission rates for leading sepsis pathogens did not differ considerably between HIV-infected and -uninfected mothers lend support to our observation that HIV-exposed and -unexposed newborns did not differ in neonatal sepsis risk. Although the number of culture-confirmed EOS cases in our cohort was small, limiting our power to detect a difference in this end point, we believe that the marginal trend in culture-confirmed EOS does not belie a true difference in risk between HIV-exposed and -unexposed newborns. However, we acknowledge that the clinical sepsis end point has lower specificity than an invasive sepsis end point, and thus, culture-confirmed trends may not match clinical sepsis trends.
Our study, however, did identify that EOS rates were higher in HIV-infected neonates. HIV exposed neonates born to severely immunocompromized mothers (CD4+ <200 cells/mm3) were at higher risk of developing EOS and LOS than those born to mothers with CD4+ >350 cells/mm3. This observation, however, requires further corroboration because CD4+ testing was very limited. The increased risk of EOS and LOS among HIV-infected compared with HEU in the total cohort and a similar trend in the matched analysis, emphasize the need to further reduce mother-to-child transmission of HIV.
Data from our study are in contrast to other studies that have reported an increase in morbidity and mortality in HIV-exposed infants, including during early infancy.13,23 Mortality rate in African HEU infants is 72.5 to 98.7/1000, compared with 48.0 to 91.0/1000 in HUU infants and 309.1 to 420.8/1000 in HIV-infected infants,14,16 with almost 30% of these deaths occurring in first 8 weeks of life.14 This is particularly pertinent because disease progression in African HIV-infected infants is rapid, including 85% of infants progressing to severe AIDS by 6 months of age.24 Our data are also not consistent with the data reported from a study in Belgium, in which the incidence of neonatal GBS sepsis, especially LOS, was greater in HIV-exposed (1.55%) than HIV-unexposed (0.08%) newborns.25
The importance of adjusting for maternal factors when assessing the association between maternal HIV infection and neonatal sepsis is evident from our observation that many risk factors commonly associated with EOS, including prolonged ROM and UTIs, were more prevalent in HIV-infected women. A major strength of our study was the utilization of propensity score methods to better evaluate associations between our comparison groups. Propensity score matching allowed us to select an appropriate referent group similar on known or suspected risk factors for each outcome evaluated while blinded to the outcome. When conclusions drawn from the entire cohort differ from the matched subset, we believe the matched subset gives the most valid comparison between the groups with regard to whether maternal-infection status, newborn HIV-exposure or HIV-infection status is responsible for the differences between groups in the applicable comparisons. The differences between the entire cohort and matched subset analyses highlights the dangers of drawing conclusions about the impact of HIV exposure on neonatal morbidity (or mortality) without appropriately adjusting for other factors.
Our study does, however, have some limitations including that CD4+ results were available for only 34.7% of HIV-infected women, and HIV testing was only undertaken in 64% of HIV-exposed infants. It is reassuring that infants with HIV testing were similar to infants without HIV testing for the variables available, although the paucity of maternal CD4+ count data prevented us from ascertaining whether this variable differed between the 2 groups. Incidentally, vertical transmission rate of HIV infection observed in our study (8.2%), where sd-NVP was recommended for the mother and newborn, was not affected by chlorhexidine interventional wipes and consistent with the transmission rate reported elsewhere with the same regimen (10.3%).24 A further limitation of our study included that our enrollment procedures and inclusion and exclusion criteria skewed enrollment away from inclusion of preterm infants. The incidence of preterm neonates in our study population was lower than in the hospital population (4% vs 15%, respectively) (S. C. Velaphi, personal communication, January 10, 2011). Consequently, because prematurity is a recognized risk factor for neonatal sepsis, the rates of EOS and LOS reported in our study may underestimate the true rates of neonatal sepsis in our setting. Nevertheless,a recent review reported that the incidence of clinical sepsis in infants <60 days in developing countries was 49 to 170 per 1000 and 5.5 per 1000 for culture-confirmed sepsis.26 Although the incidence of neonatal sepsis observed in our study was toward the lower bound of this for clinical sepsis (33.4 per 1000), we report a similar incidence of culture-confirmed sepsis (5.9 per 1000). The lower incidence of clinical sepsis rates in our study may have been influenced by more stringent criteria used in our study for diagnosing “clinical sepsis.”
Reducing neonatal morbidity and mortality has been identified as a key focus area for public health under Millennium development goal number 4, which aims at reducing under-5 mortality by two-thirds in comparison with 1990 levels.1 Although there was a high incidence of neonatal sepsis in our study population, including among the highest reported incidence rates of GBS sepsis from any developing country, maternal HIV infection does not appear to increase the risk of either early- or late-onset neonatal sepsis, except among newborns with HIV infection and possibly among newborns of severely immunocompromized mothers. Consequently, we do not predict the HIV prevention roll-out in sub-Saharan Africa will result in important reductions in neonatal sepsis in this region, and sepsis-specific interventions such as GBS vaccines may hold most promise in reducing neonatal infection-associated mortality.
Data Safety Monitoring Board: Kathryn Edwards (chair), Peter Cooper, Maria Deloria Knoll, James McIntyre. PoPS Trial Team: Scientific steering committee: Marleen Temmerman (chair), G. Justus Hofmeyr, Paolo Miotti, Robert Pattinson, Haroon Saloojee; Advisers and collaborators: Anne Schuchat, Keith Klugman, Cynthia Whitney, Avye Violari, Anne von Gottberg, Donald Rubin. SAE reviewers: Boris Jivkov, Aneetha Moodley. CDC team: Michael Thigpen, Roopal Patel. RMPRU team: Waarisa Fareed, Suzett Fourie, Veebha Gosai, Razia Hassan, Gavin Jaques, Waasila Jassat, Stephanie Jones, Marianne Kohler, Pravesh Lakha, Martin Laque, Kerry Marran, study midwives, nurses, laboratory staff, counselors, data capturers.
We thank the staff of Departments of Obstetrics, Neonatology and Paediatrics at Chris Hani Baragwanath hospital for their dedication to their patients, including our trial participants.
- Accepted May 9, 2012.
- Address correspondence to Shabir A Madhi, MBBCh, FCPaeds(SA), PhD, Chris Hani Rd, Chris Hani-Baragwanath Hospital, New Nurses Residence-11th Floor, Central-West wing, PO Box 90753, Bertsham, 2013 Soweto, South Africa. E-mail:
Drs Cutland, Schrag, Zell, Buchmann, Velaphi, and Madhi participated in the conception of the trial, study design, protocol development and amendment, study planning and implementation; Drs Cutland and Groome followed up participants and collected data; Dr Adrian managed sample processing; Dr Zell and Kuwanda analyzed the data; all authors participated in interpretation of the results; Drs Cutland, Zell, Schrag, and Madhi drafted the manuscript and all authors contributed to critical review and revision of the manuscript; all authors have seen and approved the final version of the manuscript; and Dr Madhi, as corresponding author, had full access to data and final responsibility for the decision to submit manuscript for publication.
FINANCIAL DISCLOSURE: The authors have indicated they have no financial relationships relevant to this article to disclose.
FUNDING: This work was supported by the US Agency for International Development, National Vaccine Program Office and Centers for Disease Control’s (CDC) Antimicrobial Resistance Working Group via CDC Cooperative Agreements U50/CCU021960 and 5U01CI000318, and Bill and Melinda Gates Foundation grant 39415. The article contents are solely the responsibility of the authors and do not necessarily represent the official views of the sponsors.
- Schellenberg JJ,
- Links MG,
- Hill JE,
- et al
- Spear GT,
- Sikaroodi M,
- Zariffard MR,
- Landay AL,
- French AL,
- Gillevet PM
- Chilongozi D,
- Wang L,
- Brown L,
- et al.,
- HIVNET 024 Study Team
- Kuhn L,
- Sinkala M,
- Semrau K,
- et al
- Kuhn L,
- Kasonde P,
- Sinkala M,
- et al
- ↵Progress Report on Declaration of Commitment on HIV and AIDS, Republic of South Africa. 2008. Available at: http://data.unaids.org/pub/Report/2008/south_africa_2008_country_progress_report_en.pdf. Accessed February 20, 2012
- Department of Health
- Rosenbaum PR,
- Rubin DB
- Aiken CG
- Copyright © 2012 by the American Academy of Pediatrics