Pediatric HIV has a more aggressive course than adult disease; without treatment, 50% of children will die by 2 years . Early diagnosis and initiation of antiretroviral therapy (ART) confers a strong survival benefit in infants . However, early diagnosis of infant HIV poses challenges.
Previous WHO guidelines recommended the use of either infant or maternal serology during sick child visits to determine HIV exposure, followed by virologic testing for HIV-exposed infants up to 18 months . The 2016 WHO guidelines state a preference for maternal serology in determining infant HIV exposure status at all ages between 0 and 18 months, but also note that infant serology results can reliably assess HIV exposure in infants less than 4 months of age .
Within prevention of mother-to-child transmission of HIV (PMTCT) programs, maternal HIV status is determined during pregnancy or intrapartum, allowing preferential use of maternal serology to determine infant HIV exposure. However, improved effectiveness of PMTCT interventions means that a growing proportion of new infant infections will occur due to incident maternal HIV infections in the postpartum period, often outside of the PMTCT system. These infant infections may first present at sick child visits or during hospitalization. In this symptomatic population of infants and children, enhanced for recent maternal infection, it is important to determine how well the new WHO guidelines for infant HIV exposure ascertainment perform and note how many infections would be missed with continued use of older guidelines during scale-up of newer guidelines.
In this study of hospitalized children conducted in Kenya, we characterized the performance of infant rapid serology tests to determine HIV exposure status in three age groups: 0–3, 4–8, and 9–18 months. We calculated the proportion of infant or child HIV infections that would have been missed if infant rapid serology tests were used to determine HIV exposure.
Recruitment and enrollment
This analysis was nested in the recruitment phase of the Pediatric Urgent Start of HAART (PUSH) clinical trial (NCT02063880), which was conducted to determine whether urgent antiretroviral therapy (ART) (within 48 h) would benefit hospitalized, HIV-infected children admitted with a severe co-infection, compared with early ART initiation (7–14 days).
In this analysis, sick, hospitalized children were recruited from four hospitals, two in Nairobi (Kenyatta National Hospital and Mbagathi County Hospital), and two in western Kenya (Jaramogi Oginga Odinga Referral and Teaching Hospital and Kisumu County Hospital) between April 2013 and May 2015. A subset of the children included in this analysis were subsequently enrolled in the clinical trial. Children were eligible for enrollment in the trial if they were HIV-infected, age 0–12 years, ART-naïve other than for PMTCT, did not have a suspected central nervous system infection, and were living within the study catchment area.
The parent clinical trial was approved by the University of Nairobi/Kenyatta National Hospital Ethics and Research Committee and the University of Washington Institutional Review Board. Caregivers of children provided written consent for HIV testing.
Maternal and infant HIV rapid serologic tests [Determine or KHB Colloidal Gold (KHB) followed by Unigold or First Response, per Kenyan guidelines] were performed in tandem through provider initiated testing and counseling (PITC) at hospitalization. HIV-1 DNA PCR assay was performed on dry blood spot (DBS) samples using an in-house method  in Nairobi sites and Abbott Real-Time HIV-1 qualitative assay at the Kenya Medical Research Institute laboratory in western Kenya sites.
This analysis is restricted to infants whose mothers were HIV positive by rapid serology test. All infants whose mothers were HIV positive had an HIV DNA PCR test performed, irrespective of infant serology results. We characterized the proportion of truly positive infants and children (using HIV DNA PCR as the gold standard) that would have been identified or missed if infant rapid serology tests were used to determine HIV exposure.
For the subset of infants whose maternal serology was positive, infant rapid serology test was negative, and who were categorized as truly HIV-infected by HIV DNA PCR testing, archived DBS samples were retested using fourth-generation CBios HIV 1+2 antibody/antigen (Ag/Ab), a laboratory-based immunoassay kit that detects HIV-1 glycoprotein 41 (gp41), gp120 and HIV-2 gp36 antibodies and HIV-1 protein 24 (p24) antigens (Chemus Bioscience, Inc; www.chemux.com; San Francisco, California, USA). DBS samples were processed according to the Kania et al. protocol , with the following differences: whole blood was used instead of serum, DBS samples were stored at ambient room temperature. According to the HIV-1+2 Ag/Ab ELISA test kit criteria, the results were interpreted as positive if the absorbance of the test well was greater than or equal to the cutoff value of the test plate and negative if the absorbance was less than the cutoff value. Additionally, samples with absorbance values lower than but near the cutoff value were interpreted as borderline if the ratio of the sample test well absorbance to the cutoff value of the test plate was between 0.9 and 1.1. Borderline samples were retested in duplicate to confirm initial results. These tests were performed post hoc and were not used to determine infant HIV status.
In this same subset of infants, we reviewed PMTCT history (maternal HIV test history and result during pregnancy, maternal and infant antiretroviral drug prophylaxis or treatment exposure), infant hospitalization history, and 6-month outcomes. Infants enrolled in the clinical trial had available information on viral loads, CD4+ percentage (CD4%), WHO clinical staging, and growth parameters [weight for age (WAZ) and height for age (HAZ) z scores].
Descriptive statistics were summarized using proportions, medians and interquartile ranges; 95% confidence intervals (95% CI) were estimated using a binomial distribution. We compared infant rapid serology test results in three age groups (0–3, 4–8 and 9–18 months).
We calculated the proportion of infant infections that would have been missed if infant rapid serology tests were used to determine HIV exposure status by dividing the total number of infants positive by HIV DNA PCR by the number of infants who were both positive by HIV DNA PCR and whose rapid serology test results were either positive or indeterminate.
All tests were two-sided with alpha 0.05. Data analysis was conducted using STATA 14 (Stata Corporation, College Station, Texas, USA).
There were 203 infants and children ages 0–18 months whose mothers were positive by HIV serology (HIV-exposed infants and children).
Performance of infant rapid serology tests to determine HIV exposure status by age group
Among 65 HIV-exposed infants age 0–3 months, median age was 2.0 months [interquartile range (IQR) 0.6–3.1], 32 (49%) were male and 16 (25%) were HIV-infected. Overall, 48 (74%), one (2%) and 16 (25%) had positive, indeterminate and negative infant rapid serology test results, respectively. Of those who were infant rapid serology test positive, indeterminate and negative, 12 (25%), zero (0%) and four (25%) were HIV-infected by DNA PCR, respectively. In this age group, four of 16 (25%) infant infections would have been missed had only infant rapid serology tests been used to determine HIV exposure status; 33% (95% CI 10–65%), more infant infections would have been identified in an algorithm that defined HIV exposure by maternal and not infant serology (16 versus 12, P = 0.39).
Among 71 HIV-exposed infants age 4–8 months, median age was 6.6 months (IQR 5.7–7.8), 40 (56%) were male and 23 (32%) were HIV-infected. Overall, 31 (44%), eight (11%) and 32 (45%) had positive, indeterminate and negative infant rapid serology test results, respectively. Of those who were infant rapid serology test positive, indeterminate and negative, 14 (45%), two (25%) and seven (22%) were HIV-infected by DNA PCR, respectively. In this age group, seven of 23 (30%) infant infections would have been missed had only infant rapid serology tests been used to determine HIV exposure status; 44% (95% CI 20–70%), more infant infections would have been identified in an algorithm that defined HIV exposure by maternal and not infant serology (23 versus 16, P = 0.21).
Among 67 HIV-exposed infants/children age 9–18 months, median age was 14.2 months (IQR 11.7–16.9), 33 (49%) were male and 38 (60%) were HIV-infected. Overall, 40 (60%), two (3%) and 25 (37%) had positive, indeterminate and negative infant/child rapid serology test results, respectively. Of those who were infant/child rapid serology test positive, indeterminate and negative, 36 (90%), two (100%) and two (8%) were HIV-infected by DNA PCR, respectively. In this age group, two of 40 (5%) infant/child infections would have been missed had only infant/child rapid serology tests been used to determine HIV exposure status; 5% (95% CI 0.1–18%), more infant/child infections would have been identified in an algorithm that defined HIV exposure by maternal and not infant/child serology (40 versus 38, P = 0.73) (Fig. 1).
Clinical and care-seeking history and ELISA results of rapid serology test negative, PCR-positive infants
Among 13 infants who were infant rapid serology test negative and HIV DNA PCR positive, median age was 5.7 months (range 2.4–11.3). Seven (54%) mothers reported that they tested HIV negative during pregnancy, likely representing incident infections during late pregnancy/postpartum. PMTCT antiretroviral exposure was uncommon with four of 13 mothers reporting having received any PMTCT medications and no infants having received prophylaxis. Seven of these 13 infants enrolled in the clinical trial and had baseline clinical data; most infants were WHO stage III/IV, had high viral loads, low CD4%, and were wasted and stunted (Table 1).
Upon retesting 10 available archived DBS samples by serology using fourth-generation CBios HIV-1+2 antibody/antigen ELISA assay, 9/10 infant samples were identified as HIV positive for either antibody or antigen. By 6 months following diagnosis, three infants had died, six were known to be alive, and the remaining four were either lost to follow-up or not systematically followed, as they were not enrolled in the trial (Table 1).
In this cohort of HIV-exposed, hospitalized infants and children, HIV was common with over one-third of infants and children infected. An HIV testing algorithm that relies on infant rapid serology tests to define HIV exposure would have missed 25 and 30% of HIV-infected infants in the 0–3 and 4–8 month age groups, respectively. Infant rapid serology tests performed better in older children between 9 and 18 months, missing an estimated 5% of infections. Among the infant rapid serology test negative, DNA PCR-positive infants/children, mothers frequently tested HIV negative during antenatal care, likely representing incident maternal infection during late pregnancy/postpartum.
These findings provide evidence to review the performance of the 2016 WHO guidelines among a symptomatic population. For infants 0–3 months, the permissiveness of using infant rapid serology tests to determine exposure status could result in up to one-quarter of infant infections being missed. For infants 4–8 months, the recommendation to use maternal serology to determine exposure status is well warranted, given the large proportion of infants who were HIV-infected by DNA PCR but were negative by infant rapid serology test. For infants and children 9–18 months, the use of maternal serology to determine infant exposure status performed comparably to using infant rapid serology test. Revisions to the guidelines to recommend the preferential use of maternal serology at all ages to determine infant or child exposure status for symptomatic infants and children should be considered.
There are several potential reasons that HIV-exposed, sick infants could be rapid test seronegative but HIV-infected. Rapid serology test negative infants may lack HIV antibodies. Mothers with acute HIV infections during late pregnancy/postpartum may not have produced or transferred antibodies; consequently, infants born to these mothers may be born without maternal antibodies. A majority of rapid test seronegative, HIV-infected infants had mothers who reported testing HIV negative during antenatal care, presumably having acquired HIV during late pregnancy/postpartum. Additionally, infants recently infected with HIV may not yet have mounted their own antibody response , and infants may have been identified during the window period when maternal antibodies had waned but prior to the development of a robust infant antibody response . Although HIV antibody production is relatively rapid in adults, antibody production is slower in infants . An in-vitro study of lymphocytes from HIV-infected infants measured infant antibody production and noted that just 59% of samples had detectable levels of infant antibodies by 2 months, and 86% by 6 months of age . It is also possible that sensitivity of rapid antibody tests is impaired during severe illness due to low levels of antibody production . However, this may not explain all undetected infections, as 10% of these infants were negative by both antibody and antigen on repeat testing with a fourth-generation ELISA assay.
Regardless of the mechanism, young infant/child rapid serology tests were not reliable in assessing infant HIV exposure in this study. As PMTCT coverage expands, an increasing proportion of HIV-infected infants will be due to acute maternal infections ; therefore, algorithms that recommend HIV DNA PCR tests on symptomatic HIV-exposed infants (by maternal serology), regardless of infant serostatus, may enhance timely case detection and ART initiation.
Previous studies have compared rapid HIV tests against PCR. Studies of 18 month olds or younger observed that Determine identified between 93 and 100% of HIV-infected children [11–14]. Only one study included hospitalized children. A South African study noted that virologic tests among infants who were symptomatic, regardless of infant rapid serology test results, improved sensitivity from 96.3 to 98.4% . However, the proportion of infants who were rapid test seronegative and HIV-infected was lower than in our study, possibly because less than 10% of children were acutely ill in hospital.
It is important to acknowledge that while HIV DNA testing enhances diagnostic performance, there are challenges in timely performance and return of results. Delays in turnaround time can result in infant deaths while awaiting diagnosis and interventions to decrease turnaround time are needed [15–17]. Point-of-care virologic tests decrease turnaround time for testing ; however, costs, the availability of trained personnel to operate the machinery, and high volumes of patients may challenge widespread adoption of point-of-care technologies. Cost-effectiveness and operations research studies that compare standard laboratory-based with point-of-care-based PCR tests are needed for both asymptomatic and symptomatic infants. Additionally, scale-up of Option B+ PMTCT regimens may further complicate diagnosis of infant infections due to potentially delayed detection of infant virus and antibodies .
This study had several limitations. Due to the combined antibody/antigen detection of the fourth-generation ELISA test used, we were unable to determine whether the nine infants who were positive by this test had antibodies or antigen present in their sample. We recruited from higher-level county and referral hospitals, which may have a disproportionately high level of symptomatic, undiagnosed HIV infection among infants and children; future studies should sample from a range of high and low level inpatient wards. Self-reported negative maternal HIV status during antenatal care may be overreported, which would increase the proportion of infant infections inaccurately attributed to incident maternal HIV infection during pregnancy and postpartum. We were unable to include children brought by a caregiver other than their biological mother; however, these children were rare in this age range.
Young infant infections may be missed under current WHO infant and child testing guidelines, which are permissive of using infant rapid serology tests to determine HIV exposure status. These guidelines, which recommend maternal serology preferentially, perform well in older infants and children, but poorly in younger infants, many of whom were infant rapid serology test negative, but HIV-infected. Revisions to the WHO guidelines to remove endorsement of infant rapid serology tests to determine HIV exposure status in the youngest age range for symptomatic children should be considered.
The authors thank the PUSH study participants and their families, without who this research would not be possible. We thank the PUSH administrative, clinical, and data team for their dedication and support. We thank Dr Bhavna Chohan's lab and Mr Brian Khasimwa for performing HIV PCR tests and CBios tests. We thank members of the Kizazi Working Group, UW Global Center for Integrated Health of Women, Adolescents and Children (Global WACh), Dr Grace John–Stewart's pre/postdoctoral mentorship group, and Kenya Research & Training Center (KRTC) for their support during the preparation of this article.
Its contents are solely the responsibility of the authors and do not necessarily represent the official views of the NIH. The funding sources were not involved in the analyses or interpretation of data. None of the authors were paid to write this article by a pharmaceutical company or other agency.
This research was funded by R01-HD023412, K24-HD054314 (PI for both: G.J.S.), and K12 HD000850 (L.M.C.) from the Eunice Kennedy Shriver National Institute of Child Health and Human Development (NICHD), F31MH099988 from NIMH and F32HD088204 from NICHD and R25 TW009345 from FIC (A.D.W.). This publication was also supported in part by the University of Washington CFAR (P30 AI027757), REDCap UL1TR000423 from NCRR/NIH, the UW Global Center for Integrated Health of Women, Adolescents and Children (Global WACh), and the Pediatric Scientist Development Program (PSDP) through grants from the American Academy of Pediatrics and American Pediatric Society (L.M.C.) and the Fogarty International Center of the National Institutes of Health (D43TW009783 to INN).
Conflicts of interest
There are no conflicts of interest.
Partial data presented previously at AIDS meeting in Durban, South Africa, #TUPEB079, 18–22 July 2016.
1. Newell ML, Coovadia H, Cortina-Borja M, Rollins N, Gaillard P, Dabis F, et al. Mortality of infected and uninfected infants born to HIV-infected mothers in Africa: a pooled analysis
2. Violari A, Cotton MF, Gibb DM, Babiker AG, Steyn J, Madhi SA, et al. Early antiretroviral therapy and mortality among HIV-infected infants
. N Engl J Med
3. World Health Organization. WHO recommendations on the diagnosis of HIV infection in infants and children
. Geneva, Switzerland: World Health Organization; 2010.
4. World Health Organization. Consolidated guidelines on the use of antiretroviral drugs for treating and preventing HIV infection: recommendations for a public health approach
. 2nd ed.Geneva, Switzerland: World Health Organization; 2016.
5. Chohan BH, Emery S, Wamalwa D, John-Stewart G, Majiwa M, Ng’ayo M, et al. Evaluation of a single round polymerase chain reaction assay using dried blood spots for diagnosis of HIV-1 infection in infants in an African setting
. BMC Pediatr
6. Kania D, Truong TN, Montoya A, Nagot N, Van de Perre P, Tuaillon E. Performances of fourth generation HIV antigen/antibody assays on filter paper for detection of early HIV infections
. J Clin Virol
7. Pollack H, Zhan MX, Ilmet-Moore T, Ajuang-Simbiri K, Krasinski K, Borkowsky W. Ontogeny of antihuman immunodeficiency virus (HIV) antibody production in HIV-1-infected infants
. Proc Natl Acad Sci U S A
8. Lepage P, Van de Perre P, Simonon A, Msellati P, Hitimana DG, Dabis F. Transient seroreversion in children born to human immunodeficiency virus 1-infected mothers
. Pediatr Infect Dis J
9. Martinez DR, Permar SR, Fouda GG. Contrasting adult and infant immune responses to HIV infection and vaccination
. Clin Vaccine Immunol
10. Drake AL, Wagner A, Richardson B, John-Stewart G. Incident HIV during pregnancy and postpartum and risk of mother-to-child HIV transmission: a systematic review and meta-analysis
. PLoS Med
11. De Baets AJ, Edidi BS, Kasali MJ, Beelaert G, Schrooten W, Litzroth A, et al. Pediatric human immunodeficiency virus screening in an African district hospital
. Clin Diagn Lab Immunol
12. Cape Town, South Africa
, Nuttall JJC. The performance of HIV rapid antibody detection assays in children
. 2015; University of Cape Town,
13. Menzies NA, Homsy J, Chang Pitter JY, Pitter C, Mermin J, Downing R, et al. Cost-effectiveness of routine rapid human immunodeficiency virus antibody testing before DNA-PCR testing for early diagnosis of infants in resource-limited settings
. Pediatr Infect Dis J
14. Sherman GG, Lilian RR, Coovadia AH. The performance of 5 rapid HIV tests using whole blood in infants and children: selecting a test to achieve the clinical objective
. Pediatr Infect Dis J
15. Kellerman S, Essajee S. HIV testing for children in resource-limited settings: what are we waiting for?
. PLoS Med
16. De Cock KM, Bunnell R, Mermin J. Unfinished business – expanding HIV testing in developing countries
. N Engl J Med
17. Sutcliffe CG, van Dijk JH, Hamangaba F, Mayani F, Moss WJ. Turnaround time for early infant HIV diagnosis in rural Zambia: a chart review
. PLoS One
18. Jani I, Meggi B, Loquiha O, Tobaiwa O, Mudenyanga C, Mutsaka D, Mabunda N, et al. Effect of point-of-care testing on antiretroviral-therapy initiation rates in infants
. In: Conference on Retroviruses and Opportunistic Infections (CROI)
. Seattle, WA; 2017.
19. King CC, Kourtis AP, Persaud D, Nelson JA, Ziemniak C, Hudgens MG, et al. Delayed HIV detection among infants exposed to postnatal antiretroviral prophylaxis during breastfeeding