Mother-To-Child Transmission of HIV in Kenya: A Cross-Sectional Analysis of The National Database Over Nine Years
Mother-To-Child Transmission of HIV in Kenya: A Cross-Sectional Analysis of The National Database Over Nine Years
Mother-To-Child Transmission of HIV in Kenya: A Cross-Sectional Analysis of The National Database Over Nine Years
Introduction
Without antiretroviral therapy (ART), approximately 50% of HIV-infected infants die by the
age of two[1]. However, the use of elective caesarian section, the provision of highly active anti-
retroviral therapy (HAART) during pregnancy, and the avoidance of breastfeeding have
reduced the HIV transmission rate to less than 2% and resulted in the near elimination of
mother to child transmission (MTCT) [2, 3]. Unfortunately, barriers to healthcare and sub-
optimal infant feeding and care practices still exist in resource-limited settings and continue to
propagate transmission of HIV from mothers to infants. Identifying infected infants early and
initiating them on ART as soon as possible after diagnosis is essential to slow the progression
from HIV infection to AIDS and prolong the life of the patient[1].
Despite substantial progress in efforts to reduce MTCT of HIV in Kenya, HIV transmission
through the end of breastfeeding remains at 17%, with approximately 13,000 new HIV infec-
tions among children in 2014[4]. Option B+, an approach to prevent MTCT of HIV by offer-
ing all HIV-positive pregnant women treatment for life and ART prophylaxis to all exposed
infants, free of charge, has been scaled up through a phased rollout in Kenya starting in 2014;
by October 2015 more than 90% of all sites with programs on prevention to mother-to-child
transmission of HIV (PMTCT) or maternal, newborn, and child health had adopted it[5].
UNAIDS estimates suggest that greater than 90% of all pregnant women attending ANC in
Kenya are tested for HIV[6]. However, gaps remain. Nearly half (42%) of women do not fully
access antenatal care (at least 4 visits during pregnancy)[7], and 39% deliver outside of health-
care facilities[8]. Access to care is poorest in rural areas. Due in large part to this gap, an esti-
mated 33% of HIV-infected pregnant women are not on treatment and early infant diagnosis
of HIV (EID) testing coverage by two months of age in exposed infants was 72% in 2014[4].
Better understanding of these gaps and the remaining risk factors for HIV infection among
infants who get tested is essential to eliminate MTCT in Kenya and neighboring countries.
This study used nearly a decade of nationwide data from a national HIV laboratory data-
base of over 370,000 EID samples to assess the risk factors for HIV transmission in infants and
to identify important predictors of HIV infection over time. We also assessed which facility
entry points had the greatest yield for EID case detection of HIV-infected infants to inform
resource allocation for HIV testing. While as a registry-based study, the data have some limita-
tions, this is amongst the first studies of this scale examining risk factors for HIV infection in
infants through the period of scale-up of Option B+ in sub-Saharan Africa.
was conducted at first visit for all children of unknown HIV status aged <18 months to estab-
lish HIV exposure status. If positive, an EID test was recommended. If the EID test was posi-
tive, the infant was started on ART and if negative, the infant received an HIV antibody test at
9 months (or earlier if child developed symptoms suggestive of HIV). If the antibody test was
positive at 9 months, the infant received a confirmatory EID test. If the EID test was negative,
the HIV antibody test was repeated at 18 months or 6 weeks after cessation of breastfeeding in
a child >18 months[9]. EID testing coverage by 2 months of age in HIV-exposed infants in
Kenya was estimated to be below 50% in 2011 and increased to 72% in 2014[4]. In 2015, the
algorithm was updated to recommend an EID test at 6 weeks or first contact after 6 weeks[10].
Seven laboratories form the testing network and laboratory request forms are compiled into a
national database across the laboratories. Thus, the database covers nearly all infants receiving
EID testing in the country. Data from the database can be viewed on a national dashboard
(http://eid.nascop.org/). Mean turnaround time from sample collection to results dispatch
from the laboratory has changed over time but was 17 days as of July 2015[11]. This cross-sec-
tional analysis was based on a retrospective review of all data stored in the national HIV data-
base between January 2007 and July 2015.
Study population
A total of 370,196 samples were successfully collected from infants visiting health facilities
across all regions in Kenya between January 2007 and July 2015 and tested in seven national
laboratories. 362 samples had the same patient ID number, came from the same facility, were
tested on the same date, and were from an individual of the same age and sex as another sam-
ple; these were considered to be duplicate tests from the same infant and were excluded from
the dataset. Due to improper DBS sample collection, packaging or labeling, 3228 samples were
rejected and were therefore excluded from the analysis. Patients with age listed as a negative
value or greater than or equal to 2 years (n = 765) were excluded from the main analysis; a sec-
ondary analysis was conducted setting age to missing in these individuals. Patients with miss-
ing data on other predictor variables were included in a missing indicator category. 365,841
patients had a valid test result and were included in the tables and primary analyses.
Statistical analysis
Demographic data were summarized with descriptive statistics. The primary outcome was
infant HIV status (infected vs. uninfected). We examined predictors of infant HIV status,
including breastfeeding, age, prophylaxis, antiretroviral therapy and portal of entry. As a pre-
liminary statistical analysis, continuous predictors were compared in relation to infant HIV
result using Student’s T-tests if normally distributed or Mann-Whitney U tests if skewed. Cate-
gorical data were compared using the Fisher’s exact test or the chi square test.
Logistic regression was used to calculate univariate odds ratios of potential predictors in
relation to infant HIV result and 95% confidence intervals. Variables that were significantly
associated with infant HIV infection in univariate models were included in a multivariate
logistic regression model. Finally, we combined all behavioral risk factors identified for infant
HIV infection into one variable and assessed the increased odds of infection in an infant with
multiple risk factors. The models accounted for clustering by health facilities using the vce
(cluster) function for statistical analysis. Reference groups were selected based on the most
common group and/or the standard of care in Kenya. The interaction between age and sex
was examined and a Wald p-value for interaction was calculated. Separate logistic regression
models were run by time period (2007–2010, prior to any Option B+ programs and when EID
testing was limited and targeted; 2011–2013, when EID testing was scaled up and Option B+
was piloted at some sites; and 2014–2015, when Option B+ was rolled out nationally and EID
coverage continued to expand) to assess differences over time as Kenya’s national testing and
treatment program grew and developed. Significance levels were set at p<0.05. All analysis in
this study was conducted using Stata Version 13 (StataCorp LP, College Station, Texas, USA).
Quality assurance
Kenya has a national quality assurance program for the seven molecular laboratories. The pro-
gram supports external quality assurance, training/registration of users, and harmonisation of
operating procedures. All seven labs are enrolled and participate in the CDC Atlanta profi-
ciency testing program. They also participate in a quarterly inter-laboratory EQA programme.
Ethical approval
This study was approved by the Scientific Steering Committee and Ethical Review Committee
of the Kenya Medical Research Institute (Protocol SSC No. 1066). The need for participant
consent was waived. No identifiable patient information was included in the database, but
facility-specific patient identification numbers were included. NASCOP considers the EID ser-
vice as a national HIV program and a standard of care.
Results
Patient characteristics and univariate and multivariate logistic regression analyses are pre-
sented in Table 1. Of the 365,841 infants in the dataset, 8.9% (32,441) were found to be HIV-
infected. 50.4% of samples (of those with known sex) were from females. Data on sex was
unavailable for 10.9% of samples. The median age at testing was 1.8 months (IQR: 1.5–5.0
Table 1. Predictive factors for HIV positivity in 365,841 infants testing for HIV in Kenya.
Factor Univariate Multivariate
N % HIV+ OR 95%CI OR 95%CI
Gender Male 161,682 8.4% Ref Ref
Female 164,259 8.9% 1.06 1.04– 1.08 1.05–
1.09 1.11
Missing 39,900 10.4% 1.26 1.19– 1.11 1.03–
1.32 1.20
Age group 0–6 weeks 47,968 10.1% 2.43 2.25– 1.72 1.59–
2.63 1.86
6 weeks-2 months 138,494 4.4% Ref Ref
2–6 months 86,877 9.4% 2.25 2.12– 1.76 1.68–
2.39 1.86
6–9 months 35,888 11.9% 2.93 2.73– 2.08 1.95–
3.13 2.22
9–18 months 46,834 16.3% 4.23 3.96– 3.05 2.89–
4.51 3.21
18–24 months 4,599 21.6% 5.98 5.36– 4.26 3.87–
6.67 4.69
Missing 5,181 8.3% 1.96 1.73– 1.47 1.26–
2.22 1.72
PMTCT AZT from 14 weeks of pregnancy or later; AZT+3TC+sdNVP during labor; AZT 41,394 6.5% 1.13 1.06– 1.13 1.06–
Intervention +3TC for 7 days postpartum 1.21 1.20
HAART 125,166 5.8% Ref Ref
Interrupted HAART (HAART until end of breastfeeding) 11,437 7.0% 1.22 1.09– 1.20 1.08–
1.37 1.33
None 43,837 15.2% 2.90 2.67– 1.92 1.79–
3.16 2.06
Other 12,711 8.6% 1.53 1.37– 1.23 1.10–
1.72 1.36
Single dose NVP Only 8,633 9.1% 1.62 1.45– 1.32 1.20–
1.80 1.45
Missing 122,663 10.7% 1.94 1.80– 1.53 1.43–
2.10 1.63
Infant Prophylaxis AZT+3TC for 7 days only 128 2.3% 0.57 0.17– 0.40 0.12–
1.90 1.35
NVP during breastfeeding 50,505 6.3% 1.58 1.45– 1.30 1.20–
1.73 1.41
NVP for 6 weeks (Mother on HAART or not breastfeeding) 67,429 4.0% Ref Ref
None 19,889 21.7% 6.56 5.83– 2.76 2.51–
7.39 3.05
Other 5,689 8.4% 2.18 1.82– 1.46 1.25–
2.60 1.70
Single dose NVP+AZT+3TC 3,436 8.1% 2.09 1.74– 1.45 1.22–
2.50 1.73
Single dose NVP Only 21,541 5.9% 1.49 1.34– 1.19 1.09–
1.65 1.30
Missing 197,224 10.2% 2.71 2.48– 1.56 1.43–
2.95 1.70
Breastfeeding Exclusive breastfeeding 167,166 6.2% Ref Ref
Mixed breastfeeding 23,682 15.2% 2.73 2.48– 1.39 1.30–
3.00 1.49
Breastfeeding, unspecified 32,609 11.3% 1.94 1.81– 1.13 1.05–
2.08 1.20
(Continued)
Table 1. (Continued)
Univariate and multivariable logistic regression accounting for clustering by health facility. The multivariable model was adjusted for all factors in this table.
The missing indicator method was used to account for missing data. Abbreviations: AZT: zidovudine. 3TC: lamivudine. Sd: single dose. NVP: nevirapine.
HAART: highly active antiretroviral therapy. PMTCT: prevention of mother-to-child transmission of HIV.
https://doi.org/10.1371/journal.pone.0183860.t001
months). Age at testing did not differ substantially between the sexes. 51.5% mothers were
confirmed to be on highly active antiretroviral therapy (HAART; out of those with data on
treatment regimen) and an additional 4.7% on interrupted HAART. 18.0% were on no treat-
ment. 11.8% of infants were not on any HIV prophylaxis (among those with known prophylaxis
status). Exclusive breastfeeding (EBF) was provided to 64.0% of infants (out of those with
breastfeeding information), while 14.5% of infants were not breastfed. Mixed breastfeeding was
practiced by 9.1% of infants, although only 60.4% of those infants were under 6 months of age
and it is unknown whether the older children were previously exclusively breastfed and cur-
rently complementary feeding or were previously mixed breastfed. In addition, mixed breast-
feeding decreased over time. 12.5% of infants were breastfed but did not have information
on whether it was exclusive or mixed (hereafter referred to as “unspecified breastfeeding”).
Approximately 72.3% of samples were collected via PMTCT/maternal and child health (MCH)
clinic samples, while 18.8% were collected from CCC/ PSC, 3.5% from OPD, 1.2% from the
pediatric ward and 0.9% from the maternity ward. The vast majority of samples across entry
points were from patients ages 1 to 2 months, although the entry points with the most extreme
peak in that age range were from PMTCT/MCH and CCC/PSC.
In multivariate analyses, the odds ratio for HIV infection in females was 1.08 (95% CI 1.05–
1.11) compared to males. When examining the sex differences in infection by age, there was a
statistically significant difference between positivity in females and males through 6 months
but no statistically significant difference in older children (p-value for interaction = 0.01).
10.1% of newborns less than 6 weeks of age were found to be infected with HIV. Compared to
infants being tested at 6 weeks to 2 months (the majority of infants), the odds ratio of testing
positive for HIV infection was 1.72 (95% CI: 1.59–1.86) in infants aged 0 to 6 weeks. However,
most of those infants were listed as testing at 0 months, which may have represented a data
entry error; sensitivity analyses in which these infants were considered to be missing age
showed an attenuated association in this age group [OR changed to 1.28 (95% CI: 1.18–1.40)].
The percentage testing positive for infection was the lowest at 6 weeks to 2 months when the
largest number of infants were tested, and then increased with age to 21.6% at 18 months to 2
years (76.3% of infants in this category were 18.0–18.5 months old). Compared to infants
being tested at 6 weeks to 2 months, the odds ratio of testing positive for HIV infection was
1.76 (95% CI: 1.68–1.86) in infants 2 to 6 months, increasing to 4.26 (95% CI: 3.87–4.69) in
infants 18–24 months. Infants whose mothers were not taking ART had nearly double the
odds of the HIV infection when compared to those whose mothers were on treatment (multi-
variable-adjusted OR 1.92; 95% CI 1.79–2.06). For infants not on any prophylaxis, the odds
ratio for HIV infection was 2.76 (95% CI 2.51–3.05) when compared to those who were on
nevirapine (NVP) for six weeks whose mothers were on ART or were not breastfeeding. The
percentage of infants on nevirapine for 6 weeks that tested positive was 4.0% vs. 21.7% of
infants on no prophylaxis. Among infants whose mothers were on HAART, the differences in
HIV infection by infant prophylaxis regimen were smaller but still substantial; 3.3% for those
on nevirapine for 6 weeks and mother breastfeeding while 11.8% for those on no prophylaxis.
Infants who were categorized as mixed breastfed had 1.39 (95% CI: 1.30–1.49) times the odds
of HIV infection when compared to exclusively breastfed infants. There was not a statistically
significant difference in HIV infection between exclusively breastfeeding and not breastfeed-
ing. The highest yield for detecting infected patients was observed in the pediatric ward and
the outpatient department. For infants who sought health services at the pediatric ward,
the odds ratio of HIV infection was 3.32 (95% CI 2.82–3.92), while for those seen at the outpa-
tient departments the odds ratio was 1.65 (95% CI 1.52–1.79) when compared to the MCH-
PMTCT. However, given the difference in the volume of patients tested in each entry point,
only 3.4% (5.3% in 2014–2015) of the HIV-infected infants were identified via the pediatric
ward and 5.9% (9.0% in 2014–2015) through the outpatient department, whereas 46.9%
(55.8% in 2014–2015) of patients were identified through MCH/PMTCT.
In a model combining sub-optimal behaviors for HIV transmission, the adjusted odds ratio
of HIV infection for infants who did not receive infant prophylaxis, were mixed fed, and
whose mothers did not receive any form of antiretroviral therapy (N = 2,550) was 7.10 (95%CI
5.64–8.93) compared to infants who received nevirapine for 6 weeks, were not breastfed, and
whose mothers were on HAART (N = 3,325), and the odds were similar when compared to
infants who received nevirapine, were exclusively breastfed, and mothers were on HAART
[OR: 6.91 (95% CI: 6.14–7.78); S1 Table].
Table 2 shows multivariable-adjusted model results stratified by early and recent time
period. The percentage of infants receiving an EID test who tested positive for HIV infection
Table 2. Risk factors for HIV positivity in 2007–2010 (N = 96,505; 11.1% HIV-positive) vs. 2011–2013 (N = 168,346; 8.8% HIV-positive) vs. 2014–2015
(N = 100,990; 6.9% HIV-positive) among infants testing for HIV in Kenya.
Factor 2007–2010 2011–2013 2014–2015
N % HIV OR 95%CI N % HIV OR 95%CI N % HIV OR 95%CI
+ + +
Gender Male 37,305 10.8% Ref 77,365 8.4% Ref 47,012 6.7% Ref
Female 37,639 11.4% 1.07 1.03– 78,483 9.0% 1.09 1.05– 48,137 6.9% 1.08 1.02–
1.12 1.13 1.14
Age group 0–6 weeks 23,035 11.2% 1.50 1.31– 17,115 9.9% 1.76 1.54– 7,818 7.2% 1.79 1.59–
1.71 2.01 2.03
6 weeks-2 months 23,039 6.8% Ref 63,735 4.6% Ref 51,720 3.1% Ref
2–6 months 24,858 11.1% 1.57 1.42– 39,968 9.2% 1.71 1.60– 22,051 7.9% 2.05 1.88–
1.72 1.82 2.23
6–9 months 11,959 13.6% 2.02 1.82– 17,094 11.1% 1.89 1.75– 6,835 11.0% 2.21 1.98–
2.24 2.04 2.46
9–18 months 11,455 16.6% 2.55 2.25– 25,273 15.3% 2.76 2.58– 10,106 18.5% 4.05 3.64–
2.89 2.96 4.50
18–24 months 982 20.6% 3.37 2.70– 2,632 19.3% 3.60 3.17– 985 28.7% 7.02 5.81–
4.20 4.08 8.49
PMTCT AZT from 14 weeks of pregnancy or 8,559 8.7% 1.06 0.95– 23,819 6.0% 1.15 1.05– 9,016 5.8% 1.22 1.07–
Intervention later; AZT+3TC+sdNVP during labor; 1.18 1.26 1.39
AZT+3TC for 7 days postpartum
HAART 21,979 9.1% Ref 49,017 5.9% Ref 54,170 4.4% Ref
Interrupted HAART (HAART until end 805 10.2% 0.90 0.71– 6,691 6.9% 1.24 1.09– 3,941 6.5% 1.25 1.06–
of breastfeeding) 1.14 1.41 1.48
None 15,254 12.0% 1.44 1.29– 21,448 14.8% 2.06 1.86– 7,135 23.3% 2.40 2.16–
1.61 2.28 2.67
Other 1,042 13.0% 1.24 1.01– 8,428 8.2% 1.19 1.04– 3,241 8.4% 1.46 1.25–
1.52 1.37 1.71
Single dose NVP Only 3,100 10.1% 1.22 1.04– 4,307 9.0% 1.50 1.31– 1,226 6.9% 1.20 0.93–
1.43 1.71 1.56
Infant AZT+3TC for 7 days only 0 - - 25 0.0% - - 103 2.9% 0.47 0.13–
Prophylaxis 1.68
NVP during breastfeeding 20 5.0% 0.59 0.03– 28,385 6.5% 1.26 1.14– 22,100 6.0% 1.26 1.13–
10.16 1.39 1.40
NVP for 6 weeks (Mother on HAART or 13 7.7% Ref 27,478 4.4% Ref 39,938 3.8% Ref
not breastfeeding)
None 35 11.4% 1.23 0.16– 11,731 20.8% 2.57 2.29– 8,123 23.0% 2.30 2.04–
9.34 2.89 2.58
Other 6 16.7% 1.52 0.08– 2,567 10.4% 1.51 1.22– 3,116 6.8% 1.28 1.04–
30.12 1.87 1.57
Single dose NVP+AZT+3TC 36 0.0% - - 2,349 6.8% 1.19 0.97– 1,051 11.2% 2.05 1.55–
1.45 2.71
Single dose NVP Only 19 0.0% - - 13,256 6.4% 1.23 1.10– 8,266 5.0% 1.05 0.92–
1.37 1.20
Breastfeeding Exclusive breastfeeding 13,199 12.4% Ref 83,080 6.5% Ref 70,887 4.7% Ref
Mixed breastfeeding 5,112 9.7% 0.74 0.65– 12,330 14.6% 1.39 1.27– 6,240 21.1% 1.84 1.67–
0.83 1.52 2.03
Breastfeeding, unspecified 18,491 10.3% 0.79 0.71– 9,832 13.0% 1.25 1.13– 4,286 12.2% 1.17 1.03–
0.89 1.37 1.33
No breastfeeding 11,134 9.4% 0.66 0.59– 19,213 9.9% 1.02 0.93– 7,509 9.8% 1.03 0.88–
0.74 1.12 1.20
Entry point Comprehensive clinic care/patient 11,539 12.7% 1.22 1.12– 26,043 8.9% 1.03 0.94– 16,675 6.9% 1.01 0.91–
support center 1.32 1.13 1.11
(Continued)
Table 2. (Continued)
Multivariable logistic regression accounting for clustering by health facility. A separate model was run for each time period. The multivariable model was
adjusted for all factors in this table plus test year. The missing indicator method was used to account for missing data; to save space, missing categories are
not shown. Abbreviations: AZT: zidovudine. 3TC: lamivudine. Sd: single dose. NVP: nevirapine. HAART: highly active antiretroviral therapy. PMTCT:
prevention of mother-to-child transmission of HIV
https://doi.org/10.1371/journal.pone.0183860.t002
decreased over time; in 2007–2010, positivity was 11.1% while in 2014–2015, positivity was
6.9%. Most associations were consistent over time, although some associations varied in their
strength.
The proportion of mothers not on ART during pregnancy (among those with treatment
data) decreased dramatically over time, from 30.1% in 2007–2010 to 9.1% in 2014–2015. The
percentage of infants born to mothers on HAART who were HIV-infected decreased from
9.1% in 2007–2010 to 4.4% in 2014–2015. Infant prophylaxis was not widely available in 2007–
2010 so data was not routinely collected, but between 2011–2013 and 2014–2015 the propor-
tion of infants receiving no prophylaxis decreased from 13.7% to 9.8%.
The higher prevalence of HIV infection among older children compared to younger chil-
dren was stronger after scale-up of Option B+ (adjusted OR comparing children aged 18–24
months to children 6 weeks to 2 months in 2014–2015: 7.02; 95% CI: 5.81–8.49).
Mixed breastfeeding decreased from 10.7% of infants in 2007–2010 (plus 38.6% unspecified
breastfeeding) to 7.0% (plus 4.8% unspecified breastfeeding) in 2014–2015. There may have
been differences in the way that the breastfeeding categories were defined/reported over the
years, as in 2007–2010”mixed breastfeeding” appeared to be protective compared to exclusive
breastfeeding whereas in more recent years it was harmful. Not breastfeeding appeared protec-
tive compared to exclusive breastfeeding in 2007–2010 (OR: 0.66; 95% CI: 0.59–0.74) while
there was no association in 2014–2015. Finally, associations between entry point and HIV
have changed over time, particularly in that the pediatric ward in more recent years had an
extremely high yield for identifying infected patients.
There were no substantial differences in the study findings when patients with age listed as
a negative value or greater than or equal to 2 years were included in the analysis with age listed
as missing.
Discussion
For this study, we analysed a comprehensive national dataset from Kenya covering nearly a
decade to describe the determinants of HIV status in infants below 2 years of age at testing.
We observed greater odds of HIV infection in females, older children, infants whose mothers
received no PMTCT intervention, infants receiving no HIV prophylaxis, and infants who were
mixed breastfed. Exposure to risky practices (mother not on HAART, no infant prophylaxis,
mixed breastfeeding) was associated with a seven-fold higher odds of HIV-positivity compared
to exposure to recommended practices (mother on HAART, infant on nevirapine for six
weeks, no breastfeeding). With PMTCT/EID program growth and more patients in care, there
has been a decrease in risky practices over time, but even the recent data suggests that some
infants are not reached in time for prevention; these findings would likely be more striking in
other countries in sub-Saharan Africa where PMTCT programs are not as strong. Finally, the
pediatric ward and outpatient department had the highest testing yield across entry portals,
but most HIV-infected infants were identified through PMTCT/MCH programs.
Most importantly, many infants were not identified as infected until older ages and received
sub-optimal prevention practices (lack of HAART for mother, infant prophylaxis, exclusive
breastfeeding or exclusive replacement feeding). Option B+ has been scaled up in recent years
and the number of new pediatric infections in Kenya has decreased by 29% from 2009 to 2014
[4]. Yet hundreds to thousands of infants are still becoming infected each year because their
mothers are not enrolled in care. Barriers to testing and care include distance to health facility
and transportation costs, facility inefficiencies, such as stock-outs and long wait times, and per-
sistent shamefulness and stigma[15]. Expanded testing efforts in adolescent and early adult
women and increased HIV self-testing availability may help to expand case-finding in women
of childbearing age.[16] Many pregnant women do not come to health facilities for antenatal
care until late in pregnancy[17] if at all; a high proportion of women deliver outside of facili-
ties. HIV-infected women who deliver outside of facilities tend to have lower income and be
less educated and less likely to be on treatment[18], meaning that their infants are at especially
high risk. Pilot interventions that have been shown to improve PMTCT program coverage,
retention, and quality include mHealth tools[19] such as SMS[20], rapid results initiatives[21],
systems engineering approaches[22], and efforts to reduce health provider absenteeism[23];
these could be considered in areas struggling with program performance. In addition, there
likely remain inconsistencies with health provider care among women and infants enrolled in
PMTCT programs. A case-control study in Western Kenya observed that Infants were more
likely to be infected with HIV if their provider did not follow maternal and infant ART guide-
lines[24]. Ensuring reinforcement of guidelines through periodic re-trainings and supportive
supervision, as well as strong supply chain management systems will be important to continue
to strengthen the PMTCT program.
We observed HIV infections in some infants despite their mothers being on HAART; the
percentage of infected infants in this category declined over time as drug regimens changed to
4.4% in 2014–2015. Possible reasons for transmission in these infants include late treatment
initiation in mothers, treatment nonadherence, lower efficacy on certain treatments, drug
resistance, and stock-outs. This database does not contain information on these factors and
therefore we are only able to speculate on the relative contribution of each one. Some mothers
had interrupted HAART, indicating that non-adherence to treatment may be another barrier
to care and testing. Strong adherence support programs are also important to minimize trans-
mission during pregnancy. Innovative methods including community adherence groups[25]
and transport reimbursement for low-income patients[26] may help to improve drug adher-
ence and retention in PMTCT programs. New treatment regimens for mothers and prophy-
laxis options for infants will likely influence transmission rates in the future.
Thousands of infants are still exposed to the sub-optimal practice of mixed breastfeeding
each year in Kenya, and these infants had worse outcomes compared to exclusively breastfed
infants and infants not given any breastmilk. The dangers posed by mixed feeding have been
well-described in the literature[27–31]. Besides pointing to a gap in program coverage, these
findings illustrate a need for enhanced education of the public that this practice is detrimental.
The Kenyan National Demographic and Health Survey noted that by 2014 exclusive breast-
feeding had increased to 61% in infants less than 6 months, a near doubling from 32% in 2009
[32]. Yet a recent study among women in Nairobi found that fear of stigma and discrimination
around HIV was a persistent barrier to exclusive breastfeeding[33]. In this study, authors con-
cluded that health education and counseling as well as male partner support helped to mitigate
these issues. Other studies have shown that intensive counseling is a viable way to increase
EBF in resource-limited settings[34], although in an intervention study in Nairobi, breastfeed-
ing counseling alone did not increase EBF[35]. Regardless, there is more work to be done to
reduce mixed breastfeeding in infants younger than 6 months.
In this study we found that HIV positivity did not differ significantly between EBF and
exclusive replacement feeding (ERF) after adjustment for other factors, even when adjusted for
age. This was surprising, as exclusive replacement feeding should significantly reduce HIV
transmission. It is possible that due to stigma, or the costs associated with ERF, some of those
who said they provided replacement feeding for their infants also breastfed from time to time,
leading to some measurement error in this category.
Infants who entered care and treatment through outpatient departments or the pediatric
wards were more likely to be HIV-infected than those entering through any other service
point. This finding is not surprising, as those infants come to hospital due to illnesses which
may include, or could be related to, HIV, and supports the latest Kenyan guidelines stating
that all children presenting for care should be offered an HIV test[36]. In addition, testing in
these infants was likely offered in a targeted fashion; for example, symptomatic infants were
more likely to be tested. The high proportion of infected patients suggests that these entry
points may be important for expanded infant testing to identify more HIV-exposed infants
that are not reached by routine testing through PMTCT programs. In the future, EID detection
efforts could be expanded further to better cover patients in alternative entry points such as
OPD and pediatric wards to identify those who are missed by PMTCT testing programs, given
the high yield in these areas. This is in line with current WHO recommendations, which state
that in settings with generalized epidemics, HIV testing should be performed on infants with
unknown HIV status who are admitted to inpatient and nutrition wards and offered to infants
with unknown HIV status at outpatient and immunization clinics[37]. Increased resources for
testing at these portals may be important. However, it is important to note that the greatest
numbers of cases are still detected through MCH/PMTCT programs, despite the lower yields
in these portals, and these programs should continue this important work. Of note, a number
of studies on improving Kenya’s EID program are recently completed or still ongoing[38–41].
Through continued national monitoring efforts in Kenya and scaling up such efforts in other
countries, we can continue to work to minimize transmission of HIV to infants.
Overall, we observed slightly higher odds of positivity in female infants compared to male
infants from birth to six weeks. While this small difference could be due to chance, this phe-
nomenon has been described previously and there are theories that female infants may be
more susceptible to HIV infection in the in-utero and peripartum period [42–44]. While the
exact mechanism is not known, hypotheses include chromosomal affinity and a higher rate of
male intrauterine death among HIV-infected fetuses, leading to females being more likely to
be HIV infected at birth. In this database, females did not appear to be testing later than males,
so age at testing did not seem to explain the imbalance. Confirmation in other studies and fur-
ther investigation is needed to better understand this phenomenon.
This study has a number of limitations. Firstly, it is observational and cross-sectional; thus,
causal relationships cannot be inferred. We did not have access to data on all predictors of
infant HIV transmission and all potential confounders. In addition, the study data come from
a national health registry system and were not collected for the purpose of a research study;
thus, there were large amounts of missing data for some indicators and some inaccuracies are
possible. Along those lines, given the vagueness of the patient identification numbers at some
facilities, we were unable to assess tests from the same individuals or calculate a seroconversion
rate among patients with multiple tests. Strengthening patient identification systems and data
entry would further maximize the utility of this database. This analysis did not include out-
come data on HIV-infected infants following diagnosis; linkage to care and retention in care
for these infants is an important area for future research. Finally, the data is based on infants
who were tested for HIV, which, while covering >70% of exposed infants, does not include the
infants outside of care who are at highest risk of HIV transmission. This also means that com-
parisons between groups cannot be interpreted as nationally representative, do not necessarily
represent transmission, and are in part a product of testing coverage. The strengths of this
study include its large size, numerous risk factors examined, and comprehensiveness across
Kenya over nearly a decade.
Conclusions
This study emphasizes the importance of national health registry systems to monitor the suc-
cess of programs such as Option B+ and to inform testing for EID. Given the relative success
of Kenya’s national program, the associations are likely to be even more striking in other coun-
tries in sub-Saharan Africa with less successful PMTCT programs. We have confirmed risk
factors previously seen on a smaller scale in this national dataset, and illustrated the impact of
exposure to multiple risk factors overall and over time. The findings point to a need for
resources to continue to reinforce best treatment and prophylaxis prescribing behavior, pro-
mote exclusive breastfeeding, and test infants early in life and at testing portals with greater
case-finding. Routine point of care testing will be important to dramatically reduce time to
treatment initiation for HIV-infected infants. Most importantly, efforts must continue to
reach patients that are not coming for PMTCT programs.
Supporting information
S1 Table. Sub-optimal behaviors in relation to HIV positivity in 365,841 infants testing for
HIV in Kenya. Univariate and multivariate logistic regression accounting for clustering by
health facility. The multivariable model was adjusted for sex, age, entry point, and year. The
missing indicator method was used to account for missing data. Abbreviations: NVP: nevira-
pine. HAART: highly active antiretroviral therapy. PMTCT: prevention of mother-to-child
transmission of HIV.
(DOCX)
Acknowledgments
The authors wish to thank the Centers for Disease Control and Prevention (CDC), USAID/
PEPFAR, and UNITAID for funding HIV diagnostics in Kenya, as well as the Clinton Health
Access Initiative for technical support.
Author Contributions
Conceptualization: Matilu Mwau, Martin Sirengo.
Data curation: Matilu Mwau, Priska Bwana, Lucy Kithinji, Francis Ogollah, Samuel Ochieng,
Catherine Akinyi, Maureen Adhiambo, Fred Ogumbo.
Formal analysis: Matilu Mwau, Lucy Kithinji, Caroline Boeke.
References
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. Lancet. 2004; 364
(9441):1236–43. https://doi.org/10.1016/S0140-6736(04)17140-7 PMID: 15464184.
2. UNAIDS/WHO. UNAIDS Report on the Global AIDS Epidemic 2013.
3. Cooper ER, Charurat M, Mofenson L, Hanson IC, Pitt J, Diaz C, et al. Combination antiretroviral strate-
gies for the treatment of pregnant HIV-1-infected women and prevention of perinatal HIV-1 transmis-
sion. J Acquir Immune Defic Syndr. 2002; 29(5):484–94. Epub 2002/05/01. PMID: 11981365.
4. UNAIDS. Progress report on the global plan towards the elimination of new hiv infections among chil-
dren and keeping their mothers alive. 2015.
5. IATT. Interagency Task Team on Prevention and Treatment of HIV Infection in Pregnant Women, Moth-
ers and their Children. http://emtct-iatt.org/b-countries-and-pmtct-regimen/. 2016.
6. Kenya AIDS Response Progress Report. Kenya National AIDS Control Council; 2014.
7. UNICEF Antenatal Care Coverage Database [Internet]. UNICEF (Original Source: DHS 2014). 2016.
8. UNICEF Institutional Deliveries Database [Internet]. UNICEF (Original Source: DHS 2014). 2016.
9. NASCOP. Guidelines for the prevention of mother-to-child transmission (PMTCT) of HIV/AIDS in
Kenya. 2012.
10. NASCOP. The Kenya HIV Testing Services Guidelines. 2015.
11. Tao Z, Rosental N, Kanner BI, Gameiro A, Mwaura J, Grewer C. Mechanism of cation binding to the glu-
tamate transporter EAAC1 probed with mutation of the conserved amino acid residue Thr101. J Biol
Chem. 2010; 285(23):17725–33. https://doi.org/10.1074/jbc.M110.121798 PMID: 20378543; PubMed
Central PMCID: PMCPMC2878536.
12. Kageha S, Okoth V, Kadima S, Vihenda S, Okapesi E, Nyambura E, et al. Discrepant test findings in
early infant diagnosis of HIV in a national reference laboratory in Kenya: challenges and opportunities
for programs. J Trop Pediatr. 2012; 58(4):247–52. https://doi.org/10.1093/tropej/fmr076 PMID:
22052701.
13. Chang J, Omuomo K, Anyango E, Kingwara L, Basiye F, Morwabe A, et al. Field evaluation of Abbott
Real Time HIV-1 Qualitative test for early infant diagnosis using dried blood spots samples in compari-
son to Roche COBAS Ampliprep/COBAS TaqMan HIV-1 Qual test in Kenya. J Virol Methods. 2014;
204:25–30. https://doi.org/10.1016/j.jviromet.2014.03.010 PMID: 24726703; PubMed Central PMCID:
PMCPMC4720144.
14. Lofgren SM, Morrissey AB, Chevallier CC, Malabeja AI, Edmonds S, Amos B, et al. Evaluation of a
dried blood spot HIV-1 RNA program for early infant diagnosis and viral load monitoring at rural and
remote healthcare facilities. AIDS. 2009; 23(18):2459–66. https://doi.org/10.1097/QAD.
0b013e328331f702 PMID: 19741481; PubMed Central PMCID: PMCPMC2890230.
15. Turan JM, Onono M, Steinfeld RL, Shade SB, Owuor K, Washington S, et al. Implementation and Oper-
ational Research: Effects of Antenatal Care and HIV Treatment Integration on Elements of the PMTCT
Cascade: Results From the SHAIP Cluster-Randomized Controlled Trial in Kenya. J Acquir Immune
Defic Syndr. 2015; 69(5):e172–81. https://doi.org/10.1097/QAI.0000000000000678 PMID: 25967269;
PubMed Central PMCID: PMCPMC4501892.
16. Sharma M, Ying R, Tarr G, Barnabas R. Systematic review and meta-analysis of community and facil-
ity-based HIV testing to address linkage to care gaps in sub-Saharan Africa. Nature. 2015; 528(7580):
S77–85. https://doi.org/10.1038/nature16044 PMID: 26633769; PubMed Central PMCID:
PMCPMC4778960.
17. Obermeyer CM, Neuman M, Hardon A, Desclaux A, Wanyenze R, Ky-Zerbo O, et al. Socio-economic
determinants of HIV testing and counselling: a comparative study in four African countries. Trop Med Int
Health. 2013; 18(9):1110–8. https://doi.org/10.1111/tmi.12155 PMID: 23937702; PubMed Central
PMCID: PMCPMC3808878.
18. Kinuthia J, Kohler P, Okanda J, Otieno G, Odhiambo F, John-Stewart G. A community-based assess-
ment of correlates of facility delivery among HIV-infected women in western Kenya. BMC Pregnancy
Childbirth. 2015; 15:46. https://doi.org/10.1186/s12884-015-0467-6 PMID: 25885458; PubMed Central
PMCID: PMCPMC4344995.
19. Mushamiri I, Luo C, Iiams-Hauser C, Ben Amor Y. Evaluation of the impact of a mobile health system
on adherence to antenatal and postnatal care and prevention of mother-to-child transmission of HIV
programs in Kenya. BMC Public Health. 2015; 15:102. https://doi.org/10.1186/s12889-015-1358-5
PMID: 25886279; PubMed Central PMCID: PMCPMC4328364.
20. Jennings L, Ong’ech J, Simiyu R, Sirengo M, Kassaye S. Exploring the use of mobile phone technology
for the enhancement of the prevention of mother-to-child transmission of HIV program in Nyanza,
Kenya: a qualitative study. BMC Public Health. 2013; 13:1131. https://doi.org/10.1186/1471-2458-13-
1131 PMID: 24308409; PubMed Central PMCID: PMCPMC4234194.
21. Dillabaugh LL, Lewis Kulzer J, Owuor K, Ndege V, Oyanga A, Ngugi E, et al. Towards Elimination of
Mother-to-Child Transmission of HIV: The Impact of a Rapid Results Initiative in Nyanza Province,
Kenya. AIDS Res Treat. 2012; 2012:602120. https://doi.org/10.1155/2012/602120 PMID: 22548155;
PubMed Central PMCID: PMCPMC3324880.
22. Rustagi AS, Gimbel S, Nduati R, Cuembelo MF, Wasserheit JN, Farquhar C, et al. Impact of a systems
engineering intervention on PMTCT service delivery in Cote d’Ivoire, Kenya, Mozambique: a cluster
randomized trial. J Acquir Immune Defic Syndr. 2016. https://doi.org/10.1097/QAI.0000000000001023
PMID: 27082507.
23. Goldstein M, Zivin JG, Habyarimana J, Pop-Eleches C, Thirumurthy H. The Effect of Absenteeism and
Clinic Protocol on Health Outcomes: The Case of Mother-to-Child Transmission of HIV in Kenya. Am
Econ J Appl Econ. 2013; 5(2):58–85. https://doi.org/10.1257/app.5.2.58 PMID: 24163722; PubMed
Central PMCID: PMCPMC3806719.
24. Onono M, Owuor K, Turan J, Bukusi EA, Gray GE, Cohen CR. The role of maternal, health system, and
psychosocial factors in prevention of mother-to-child transmission failure in the era of programmatic
scale up in western Kenya: a case control study. AIDS Patient Care STDS. 2015; 29(4):204–11. https://
doi.org/10.1089/apc.2014.0181 PMID: 25738870; PubMed Central PMCID: PMCPMC4378858.
25. Luque-Fernandez MA, Van Cutsem G, Goemaere E, Hilderbrand K, Schomaker M, Mantangana N,
et al. Effectiveness of patient adherence groups as a model of care for stable patients on antiretroviral
therapy in Khayelitsha, Cape Town, South Africa. PLoS One. 2013; 8(2):e56088. https://doi.org/10.
1371/journal.pone.0056088 PMID: 23418518; PubMed Central PMCID: PMCPMC3571960.
26. Lowther K, Harding R, Ahmed A, Gikaara N, Ali Z, Kariuki H, et al. Conducting experimental research in
marginalised populations: clinical and methodological implications from a mixed-methods randomised
controlled trial in Kenya. AIDS Care. 2016; 28 Suppl 1:60–3. https://doi.org/10.1080/09540121.2016.
1146214 PMID: 26916738; PubMed Central PMCID: PMCPMC4828598.
27. Fowler MG. Further evidence that exclusive breast-feeding reduces mother-to-child HIV transmission
compared with mixed feeding. PLoS Med. 2008; 5(3):e63. https://doi.org/10.1371/journal.pmed.
0050063 PMID: 18336065; PubMed Central PMCID: PMCPMC2265760.
28. Becquet R, Ekouevi DK, Menan H, Amani-Bosse C, Bequet L, Viho I, et al. Early mixed feeding and
breastfeeding beyond 6 months increase the risk of postnatal HIV transmission: ANRS 1201/1202
Ditrame Plus, Abidjan, Cote d’Ivoire. Prev Med. 2008; 47(1):27–33. https://doi.org/10.1016/j.ypmed.
2007.11.014 PMID: 18190955.
29. Piwoz EG, Humphrey J. Increased risk of infant HIV infection with early mixed feeding. AIDS. 2005; 19
(15):1719–20; author reply 20–1. PMID: 16184057.
30. Wise J. Breast feeding safer than mixed feeding for babies of HIV mothers. BMJ. 2001; 322(7285):511.
PMID: 11230060; PubMed Central PMCID: PMCPMC1119730.
31. Mixed feeding may be the most dangerous option for babies of HIV positive mothers in africa. BMJ.
2000; 320(7250):G. PMID: 10856091.
32. KNBS. Kenya Demographic and Health Survey 2014. Kenya National Bureau of Statistics. 2014.
33. Odeny BM, Pfeiffer J, Farquhar C, Igonya EK, Gatuguta A, Kagwaini F, et al. The Stigma of Exclusive
Breastfeeding Among Both HIV-Positive and HIV-Negative Women in Nairobi, Kenya. Breastfeed Med.
2016. https://doi.org/10.1089/bfm.2016.0014 PMID: 27093583.
34. Okanda JO, Borkowf CB, Girde S, Thomas TK, Lecher SL. Exclusive breastfeeding among women tak-
ing HAART for PMTCT of HIV-1 in the Kisumu Breastfeeding Study. BMC Pediatr. 2014; 14:280.
https://doi.org/10.1186/1471-2431-14-280 PMID: 25380718; PubMed Central PMCID:
PMCPMC4326202.
35. Bosire R, Betz B, Aluisio A, Hughes JP, Nduati R, Kiarie J, et al. High Rates of Exclusive Breastfeeding
in Both Arms of a Peer Counseling Study Promoting EBF Among HIV-Infected Kenyan Women. Breast-
feed Med. 2016; 11:56–63. https://doi.org/10.1089/bfm.2015.0071 PMID: 26885769; PubMed Central
PMCID: PMCPMC4782034.
36. The Kenya HIV Testing Services Guidelines. Kenya National AIDS and STI Control Programme (NAS-
COP), 2015.
37. Consolidated guidelines on the use of antiretroviral drugs for treating and preventing HIV infection: Rec-
ommendations for a public health approach. Geneva: World Health Organization (WHO), 2016.
38. Goggin K, Wexler C, Nazir N, Staggs VS, Gautney B, Okoth V, et al. Predictors of Infant Age at Enroll-
ment in Early Infant Diagnosis Services in Kenya. AIDS Behav. 2016. https://doi.org/10.1007/s10461-
016-1404-z PMID: 27108002.
39. Ambia J, Mandala J. A systematic review of interventions to improve prevention of mother-to-child HIV
transmission service delivery and promote retention. J Int AIDS Soc. 2016; 19(1):20309. https://doi.org/
10.7448/IAS.19.1.20309 PMID: 27056361; PubMed Central PMCID: PMCPMC4824870.
40. Finocchario-Kessler S, Gautney BJ, Khamadi S, Okoth V, Goggin K, Spinler JK, et al. If you text them,
they will come: using the HIV infant tracking system to improve early infant diagnosis quality and reten-
tion in Kenya. AIDS. 2014; 28 Suppl 3:S313–21. https://doi.org/10.1097/QAD.0000000000000332
PMID: 24991904; PubMed Central PMCID: PMCPMC4226133.
41. Hassan AS, Sakwa EM, Nabwera HM, Taegtmeyer MM, Kimutai RM, Sanders EJ, et al. Dynamics and
constraints of early infant diagnosis of HIV infection in Rural Kenya. AIDS Behav. 2012; 16(1):5–12.
https://doi.org/10.1007/s10461-010-9877-7 PMID: 21213034; PubMed Central PMCID:
PMCPMC3254874.
42. Brahmbhatta H, Kigozi G, Serwadda D, Wabwire-Mangen, Nelson Sewankamboe F, Wawera M, et al.
Is the risk of mother-to-child transmission of HIV higher among female compared with male infants? A
case of Rakai, Uganda. J Pediatr Infect Dis. 2009; 04(03):275–9. https://doi.org/10.3233/JPI-2009-
0168
43. Biggar RJ, Taha TE, Hoover DR, Yellin F, Kumwenda N, Broadhead R. Higher in utero and perinatal
HIV infection risk in girls than boys. J Acquir Immune Defic Syndr. 2006; 41(4):509–13. https://doi.org/
10.1097/01.qai.0000191283.85578.46 PMID: 16652061.
44. Galli L, Puliti D, Chiappini E, Gabiano C, Tovo PA, Pezzotti P, et al. Lower mother-to-child HIV-1 trans-
mission in boys is independent of type of delivery and antiretroviral prophylaxis: the Italian Register for
HIV Infection in Children. J Acquir Immune Defic Syndr. 2005; 40(4):479–85. PMID: 16280705.