High-Risk Enteric Pathogens Associated with HIV-Infection and HIV-Exposure in Kenyan Children with Acute Diarrhea

Abstract

Objective

HIV-infection is an established risk for diarrheal severity, less is known about specific enteric pathogens associated with HIV status. We determined associations of selected enteric pathogens with HIV-infection and HIV-exposure among Kenyan children.

Design

Cross-sectional study among 6 months to 15 year olds presenting to two Western Kenya District hospitals with acute diarrhea between 2011–2013.

Methods

Stool was tested using standard bacterial culture and microscopy for ova and parasites. HIV testing was obtained on children and mothers. Enteric pathogen prevalence was compared between HIV-infected and HIV-uninfected children and between HIV-exposed uninfected (HEU) and HIV-unexposed. Unadjusted and adjusted prevalence ratios (PR) for selected pathogens by HIV-status were estimated using relative risk (RR) regression and P-values. Age, site, income, household crowding, water source/treatment, anthropometrics, cotrimoxazole use, and breastfeeding history were accounted for in multivariable models.

Results

Among 1,076 children, median age was 22 months (interquartile range: 11–42), 56 (5.2%) were HIV-infected, and 10.3%(105/1020) of HIV-uninfected children were HIV-exposed. The following organisms were most frequently isolated from stool: enteroaggregative Escherichia coli (13.3%), Giardia spp. (11.1%) Campylobacter (6.3%), enteropathogenic Escherichia coli (EPEC) (6.1%) and Cryptosporidium spp. (3.7%). Accounting for age, HIV-infection was associated with EPEC infection (PR: 3.70, P=0.002) while HIV-exposure was associated with Cryptosporidium among HIV-uninfected children (PR: 2.81, P=0.005).

Conclusion

EPEC and Cryptosporidium infections were more common in HIV-infected and HIV-exposed children, respectively. This could explain the increased mortality attributed to these pathogens in other studies. Interventions targeting EPEC and Cryptosporidium may reduce morbidity and mortality in high HIV-prevalence settings.

INTRODUCTION

Diarrheal disease remains a leading cause of death in children under 5 years of age and in resource-limited settings, most diarrhea is attributed to enteric pathogens [1, 2]. In addition to the acute morbidity and mortality attributable to diarrhea, the enteric mucosal damage that occurs in diarrhea leads to decreased nutrient absorptive capacity, growth failure, and cognitive delay [3–7]. HIV-infected children experience more frequent and severe diarrhea episodes and are at higher risk of malnutrition and cognitive impairment than their uninfected counterparts [8–12]. As prevention of mother-to-child HIV transmission (PMTCT) programs expand, pediatric HIV infections are declining, however, there is a growing population of HIV-exposed uninfected (HEU) children [13]. HEU children experience greater risk of death, hospitalization and neurodevelopmental delays compared to HIV-unexposed children [14–17]. The increased morbidity and mortality observed among HEU children may be due to more frequent enteric infections, earlier weaning, reduced breast milk exposure, decreased immunologic development during infancy, poor socioeconomic status, or diminished parental caretaking capacity [12, 18, 19].

Guidelines for syndromic management of diarrheal illness in low-resource settings do not differentiate management strategies by HIV-status [20, 21]. If HIV-infected or HEU children are more likely to be infected with pathogens independently associated with poor growth and death, targeted diagnostic stool testing and/or empiric antibiotic/anti-protozoal therapy for these high risk groups may be helpful in diminishing mortality, morbidity, and transmission. We determined the prevalence of enteric pathogens among HIV-infected, HEU, and HIV-unexposed children presenting with acute diarrhea.

METHODS

Population

Between November 2011 through October 2013 children aged 6 months to 15 years presenting to Kisii Provincial or Homa Bay District Hospital with acute diarrhea (defined as ≥3 loose stools within 24 hours lasting less than 14 continuous days[22]) were enrolled in an ongoing diarrhea surveillance study. Children were excluded if they were unaccompanied by a biological parent or legal guardian, unable to provide a stool sample or rectal swab, or if the primary caregiver elected not to receive HIV counseling on behalf of the child. Study participants were recruited from both outpatient and inpatient settings. Written informed consent was obtained from primary caregivers of enrolled children and assent was obtained from children over 12 years. The University of Washington Institutional Review Board and the Kenya Medical Research Institute Ethical Review Committee approved the current study.

Data collection

Stool was collected, examined for consistency and appearance, and separated into two containers for shipment. When children could not produce stool, 3 rectal swabs were collected. Sociodemographic characteristics, possible exposures (recent antibiotic use [including cotrimoxazole (CTX)], travel history, water source and filtration, sanitation), breastfeeding and vaccination history were obtained from the primary caregiver. Study physicians measured height and weight, and assessed danger and dehydration signs according to the WHO Integrated Management of Childhood Illness (IMCI) algorithm Height for age and weight for height z-scores (HAZ & WHZ) were calculated using the 2006 and 2007 WHO reference populations for children under 5 and 5 or over, respectively and stunting and wasting defined as HAZ <−2 and WHZ <−2, respectively [23, 24]. A child was classified as having a severe illness if one or more IMCI danger signs (unable to drink or breastfeed, convulsions, continuous vomiting, and/or lethargy/unconsciousness) were identified[20]. Children were classified as having MSD if they had sunken eyes, loss of skin turgor, visible blood in stool, or required intravenous hydration or hospital admission based on diarrhea or dysentery[2].

Children were tested for HIV using antibody testing (Abbott Determine^™ rapid test kit and confirmed using Uni-Gold^™) or HIV DNA polymerase chain reaction (PCR) assays if <18 months. Malaria parasitemia was assessed by both rapid testing (Paracheck Pf® Orchid Biomedical Services, India) and microscopy. Maternal HIV-status was ascertained by antibody testing or by self-report if the mother reported being HIV-infected.

Stool Specimen Processing

All stool/rectal swab specimens were shipped to the United States Army Research Unit Microbiology Hub in Kericho, Kenya, within 24–48 hours of collection. Selective media was used for bacterial identification-BAP (blood agar plate) for hemolysis and oxidase test, Sorbitol-MacConkey agar to select for non-sorbitol fermenting (NSF) Escherichia coli (E. coli), MacConkey agar for E. coli colonies, Hektoen or xylose lysine deoxycholate agar for Salmonella and Shigella spp., cefsulodin irgasan novobiocin agar for Yersinia spp., andcefoperazone vancomycin amphotericin (CVA) agar for Campylobacter spp. Pure colonies exhibiting the proper characteristics on the various media above were further processed using MicroScan WalkAway 40 Plus automated platform.

NSF E. coli isolates were batch tested using multiplex PCR to identify virulent forms of E. coli: enterotoxigenic E. coli (ETEC)- heat labile enterotoxin (elt) and/or heat stable enterotoxin (est), enteroaggregative E. coli (EAEC)- aatA; enteroinvasive E. coli (EIEC)- invasion plasmid antigen H (ipaH); or enterohemmorhagic E. coli (EHEC)- Shiga toxin 1, 2 and variants (stx); and enteropathogenic E. coli (EPEC)- bundle forming pilus (bfpA) [25]. Starting in March 2013, additional gene targets for EPEC- intimin (eae) and for EAEC- aaiC were incorporated. A classification of EAEC was therefore the identification of aatA and/or aaiC. EPEC was disaggregated into two categories: typical EPEC (bfpA with or without eae) and atypical EPEC (eae without either bfpA or stx).

For parasitic identification, the stool was concentrated using the Mini Parasep® Solvent Free concentration kit (DiaSys, Berkshire, England) and then microscopy used to identify parasitic forms. Parasite testing was not performed on rectal swabs.

Statistical Analysis

Two sets of analyses were performed; the first compared risk factors and enteric pathogens between HIV-infected and HIV-uninfected children, and the second compared risk factors and pathogens between HEU and HIV-unexposed children. Children who were missing HIV-status information were excluded and children without maternal HIV status available were excluded from the second analysis.

Sociodemographic and clinical characteristics of included children were compared using Chi-square or Fisher’s exact tests for categorical variables and t-tests for continuous variables. Enteric pathogens were compared using prevalence ratios (PR) and associated P-values estimated using relative risk (RR) regression and associated chi-square or Fisher’s exact tests. For the comparison of enteric pathogens, we adjusted for multiple comparisons using the Benjamini and Hochberg method using a false discovery proportion of 0.05 [26]. For all pathogens associated with HIV-infection and/or HIV-exposure in the univariate analyses, we constructed multivariable models to account for confounding and mediating variables. Because of strong evidence of an age association with specific enteric infections, age was a priori retained as a confounding variable in all multivariable models [2]. Potential confounders, including site, year of enrollment (2011, 2012, 2013), household income, number persons/room, and drinking water source and treatment were included stepwise in multivariable models and maintained in the final model if they changed the PR for the main exposure variable of interest (HIV or HEU) by more than 10%. Variables that we considered to plausibly be on the causal pathway between HIV and enteric pathogens included breastfeeding history (exclusive breastfeeding duration, current breastfeeding), nutritional status indicators (HAZ and WHZ) and recent CTX use [27–30]. Potential mediators were individually added to age-adjusted models. Subgroup analyses were performed among children <24 months and among children with MSD. Confounder, mediator, and subgroup analyses were considered exploratory analyses and therefore an alpha of 0.05 was used to determine statistical significance.

RESULTS

Among 1,512 children presenting with diarrhea, 1,116 met inclusion criteria, 1,099 were enrolled and 1,076 had known HIV-status and therefore were included in analysis I (HIV-infected 5.2% [N=56] and HIV-uninfected 94.8% [N=1,020]) (Figure 1 & Table 1). Among the 56 HIV-infected children, almost half (44.6%) were newly diagnosed as part of this study and among those already diagnosed, 27 (87.1%) were reportedly enrolled in HIV care and 7 (22.6%) reported current ART use. Among the 53 HIV-infected children for whom CD4% or CD4 count was available, 19 (35.9%) were immunosuppressed based on WHO age-specific CD4 cut-offs [31]. Maternal HIV status was ascertained for 926 (90.1%) of the 1,020 HIV-uninfected children, of which 105 (11.3%) and 821 (88.7%) were HEU and HIV-unexposed, respectively. These 926 children were included in analysis II (Figure 1). Only 36 (34.3%) HIV-infected mothers self-reported their last CD4 count (median was 483 cells per cells/mm³ (IQR: 301–834)).

Figure 1.

Participant Inclusion Flow for Analyses I and II

Table 1.

Characteristics of children in Analysis I

Characteristic	Enrolled N=1,076
	n (%a) Median (IQR)
Sociodemographic
Site
Kisii	513 (47.7%)
Homa Bay	563 (52.3%)
≥1 hour to get to hospital	177(16.5%)
Child accompanied by
Biological mother	998 (92.8%)
Biological father	53 (4.9%)
Legal guardian	25 (2.3%)
Monthly household income <5,000 Kenyan Shillings	429(40.0%)
Household owns ≥ cow	475(44.3%)
# persons/room	2(1–3)
Water source
Piped in house or yard or public tap	391 (36.5%)
Protected well/spring	456 (42.5%)
Unprotected well/spring/surface water	149 (13.9%)
Other b	76 (7.1%)
Household treats drinking water	795(73.9%)
Flush toilet	75(7.0%)
Male	576(53.5%)
Age
6m–2yr	585 (54.4%)
>2yr–5yr	367 (34.1%)
>5yr	124(11.5%)
Clinical Presentation
1 or more IMCI General Danger Sign c	335(31.3%)
Moderate to severe diarrhea d	315(29.3%)
Blood observed in stool	14(1.3%)
Mucous in stool	509(47.3%)
Stunted (HAZ<−2)	171(16.9%)
Wastede (WHZ<−2)	202(21.4%)
MUAC <12.5	86(8.0%)
Malariaf	106(9.9%)
Current fever (≥37.5°C)	362(33.6%)
Clinical History
Child ever breastfed	1086(98.9%)
Child currently breast-feedingamong children less than 2 years	452(77.4%)
# months exclusively breastfed	6.0(4–6)
HIV positive	56(5.2%)
New diagnosis	25
Known status	31
Immunosuppressed g	19(35.9%)
HIV-exposed uninfected h	105(11.3%)

^a% among those with non-missing data

^bTube well or borehole (n=18), rainwater (n=55), cart with small tank (n=1), bottled water (n=2)

^cDefined as not able to drink or breastfeed, convulsions, vomits everything, lethargic or unconscious [22]

^dOne or more of the following; sunken eyes, loss of skin turgor, intravenous hydration administered or prescribed, visible blood in stool, or hospital admission based on diarrhea or dysentery [2]

^eOnly calculated for children 5 years and younger

^fPositive result on microscopy alone (n=8), on RDT alone (n=5) or both RDT and microscopy (n=93)

^gDefined in terms of CD4% (age ≤11 months: <25%, 12 months–35 months: <20%, 36⁺ months: <15%) or, in absence of CD4 % data, in terms of CD4 count (age ≤11 months: <1500 cells/mm³, 12 months–35 months: <750 cells/mm³, 36⁺ months <350 cells/mm³)

^hAmong 926 HIV-uninfected children who were accompanied by the biological mother who was HIV-infected

Bacteria and Parasite Frequency

At least one potential pathogen (bacterial or parasitic) was identified in 45.8% (493/1,076) of the children. Nearly 10% (105/1,076) had 2 organisms identified, 1.4% (15/1076) with 3, <0.5% (2/1076) with 4, and 1 had 5 distinct isolates (Supplementary Figure 1). There were no differences in frequency of bacteria isolation between the 981 children who provided whole stool samples and the 95 who provided rectal swab only (35.1% vs. 31.6%, P=0.50).

One or more diarrheagenic E. colis were identified in the stools of 262/1,076 (24.4%) children. EAEC was identified in 143 (13.3%) of the 1,076 children, EPEC in 66 (6.1%) (4.0% were typical and 2.1% atypical), ETEC in 47 (4.4%), EIEC in 32 (3.0%), and EHEC in 4 (0.4%) stools. Among the 262 children with any identified diarrheagenic E. coli, infection with multiple E. coli serotypes was common (11.1%) (Supplementary Figure 1). Other commonly isolated bacteria included Campylobacter spp. (68 [6.3%]) and Shigella spp. (49 [4.6%]).

Almost a quarter (24.2%) of the 981 children that provided a whole stool sample had at least one parasite identified. The most frequently identified parasites were Giardia spp. (109 children [11.1%]) and Cryptosporidium spp. (36 [3.7%]). Other likely non-pathogenic parasites were also identified, including Blastocystis hominis (73 [7.4%]), Chilomastix spp. (4 [0.40%]) and Endolimax spp. (1 [0.10%]).

HIV-infected vs. HIV-uninfected Children

Compared to HIV-uninfected children, HIV-infected children were older (mean age 50 vs. 31 months, p<0.001), more likely to be enrolled at the Homa Bay site (67.9% vs. 51.5%, P=0.017), more likely to come from low-income households (52.7% vs. 39.3%, P=0.048), and less likely to be accompanied by their biologic mother (85.7% vs. 93.1%, P=0.037) (Table 2a). The mean reported number of months of exclusive breastfeeding was similar between HIV-infected and. HIV-uninfected children (4.8 vs. 5.0 months, P=0.33). However, among the children under 24 months old, HIV-infected children were less likely to be breastfeeding at the time of enrollment (50% vs. 80.5%, P=0.002) even after accounting for age (P=0.018), more likely to have taken CTX within the preceding week (23.2% vs. 5.4%, P <0.001), more likely to be stunted (34.6% vs. 15.8%, P=0.001), and more likely to have MSD (41.1% vs. 28.6%, P=0.046).

Table 2a.

Demographic and clinical differences between HIV-infected and -uninfected children

	HIV-Infected N=56	HIV-Uninfected N=1020
Selected Factors	N/mean (%/SD)	N/mean (%/SD)	P-value
Homa Bay Site	38 (67.9%)	525 (51.5%)	0.017
Age in months	49.7 (40.4)	31.3 (31.1)	<0.001
Income <5,000 KSH	29 (52.7%)	400 (39.3%)	0.048
# persons/room in house	2.4 (1.2)	2.4 (1.4)	0.73
Unprotected water source a	16 (29.1%)	133 (13.1%)	0.001
Household treats drinking water	46 (82.1%)	749 (73.4%)	0.15
Accompanied by biological mother	48 (85.7%)	950 (93.1%)	0.037
Child <24 months and currently breast-feeding b	9 (50%)	437 (80.5%)	0.002
# months exclusively breastfed c	4.8 (1.9)	5.0 (1.8)	0.329
Child took cotrimoxazole in last 7 days	13 (23.2%)	55 (5.4%)	<0.001
Stunted (HAZ<−2)	18 (34.6%)	153 (15.9%)	<0.001
Wasted (WHZ<−2)	12 (29.3%)	190 (21.1%)	0.212
Moderate to Severe Diarrhea	23 (41.1%)	292 (28.6%)	0.046
Rectal swab taken	8 (14.3%)	87 (8.5%)	0.140
Blood in stool	1 (1.8%)	13 (1.3%)	0.529

^aUnprotected well, unprotected spring, or surface water

^bAmong 18 HIV-infected children and 544 HIV-uninfected children

^cWhen considering only the subset of children < 24 months of age: mean # of months exclusively breastfed were 5.2 (SD: 1.8) vs. 5.1 (SD: 1.8), p=0.861.

The prevalence of enteric bacteria and parasites by HIV-infection status are reported in Table 2b. In univariate analysis, HIV-infected children were nearly 3 times more likely to have typical EPEC identified in their stools compared to HIV-uninfected children (PR=2.95; P=0.008); this association persisted after adjusting for age (adjusted [a]PR: 3.70 [95%CI: 1.6–8.4, P=0.002]) and no additional measured confounders were identified. HIV-status remained associated with typical EPEC in analyses further adjusted for duration (months) of exclusive breastfeeding (aPR: 3.81 [1.68–8.66, P=0.001]), current breastfeeding (aPR: 3.46 [1.51–7.90, P=0.003]), WHZ (aPR: 2.9 [95%CI: 1.1–7.7, P=0.036]), HAZ (aPR: 3.9 [95%CI; 1.7–8.9, P=0.001]), and CTX use (aPR: 3.5 [95%CI: 1.5–8.0, P=0.004]). Among HIV-infected children in whom CD4 data were available (n=53), typical EPEC was identified more commonly among the immunosuppressed than non-immunosuppressed children (15.8% vs. 8.8%) but the difference was not significant (P=0.655).

Table 2b.

Enteric pathogen differences between HIV-infected and -uninfected children

	HIV-Infected N=56	HIV-Uninfected N=1020	Prevalence Ratio	P-value
Organism Identified	N %	N %
Bacteria
Campylobacter speciesa	6 (10.7%)	62 (6.1%)	1.76	0.165
>1 E. coli serotypeb	2 (11.1%)	27 (11.1%)	1.00	1.00
EAEC	6 (10.7%)	137 (13.2%)	0.81	0.560
EIEC	1 (1.8%)	31 (3.0%)	0.59	1.00
EHEC	0 (0%)	4 (0.4%)	--	1.00
EPEC-atypicalc	2 (3.6%)	21 (2.1%)	2.74	0.178
EPEC-typical	6 (10.7%)	37 (3.6%)	2.95	0.008
ETEC	5 (8.9%)	42 (4.1%)	2.17	0.092
Salmonella speciesd	1 (1.8%)	13 (1.3%)	1.40	0.529
Shigella speciese	1 (1.8%)	46 (4.5%)	0.40	0.508
Other bacteriaf	1 (1.8%)	22 (2.2%)	0.83	1.00
Parasitesg
Giardia species	5 (10.4%)	104 (11.2%)	0.93	0.875
Cryptosporidium species	1 (2.1%)	35 (3.8%)	0.56	1.00
Entaeombea species	0 (0%)	4 (2.6%)	--	1.00
Ascaris lumbricoides	1 (2.1%)	23 (2.5%)	0.85	1.00
Blastocystis hominis	4 (8.3%)	69 (7.4%)	1.13	0.809
Other parasiteh	1 (2.1%)	5 (0.5%)	3.89	0.261
No organism identifiedi	19 (39.6%)	456 (47.4%)	0.84	0.209

^aHIV-infected: Campylobacter jejuni (n=6), HIV-uninfected: Campylobacter jejuni (n=39), Campylobacter spp. other than jejuni (n=23)

^bAmong children with at least one diarrheagenic E. coli serovar (18 HIV-infected and 244 HIV-uninfected)

^cAmong 417 patients enrolled since March 2013 who had the gene eae tested

^dHIV-infected: non-typhoidal species (n=1), HIV-uninfected: Salmonella typhi (n=3), Salmonella paratyphi (n=1), Salmonella non-typhoidal species (n=9)

^eHIV-infected: species not determined (n=1); HIV-uninfected: Shigella flexneri (n=15), Shigella sonnei (n=18), Shigella dysentariae (n=2), species not determined (n=11)

^fHIV-infected: Plesiomonas shigelloides (n=1); HIV-uninfected: Providencia alcalifaciens(n=6), Providencia stuartii (n=1), Providencia rettgeri (n=1) Citrobacter freundii (n=3), Citrobacter amalonaticus (n=1), Enterobacter agglomerans (n=2), Enterobacter cloacae (n=1), Kluyvera ascorbata (n=2), Escherichia vulneris (n=1), Yersinia enterocolitica (n=1), Aeromonas hydrophila (n=1), Pseudomonas aeruginosa (n=1), Edwardsiella tarda (n=1)

^gAmong 981 children in whom whole stool was collected (HIV-infected: n=48, HIV-uninfected: n=933)

^hHIV-infected: Chilomastix mesnili (n=1); HIV-uninfected: Chilomastix mesnili (n=3), Cystoisospora belli (n=1), Endolimax nana (n=1)

ⁱAmong the 981 children who had both bacteria and parasite testing performed (HIV-infected: n=48, HIV-uninfected: n=933)

Subgroup analyses performed among the 562 children aged <24 months (18 (3.2%) HIV-infected and 544 (96.8%) HIV-uninfected) demonstrated no significant differences in the prevalence of typical EPEC between HIV-infected and HIV-uninfected children (11.1% vs. 5.0%, PR: 2.2 [95%CI: 0.6–8.7, P=0.25], age-adjusted PR: 2.6 [95%CI: 0.7–10.2, P=0.17]). In subgroup analyses among 315 children with MSD (23 [7.3%] HIV-infected and 292 [92.7%] HIV-uninfected) HIV-infected children were much more likely to have typical EPEC identified compared to HIV-uninfected children (21.7% vs. 4.1%, PR: 5.3 [95%CI: 2.0–13.7, P<0.001], age-adjusted PR: 6.1 [95%CI: 2.28–16.1, P<0.001]). No other differences in prevalence of other pathogens were observed in these two subgroups.

HEU vs. HIV-Unexposed Children

Compared to HIV-unexposed children (n=821), HEU children (n=105) were more likely to present in Homa Bay than Kisii (87.6% vs. 48.5%, P <0.001) (Table 3a), to live in households with lower incomes (64.8% vs. 36.8%, P<0.001) and were more likely to report having an unprotected water source (28.6% vs. 11.1%, P<0.001) but more likely to treat their drinking water (87.6% vs. 71.8%, P<0.001). HEU and HIV-unexposed children had similar mean number of exclusive breastfeeding months (4.8 vs. 5.0, p=0.329) but among the children under 24 months, HEU children were less likely to be currently breastfeeding (50% vs. 80.5%, p=0.002) despite no difference in mean age among those under 24 months (mean ages in HEU vs HIV-unexposed: 12.0 months vs. 11.1 months, P=0.14). Finally, HEU children were more likely to report having taken CTX in the preceding week (18.3% vs. 4.0%, P<0.001) and were more likely to be stunted (24.5% vs. 15.2%, P=0.016).

Table 3a.

Demographic and clinical differences between HEU and HIV-unexposed children

	HEU N=105	HIV-unexposed N=821
Characteristic	N/Mean (%/SD)	N/mean (%/SD)	P-value
Homa Bay Site	92 (87.6%)	398 (48.5%)	<0.001
Age in months	25.2 (20.1)	30.3 (29.6)	0.092
Income <5,000 KSH	68 (64.8%)	301 (36.8%)	<0.001
# persons/room in house	4.9 (1.6)	4.8 (1.7)	0.6305
Unprotected water source	30 (28.6%)	91 (11.1%)	<0.001
Household treats drinking water	92 (87.6%)	589 (71.8%)	<0.001
Child <24 months and currently breast-feeding a	32 (53.3%)	380 (85.8%)	<0.001
# months exclusively breastfed b	5.4 (1.6)	5.0 (1.9)	0.096
Child took cotrimoxazole in last 7 days	19 (18.1%)	33 (4.0%)	<0.001
Stunted (HAZ<−2)	25 (24.5%)	117 (15.2%)	0.016
Wasted (WHZ<−2)	26 (26.5%)	148 (20.2%)	0.145
Moderate to Severe Diarrhea	35 (33.3%)	224 (27.3%)	0.193
Rectal swab taken	14 (13.3%)	67 (8.2%)	0.077
Blood in stool	0 (0%)	10 (1.2%)	0.614

^aAmong 60 HEU and 443 HIV-unexposed children

^bWhen considering only the subset of children < 24 months of age the mean # of months exclusively breastfed were 5.3 (SD: 1.5) vs. 5.1 (SD: 1.9), p=0.235

The prevalence of bacterial and parasitic infections by HIV-exposure category are presented in Table 3b. Cryptosporidium spp. was more common in HEU children (PR: 3.0, P=0.003). The association between HIV-exposure and Cryptosporidium spp. was independent of age-alone (aPR: 2.81 [95%CI: 1.36–5.80, P=0.005]) and age and site, the only identified confounder (aPR: 2.09 [95%CI: 1.01–4.50, P=0.046]). When we accounted for duration (months) of exclusive breastfeeding and current breastfeeding in age-adjusted models, the association remained significant (aPR: 2.70 [95%CI: 1.31–5.58, P=0.007]; aPR: 2.80 [95%CI: 1.32–5.92, P=0.007]). Similarly, the association between maternal HIV-status and Cryptosporidium spp. infection persisted after including WHZ, HAZ, and CTX use in age-adjusted models (WHZ-aPR: 2.8 [95%CI: 1.4–5.7, P=0.005]; HAZ-aPR: 2.7 [95%CI 1.3–5.6, P=0.008]; CTX-aPR 2.5 [95%CI: 1.2–5.4, P =0.020], respectively). Only 36 of the 105 HIV-infected mothers of HIV-uninfected children knew their most recent CD4 count. Although, the mean CD4 count among the mothers of HEU children with Cryptosporidium spp. appeared to be lower (378.5 cells/mm³) than the maternal CD4 count among those without Cryptosporidium spp. identified (647.9 cells/mm³), this difference was not statistically significant (P=0.300).

Table 3b.

Enteric pathogen differences between HEU and HIV-unexposed children

	HEU N=105	HIV-unexposed N=821	Prevalence Ratio	P-value
Organism Identified	N (%)	N (%)
Bacteria
Campylobacter species	8 (7.6%)	47 (5.7%)	1.33	0.439
>1 E. coli serotypea	2 (8.0%)	21 (10.7%)	0.75	1.00
EAEC	14 (13.3%)	111 (13.5%)	0.99	0.958
EIEC	4 (3.8%)	24 (2.9%)	1.3	0.548
EHEC	0 (0%)	2 (0.24%)	--	1.00
EPEC-atypicalb	4 (3.8%)	15 (1.8%)	1.96	0.264
EPEC-typical	3 (2.9%)	29 (3.5%)	0.81	1.00
ETEC	2 (1.9%)	37 (4.5%)	0.42	0.302
Salmonella species	0 (0%)	12 (1.5%)	--	0.380
Shigella species	2 (1.9%)	36 (4.4%)	0.43	0.302
Other bacteria	4 (3.8%)	14 (1.7%)	2.23	0.137
Parasitesc
Giardia species	13 (14.3%)	80 (10.6%)	1.35	0.290
Cryptosporidium species	9 (9.9%)	25 (3.3%)	2.98	0.003
Entaeombea species	0 (0%)	4 (0.5%)	--	1.00
Ascaris lumbricoides	0 (0%)	21 (2.8%)	--	0.154
Blastocystis species	8 (8.8%)	55 (7.3%)	1.21	0.608
Other parasite	2 (2.2%)	3 (0.40%)	5.52	0.092
No organism identifiedd	41 (45.1%)	377 (50.0%)	0.90	0.597

^aAmong children with at least one diarrheagenic E. coli serovar (25 HEU and 196 HIV-unexposed)

^bAmong 367 patients enrolled since March 2013 who had the gene eae tested

^cAmong 845 children in whom whole stool was collected (HEU: n=91, HIV-unexposed: n=754)

^dAmong the 845 children who had both bacteria and parasite testing performed (HEU: n=91, HIV-unexposed: n=754)

Cryptosporidium spp. was also more commonly identified in the stool of the 60 HEU children <24 months of age who were tested for parasites (12.2%) as compared to the 444 HIV-unexposed children <24 months who were tested for parasites (4.7%) (PR 2.59 [95%CI: 1.1–6.2 P=0.033], age-adjusted PR: 2.68 [95%CI: 1.1–6.4, P=0.03]). Among children with MSD (35 HEU & 224 HIV-unexposed), the prevalence of Cryptosporidium spp. infection was also significantly higher in HEU children compared to HIV-unexposed children (PR 4.1 [95%CI: 1.4–11.7, P=0.008], age-adjusted PR: 4.1 [95%CI: 1.5–11.7, P=0.008]). No other differences in pathogen distribution were observed between HEU and HIV-unexposed children in these subgroups.

DISCUSSION

In this cross-sectional study of Kenyan children with acute diarrhea, HIV-infection and HIV-exposure status were both associated with specific enteric pathogens. Typical EPEC was over three times more common in HIV-infected compared to HIV-uninfected children and Cryptosporidium spp. three times more common in HEU compared to HIV-unexposed children. These associations were independent of measured potential confounding and mediating factors. Additionally, we found magnitude of associations to be particularly strong in the subgroup of children with MSD. This finding further supports data from the multi-site GEMS study which found typical EPEC and Cryptosporidium spp. infection during diarrhea episodes to be independently associated with higher case-fatality, particularly in sub-Sahahan African countries where HIV prevalence was highest [2]. If typical EPEC and Cryptosporidium spp. are disproportionally affecting HIV-infected and HEU children, and if HIV-testing rates in health care settings continue to increase, then health workers seeing children with acute diarrhea might consider child and maternal HIV-infection to be an indicator for more intense follow-up or empiric antibiotic/antiprotozoal therapy.

The association between typical EPEC and HIV-infection in children with acute diarrhea suggests a possible immune-modulated mechanism of action; it could be that the attaching and effacing (A/E) lesions characteristic of EPEC exploit deficiencies in the intestinal immune system or epithelial barrier, both common pathologies in HIV-associated intestinal dysfunction [32, 33]. Current Kenyan guidelines for management of diarrhea do not specify antimicrobial treatment of EPEC [34]. In other settings, the recommended treatment for EPEC is either CTX or a floroquinolone [35]. We did not observe differences in the prevalence of other pathogens between HIV-infected and HIV-uninfected children, a finding consistent with reports from other studies [8, 9]. However, among a subset of the GEMS study population in Western Kenya, higher prevalences of ETEC, Cryptosporidium, EPEC and astrovirus in HIV-infected children were reported [36].

We also observed a strong association between HIV-exposure and Cryptosporidium spp. infection among HIV-uninfected children. HIV-infected persons are at greater risk for Cryptosporidium spp. infection and likely shed Cryptosporidium oocysts in greater quantities and for longer periods of time than their HIV-uninfected counterparts [37, 38]. As a result, children of HIV-infected mothers may be exposed to the parasite more frequently than HIV-unexposed children. HEU children may also be less likely to acquire Cryptosporidium-specific antibodies from breast milk [39]. Although we did not observe differences in exclusive breastfeeding duration, we did find that among the younger children (<24 months), HIV-infected and HEU children were less likely to be currently breastfed even after accounting for age differences within the younger age group. Finally HEU children were more likely to have recently taken CTX than HIV-unexposed children. Recent exposure to antibiotics may result in treatment or suppression of bacterial pathogens, resulting in the preferential identification of parasitic infections such as Cryptosporidium spp.

Cryptosporidium spp. infections are associated with nutrient malabsorption, growth faltering, and cognitive disabilities, even in the absence of diarrhea, and these outcomes are common among HEU children [40–42]. It is plausible that some of the failure to thrive observed among HEU children may be the result of increased risk of Cryptosporidium spp. infection in this group. Nitazoxanide has shown efficacy, including among HIV-infected individuals, and could be considered in HEU children with Cryptosporidium spp. infection [43, 44]. Antiretroviral treatment (ART) has also been shown to decrease susceptibility to Cryptosporidium spp. and earlier and/or better coverage of ART among HIV-infected mothers may reduce exposure to Cryptosporidium spp. infection in HEU children [45].

Our study had several strengths and limitations. Strengths include the large cohort with detailed characterization of bacterial and parasitic pathogens associated with diarrhea and consideration of the type 1 error rate from testing for associations among multiple pathogens. Limitations include the lack of non-diarrhea controls which limited our ability to estimate prevalence of asymptomatic carriage of each organism and subsequently estimate the proportion of diarrheal cases attributed to a given organism. However data from a large multi-country case-control study, the GEMS study, has addressed this issue among children with MSD [2]. While our study could not conclusively attribute the cause of diarrhea to the organisms identified, many of the organisms isolated are known to be associated with poor weight gain and linear growth, even in the absence of diarrhea [46–49]. We were also not able to isolate enteric viruses in this study and rotavirus is the leading cause of diarrheal illness in children [2, 36]. However, the introduction of rotavirus vaccination in high burden countries is expected to substantially reduce diarrheal illness due to rotavirus and thus bacterial and parasitic causes of diarrhea may become increasingly important [50, 51].

Although we measured exclusive breastfeeding duration and current breastfeeding status, we did not ascertain the age of weaning and thus could not evaluate how this variable might impact our findings. We also did not collect previous diarrhea morbidity information from the child or mother, and doing so might have strengthened a proposed causal relationship between maternal HIV, maternal Cryptosporidium infection, and Cryptosporidium in HEU children. Future studies that include surveillance stool sampling, ART-status of HIV-infected mothers, detailed weaning and feeding data, and extensive clinical information from the mothers will help us understand why typical EPEC and Cryptosporidium spp. were independently associated with HIV-infection and exposure, respectively, in this study population.

Intriguingly, although we detected an association between HEU and Cryptosporidium spp., we did not find an association with HIV-infection. The lack of association may be due to lack of power-- we had only 56 children with HIV, 8 of whom were not tested for parasites because they could not produce fresh stool. The cohort prevalence of Cryptosporidium spp. was 3.7%, a prevalence similar to studies which utilized microscopy, but lower than studies using PCR-based methods (prevalences as high as 31.3%) [52–54]. In addition, Cryptosporidium spp. infection is more common in children under 2 years old and the average age of HIV-infected children in our cohort was higher (4 years) [2, 55].

Despite these limitations, this study found important and novel relationships between HIV status and two enteropathogens that are significant causes of morbidity and mortality in sub-Saharan Africa. In current management guidelines for acute diarrhea, EPEC and Cryptosporidium are not specifically considered. In the absence of laboratory-based stool testing, ascertaining HIV-infection status of the child and the mother may help clinicians determine optimal empiric treatment for acute diarrhea. If HIV-infected mothers or other HIV-infected household members are indeed exposing children to Cryptosporidium spp. more frequently, then improving HIV care and treatment of mothers and other HIV-infected household members may have indirect benefits such as reducing childhood diarrhea incidence, growth failure, and cognitive delay. Finally, this study suggests that efforts to increase coverage of water, sanitation, and hygiene (WASH) programs are particularly important in high HIV-prevalence settings.