Cytomegalovirus Viremia and Clinical Outcomes in Kenyan Children Diagnosed With Human Immunodeficiency Virus (HIV) in Hospital

Abstract

Background

Cytomegalovirus (CMV) viremia is common in human immunodeficiency virus (HIV) infection and is associated with worse long-term outcomes. To date, no studies have assessed CMV viremia in children diagnosed with HIV in hospital.

Methods

We studied CMV viremia and clinical outcomes in 163 Kenyan children aged 2 months to 12 years, diagnosed with HIV in hospital. CMV DNA levels in plasma were measured using quantitative polymerase chain reaction (PCR). Regression models were used to assess associations between CMV viremia ≥1000 IU/mL and the risk of continued hospitalization or death at 15 days, duration of hospitalization, and 6-month mortality.

Results

At enrollment, 62/114 (54%) children had CMV viremia, and 20 (32%) were ≥1000 IU/mL. Eleven CMV reactivations were observed after admission. The prevalence and level of CMV viremia were highest in children <2 years and lowest in children ≥5 years old. CMV viremia ≥1000 IU/mL was independently associated with age <2 years (P = .03), higher log₁₀ HIV RNA level (P = .01), and height-for-age z score >−2 (P = .02). Adjusting for age and log₁₀ HIV RNA, the relative risk of death or continued hospitalization at 15 days was 1.74 (95% confidence interval [CI] = 1.04, 2.90), and the hazard ratio of 6-month mortality was 1.97 (95% CI = .57, 5.07) for children with CMV DNA ≥1000 IU/mL compared to lower-level or undetectable CMV DNA. Children with CMV DNA ≥1000 IU/mL were hospitalized a median ~5 days longer than children with lower-level or undetectable CMV DNA (P = .002).

Conclusions

In this nested observational study, CMV viremia was common in hospitalized children with HIV, and levels ≥1000 IU/mL were associated with increased risk of mortality and longer hospitalization.

In Kenyan children ≤ 12 years old diagnosed with human immunodeficiency virus (HIV) in hospital, cytomegalovirus (CMV) viremia at ≥1000 IU/mL was found in 32% of children and associated with higher risk of death or continued hospitalization at 15 days, and longer duration of hospitalization.

The association between cytomegalovirus (CMV) coinfection and poor human immunodeficiency virus (HIV) outcomes is well established in both children and adults [1, 2]. In addition to being an opportunistic infection, CMV likely contributes to non-AIDS morbidity and mortality risk even during effective HIV suppression on antiretroviral therapy (ART) [3]. CMV infection is associated with greater expansion of activated and exhausted T cells, and augmented host inflammation and microbial translocation [4–7], which are hallmarks of HIV pathogenesis. Clinical trials of CMV suppression in the pre-ART era were designed to use ganciclovir, high-dose acyclovir, or valacyclovir over many months [8–10]; but because many of these were terminated early due to adverse events, there are limited data on the benefit of CMV suppression. Short-course oral CMV antivirals were found to reduce frequencies of activated CD8 T cells in blood of ART-treated adults [11] and in the semen of ART-naive men [12], suggesting that CMV therapy even in the absence of apparent disease could reduce host immune activation.

Despite a wealth of research focused on CMV and long-term HIV outcomes, there are virtually no data on the impact of CMV viremia in children who are diagnosed with HIV while hospitalized. This question is highly relevant as the yield of HIV testing in African hospitals is now highest compared to other testing venues [13]. In sub-Saharan Africa, hospitalized children with HIV have extremely high mortality (16–39% [14–16]), despite rapid initiation of ART [14, 15, 17].

In the recent Pediatric Urgent Start of Highly Active Antiretroviral Therapy (HAART) (PUSH) Study, starting ART within 48 hours of hospital HIV diagnosis in hospitalized children was found to be safe but similar to post-stabilization ART initiation for survival [14]. We used archived specimens and data from PUSH to evaluate the association between CMV viremia and outcomes, to determine whether CMV suppression during ART initiation should be studied as a novel interventional approach to improve outcomes in this population.

METHODS

Recruitment and Follow-up

This study was nested in the PUSH randomized clinical trial (RCT) of urgent versus post-stabilization ART initiation among children diagnosed with HIV infection during hospitalization (NCT02063880). Human subjects’ approvals were obtained from University of Nairobi/ Kenyatta National Hospital (UoN/KNH) Ethics Research Committee (ERC) and the University of Washington Institutional Review Board (IRB).

The study was conducted in 2 hospitals in Nairobi (KNH and Mbagathi County Hospital) and 2 in western Kenya (Jaramogi Oginga Odinga Teaching and Referral Hospital and Kisumu County Hospital). Details on recruitment, interventions, and main study outcomes have been reported previously [14]. Eligibility criteria for children enrolled in the PUSH trial included: hospitalized, age 0–12 years, HIV-positive, antiretroviral therapy (ART) naive, and no clinical evidence of central nervous system infection. Hospital staff tested children older than 18 months for HIV using 2 different rapid tests in compliance with Kenya National Guidelines; those with confirmed HIV infection were referred to the study team for screening. Children younger than 18 months who met Kenya National Guidelines as “HIV-exposed” received an HIV polymerase chain reaction (PCR) test; PCR result turnaround time was 48 hours after sample collection. Enrolled children were randomized to receive either ART within 48 hours (urgent) or 7–14 days (post-stabilization), with regimens selected according to contemporaneous Kenya and World Health Organization (WHO) guidelines.

Children were followed for 6 months after ART initiation. At 8 scheduled visits (enrollment, week 1, 2, and monthly from 1 to –6 months post-ART), detailed sociodemographic and clinical data were collected. Laboratory samples were collected at 5 timepoints (enrollment, week 2, month 1, month 3, month 6). Plasma samples were stored at all sampling visits when possible.

Clinical Data

Reason for hospital admission was recorded at enrollment, with each case diagnosed by standard hospital protocols. At enrollment and each follow-up visit, growth indicators (weight, height, mid-upper arm circumference [MUAC]) were collected and used to calculate Z-scores using the WHO reference population. Severe acute malnutrition (SAM) was defined as WHZ <−3 and visible wasting or edema. Baseline laboratory values included neutrophil count, CD4, and log₁₀ HIV viral load. Clinical stage and severe immunosuppression were defined using CD4% thresholds by age (<12 months <30%, 12–35 months <25%, 36–59 months <20%, over 5 years <350 cells/uL) according to WHO criteria [18, 19].

CMV DNA Measurements and Definitions

CMV DNA levels were measured from stored plasma specimens using real-time quantitative PCR as previously described [26]. Briefly, viral nucleic acids were extracted from 200 uL of thawed plasma using a QIAamp Ultrasens Virus kit (Qiagen, Inc., Valencia, California, USA) according to manufacturer’s instructions. TaqMan PCR was performed targeting the UL55/UL123-exon 4 regions of CMV, and cycle thresholds compared to a standard curve to determine viral copies/mL; primers and probes are provided in reference [26]. Values were transformed to express measurements in international units (IU)/mL by dividing by 1.4. The PCR limit of detection was 1 copy per reaction (35.7 IU/mL). CMV reactivation was defined as CMV DNA detection > 1 copy/rection following one or more initially negative CMV DNA tests.

Statistical Analysis

Stata v.14 (StataCorp, College Station, Texas, USA) was used for all analyses, and all comparisons were 2 sided with alpha = 0.05. The Wilcoxon rank sum test was used to compare the distribution of CMV levels between groups. Potential correlates of CMV viremia were selected from previous literature in adult critical care settings (corticosteroids, clinical acuity, clinical laboratory values) [20] and expected confounders in HIV-infected children (HIV RNA, CD4 percent, age, malnutrition) as well as site of recruitment (Nairobi/Kisumu). Correlates of CMV viremia were identified using Poisson regression to estimate prevalence ratios and 95% confidence intervals (CI), adjusting for baseline log₁₀ HIV RNA level and age.

Primary outcomes defined a priori included death within 6 months, the combined outcome of death or continued hospitalization on day 15, and duration of hospitalization among survivors. Fifteen days were chosen for assessment of the combined endpoint because this is the median duration of hospitalization in the high-dependency acute wards and intensive care unit at Kenyatta National Hospital (Saini et al, manuscript in preparation). The primary exposures investigated were CMV viremia ≥1000 IU/mL and baseline CMV viral load (log IU/mL). There is no universally agreed-upon plasma CMV level of clinical significance, however levels ≥1000 IU/mL have higher statistical reliability than lower levels [21]. This level was predictive of mortality in immunocompetent adults admitted to intensive care in the United States [22] and has been associated with decreased lung function and stunting in Zimbabwean youth with HIV infection [23].

Generalized linear models were used to compare the relative risk of the combined endpoint, and Kaplan-Meier survival analysis was used to compare 6-month mortality risk between groups. Variables that were independently associated with both CMV viremia at enrollment and mortality in our earlier report [24] at P < .05 (child age and baseline log₁₀ HIV RNA level) were assessed as confounders between CMV viremia and survival outcomes. We assessed potential effect modification by age on 6 month survival, stratifying by ≥2 years and < 2 years and present these analyses as Supplementary data. We used the Benjamini-Hochberg method at a false discovery rate of 0.05 to control for multiple testing for the correlates analysis.

Among children surviving to 6 months, the Wilcoxon rank sum test was used to compare duration of hospitalization between children who were CMV viremic and aviremic at enrollment.

Because this study was nested into an RCT, we additionally assessed potential effect modification by randomization arm (urgent vs post-stabilization initiation of ART) for each analysis but found no evidence that results differed by randomization arm (data not shown).

RESULTS

Population Characteristics

A total of 163 (88%) of 181 children enrolled in PUSH had plasma available for CMV PCR. Among these, median age was 2 years (interquartile range [IQR] = 1.0, 5.4), 47% were female, and 93% were under the care of their biological mother (Table 1). Most (78%) were severely immunosuppressed at admission with very high HIV RNA levels. Many children were classified as stunted (59%) or wasted (45%), and a quarter had severe acute malnutrition. The most common admission diagnosis was pneumonia (65%). Median time between admission and enrollment was 4 days (IQR = 2, 6).

Table 1.

Baseline Characteristics of Kenyan Children Diagnosed With Human Immunodeficiency Virus (HIV) in Hospital Who Were Tested for Cytomegalovirus (CMV) DNA

	N	Median (IQR) / n (%)
Demographics
Age	163	2.0 (1.0, 5.4)
Age < 1 years old	163	40 (25%)
Age ≥1to <2 years old	163	42 (26%)
Age ≥2 to <5 years	163	39 (24%)
Age ≥5 to <13 years	163	42 (26%)
Female	163	76 (47%)
Primary caregiver mother	163	152 (93%)
Enrolled at Kisumu site	163	90 (55%)
Growth
WAZ < −2SD	158	98 (62%)
WHZ < −2SD	121	55 (45%)
HAZ < −2SD	159	94 (59%)
Severe acute malnutrition	119	31 (26%)
Enrollment diagnosis and disease stage
Pneumonia	163	106 (65%)
Pneumonia with hypoxia	94	22 (23%)
Suspected pulmonary TB	161	26 (16%)
Malaria	162	36 (22%)
Gastroenteritis	163	29 (18%)
Persistent diarrhea	163	17 (10%)
Baseline WHO stage III/IV	162	109 (67%)
Laboratory and treatment
Total neutrophil count	157	3.4 (2.3, 5.1)
CD4%	162	15 (9, 23)
CD4% ≥15%	162	83 (51%)
Severe immunosuppression (WHO age and CD4% criteria)	162	127 (78%)
Log10 HIV RNA copies/ml	154	5.7 (5.0, 6.3)
Lab confirmed TB	163	10 (6%)
Received steroids	155	8 (5.0%)
Key outcomes
Died before 6 months	163	25 (15%)
Died or continued hospitalization at 15 days	163	61 (37%)
Continued hospitalization among survivors (day 15)	138	36 (26%)
Median duration of hospitalization (days)a	151	10 (7, 16)

WAZ, HWZ, HAZ = weight for age, height for weight, and height for age z-score, respectively. Severe acute malnutrition was defined as WHZ < −3 SD, visible wasting, ± edema.

Abbreviations: IQR, interquartile range; MUAC, mid-upper arm circumference; SD, standard deviation; TB, tuberculosis; WHO, World Health Organization.

^aAmong children surviving to 6 months.

Twenty-five (15%) of 163 children included in this analysis died, and 61 (37%) died or were still hospitalized on day 15 since admission. Among 138 survivors at 6 months, 36 (26%) were still hospitalized 15 days after admission.

Prevalence of CMV Viremia

Of the possible 5 sample collection time points, the majority of children had between 3 and 5 CMV DNA measurements (data not shown). Overall, 72% (118/163) of children had CMV DNA detected at least once during follow-up. Sixty-two (54%) of 114 children were CMV viremic pre-ART at enrollment, and of these, 32% (20/62) had CMV levels ≥1000 IU/mL. Figure 1A shows the prevalence of CMV detection in children grouped by age; the prevalence of viremia was highest for children <1 year old (88%) and lowest in the children aged 5–12 years (21%). Of 72 children with CMV viremia at any time during the study, 11 (9%) were CMV reactivations, and these children were mostly older (7 of these children were aged 2–5 or 5–12 years, Figure 1B).

Figure 1.

Proportion of children with CMV viremia at hospital admission or reactivation during follow-up, by age at enrollment. A, Proportion of children with CMV viremia at enrollment. N = 114 children with admission samples screened in children aged ≤1 year (n = 26), 1–2 years (n = 33), 2–5 years (n = 27), and 5–12 years (n = 28). Black bars = CMV viremia ≥35.7 IU/mL (above the limit of detection) detected at enrollment, gray bars = CMV viremia ≥1000 IU/mL detected at enrollment. B, Proportion of children with CMV reactivation. N = 118 children first screened at enrollment or their next visit; 11 were initially negative then experienced CMV viremia at a later visit. Reactivations occurred in 2 children aged ≤1 year, 2 children aged 1–2 years, 3 children aged 2–5 years, and 4 children aged 5–12 years. A, Proportion CMV viremic at admission. B, Proportion reactivating CMV after admission. Abbreviation: CMV, cytomegalovirus.

CMV Viral Load Trajectories From Admission to 6 Months

The overall median CMV level of children who were viremic at baseline was 2.41 log₁₀ IU/mL (IQR = 2.06, 3.21), and their peak was 2.82 log₁₀ IU/mL (IQR = 2.18, 3.41); CMV viral load trajectories generally decreased following ART initiation but remained detectable in 21/75 (28%) at 6 months (Figure 2A). Follow-up time was much shorter in children who died, and there was high variability in their CMV viral load trends over time (Figure 2B). CMV levels followed similar trajectories for younger (<2 years) and older (≥2 years), with both groups declining over time (Figure 2C). Children aged <2 years old had higher median CMV levels compared to children ≥2 years old at enrollment (median 3.0 vs 2.2, P = .0001).

Figure 2.

Longitudinal CMV levels in children with CMV viremia at admission. Longitudinal trajectories of children who were CMV viremic, who (A) survived or (B) died during follow-up. C, CMV IU/mL of plasma is shown longitudinally at each study timepoint (W = week) for children who had CMV viremia at hospital admission (Admit), categorized by age under 2 years and 2 years and older. D, Longitudinal trajectories of 11 children who reactivated CMV while in hospital. A, Survivors with enrollment CMV viremia. B, Mortalities with CMV viremia. C, All children CMV viremic at enrollment. D, CMV reactivations after admission. Abbreviation: CMV, cytomegalovirus.

The median first CMV level measurement of children reactivating CMV was 1.71 log₁₀ IU/mL (IQR = 1.34, 2.07), and their peak was 1.84 (IQR = 1.71, 2.28); these were both significantly lower than the baseline and peak levels of children who were viremic at admission (P < .001 for each comparison). There was limited follow-up time post-reactivation to observe longitudinal changes, but generally their CMV levels remained low through follow-up (<1000 IU/mL, Figure 2D).

Correlates of CMV Viremia at Enrollment

Clinical and laboratory characteristics were compared between children with CMV viremia ≥1000 IU/mL at enrollment (hospital admission) versus children who were CMV undetectable or had CMV DNA levels <1000 IU/mL (Table 2). Each variable was assessed individually and then adjusted for age and/or HIV RNA level. Age <2 years (adjusted prevalence ratio [aPR] = 3.71 [95% CI = 1.18, 11.8], P = .026), HAZ < -2 (aPR = 4.74 [95% CI = 1.23, 18.3], P = .024) and HIV RNA level (aPR = 1.91 [95% CI = 1.16, 3.16], P = .012) were independently associated with CMV viremia ≥1000 IU/mL. Covariates that were significantly different in univariate analysis but did not retain significance after adjustment included receipt of steroids, WHO clinical stage III and C-reactive protein level. After adjustment for multiple testing, none of the correlates remained significantly associated with CMV viremia ≥1000.

Table 2.

Correlates of Cytomegalovirus (CMV) Viremia ≥1000 IU/mL at Admission in Hospitalized, ART-Naive Kenyan Children With Human Immunodeficiency Virus (HIV)

	N	n (%) or median (IQR) in Category	CMV ≥1000 IU/mL n (%) or median (IQR)	Crude PR (95% CI)	P value	Adjusted PR (95% CI)	Adjusted P value
Demographics
Age < 2 years	114	59 (52%)	17 (29%)	5.28 (1.63, 17.1)	.006	3.71 (1.17, 11.8)	.03a
Age ≥2 years		55 (48%)	3 (5%)
Recruitment site Kisumu	114	78 (68%)	11 (14%)	0.56 (.26, 1.24)	.2	0.76 (.37, 1.56)	.5
Recruitment site Nairobi		36 (32%)	9 (25%)
Nutritional status
Underweight	110	70 (64%)	15 (21%)	1.71 (.67, 4.38)	.3	1.26 (.50, 3.20)	.6
Not underweight		40 (36%)	5 (13%)
Wasted	86	40 (47%)	10 (25%)	1.44 (.63, 3.30)	.4	1.17 (.52, 2.62)	.7
Not wasted		46 (53%)	8 (17%)
Stunted	111	65 (59%)	16 (25%)	5.66 (1.36, 23.6)	.02	4.74 (1.23, 18.3)	.02
Not stunted		46 (41%)	2 (4%)
SAMc	85	21 (25%)	6 (38%)	1.83 (.75, 4.45)	.2	1.79 (.79, 4.06)	.2
Not SAM		64 (75%)	10 (16%)
Immune status
WHO stage IV	113	36 (32%)	5 (45%)	5.45 (1.53, 19.4)	.009	2.44 (.69, 8.68)	.2b
WHO stage III		66 (58%)	12 (18%)	2.18 (.65, 7.27)	.2	2.00 (.62, 6.43)	.2b
WHO stage I/II		36 (32%)	3 (8%)	Ref
Severe immunosuppressiond	113	85 (75%)	18 (90%)	2.96 (.73, 12.1)	.1	1.08 (.27, 4.38)	.9b
Not severe immunosuppressiond		28 (25%)	2 (7%)
On steroids at enrollment	108	5 (5%)	3 (60%)	4.12 (1.75, 9.73)	.001	2.05 (.86, 4.92)	.1
Not on steroids		103 (95%)	15 (15%)
Urgent RCT ART arm	114	57 (50%)	12 (21%)	1.50 (.66, 3.40)	.3	1.53 (.70, 3.33)	.3
Early RCT ART arm		57 (50%)	8 (14%)
Laboratory
Total neutrophil count	112	3.34 (2.12, 4.92)	3.02 (1.73, 6.81)	1.01 (.95, 1.06)	.9	1.04 (.98, 1.11)	.2
C-reactive protein	71	14 (3, 59)	3.8 (3, 36)	0.98 (.97, .99)	.01	0.99 (.97, 1.00)	.06
Hemoglobin level (g/dL)	114	8.8 (7.5, 9.6)	8.7 (7.9, 9.8)	1.15 (.94, 1.40)	.2	1.14 (.95, 1.38)	.2
CD4%	113	16.1 (10.0, 22.9)	11.6 (8.95, 19.6)	0.97 (.94, 1.01)	.1	0.97 (.93, 1.02)	.2
Log10 HIV RNA copies/mL	110	5.71 (5.02, 6.25)	6.29 (5.94, 6.70)	2.51 (1.44, 4.38)	.001	1.91 (1.16, 3.16)	.01b

Abbreviations: ART, antiretroviral therapy; CI, confidence interval; PR, prevalence ratio; RCT, randomized clinical trial; SAM, severe acute malnutrition; WHO, World Health Organization.

^aAdjusted for viral load only.

^bAdjusted for age only.

^cSevere acute malnutrition, defined as WHZ < −3 S, visible wasting, ± edema.

^dSevere immunosuppression defined by age and CD4 count according to WHO criteria. After the Benjamini-Hochberg false discovery rate correction, none of the correlates remained significantly predictive of CMV viremia.

Correlates of any CMV DNA viremia above versus below the limit of detection (35.7 IU/mL) are provided in Supplementary Table 1. Variables independently associated with CMV detection included younger age (aPR = 0.85 [95% CI = .77, .93], P = .001) and higher baseline HIV viral load (aPR = 1.25 [95% CI = 1.03, 1.53], P = .026). Covariates that were significantly different in univariate analysis but did not retain significance after adjustment included clinical diagnosis of malaria and total neutrophil count. After adjustment for multiple testing, none of the correlates remained significantly associated with CMV DNA viremia.

Admission Diagnosis and CMV Viremia

The prevalence of CMV viremia and CMV levels are shown for children by admission diagnosis in Table 3, ranked by median CMV level at enrollment. No children were diagnosed with active CMV disease during the study period and no children with treated with anti-CMV medications. CMV was most prevalent (77%) and detected at the highest level in children with diarrhea or gastroenteritis (median log₁₀ 2.6 IU/mL). CMV viremia was detected in a fifth of children with pneumonia, at a median of 2.4 log₁₀ IU/mL. Median oxygen saturation levels (SpO₂) among children admitted for pneumonia with hypoxia were 86% (IQR = 80, 88) in CMV aviremic children, 82% (IQR = 75, 86) in children with CMV viremia, and 78% (IQR = 72, 84) in children with CMV ≥ 1000IU/mL (comparing SpO₂ CMV ≥ 1000IU/mL vs CMV < 1000 mL or aviremic P = .17). CMV levels were higher with more advanced WHO clinical stage at enrollment (median 1.9 log₁₀ IU/mL [IQR 1.6, 2.5] for those in stage I/II, 2.1 [IQR 1.5, 2.9] for stage III, and 2.9 [IQR 1.5, 3.9] for stage IV). Clinical characteristics of the 11 children who reactivated CMV during the study are described in detail in Supplementary Table 2.

Table 3.

Prevalence and Levels of Cytomegalovirus (CMV) Viremia at Hospital Admission, by Diagnosis

	n/N (%)	Prevalence of CMV Viremia	Prevalence of CMV > 1000 IU/mL	Median log10 CMV IU/mL in Children With CMV Viremia
Diarrhea and/or gastroenteritis	37/114 (23%)	21 (70%)	8 (27%)	2.6 (2.1, 3.6)
Persistent diarrhea	15/114 (13%)	10 (67%)	5 (33%)	3.0 (2.2, 3.7)
Gastroenteritis	22/114 (19%)	17 (77%)	6 (18%)	2.6 (2.1, 3.4)
Pneumonia	74/114 (65%)	42 (57%)	14 (19%)	2.4 (2.1, 3.2)
Pneumonia with hypoxia	19/70 (27%)	11 (58%)	4 (21%)	2.3 (2.0, 3.6)
Malaria	27/114 (24%)	9 (33%)	2 (7%)	2.1 (1.8, 2.6)
Confirmed tuberculosisa	5/114 (5%)	1 (20%)	0	2.1 (2.1, 2.1)

Five most frequent admission diagnoses in Pediatric Urgent Start of Highly Active Antiretroviral Therapy (PUSH). Some children had multiple diagnoses and appear in more than 1 category.

^aMicrobiologically confirmed tuberculosis.

CMV Viremia and Outcomes Over 6 Months Follow-up

Figure 3A shows Kaplan-Meier survival curves over the 6 months of follow-up by CMV level in the 114 children with enrollment CMV testing. Mortality rates in CMV aviremic children and children with low level CMV viremia (< 1000 IU/mL) had similar time to death (P = .5). Children with CMV viremia ≥1000 IU/mL had a significantly shorter time to death compared to children with lower CMV levels (P = .02) and aviremic children (P = .002). Children who were aviremic or had low-level viremia (<1000 IU/mL) were grouped together for further analyses (Figure 3B). Kaplan-Meier survival curves for children stratified by age <2 years and ≥2 years are shown in Supplementary Figure 1.

Figure 3.

Survival probabilities for hospitalized children by admission CMV viremia. N = 114 children with CMV assessed at enrollment. A, Twenty mortalities occurred: 5 in CMV aviremic children (blue line), 5 in CMV viremic children with levels < 1000 IU/mL (red line), and 10 in children with CMV levels ≥1000 IU/mL (green line) (log-rank P = .002 for trend). Log-rank for CMV viremic < 1000 IU/mL group versus aviremic: P = .5, CMV viremic ≥1000 IU/mL vs aviremic: P = .002, CMV viremic ≥1000 IU/mL vs viremic < 1000 IU/mL, P = .02. B, log-rank P = .02. Abbreviation: CMV, cytomegalovirus.

Crude and adjusted point estimates for CMV viremia and clinical outcomes are shown in Table 4. CMV levels ≥1000 IU/mL were associated with a 74% increased risk of attaining the combined endpoint of death or continued hospitalization at 15 days, independent of log₁₀ HIV RNA level and age (aRR = 1.74 [95% CI = 1.04, 2.90]). Enrollment CMV DNA level was associated with increased risk of the combined endpoint but did not retain significance after adjusting for age and baseline log₁₀ HIV viral load (aRR = 1.17 [95% CI = .87, 1.58]).

Table 4.

Survival Outcomes by Enrollment Cytomegalovirus (CMV) Viremia in 114 Kenyan Children Diagnosed With Human Immunodeficiency Virus (HIV) Infection in Hospital

	Total Events	Events (%) or Median (IQR) in CMV Aviremic or Low-level Viremia n = 62	Events (%) or Median (IQR) in CMV Viremic ≥ 1000 IU/mL n = 20	CMV Aviremic or Low-Level Viremia vs Viremic ≥ 1000 IU/mL, crude	CMV Aviremic or Low-Level Viremia vs Viremic ≥ 1000 IU/mL, Adjusted for HIV RNA and Age	Log10 CMV IU/mL, Crude	Log10 CMV IU/mL, Adjusted for HIV RNA and Age
Death or continued hospitalization at day 15	40/114	28 (45%)	13 (65%)	RR = 2.26 (1.44, 3.56) P = <.001	aRR = 1.74 (1.04, 2.90) P = .04	RR = 1.34 [95% CI = 1.05, 1.69], P = .02	aRR = 1.17 [95% CI = .87, 1.58], P = .3
6-month mortality	20/114	15 (24%)	8 (40%)	HR = 3.78 (1.54, 9.25) P = .004	aHR = 1.97 (.77, 5.07) P = .2	HR = 1.42 [95%CI = .98, 2.07], P = .06	aHR = 1.05 [95% CI = .66, 1.68], P = .8

Compares children with CMV ≥ 1000 IU/mL to children who are CMV DNA undetectable or have CMV detection <1000 IU/mL. Abbreviations: aHR, adjusted hazard ratio; aRR, adjusted risk ratio; CI, confidence interval; HR, hazard ratio; IQR, interquartile range; RR, risk ratio.

Among children surviving to 6 months post-admission, the median duration of hospitalization was approximately 5 days longer (14.5 days [IQR = 13, 16]) in the children with CMV level ≥1000 IU/mL compared to children who were aviremic or had a CMV level <1000 IU/mL (9 days [IQR = 5, 12], P = .002). Enrollment CMV DNA level was not associated with longer duration of hospitalization (P = .3). We did not adjust for age and HIV RNA as confounders in this analysis because they were not associated with duration of hospitalization.

CMV level ≥1000 IU/mL was associated with increased hazard of death over 6 months in crude analyses (hazard ratio [HR] = 3.78, [95% CI = .54, 9.25]) but did not retain significance when adjusting for HIV RNA level and age. Enrollment CMV DNA level was not significantly associated with higher 6-month mortality.

DISCUSSION

In this study of Kenyan children diagnosed with HIV at hospital admission, we found a very high rate of CMV viremia, which was inversely associated with child age. Most children were viremic at admission, but a substantial number were also observed to reactivate CMV infection during study follow-up. We found that children with CMV DNA levels ≥1000 IU/mL in plasma had a higher risk of continued hospitalization or mortality at 15 days post-admission, and longer duration of hospitalization compared to children who were CMV aviremic or had low-level CMV viremia. These data suggest CMV levels above 1000 IU/mL identify children at high risk for poor outcomes while starting ART during hospitalization.

CMV viremia was present in more than half of children at baseline and overall was detected in nearly three quarters of children during follow-up. To date, the only similar data to ours are from a study in Zambia where 55% of hospitalized children with HIV who were <2 years old were CMV viremic at admission [25]. Consistent with our study, Tembo et al found that CMV viremia was associated with being underweight and younger age; however, they did not report outcomes by CMV viremia. Younger age and higher HIV RNA level were independently associated with an increased risk of having baseline CMV viremia, but surprisingly CD4 percent and WHO clinical stage were not; we hypothesize that this is because host inflammation and global immune dysregulation also contribute to CMV viremia. Given age-related changes in CD4 numbers over infancy, there may also be residual confounding by age. Children under 2 years old comprised the majority of mortalities in the PUSH Cohort, suggesting that quantitative CMV PCR testing in this age group may be important for clinical prognosis and future evaluation of the role of anti-CMV therapeutics.

We previously reported detection of CMV viremia by 6 months of age in ~80% of ART-naive Kenyan children with in utero or very early (<1 month) HIV acquisition, with many children remaining CMV viremic for a year or more [26]. Because the PUSH study included infants and children across a wide age range, the viremia we found is most likely a mix of primary CMV infections and reactivations. Children with CMV viremia at admission had higher levels than those with CMV observed after enrollment, and none of the children with observed CMV reactivation died. Given our sampling intervals and the small number of reactivations observed, we cannot conclude whether reactivations occurred as a manifestation of treatments received in hospital. The observed reactivations occurred across diverse clinical diagnoses and did not appear to be related to transfusion or receipt of steroids. None of the children with CMV reactivation had evidence of immune reconstitution inflammatory syndrome (IRIS). Because of the small number and short follow-up time post-CMV detection, our study cannot determine the potential clinical relevance of these CMV reactivation events.

CMV level ≥1000 IU/mL was associated with an increased risk of continued hospitalization or death by 15 days, independent of log₁₀ HIV viral load and age. In survivors, high-level CMV viremia was associated with 5 days longer hospitalization. Although we cannot ascertain if CMV viremia is causally related to these poor outcomes due to the observational nature of our study, our findings are consistent with previous literature demonstrating an association between CMV viremia and poor long-term outcomes (growth, morbidity, mortality) in children and adults living with HIV [3, 27–29] and short-term hospitalization outcomes (including increased mortality, increased oxygen dependency, and longer hospital stay) in HIV-negative, immunocompetent adults [20, 30].

The most common hospital admission diagnoses in the PUSH trial were pneumonia or diarrhea; these are also the 2 most frequent causes of hospitalization in Kenya and children from areas with high rates of malnutrition globally [31]. CMV viremia ≥1000 IU/mL was especially common in children with diarrhea and pneumonia and could reflect the primary source of CMV viral replication in the lung or gastrointestinal tract. Our data suggest SpO₂ levels may be lower in the CMV viremic children admitted for severe pneumonia and warrants further study. CMV has wide tropism and can cause pneumonia and colitis. Immune modulation/increased susceptibility to other infections, and CMV-induced aggravation of lung inflammation is hypothesized to precipitate clinical decline in critically ill immunocompetent adults [20], and a similar mechanism could be proposed in the gastrointestinal tract. The association we found between CMV viremia and stunting, a marker of chronic malnutrition, also supports a potential role for CMV in the gastrointestinal tract of these children. Malnutrition contributes to immune dysregulation and gut dysbiosis, and both animal and human studies have found that CMV replication disrupts the gut microbiome [32, 33]. Systematic evaluation of CMV replication in fluids and tissues from the gastrointestinal tract and lung would be informative but present unique implementation challenges in children.

Our study has several strengths and limitations. Strengths include the large and unique hospital cohort, with systematic and detailed sampling and clinical data collection. Despite this, the children had diverse clinical experiences and investigations during hospitalizations that limit our ability to analyze the clinical course of children, and we do not have detailed data on cause of death. We were not able to perform histology on lung or gastrointestinal tract samples, which would be mechanistically informative, and we did not perform autopsies on children who died. We also had low statistical power for mortality. Because some specimens had been used for the parent study, we did not have complete sampling at baseline on all participants, further limiting statistical power. Finally, because our study was observational, we cannot conclude whether a causal relationship exists between CMV and the outcomes we studied.

In summary, CMV viremia was found in the majority of ART-naive children diagnosed with HIV at hospital admission, and levels ≥1000 IU/mL were associated with worse clinical outcomes, including longer duration of hospitalization and a higher risk of continued hospitalization or death at 15 days. Together, these data support further research into the relevance of CMV viremia in this population and CMV suppression as a potential novel intervention.

Supplementary Data

Supplementary materials are available at Clinical Infectious Diseases online. Consisting of data provided by the authors to benefit the reader, the posted materials are not copyedited and are the sole responsibility of the authors, so questions or comments should be addressed to the corresponding author.

Notes

Data availability statement. The data that support the findings of this study are available from the corresponding author upon reasonable request.

Disclaimer. The publication’s contents are solely the responsibility of the authors and do not represent the official views of the funders.

Financial support. This work was funded by grant number HD089821 from the Eunice Kennedy Shriver National Institute of Child Health and Human Development (National Institutes of Health [NIH]) to J. A. S. and grant number HD023412 to G. J. S. This publication was made possible with support from the University of Washington Center for AIDS Research (grant number P30 AI027757) and the Global Center for Integrated Health of Women, Adolescents, and Children (Global WACh).

Potential conflicts of interest. D. W. reports receiving NIH Grant for the parent study, outside the conduct of the study. G. J. S. reports grants from NIH, during the conduct of the study; grants from NIH, grants from Centers for Disease Control and Prevention other support from Thrasher, personal fees from UpToDate, other support from Malaika, personal fees from UW, and grants from IMPAACT, outside the submitted work. B. R. reports grants from NIH, during the conduct of the study; personal fees from Gilead, personal fees from UNC, personal fees from PATH, all for serving on DSMB, outside the submitted work. M. B. reports grants and personal fees from Merck and Gilead for research support and consulting; grants and personal fees from GlaxoSmithKline for consulting; personal fees from Chimerix, Helocyte, EvrysBio, Moderna, Allovir, and Symbio for consulting; grants from Astellas for research support; participated in an ad hoc advisory board meeting and received no compensation for Takeda; all outside the submitted work. L. C. reports grants from NIH/National Institute of Allergy and Infectious Diseases (K23AI143479 (PI Cranmer); UM1 AI068632 (PI Nachman; Cranmer Protocol chair IMPAACT 2035); and R01AI142647 (MPI John-Stewart and Day) and grants from Emory Global Health Office of Pediatrics (Pilot grant; MPI Cranmer and Quincer), outside the submitted work. All other authors report no potential conflicts. All authors have submitted the ICMJE Form for Disclosure of Potential Conflicts of Interest. Conflicts that the editors consider relevant to the content of the manuscript have been disclosed.