Acyclovir and ganciclovir exposure and short-term outcomes in hospitalized children with EBV-associated mononucleosis-like illness: a retrospective cohort study using inverse probability of treatment weighting

Abstract

Background

Epstein-Barr virus (EBV)-associated mononucleosis-like illness in children involves marked immune activation and variable clinical severity. Although supportive care remains standard for immunocompetent children, acyclovir and ganciclovir are frequently used in hospitalized patients with severe manifestations. However, their real-world effectiveness remains uncertain. This study aimed to characterize real-world antiviral treatment patterns and evaluate the associations of acyclovir and ganciclovir exposure with short-term clinical outcomes in hospitalized children with EBV-associated mononucleosis-like illness.

Methods

This single-center retrospective cohort study included 1,059 hospitalized children with EBV-associated mononucleosis-like illness at the Children’s Hospital of Soochow University from January 2019 to December 2023. Patients were categorized into no antiviral therapy (n=371), acyclovir (n=526), and ganciclovir (n=162) groups. Eligibility required compatible symptoms, EBV serologic evidence, detectable EBV DNA, atypical lymphocytosis, and acute-phase immunophenotyping. Antiviral exposure was analyzed as a time-varying variable, and stabilized inverse probability of treatment weighting (IPTW) was applied using baseline clinical, laboratory, virological, and immunological variables.

Results

Before weighting, antiviral-treated children had greater disease severity, including higher alanine aminotransferase (ALT), aspartate aminotransferase (AST), EBV DNA levels, and CD4/CD8 ratio inversion. After IPTW, covariate balance was achieved [all standardized mean difference (SMD) <0.10]. Median time to defervescence was 5.2, 6.1, and 5.6 days in the no-antiviral, acyclovir, and ganciclovir groups, respectively. IPTW-weighted Cox analysis showed lower fever-resolution hazard with acyclovir [hazard ratio (HR) 0.78, 95% confidence interval (CI): 0.69–0.89], whereas ganciclovir showed no significant association (HR 0.91, 95% CI: 0.76–1.09). The prediction model integrating clinical and immune indicators achieved an area under the curve (AUC) of 0.69.

Conclusions

Antiviral use mainly reflected disease severity and was not associated with improved short-term outcomes after adjustment. Supportive care and individualized risk stratification remain important.

Highlight box.

Key findings

• Children receiving antiviral therapy generally had greater baseline disease severity and longer hospitalization, reflecting severity-based treatment selection.

• After adjustment for measured confounders, acyclovir exposure was not associated with faster fever resolution.

• Clinical and immunologic indicators provided better risk stratification than Epstein-Barr virus (EBV) DNA levels alone.

What is known and what is new?

• Routine antiviral therapy is not generally recommended for uncomplicated primary EBV infection in immunocompetent children because evidence of clinical benefit remains limited. Nevertheless, antiviral agents are still used in real-world practice, particularly in hospitalized pediatric patients with more severe clinical presentations.

• In this large retrospective cohort, acyclovir exposure was not associated with faster fever resolution after adjustment for measured confounders, while ganciclovir showed no significant association with improved short-term outcomes. In addition, combined clinical and immunologic indicators provided better risk stratification than EBV DNA load alone.

What is the implication, and what should change now?

• Supportive care should remain the primary management strategy for immunocompetent children with EBV-associated mononucleosis-like illness.

• Antiviral prescribing should incorporate stewardship principles and avoid escalation based solely on EBV DNA levels.

• Early risk assessment using clinical and immune indicators may improve hospitalization and treatment decisions.

Introduction

Epstein-Barr virus (EBV) is one of the most common human herpesviruses, and most individuals acquire the infection asymptomatically or present with clinically nonspecific symptoms during childhood or adolescence (1). Primary infection in young children is frequently asymptomatic or clinically nonspecific, whereas delayed primary infection in adolescents and young adults more often presents as classical infectious mononucleosis (IM) with fever, pharyngitis, cervical lymphadenopathy, and hepatosplenomegaly (2). After primary infection, EBV establishes lifelong latency in memory B cells, with intermittent reactivation controlled primarily by cellular immune responses (3).

Acute EBV-associated IM is characterized by expansion of EBV-specific CD8+ T cells and broad lymphocyte-subset changes, including reduced CD4/CD8 ratio and activation of natural killer (NK) cells (4,5). CD4/CD8 inversion, defined as a ratio below 1.0, reflects a predominance of CD8+ effector responses during acute infection rather than the approximate 2:1 CD4/CD8 distribution commonly observed in healthy individuals (6). NK cells also contribute to early antiviral control through recognition of lytically infected cells, and expansion of early differentiated CD56dim NKG2A+KIR- NK-cell populations has been described during IM (7-9).

The therapeutic rationale for nucleoside analogues in EBV infection remains biologically limited. Although acyclovir and ganciclovir suppress lytic EBV DNA replication, the clinical manifestations of primary EBV-associated mononucleosis-like illness are predominantly driven by host immune activation rather than uncontrolled viral replication. Previous studies have not demonstrated consistent benefits of antiviral therapy in shortening symptom duration or preventing complications among immunocompetent patients (10,11). Therefore, supportive care remains the recommended management strategy in current clinical guidelines and pediatric expert consensus statements (10-12).

Despite the lack of evidence, however, a clinically relevant question remains. In real-world practice, hospitalized children with EBV-associated mononucleosis-like illness frequently receive antiviral agents such as acyclovir or ganciclovir, driven by concerns over airway compromise, marked hepatic involvement, persistent fever, or suspected immune vulnerability. In these situations, antiviral prescription often reflects physicians’ assessment of disease severity rather than evidence-based indications. This practice is particularly relevant in Asian pediatric settings, including China, where primary EBV infection occurs earlier than in classical adolescent IM, and hospitalized children are often classified as having EBV-associated mononucleosis-like illness when clinical, serologic, virologic, and hematologic criteria converge (13-15).

To reflect this local clinical context while maintaining diagnostic specificity, the study required concurrent evidence of compatible symptoms, EBV serology, EBV DNA detection, atypical lymphocytosis, and acute-phase immunophenotyping. This broader but still multi-component definition was intended to capture pediatric EBV-associated mononucleosis-like illness rather than to equate all cases with classical adolescent IM.

Although previous Chinese cohorts have described clinical characteristics and antiviral utilization among hospitalized children with EBV infection (13), few studies have simultaneously integrated acute immunophenotypic profiling with real-world treatment-pattern analysis and adjusted outcome comparisons while accounting for non-randomized, severity-driven treatment selection.

Therefore, this retrospective cohort study aimed to characterize real-world antiviral prescribing patterns and evaluate the association between antiviral exposure (acyclovir or ganciclovir) and short-term hospital outcomes among children with EBV-associated mononucleosis-like illness. We hypothesized that antiviral administration would primarily reflect baseline clinical severity and immune activation rather than provide measurable improvements in early clinical recovery. By integrating clinical features, laboratory parameters, and immunophenotypic indicators, this study sought to provide a more comprehensive framework for treatment decision-making and risk stratification in hospitalized pediatric EBV infection. We present this article in accordance with the STROBE reporting checklist (available at https://tp.amegroups.com/article/view/10.21037/tp-2026-0517/rc).

Methods

Study design and setting

This retrospective cohort study was conducted at the Children’s Hospital of Soochow University, a tertiary pediatric referral center in Suzhou, China, between January 1, 2019 and December 31, 2023. Consecutive hospitalized children aged 1–18 years who were diagnosed with EBV-associated mononucleosis-like illness were screened from the electronic medical record system. Clinical characteristics, laboratory findings, immunophenotyping results, antiviral exposure, safety outcomes, and hospitalization information were systematically extracted from electronic medical records.

Ethical approval

The study was conducted in accordance with the Declaration of Helsinki and its subsequent amendments. This study was approved by the Ethics Committee of Children’s Hospital of Soochow University (No. 2025CS175). Because only de-identified retrospective data were analyzed and no patient contact or research-directed treatment occurred, the requirement for informed consent was waived by the Ethics Committee.

Study population

During the study period, a total of 1,312 pediatric patients diagnosed with EBV-associated mononucleosis-like illness were initially screened. Of these, 193 patients were excluded due to incomplete medical records (n=97), receipt of mixed antiviral regimens or other antivirals (n=53), and missing key baseline data (n=43), leaving 1,119 eligible patients. Subsequently, 60 patients were excluded because they were lost to follow-up before outcome ascertainment (lack of documented fever resolution or discharge date). Ultimately, a final analytic cohort of 1,059 children was included in the study. Finally, 1,059 children fulfilled the eligibility criteria and were included in the final analysis (Figure 1).

Figure 1.

Flow diagram of patient selection and study cohort construction. EBV, Epstein-Barr virus.

Children were eligible only when all of the following criteria were fulfilled: (I) serological evidence compatible with acute or recent EBV infection, including viral capsid antigen-immunoglobulin M (VCA-IgM) positivity (with or without VCA-IgG), and Epstein-Barr nuclear antigen (EBNA)-IgG negativity or low-avidity VCA-IgG when available; (II) detectable EBV DNA in peripheral blood by quantitative polymerase chain reaction (PCR); (III) compatible clinical manifestations, including fever, pharyngitis, lymphadenopathy, hepatosplenomegaly, or prolonged systemic symptoms; (IV) atypical lymphocytes accounting for >10% of total leukocytes; and (V) available acute-phase immunophenotyping data. Therefore, diagnosis was based on integrated clinical, serological, virological, hematological, and immunological evidence rather than a single laboratory parameter. Thus, a single positive criterion alone was insufficient for study inclusion.

In China and other Asian pediatric settings, primary EBV infection often occurs at younger ages than classical adolescent IM in Western cohorts. Younger children may have nonspecific febrile presentations, less typical pharyngitis, and variable heterophile-antibody responses, and hospital admission can reflect local evaluation pathways for prolonged fever or organ involvement. For this reason, the cohort was described as EBV-associated mononucleosis-like illness rather than classical adolescent IM.

Children with primary or secondary immunodeficiency, concurrent systemic infection, active chronic EBV infection, EBV-associated hemophagocytic lymphohistiocytosis, malignancy, chemotherapy within the preceding six months, systemic corticosteroid therapy during the index illness, intravenous immunoglobulin use during the index illness, or insufficient clinical or laboratory data were excluded. Therefore, none of the included children received systemic corticosteroids or intravenous immunoglobulin as treatment for the indexed EBV illness. Antipyretics and non-steroidal anti-inflammatory drugs used for symptomatic fever control were permitted as supportive care.

EBV diagnostics

EBV-specific serology was performed using chemiluminescent immunoassays (LIAISON, DiaSorin, Saluggia, Italy) to detect viral capsid antigen IgM, viral capsid antigen IgG, and EBV nuclear antigen antibodies according to the manufacturer’s instructions.

Quantification of EBV DNA targeted the BALF5 gene using real-time PCR following automated magnetic bead-based DNA extraction (MagNA Pure 96, Roche Diagnostics). Amplification was performed on LightCycler 480 II instruments, with results expressed as EBV DNA copies/mL. The assay displayed a dynamic detection range from 500 to 10⁷ copies/mL with <5% inter-assay variation. External World Health Organization (WHO) reference standards were included for quality assurance.

A prespecified subgroup analysis evaluated patients with plasma or serum EBV DNA loads when these samples were available. When plasma or serum samples were unavailable, whole-blood EBV DNA values were used.

Immunophenotyping

Flow cytometry was conducted on ethylenediaminetetraacetic acid (EDTA)-anticoagulated peripheral blood within 24 hours of collection. Red blood cells were lysed, followed by staining with antibodies including CD3, CD4, CD8, CD19, CD16, CD56, CD45, CD38, and human leukocyte antigen (HLA)-DR. Acquisition was performed on a BD FACSCanto II cytometer.

T lymphocytes were identified as CD3+; subsets included CD3+CD4+ and CD3+CD8+ T cells. B cells were defined as CD19+CD3–, while NK cells were defined as CD3–CD16+CD56+. Activation markers included HLA-DR and CD38 expression. Absolute lymphocyte counts were calculated using dual-platform methodology combined with automated haematology results.

Clinical chemistry

Biochemical analyses were performed using Cobas 8000 systems (Roche Diagnostics). Alanine aminotransferase (ALT) and aspartate aminotransferase (AST) were assessed using enzymatic assays; lactate dehydrogenase via ultraviolet (UV) enzymatic method; gamma-glutamyl transferase via colourimetry; total and direct bilirubin by diazo method. C-reactive protein was quantified using high-sensitivity immunoturbidimetric assays, and immunoglobulin levels were evaluated via nephelometry.

Treatment allocation

Antiviral exposure was defined as administration of acyclovir or ganciclovir during hospitalization after admission. Patients were categorized into three mutually exclusive groups: no antiviral therapy, acyclovir therapy, and ganciclovir therapy. Antiviral initiation time, cumulative dose, administration route, and treatment duration were extracted from medication records. Patients receiving both acyclovir and ganciclovir sequentially or concurrently were excluded from the primary comparative analysis and analyzed separately in sensitivity analyses. Antiviral initiation was not assigned by protocol and was not determined by ALT alone. In routine practice, attending physicians considered the overall clinical picture, including hepatobiliary involvement, persistent fever, airway risk, age, hydration status, and perceived disease severity. Because liver enzyme elevation could influence clinical decisions, baseline ALT and related biomarkers were treated as descriptive severity variables rather than as independent evidence of treatment-group differences.

Consequently, treatment selection was severity-based and non-randomized, creating confounding by indication. The statistical analysis therefore emphasized adjusted associations and avoided interpreting between-group liver biomarker differences as antiviral effects.

Patients in the acyclovir group received 10 mg/kg every 8 hours intravenously. Patients in the ganciclovir group received 5 mg/kg every 12 hours intravenously. The median antiviral treatment duration was 5 days [interquartile range (IQR), 3–7 days] for acyclovir and 6 days (IQR, 4–8 days) for ganciclovir. Dose adjustments were made according to renal function. Monitoring included serum creatinine, hepatic enzymes, hematologic indices, and neurologic status. Documented adverse events of interest included renal dysfunction, neutropenia, thrombocytopenia, neurologic symptoms, and antiviral discontinuation due to suspected toxicity. Standard supportive care was provided to all patients, including antipyretics, hydration, gradual return to activity, and monitoring for complications.

Primary outcomes

Primary endpoints included fever duration and hospital length of stay. Fever duration was defined as days from admission to the first afebrile interval of ≥48 hours, with temperature measured every 4 hours during hospitalization. Fever recurrence was defined as temperature ≥38.0 ℃ after the initial 48-hour afebrile period. Hospital length of stay was measured from admission to discharge after clinical criteria were met; transfers to intensive care unit (ICU) or other hospitals were censored at transfer date and analyzed as competing risks. Length of stay reflects both disease severity and local discharge practices and should not be interpreted as a direct measure of antiviral effectiveness.

Secondary outcomes

Secondary endpoints included serial immunophenotyping trajectories, time-dependent EBV DNA kinetics, time to at least 1-log viral decline, time to ALT improvement, and documented adverse events during hospitalization.

Activation markers (CD38 and HLA-DR) were characterized longitudinally on days 1, 3, 5, and 7 when available. Viral clearance rate was calculated as the decline in log₁₀ EBV-DNA per day. The time to normalization of hepatic transaminases and bilirubin was evaluated.

A fever area under the curve (AUC) metric was constructed by integrating temperature measurements over time to capture cumulative febrile burden.

Time origin and time-varying exposure

To reduce immortal-time and late-start biases, the primary time origin was hospital admission. Antiviral therapy was modeled as a time-varying exposure in Cox-based analyses.

Causal-inference framework

Because treatment assignment was not randomized, confounding by indication was addressed using propensity-score inverse probability of treatment weighting (IPTW) (16,17). Propensity scores were estimated using multinomial logistic regression including prespecified baseline variables associated with antiviral decision-making, including age, sex, duration of fever before admission, pharyngeal/tonsillar involvement, airway-risk indicators, hydration status, PRESS score, AST, ALT, bilirubin, C-reactive protein (CRP), platelet count, coagulation-related parameters when available, log10 EBV DNA load, CD4/CD8 ratio, NK-cell proportion, and baseline immunophenotypic markers. ALT was not used to define exposure groups in the analytic model; when liver injury was evaluated, it was treated as a baseline severity covariate or sensitivity variable. Stabilized inverse probability treatment weights were calculated from multinomial propensity scores. Extreme weights were truncated at the 1st and 99th percentiles to reduce the influence of outliers. Covariate balance before and after weighting was assessed using standardized mean differences (SMDs), with SMD <0.10 considered indicative of adequate balance. Effective sample size after weighting was calculated using the formula:

$$N_{\text{effective}}=\frac{({\sum }_i{w}_i{)}^2}{{\sum }_i{w}_i^2}$$ [1]

Time-to-event analyses

Weighted Cox proportional hazards models and restricted mean survival time (RMST) at 10 days were used to compare time to fever resolution and hospital discharge among exposure groups (18).

Outcome coding

Binary clinical outcomes were early defervescence within 5 days and short hospital stay within 9 days. Outcomes were coded so that odds ratios below 1 represented a lower likelihood of the favorable early outcome. The proportional hazards assumption was evaluated using Schoenfeld residual tests and graphical inspection of log-minus-log survival plots. No significant violation of the proportional hazards assumption was identified (all P>0.05). Patients were censored at discharge or last documented follow-up. Because discharge decisions may depend on clinical recovery, RMST analyses were additionally performed to reduce reliance on the proportional hazards assumption.

Model specification and validation

Multivariable models included predictors specified a priori. Continuous variables were scaled when needed to avoid clinically meaningless odds-ratio rounding, including ALT per 10-U/L increments and NK cells per 100-cell/μL increments. Odds ratios, 95% confidence intervals (CIs), model intercepts, and performance metrics were reported. Discrimination was assessed using receiver-operating-characteristic AUC with 1,000 bootstrap-derived CIs, and calibration was assessed with calibration curves and Brier scores.

The index time was defined as hospital admission. Because antiviral initiation occurred after admission and could introduce immortal time bias, antiviral exposure was modeled as a time-dependent variable in Cox proportional hazards analyses. Patients contributed person-time to the untreated state until antiviral initiation and subsequently contributed time to the corresponding treatment exposure group.

Missing data

Missing values occurred in less than 5% of key laboratory parameters and were addressed using multiple imputation by chained equations to generate ten datasets. Sensitivity analyses comparing complete-case and imputed-data models showed consistent direction and magnitude of estimates.

Sensitivity analyses were performed to evaluate robustness of the findings:

1. Restricting analyses to patients with complete laboratory-confirmed EBV infection;

2. Excluding patients with any potential immune impairment;

3. Excluding patients receiving corticosteroids or immunomodulatory therapies;

4. Restricting antiviral exposure to treatments initiated within the first 48 hours after admission;

5. Excluding patients receiving combined or sequential antiviral regimens;

6. Comparing conventional multivariable regression, IPTW-adjusted models, and propensity-score–matched analyses. For the propensity-score-matched sensitivity analysis, patients receiving antiviral therapy were matched to patients receiving no antiviral therapy using nearest-neighbor matching based on the prespecified baseline propensity-score covariates, with a caliper width of 0.2 standard deviations of the logit of the propensity score. Covariate balance after matching was evaluated using SMDs, with an SMD <0.10 indicating adequate balance.

Consistency of effect estimates across these analyses was interpreted as evidence of robustness.

Statistical analysis

All statistical analyses were performed using R version 4.3.0. Two-sided P values <0.05 were considered statistically significant. Baseline comparisons were interpreted as descriptive because exposure selection was non-randomized and severity-based, consistent with recommendations to avoid circular or double-dip interpretation of variables that may also inform clinical decisions (19).

Results

Patient characteristics and cohort construction

A total of 1,312 children with EBV-associated mononucleosis-like illness were screened between January 1, 2019, and December 31, 2023. Following the initial exclusion of 193 patients—comprising those with incomplete medical records (n=97), mixed antiviral regimens or other antivirals (n=53), and missing key baseline data (n=43)—1,119 eligible patients were identified. After further excluding 60 patients lost to follow-up before outcome ascertainment, 1,059 children were successfully enrolled in the final analytic cohort (Figure 1). Among these patients, 371 received supportive care without antiviral therapy, 526 received acyclovir, and 162 received ganciclovir.

Before weighting, antiviral-treated children showed greater baseline disease severity. Compared with the non-antiviral group, children receiving acyclovir or ganciclovir had higher ALT levels (median 61.7 and 58.9 vs. 42.4 U/L, P<0.001), higher AST levels (60.6 and 64.4 vs. 50.3 U/L, P<0.001), greater CD4/CD8 ratio inversion (median 0.20 and 0.21 vs. 0.27, P<0.001), and higher EBV DNA load (419.0 and 313.0 vs. 303.0 ×10³ copies/mL, P=0.01) (Table 1). The distributions of key clinical, immunological, and virological parameters across treatment groups are illustrated in Figure 2.

Table 1. Baseline demographic, clinical, laboratory, immunological, and virological characteristics before IPTW weighting.

Characteristic	No antiviral (n=371)	Acyclovir (n=526)	Ganciclovir (n=162)	P value†
Demographics
Age, years	4.3 (2.8–6.4)	5.0 (3.5–6.8)	4.8 (3.3–7.1)	0.08
Sex, male	208 (56.1)	298 (56.7)	81 (50.0)	0.32
Hematological parameters
White blood cells, ×109/L	14.2 (11.5–17.8)	14.5 (11.4–19.1)	13.4 (11.0–17.9)	0.14
Hemoglobin, g/L	120.0 (114.0–125.0)	119.0 (112.0–125.0)	121.0 (115.0–126.0)	0.27
Platelets, ×109/L	215.0 (181.0–256.0)	189.0 (154.0–229.0)	201.0 (154.0–245.0)	<0.001
Biochemical parameters
C-reactive protein, mg/L	8.4 (3.7–17.0)	7.1 (3.5–14.9)	7.4 (3.3–14.0)	0.09
Alanine aminotransferase, U/L	42.4 (22.7–102.1)	61.7 (29.3–129.9)	58.9 (28.7–143.2)	<0.001
Aspartate aminotransferase, U/L	50.3 (38.2–76.4)	60.6 (42.2–103.5)	64.4 (42.5–109.6)	<0.001
Lactate dehydrogenase, U/L	532.0 (446.0–606.0)	546.0 (464.0–641.0)	561.0 (468.0–661.0)	0.02
γ-glutamyl transferase, U/L	16.6 (11.0–41.6)	21.4 (11.6–63.1)	22.6 (12.0–71.1)	<0.001
Immunological parameters
Immunoglobulin M, g/L	1.64 (1.34–2.06)	1.69 (1.30–2.14)	1.71 (1.38–2.22)	0.43
Immunoglobulin G, g/L	11.5 (9.9–13.3)	11.7 (10.1–13.8)	11.8 (10.0–13.4)	0.57
CD4+/CD8+ T cell ratio	0.27 (0.20–0.40)	0.20 (0.18–0.30)	0.21 (0.17–0.32)	<0.001
NK cells, cells/μL	80.1 (51.5–118.0)	73.6 (50.7–114.1)	75.7 (53.4–121.7)	0.09
Virological parameter
EBV DNA load, ×103 copies/mL	303.0 (90.4–966.0)	419.0 (130.0–1538.0)	313.0 (82.3–1395.0)	0.01

Data are presented as median (IQR) or n (%). ^†, P values represent descriptive comparisons before weighting and should not be interpreted as causal treatment effects because antiviral allocation was non-randomized and severity-driven. EBV, Epstein-Barr virus; IPTW, inverse probability of treatment weighting; IQR, interquartile range; NK, natural killer.

Figure 2.

Composite baseline violin plots. Distribution of key immunological and clinical parameters across treatment groups. Violin plots show the distribution of age, CD4/CD8 ratio, NK cell percentage, EBV DNA load (log10), ALT, CRP, WBC, and platelet count for patients receiving no antiviral therapy, acyclovir, or ganciclovir. These baseline distributions describe illness severity and are not interpreted as treatment effects. ALT, alanine aminotransferase; CRP, C-reactive protein; EBV, Epstein-Barr virus; NK, natural killer; WBC, white blood cell count.

Baseline severity and covariate balance

Stabilized IPTW substantially improved covariate balance across treatment groups. The effective sample sizes after weighting were 365.8, 517.6, and 158.4 for no antiviral therapy, acyclovir, and ganciclovir groups, respectively. All prespecified covariates achieved SMD values below 0.10 after weighting (Table 2). Residual confounding from unmeasured clinical decision factors could not be completely excluded.

Table 2. Covariate balance before and after inverse probability of treatment weighting.

Variable	Before weighting SMD	After weighting SMD
Age	0.09	0.03
Sex, male	0.07	0.02
Fever duration before admission	0.21	0.05
ALT	0.24	0.05
AST	0.27	0.06
CRP	0.08	0.03
Platelet count	0.22	0.04
EBV DNA load	0.18	0.04
CD4/CD8 ratio	0.31	0.06
NK-cell proportion	0.07	0.02
Airway-risk indicator	0.19	0.05

Effective sample sizes after weighting were 365.8, 517.6, and 158.4 for the no-antiviral, acyclovir, and ganciclovir groups, respectively. All covariates achieved adequate balance after IPTW (SMD <0.10). ALT, alanine aminotransferase; AST, aspartate aminotransferase; CRP, C-reactive protein; EBV, Epstein-Barr virus; IPTW, inverse probability of treatment weighting; NK, natural killer; SMD, standardized mean difference.

Weighted cohort characteristics and balance diagnostics are presented in Table S1.

Clinical outcomes before adjustment

Before adjustment, fever duration differed significantly among treatment groups. Median fever duration was 5.0 days (IQR, 3.0–6.0 days) in the no-antiviral group, 7.0 days (IQR, 5.0–9.0 days) in the acyclovir group, and 5.0 days (IQR, 3.0–7.8 days) in the ganciclovir group (P<0.001). Hospital length of stay was 8.0, 10.0, and 13.0 days, respectively (P<0.001). These unadjusted differences were interpreted as reflecting baseline disease severity rather than treatment effects (Table 3).

Table 3. Clinical outcomes by treatment group.

Outcome	No antiviral (n=371)	Acyclovir (n=526)	Ganciclovir (n=162)	P value	q†
Fever duration, days	5.0 (3.0–6.0)	7.0 (5.0–9.0)	5.0 (3.0–7.8)	<0.001	No-ACV <0.001; no-GCV 0.72; ACV-GCV 0.015
Hospital length of stay, days	8.0 (7.0–9.0)	10.0 (9.0–12.0)	13.0 (11.0–15.0)	<0.001	No-ACV <0.001; no-GCV <0.001; ACV-GCV <0.001

Data are presented as median (interquartile range). ^†, P values were calculated using the Kruskal-Wallis test. Pairwise q values were derived using Dunn-Bonferroni correction. Outcome differences are descriptive and should not be interpreted as causal treatment effects. ACV, acyclovir; GCV, ganciclovir.

IPTW-adjusted time-to-event analyses

After IPTW adjustment, differences in fever resolution were attenuated. Median time to defervescence was 5.2 days (95% CI, 4.9–5.5) for no antiviral therapy, 6.1 days (95% CI, 5.7–6.5) for acyclovir, and 5.6 days (95% CI, 5.0–6.2) for ganciclovir. Kaplan-Meier curves for time to fever resolution are shown in Figure 3.

Figure 3.

Kaplan-Meier curves for fever resolution. Time to fever resolution stratified by treatment group. Curves show earlier fever resolution in the no-antiviral group than in the acyclovir and ganciclovir groups, reflecting lower baseline severity. Log-rank P<0.001.

IPTW-weighted Cox proportional hazards models were subsequently performed to compare time to fever resolution and hospital discharge between treatment groups. The proportional hazards assumption was assessed using Schoenfeld residual tests, and no significant violation was observed (all P>0.05). The results of the weighted Cox models are summarized in Table 4. Kaplan-Meier curves for time to hospital discharge are shown in Figure 4.

Table 4. IPTW-weighted Cox regression analyses for time-to-event outcomes.

Outcome	Comparison	HR (95% CI)	P
Fever resolution	Acyclovir vs. no antiviral	0.78 (0.69–0.89)	<0.001
Fever resolution	Ganciclovir vs. no antiviral	0.91 (0.76–1.09)	0.31
Hospital discharge	Acyclovir vs. no antiviral	0.82 (0.71–0.94)	0.004
Hospital discharge	Ganciclovir vs. no antiviral	0.74 (0.60–0.92)	0.006

PH assumption: Schoenfeld residual testing showed no violation of the proportional hazards assumption (all P>0.05). CI, confidence interval; HR, hazard ratio; IPTW, inverse probability of treatment weighting; PH, proportional hazard.

Figure 4.

Kaplan-Meier curves for hospital discharge. Time to hospital discharge stratified by treatment group. Patients in the no-antiviral group had shorter hospital stays than those receiving acyclovir or ganciclovir, consistent with lower baseline severity and severity-based antiviral selection. Log-rank P<0.001.

Regression models

Adjusted logistic regression models showed lower odds of early defervescence among children receiving acyclovir [adjusted odds ratio (OR) 0.34, 95% CI: 0.25–0.45, P<0.001], whereas ganciclovir was not significantly associated with early fever resolution (adjusted OR 0.84, 95% CI: 0.56–1.25, P=0.39). The prediction model achieved an AUC of 0.69 (95% CI: 0.65–0.73), exceeding EBV DNA alone (AUC 0.52) and CD4/CD8 ratio alone (AUC 0.59). The comparative discrimination of these predictive models is presented in Figure 5.

Figure 5.

ROC curves for the prediction of fever ≤5 days. Receiver operating characteristic curves comparing the predictive performance of different models for early fever resolution within 5 days. The multivariable model incorporating clinical and immunological parameters (AUC 0.69) demonstrated better discrimination than EBV DNA load alone (AUC 0.52) or CD4/CD8 ratio alone (AUC 0.59). The diagonal line represents random chance (AUC 0.50). AUC, area under the curve; EBV, Epstein-Barr virus; ROC, receiver operating characteristic.

Detailed multivariable logistic regression results for early fever resolution within 5 days are presented in Table 5. After adjustment for age, EBV DNA load, CD4/CD8 ratio, NK-cell count, and baseline ALT level, acyclovir exposure remained associated with a lower probability of fever resolution within 5 days (adjusted OR 0.34, 95% CI: 0.25–0.45, P<0.001), whereas ganciclovir exposure was not significantly associated with early defervescence (adjusted OR 0.84, 95% CI: 0.56–1.25, P=0.39).

Table 5. Multivariable logistic regression analysis of factors associated with fever resolution within 5 days.

Variable	Adjusted OR (95% CI)	P value
Treatment group†
No antiviral	1.00 (reference)	–
Acyclovir	0.34 (0.25–0.45)	<0.001
Ganciclovir	0.84 (0.56–1.25)	0.39
Age, per year increase	0.88 (0.83–0.93)	<0.001
Log10 EBV DNA load	0.98 (0.83–1.16)	0.81
CD4+/CD8+ ratio	1.76 (0.81–3.84)	0.15
NK cells, per 100-cell/μL increase	1.01 (0.99–1.03)	0.16
Alanine aminotransferase, per 10-U/L increase	1.00 (0.990–1.010)	0.86

Model intercept: 3.55 (95% CI, 1.72–7.34; P<0.001). ^†, reference category = no antiviral therapy. Continuous variables were scaled as specified in the table to avoid rounding artifacts. CI, confidence interval; EBV, Epstein-Barr virus; NK, natural killer; OR, odds ratio.

In addition, a multivariable logistic regression model was constructed to identify factors associated with shorter hospitalization (defined as hospital stay ≤9 days). After adjustment for treatment exposure, age, EBV DNA load, CD4/CD8 ratio, NK-cell count, and baseline ALT level, antiviral exposure remained inversely associated with the probability of short hospitalization. Specifically, children receiving acyclovir (adjusted OR 0.12, 95% CI: 0.08–0.17, P<0.001) and ganciclovir (adjusted OR 0.02, 95% CI: 0.01–0.04, P<0.001) were less likely to achieve hospitalization within 9 days compared with children receiving supportive care alone. These findings were interpreted as reflecting severity-based antiviral selection rather than a causal effect of antiviral therapy on hospitalization duration (Table 6).

Table 6. Multivariable logistic regression analysis of factors associated with hospital stay ≤9 days.

Variable	Adjusted OR (95% CI)	P value
Treatment group†
No antiviral	1.00 (reference)	–
Acyclovir	0.12 (0.08–0.17)	<0.001
Ganciclovir	0.02 (0.01–0.04)	<0.001
Age, per year increase	1.02 (0.96–1.08)	0.56
Log10 EBV DNA load	0.76 (0.62–0.92)	0.005
CD4+/CD8+ ratio	1.04 (0.43–2.55)	0.93
NK cells, per 100-cell/μL increase	1.01 (0.99–1.03)	0.51
Alanine aminotransferase, per 10-U/L increase	0.99 (0.99–1.01)	<0.001

Model intercept: 15.45 (95% CI, 6.59–36.22; P<0.001). ^†, reference category = no antiviral therapy. NK cells were scaled per 100-cell/μL increase and ALT per 10-U/L increase to avoid rounding artifacts. ALT, alanine aminotransferase; CI, confidence interval; EBV, Epstein-Barr virus; NK, natural killer; OR, odds ratio.

Secondary outcomes

The median time to ALT improvement was 6, 7, and 7 days for no antiviral, acyclovir, and ganciclovir groups, respectively (P=0.02). After IPTW adjustment, neither acyclovir nor ganciclovir was associated with faster ALT recovery.

Sensitivity analyses

Sensitivity analyses yielded effect estimates that were directionally consistent with the primary findings. Restricting the cohort to children with complete laboratory-confirmed acute EBV infection, excluding patients with potential immune impairment, excluding corticosteroid or immunomodulatory exposure, restricting antiviral initiation to within 48 hours after admission, and excluding combined or sequential antiviral regimens did not materially alter the observed associations. In addition, conventional multivariable regression, IPTW-adjusted analyses, and propensity-score–matched analyses produced broadly comparable effect estimates. Across these analyses, acyclovir remained associated with a lower probability of early defervescence relative to supportive care, whereas the association for ganciclovir was consistently weaker and generally not statistically significant. Findings for short hospital stay similarly remained directionally consistent across adjustment strategies. Detailed sensitivity-analysis results are presented in Table S2.

Discussion

Young hospitalized children with EBV-associated mononucleosis-like illness differ from classical adolescent IM in several important ways. Age influences both clinical expression and diagnostic ascertainment. Primary infection in early childhood is often asymptomatic or nonspecific, heterophile antibody responses may be less reliable, and hospitalization in pediatric settings frequently reflects prolonged fever, hepatobiliary involvement, airway concerns, or local evaluation pathways rather than the classic triad of fever, pharyngitis, and lymphadenopathy (1,2,14,15). For these reasons, this cohort should be interpreted as representing pediatric EBV-associated mononucleosis-like illness rather than classical adolescent IM.

The rationale for examining antiviral exposure was not that acyclovir or ganciclovir is expected to improve uncomplicated EBV infection. Current evidence and guidance emphasize supportive care and do not recommend routine antiviral therapy for immunocompetent patients with EBV-associated IM (2,10-12). The clinical question addressed here is narrower: when antivirals are used in real-world hospitalized children selected by perceived severity, what immunophenotypic and outcome patterns are observed?

The biological limitations of antiviral therapy help frame the findings. Nucleoside analogues act mainly during lytic viral replication, whereas the clinical manifestations of acute EBV infection are largely driven by host immune activation, cytokine-mediated inflammation, and tissue infiltration by EBV-specific cytotoxic lymphocytes and NK cells (3-9). This distinction is clinically relevant: elevated transaminases and systemic symptoms should not be interpreted simply as direct lytic viral injury in hepatocytes. Although acyclovir and ganciclovir can inhibit EBV DNA polymerase in vitro, they are unlikely to reverse an already established host immune cascade. This pathophysiologic reality helps explain why higher transaminases and prolonged symptoms may lead clinicians to prescribe antivirals while not necessarily identifying children likely to benefit.

The longer fever duration observed in the acyclovir group therefore should not be interpreted as evidence of antiviral related harm. The more plausible explanation is confounding by indication. Children with persistent fever, greater hepatic involvement, airway concerns, or more pronounced immune activation were more likely to receive antiviral therapy. Although IPTW and time-varying exposure analyses reduced measurable baseline imbalance, residual differences in clinical severity may still have influenced treatment allocation and outcomes (3-9,20-24).

The immunophenotypic findings support this interpretation. The cohort showed severe CD4/CD8 inversion, with median ratios of 0.20 to 0.27, along with broad NK cell expansion. The more prominent inversion in antiviral treated children likely reflected a stronger cellular immune response at baseline rather than an antiviral responsive state. In clinical terms, an exaggerated host immune response, not viral load alone, appears to be a major driver of persistent fever and longer hospitalization (6,20,21,25).

The ROC analysis adds practical value beyond statistical description. A model combining age, immune indicators, and clinical characteristics outperformed EBV DNA alone. This suggests that pediatric risk assessment should not rely solely on viral load thresholds. Instead, early clinical and immunologic risk stratification may help identify children who require closer supportive monitoring while reducing unnecessary escalation in lower risk children.

These findings align with domestic and international guidance that discourages routine antiviral therapy for immunocompetent patients with EBV associated IM (10-12,20). At the same time, they highlight a gap between guidelines and bedside practice. Clinicians may prescribe antivirals out of concern for liver injury or prolonged symptoms, but this empiric approach may increase intravenous therapy, toxicity monitoring, hospitalization time, and resource use without clear clinical benefit. A stricter antiviral stewardship pathway for pediatric EBV infection could help standardize care and preserve hospital resources.

Nevertheless, these findings should not be taken as evidence that antiviral therapy is ineffective in all circumstances or in every high-risk subgroup. Within this immunocompetent hospitalized cohort, routine antiviral exposure was not associated with measurable short-term clinical improvement after adjustment for available confounders. Future studies should focus on identifying specific patient subsets, such as children with severe airway compromise, progressive organ involvement, or impaired immune function, who might derive potential benefit from targeted antiviral intervention.

Several limitations deserve consideration. First, the retrospective single-center design and non-randomized treatment allocation inherently limit causal inference. Although IPTW, time-varying exposure modeling, RMST analyses, and multiple sensitivity analyses were applied, unmeasured confounding by indication cannot be completely excluded. Clinical factors influencing antiviral prescription, such as physician perception of disease trajectory, subtle airway compromise, or changes in clinical status after admission, may not have been fully captured. Second, antiviral initiation timing, treatment duration, and dose adjustments were determined by routine clinical practice rather than a standardized protocol, introducing treatment heterogeneity. Third, although immunophenotyping and virological measurements provided detailed biological characterization, missing longitudinal measurements in some patients limited evaluation of dynamic immune and viral changes. Fourth, the single-center design and pediatric population may restrict generalizability to other healthcare settings or adult EBV infection. Although time-varying exposure modeling, IPTW, and RMST analyses were used to reduce bias (16-18), residual confounding by indication remains possible, and liver enzyme differences should be interpreted only as baseline severity markers rather than as evidence of treatment effects (19). Retrospective safety ascertainment also cannot provide a definitive comparative safety assessment. Future prospective multicenter studies should use standardized pediatric case definitions, prespecified severity criteria, systematic adverse event capture, and transparent causal inference frameworks.

Conclusions

In this large real-world retrospective cohort of children hospitalized with EBV-associated mononucleosis-like illness, antiviral therapy was commonly used in children with greater baseline severity, including hepatic involvement and marked immune activation, but it was not associated with improved clinical recovery. Severe CD4/CD8 inversion and the multivariable clinical-immunologic model provided more informative risk stratification than viral load alone. These findings do not support routine antiviral treatment for immunocompetent children with EBV-associated mononucleosis-like illness in the absence of specific clinical indications. Clinical practice should emphasize supportive care, early identification of high-risk children using combined clinical and immune indicators, and judicious antiviral stewardship to avoid unnecessary treatment while preserving individualized decision-making for children with severe or complicated presentations.

Supplementary

The article’s supplementary files as

Ethical Statement: The authors are accountable for all aspects of the work in ensuring that questions related to the accuracy or integrity of any part of the work are appropriately investigated and resolved. The study was conducted in accordance with the Declaration of Helsinki and its subsequent amendments. This study was approved by the Ethics Committee of Children’s Hospital of Soochow University (No. 2025CS175). Because only de-identified retrospective data were analyzed and no patient contact or research-directed treatment occurred, the requirement for informed consent was waived by the Ethics Committee.

Data Sharing Statement

Available at https://tp.amegroups.com/article/view/10.21037/tp-2026-0517/dss

tp-15-9-384-dss.pdf ^{(80.7KB, pdf)}

DOI: 10.21037/tp-2026-0517