The rationale for immunomodulation is biologically plausible. In herpes simplex virus encephalitis, much of the neurological damage is attributed not only to direct viral cytolysis but also to the host immune response, including cytokine release, edema, and secondary inflammatory cascades that can extend injury well beyond infected neurons. Corticosteroids, by suppressing this inflammation, could in theory reduce swelling, limit tissue damage, and improve outcomes. The idea is not new; discussions of dexamethasone in herpes simplex encephalitis appeared in the medical literature as early as the 1970s, and the approach has been incorporated into some clinical guidelines as an option for patients with life-threatening cerebral edema. Yet the clinical evidence base has remained fragmented, small, and methodologically uneven.

To resolve the uncertainty, Zhang and colleagues searched PubMed/MEDLINE, Embase, and the Cochrane Central Register of Controlled Trials from their inception through 4 July 2026, and also examined trial registries and reference lists of relevant publications. They included randomized controlled trials and comparative observational studies that evaluated systemic corticosteroids added to background antiviral therapy against the same antiviral therapy alone or with placebo. Acyclovir was the background antiviral agent in every study that specified the regimen, ensuring that the comparison isolated the effect of the steroid rather than differences in antiviral care. The co-primary outcomes were study-defined unfavorable global neurological or functional outcome and all-cause mortality, with secondary endpoints covering serious adverse events, seizures during follow-up, cognition, persistent herpes simplex virus DNA in cerebrospinal fluid, relapse, and functional independence measured by the Barthel Index.

The yield from more than five decades of literature was strikingly thin. Only four comparative studies, encompassing 209 patients, met the inclusion criteria: two randomized trials and two retrospective studies. The randomized trials were pooled using Mantel-Haenszel random-effects methods for risk ratios, while mean differences were combined with inverse-variance random-effects approaches. Because non-randomized treatment allocation introduces substantial clinical heterogeneity and confounding by indication, the observational evidence was analyzed and interpreted separately from the randomized data, a decision that reflects contemporary standards for evidence synthesis.

The pooled results from the randomized trials were uniformly inconclusive. For unfavorable global outcome, the pooled risk ratio was 1.00 with a 95 percent confidence interval of 0.68 to 1.47, a point estimate of exactly no effect flanked by intervals wide enough to accommodate both meaningful benefit and meaningful harm. Mortality showed a risk ratio of 0.92 (95 percent confidence interval 0.35 to 2.40), serious adverse events a risk ratio of 1.14 (95 percent confidence interval 0.54 to 2.44), and seizures during follow-up a risk ratio of 0.73 (95 percent confidence interval 0.33 to 1.62). None of these estimates approached conventional thresholds for statistical significance, and all were imprecise, reflecting the small number of patients contributing to each analysis.

Functional and cognitive endpoints fared no better. Barthel Index scores, a standard measure of activities of daily living, showed a mean difference of 3.05 points (95 percent confidence interval, minus 6.99 to 13.09) at approximately six months and 0.30 points (95 percent confidence interval, minus 13.84 to 14.45) at discharge or day 30, differences that are clinically negligible and statistically indistinguishable from zero. Cognitive findings across studies were not consistently favorable. In short, on every endpoint the meta-analysis examined, the randomized evidence could neither confirm benefit nor exclude harm, a pattern the authors attribute to the limited and imprecise nature of the underlying data rather than to demonstrated equivalence.

The observational studies added little clarity and, in some respects, raised concerns. Their estimates conflicted with one another and were judged to be seriously compromised by confounding, since clinicians tend to give steroids to the sickest patients, biasing any naive comparison. The randomized trials did, however, generate two signals worth noting. In the DexEnceph trial, a multicenter, observer-blind, phase 3 study of adjunct dexamethasone in adults with herpes simplex virus encephalitis conducted in the United Kingdom, herpes simplex virus DNA remained detectable in cerebrospinal fluid at approximately day 14 in 4 of 36 dexamethasone recipients compared with 9 of 43 controls, a numerical difference of uncertain significance. More strikingly, five relapses were reported among dexamethasone-treated patients and none among controls, although the authors caution that so few events do not permit a causal conclusion. The possibility that corticosteroid-induced immunosuppression might prolong viral clearance or provoke relapse has long been a theoretical worry, and these findings keep it alive without settling it.

The meta-analysis also highlights a critical gap. The one scenario in which corticosteroids are most often deployed in practice, life-threatening cerebral edema with impending herniation, was never specifically evaluated in any included study. The evidence is therefore insufficient to determine whether steroids help or harm this selected subgroup, leaving clinicians to make high-stakes decisions on physiological reasoning alone. The authors call for future adequately powered, multicenter randomized trials that use standardized treatment protocols and prespecify severity- and edema-defined subgroups, so that the question of immunomodulation can be answered separately for the broad population and for the narrow emergency indication.

For now, the practical message is one of restraint. With roughly 209 randomized and comparative patients studied across half a century, the field has generated data too sparse to justify routine adjunctive corticosteroids, yet too imprecise to close the door entirely. Clinicians managing herpes simplex virus encephalitis should continue to prioritize early acyclovir, the intervention with proven mortality benefit, and reserve corticosteroids for circumstances where cerebral edema threatens immediate deterioration, ideally within the framework of clinical trials or prospective registries. The new synthesis in the Journal of Neurology does not declare steroids useless; it declares the question still open, and in doing so it sets a clear agenda for the trials the field has, until now, failed to deliver.

Subject of Research: Adjunctive corticosteroid therapy for herpes simplex virus encephalitis

Article Title: Adjunctive corticosteroids in herpes simplex virus encephalitis: a meta-analysis of randomized and observational studies

Article References: Zhang, D., Yang, M., Huang, Y., Liang, M., Yao, L., Liang, J., Deng, M., & Shen, R. (2026). Adjunctive corticosteroids in herpes simplex virus encephalitis: a meta-analysis of randomized and observational studies. Journal of Neurology, 273(10), Article 567. https://doi.org/10.1007/s00415-026-14128-x

Image Credits: AI Generated

DOI: 10.1007/s00415-026-14128-x

Keywords: herpes simplex virus encephalitis, corticosteroids, dexamethasone, acyclovir, meta-analysis, neurological outcome, cerebral edema, randomized controlled trials, Barthel Index, relapse, cerebrospinal fluid HSV DNA, Journal of Neurology