HSV reactivation on bronchoalveolar lavage sequencing does not prove HSV pneumonia, and it does not mean the patient needs acyclovir. Across 350 ventilated adults with severe pneumonia in five intensive care units, HSV-1 reactivation reached 35.4 percent in immunocompromised patients and 17.0 percent in immunocompetent ones. Twenty-eight-day mortality was 62.8 percent against 36.6 percent, but only in the immunocompromised group.
Read that last clause twice. In immunocompetent patients the same positive result carried no mortality signal, and the numbers leaned the other way. A finding that predicts death in one population and predicts nothing in another is reporting on the host, not on a virus eating lung tissue.
Now think about 2 a.m., a sequencing report, and a patient doing worse. The reflex is acyclovir: the virus has a name and the drug exists. The better move is three questions. Which population is this patient in? Does the signal clear the quantitative threshold? Is the problem the virus, or the immune collapse that let it through?
What HSV reactivation means when sequencing defines it
Herpes simplex virus hides in sensory ganglia for life and reactivates whenever host defense sags. In critical illness that happens constantly. Catecholamine stress, corticosteroids, malnutrition, sepsis-related lymphopenia and mechanical ventilation all push the same direction. Find HSV DNA in lavage fluid from a patient like that and the accurate word is reactivation, not pneumonia.
This cohort handled the ambiguity with an explicit definition instead of a clinical impression. Reactivation meant HSV-1 DNA in bronchoalveolar lavage by metagenomic sequencing, at a prespecified threshold of 3.0 viral reads per million or higher. The authors say outright that the definition does not establish HSV pneumonia. That honesty is what makes the dataset usable: a threshold on a sequencing signal is reproducible, a diagnosis of viral pneumonitis is not.
Two consequences follow. HSV is a frequent bystander in ventilated airways, so treating it as a pathogen buys nephrotoxicity and neurotoxicity with no return. And because HSV in the lower respiratory tract already tracks severity, any mortality association has to survive the objection that it is just measuring how sick the patient was on admission. That is the test this study actually runs.
How the five-center ICU cohort was built
Retrospective, multicenter, and narrow on purpose: adults ventilated for severe pneumonia, sampled on a schedule rather than when someone happened to suspect a virus.
Item | Detail |
Design | Retrospective multicenter cohort |
Centers | Five hospitals |
Patients screened | 364 |
Patients analyzed | 350 |
Enrollment window | November 2021 to November 2025 |
Specimen | Bronchoalveolar lavage within 24 hours of ICU admission |
Assay | Metagenomic next-generation sequencing |
Reactivation definition | HSV-1 DNA positive at a viral reads-per-million threshold of 3.0 or higher |
Primary endpoint | 28-day mortality from ICU admission |
Immunocompromised | 144 patients (41.14 percent) |
Immunocompetent | 206 patients (58.86 percent) |
Deaths in the full cohort | 130 |
The immunocompromise definition deserves a look because it decides who the finding applies to. Among the 51 immunocompromised patients with HSV reactivation, 29.41 percent had a hematological disease, 25.49 percent had taken at least 0.3 mg per kg per day of prednisone equivalent for three weeks or more, 17.65 percent had neutropenia or recent chemotherapy for a solid tumor, 17.65 percent were on antirheumatic drugs or biological immunomodulators, and 15.69 percent had received a stem cell or solid organ transplant.
That is a wide net. It mixes marrow suppression with iatrogenic T-cell suppression, which is a genuine weakness and also the reason the result travels beyond a transplant unit into an ordinary medical intensive care unit.
HSV reactivation was twice as common in the immunocompromised group
The prevalence gap is large, consistent, and built on clean denominators.
Group | HSV reactivation | No reactivation | Rate |
Immunocompromised (n = 144) | 51 | 93 | 35.42 percent |
Immunocompetent (n = 206) | 35 | 171 | 16.99 percent |
Whole cohort (n = 350) | 86 | 264 | 24.57 percent |
A 35.4 percent detection rate in immunocompromised ICU patients resets expectations. In a neutropenic or steroid-soaked patient with a worsening infiltrate, HSV positivity is closer to routine than exceptional, and routine findings carry little diagnostic weight. That is exactly the situation where a test loses specificity in practice, no matter how good the assay is.
The two arms were not identical. Immunocompromised patients with HSV reactivation had more chronic pulmonary disease at 29.41 percent against 13.98 percent (P = 0.030) and more liver disease at 17.65 percent against 5.38 percent (P = 0.036). Their C-reactive protein ran higher at 162.2 against 108.28 mg/L (P = 0.028) and hemoglobin lower at 79 against 96 g/L (P = 0.023). Age, sex, diabetes, cardiac disease, renal disease, hematological malignancy, solid tumor and severity scores were balanced.
Those imbalances are the price of an observational design. Chronic lung disease and liver disease both plausibly worsen outcome by themselves, which is why the analysts added them into sensitivity models instead of claiming the arms were comparable.
Mortality separates cleanly by immune status
Here is the core of the paper. The outcomes split along the immune-status line, not the virus line.
Outcome | Immunocompromised without HSV (n = 93) | Immunocompromised with HSV (n = 51) | P value |
28-day mortality | 34 (36.56 percent) | 32 (62.75 percent) | 0.003 |
ICU mortality | 28 (30.11 percent) | 29 (56.86 percent) | 0.002 |
Ventilation days | 8 | 14 | 0.019 |
ICU length of stay in days | 10 | 14 | 0.039 |
Outcome | Immunocompetent without HSV (n = 171) | Immunocompetent with HSV (n = 35) | P value |
28-day mortality | 56 (32.75 percent) | 8 (22.86 percent) | 0.318 |
ICU mortality | 34 (19.88 percent) | 3 (8.57 percent) | 0.148 |
Ventilation days | 11 | 12 | 0.991 |
ICU length of stay in days | 13 | 13 | 0.520 |
Put the panels side by side and the argument writes itself. In immunocompromised patients, HSV reactivation came with a 26-point absolute mortality difference and four extra days on the ventilator. In immunocompetent patients, mortality in the positive arm was numerically lower, ventilation days were identical at 12 against 11, and length of stay did not move at all. Same virus, same assay, two different meanings.
The interaction term carries the finding
Subgroup P values are fragile things. The interaction analysis is the stronger claim, and the one worth quoting.
Model | Measure | Estimate | 95 percent CI | P value |
Unadjusted, immunocompromised | Odds ratio for 28-day mortality | 2.923 | 1.441 to 5.929 | 0.003 |
Unadjusted, immunocompetent | Odds ratio for 28-day mortality | 0.608 | 0.260 to 1.425 | Not significant |
Unadjusted interaction | Ratio of odds ratios | 4.803 | 1.588 to 14.527 | 0.005 |
Adjusted interaction | Ratio of odds ratios | 4.451 | 1.449 to 13.668 | 0.009 |
Adjusted, immunocompromised | Model-derived odds ratio | 2.622 | 1.275 to 5.395 | 0.009 |
Adjusted, immunocompetent | Model-derived odds ratio | 0.589 | 0.249 to 1.394 | 0.228 |
Subgroup model, immunocompromised | Adjusted odds ratio | 2.403 | 1.155 to 5.002 | 0.019 |
The models included HSV status, immune status, their product term, age, severity score and cytomegalovirus co-detection. A ratio of odds ratios of 4.451 says the HSV-mortality association differs by more than fourfold depending on immune status. Translated: these two populations should never have been pooled, and any single overall estimate would have described neither of them.
Sensitivity analyses refused to move the immunocompromised estimate. Adding chronic pulmonary and liver disease gave 2.867 (1.319 to 6.231). Adding C-reactive protein and hemoglobin gave 2.453 (1.088 to 5.531). A center fixed-effect model gave 3.498 (1.543 to 7.931). An estimate that holds across four specifications is not a covariate artifact, though residual confounding by unmeasured severity cannot be ruled out.
What else travels with HSV reactivation
The co-detection profile is where the immune-failure story becomes visible.
Co-detected organism | With HSV reactivation (n = 51) | Without (n = 93) | P value |
Stenotrophomonas maltophilia | 27.45 percent | 11.83 percent | 0.022 |
Candida albicans | 35.29 percent | 16.13 percent | 0.013 |
Human herpesvirus 7 | 15.69 percent | 3.23 percent | 0.017 |
Cytomegalovirus | 37.25 percent | 19.35 percent | 0.028 |
The authors label these comparisons exploratory, uncorrected for multiple testing, and reflective of co-detection rather than adjudicated coinfection. Fair. Read them as a pattern rather than a set of independent findings.
The pattern is coherent anyway. A patient carrying Stenotrophomonas, Candida and a second herpesvirus alongside HSV is a patient whose defenses failed broadly, not one with a single treatable viral infection. The bacterial and fungal names matter more for antibiotic and antifungal review than the viral one matters for antiviral prescribing. That inverts the usual reading order of a positive viral result, and it is the most practical thing in the paper.
Does HSV reactivation justify acyclovir?
Mostly the honest answer is no, and these data cannot say otherwise. No antiviral exposure was recorded. No treated-versus-untreated comparison exists. The study cannot demonstrate benefit, harm or futility. What it can do is sharpen the prior probability that a positive result means disease, and that probability lives almost entirely in the host.
Situation | Reasonable interpretation | Action to consider |
HSV detected in an immunocompromised patient with worsening infiltrates and no alternative explanation | Reactivation marker with substantial mortality association | Complete the workup for bacterial and fungal drivers and discuss antiviral therapy on tissue-level grounds |
HSV detected in an immunocompromised patient who is improving | Likely passenger | Document and continue current management |
HSV detected in an immunocompetent patient | Weak severity marker with no mortality signal in this cohort | Do not treat on the sequencing result alone |
HSV detected with cytomegalovirus in an immunocompromised patient | Highest-risk combination in this dataset | Escalate monitoring and reassess both viruses |
HSV detected at a reads-per-million value below the prespecified threshold | Below the study definition of reactivation | Repeat sampling if suspicion persists |
Acyclovir and its oral prodrug valacyclovir remain the standard agents for herpes simplex disease: cheap, familiar, usually well tolerated. The case against reflexive use is not toxicity in the abstract. It is that a drug started for the wrong reason ends the diagnostic search. Acyclovir does not touch Stenotrophomonas, does not touch Candida, and will not fix an infiltrate driven by either. The prescription can manufacture the feeling that the explanation has been found.
Where treatment is genuinely indicated, timing matters. Acyclovir needs activation by viral thymidine kinase, so efficacy depends on replicating virus with intact kinase activity. Resistance through thymidine kinase mutation is a recognized problem in immunocompromised hosts after repeated or prolonged courses. That is the same population where the threshold to treat should be lowest and the threshold to confirm disease highest.
The counterintuitive result: the signal reverses in immunocompetent patients
The most useful number in this paper may be the one showing nothing. In immunocompetent patients, 28-day mortality was 22.86 percent with HSV reactivation and 32.75 percent without. The confidence interval around the odds ratio of 0.608 ran from 0.260 to 1.425, so nobody can claim protection. What the data do exclude is the claim of harm that a pooled analysis would have manufactured.
Picture the pooled version. With 130 deaths across 350 patients and 86 HSV-positive cases, the crude association would likely have leaned toward harm and been published that way. The interaction analysis explains why that figure would have been an average of two different biological situations, and wrong for both halves.
There is a second inversion. After a positive viral result, instinct says escalate. Here the group with the strongest mortality association is also the group with the heaviest comorbidity burden and the highest baseline risk, so the positive result partly restates what the clinician already knew. A test earns its keep when the patient looks better than the result suggests. That is when positivity is news rather than confirmation.
Limitations worth stating plainly
Retrospective, five centers, and the decision to lavage was made by treating teams, so patients too unstable for bronchoscopy never entered the cohort. The exposure is a sequencing threshold, not a histopathological diagnosis, and it is misclassified in both directions to an unknown degree.
The subgroups are small. Fifty-one HSV-positive immunocompromised patients and 35 HSV-positive immunocompetent patients cannot carry finely adjusted models, and the interaction estimate, significant as it is, has a confidence interval running from 1.449 to 13.668. There is no treatment data, no adjudication of HSV pneumonitis, and no follow-up past 28 days.
Two gaps matter for translation. Steroid dose was never modeled as a continuous variable, so the dose-response evidence that would strengthen a causal reading is missing. And cytomegalovirus cannot be cleanly separated from HSV, because within the HSV-positive group 16 of 19 patients died when both viruses were present against 16 of 32 when only HSV was found. Those numbers come from 19 and 32 patients. They are a hypothesis, not a result.
Bottom line
HSV reactivation on lavage sequencing means whatever the host makes it mean. In immunocompromised ICU patients with severe pneumonia it came with 62.75 percent 28-day mortality against 36.56 percent, an adjusted odds ratio near 2.4, and an interaction term showing the association was more than four times stronger than in immunocompetent patients, where it vanished. Treat the positive result as a prompt for a wider search and a narrower antiviral decision. Reserve acyclovir for patients whose tissue-level evidence supports it.
FAQ
Does HSV reactivation on BALF mean HSV pneumonia?
No. The study defined reactivation strictly as HSV-1 DNA positivity at a viral reads-per-million threshold of 3.0 or higher on metagenomic sequencing, and the authors state that this definition does not establish HSV pneumonia. Establishing viral pneumonitis requires tissue or cytological evidence of viral cytopathic effect alongside compatible imaging and clinical decline. Treating a sequencing threshold as a diagnosis exposes patients to unnecessary antiviral drugs and, more damagingly, closes the search for the bacterial or fungal driver that is actually progressing.
How common is HSV reactivation in severe pneumonia?
In this cohort of 350 ventilated adults, 86 patients met the sequencing definition, or 24.6 percent overall. The rate was not uniform: 35.4 percent among immunocompromised patients and 17.0 percent among immunocompetent ones. Prevalence that high in an intensive care population is the core interpretive problem, because a finding present in one of every three immunosuppressed patients cannot, by itself, carry much diagnostic weight when the pretest probability of disease is far lower.
Why does immune status change the meaning of the result so much?
Because the same virus signal arises from two different host states. In an immunocompromised patient, reactivation reflects a measurable collapse in T-cell control, and that collapse is itself the thing driving mortality, which is why the association is strong. In an immunocompetent patient, transient shedding during critical illness is common and host control is largely intact, so the signal carries little prognostic information. The interaction term of 4.451 quantifies how differently the same exposure behaves.
Should HSV reactivation trigger acyclovir?
Only with tissue-level evidence of disease, and this dataset cannot support a treatment threshold. No antiviral exposure or outcome comparison was recorded, so no benefit or harm estimate exists. Acyclovir and valacyclovir remain appropriate for documented herpes simplex disease, and early initiation matters because acyclovir depends on viral thymidine kinase for activation. Starting a drug on a sequencing result alone risks both drug toxicity and a false sense of diagnostic closure in a patient who needs a broader search.
What does the cytomegalovirus co-detection mean?
It flags the highest-risk subgroup in the dataset. Among 51 immunocompromised patients with HSV reactivation, 28-day mortality was 84.21 percent when cytomegalovirus was also detected and 50.00 percent when it was not, a difference that reached P = 0.018 on an exact test. That comparison rests on 19 and 32 patients respectively, so it is exploratory rather than confirmatory, and it should not be read as evidence of viral synergy.
Which organisms should be looked for after a positive HSV result?
The co-detection profile points at Stenotrophomonas maltophilia, Candida albicans, human herpesvirus 7 and cytomegalovirus, all of which appeared more often in the reactivation arm. Those comparisons were uncorrected for multiple testing and measure co-detection rather than coinfection. The practical reading is that a positive HSV result should trigger review of bacterial and fungal coverage rather than automatic antiviral escalation, because those are the organisms that plausibly explain a deteriorating course.
Does HSV reactivation prolong ventilation and ICU stay?
In immunocompromised patients it was associated with longer ventilation at 14 days against 8, and a longer ICU stay at 14 days against 10, both statistically significant. In immunocompetent patients there was no difference at all, with ventilation at 12 days against 11 and ICU stay at 13 days against 13. The resource implication follows the same immune-status split as the mortality data, which suggests the driver is prolonged critical illness rather than the virus itself.
Can a threshold value separate reactivation from contamination?
It can make the definition reproducible, which is not the same thing as clinically decisive. The prespecified threshold of 3.0 viral reads per million was fixed before analysis, and that discipline is a strength. It does not establish that patients above the line have disease or that patients below it do not. A negative result in a patient with a compatible course should prompt repeat sampling rather than closure, especially where sampling quality or timing was suboptimal.
Why did mortality look lower in immunocompetent patients with HSV?
Almost certainly because the subgroup is small and the comparison is uncontrolled. Eight deaths among 35 HSV-positive immunocompetent patients against 56 among 171 HSV-negative patients produced an odds ratio of 0.608 with a confidence interval from 0.260 to 1.425, which excludes neither harm nor benefit. The value of the observation is negative: it rules out the assumption that HSV detection is uniformly ominous, which a pooled analysis would have wrongly implied.
What are the main limitations of this evidence?
The cohort is retrospective and multi-center, with lavage performed at clinician discretion, so patients too unstable for bronchoscopy were excluded. The exposure is a sequencing threshold rather than a histopathological diagnosis. The subgroups are small, with 51 and 35 HSV-positive patients in the two arms, which limits adjustment. There is no antiviral treatment data, no adjudication of HSV pneumonitis and no follow-up beyond 28 days, so no treatment recommendation can be derived from it.
How should this change empirical management in the ICU?
It should make the positive result a trigger for widening the differential rather than narrowing it. In an immunocompromised patient with a positive lavage, the immediate tasks are to confirm bacterial and fungal coverage is adequate, review corticosteroid exposure, and decide whether tissue-level evidence supports antiviral therapy. In an immunocompetent patient the same result changes little. The single most useful habit is to ask which population the patient belongs to before reading the rest of the report.
Is valacyclovir an alternative to intravenous acyclovir here?
It can be, when a decision to treat has already been made and the patient has reliable enteral absorption. Oral valacyclovir delivers substantially higher acyclovir exposure than oral acyclovir and avoids the line and fluid burden of intravenous therapy. In a ventilated ICU patient with ileus or vasopressor dependence, absorption is unreliable enough that intravenous therapy is usually preferred. None of this resolves the prior question, which is whether the sequencing result justifies treatment at all.