HHV-6 reactivation and its effect on delirium and cognitive functioning in hematopoietic cell transplantation recipients

Active human herpesvirus 6 (HHV-6) infection occurs in approximately 40% of hematopoietic cell transplantation (HCT) recipients.^1-3  Multiple case reports document HCT recipients with HHV-6–associated encephalitis, resulting in poor outcome.^4-7  However, the extent of the effect of HHV-6 on the central nervous system (CNS) after HCT has not been fully characterized on a population basis. Many previous studies of HHV-6 after HCT have been retrospective, included small study populations, and lacked systematic assessment of CNS functioning.

We performed a large prospective study of HHV-6 and CNS dysfunction in HCT recipients, incorporating systematic assessment of CNS function using validated neuropsychiatric and neurocognitive tools.

The protocol was approved by the Fred Hutchinson Cancer Research Center Institutional Review Board.

Patients proficient in English and undergoing allogeneic HCT from January 2005 through August 2008 were included. Consent was presented to 880 patients during the pretransplantation evaluation. Of the 474 preliminarily enrolled, a total of 152 withdrew or were deemed ineligible before contributing data, leaving 322 patients who contributed data (Figure 1). Six patients with HHV-6 DNA levels suggestive of chromosomal integration⁸  (determined a priori as increasing HHV-6 plasma DNA levels during the first 2 weeks after transplantation and persistent levels ≥ 100 copies/mL in ≥ 80% of subsequent plasma samples) were excluded from analyses. One additional patient, who contributed only baseline data, was also excluded. The 315 included patients were representative of the total population of 958 patients receiving allogeneic HCT during the study period with regards to demographic and baseline clinical characteristics (data not shown). Of the 315 included patients, 308 (98%) were followed for 4 weeks or more after transplantation or until death.

Neuropsychiatric assessments for delirium were obtained at approximately the same time each day at baseline, every Monday, Wednesday, and Friday through day 56, and once weekly days 57 to 84 after HCT. Pain was assessed at each delirium assessment. Neuropsychologic testing was administered at baseline and approximately 84 days after HCT. Children younger than 3 years old did not participate in the delirium assessments, and those younger than 5 years did not participate in the neuropsychologic assessments. Trained research personnel administered the neuropsychiatric and neuropsychologic testing, and study personnel and investigators were blinded to patients' HHV-6 results during assessments.

Baseline and twice-weekly plasma specimens were collected through day 84 after HCT for HHV-6 testing. A total of 6255 specimens were obtained (85% of planned). Duration of HHV-6 detection was defined as days from first to last positive without intervening negative tests.

The research protocol did not include brain imaging and cerebrospinal fluid (CSF) testing; when obtained clinically, results were collected for the study.

Demographic, clinical, and laboratory information was collected from clinical records and databases. Receipt of medications known to cause delirium (opioids, anticholinergic drugs, benzodiazepines, and glucocorticoids) and antivirals that might affect HHV-6 activity were collected.

Underlying disease was categorized as “less advanced” (acute myeloid leukemia, acute lymphoblastic leukemia, or lymphoma in first remission, chronic myeloid leukemia in chronic phase, and refractory anemia without excess blasts) or “more advanced” (all other diagnoses).⁹ 

Medical comorbidity was defined and categorized using a validated scale.¹⁰ 

Conditioning regimen was categorized as “myeloablative” (carmustine/etoposide/arabinoside/melphalan, busulfan/cyclophosphamide with or without antithymocyte globulin, or any of the previous regimens with or without total body irradiation > 800 cGy), “nonmyeloablative” (fludarabine 90 mg/m² with or without total body irradiation ≤ 300 cGy), and “reduced intensity” (all other regimens).

Acute graft-versus-host disease was graded as previously described.¹¹ 

Delirium was assessed using the Delirium Rating Scale (DRS), a 10-item scale assessing delirium symptoms over 24 hours using information from patient interview, mental status examination, medical history and tests, nursing observations, and family reports.¹²  It has good construct, discriminant, and criterion validity¹³  and has been used previously in studies of HCT recipients.^14,15  The DRS has been used in children and was modified for patients 3 to 7 years old by including age-appropriate memory screening from the Children's Orientation and Amnesia Test.^16,17  In the few assessments (1%) for which some DRS items could not be scored, mostly because of severe illness, the completed items were averaged to assign scores to the missing items. Patients too ill to undergo DRS assessment (eg, severely impaired consciousness, mechanically ventilated) or when hospitalized patients refused, they were assessed using Diagnostic and Statistical Manual of Mental Disorders, Fourth Edition (DSM-IV) criteria for delirium.¹⁸  Of the 7154 possible delirium assessments, 6162 (86%) were obtained using DRS, 411 (6%) using DSM-IV criteria, and 581 (8%) were missed. Interrater reliability was measured on 228 (4%) of the 6190 assess-ments obtained during the portion of the study when more than one research staff performed assessments with 100% agreement. A delirium episode was defined as a DRS score more than 12 or delirium based on the DSM-IV checklist on at least 2 of 3 consecutive assessments.^14,19-21 

Neuropsychologic testing assessed the following domains: (1) cognitive flexibility and divided attention (Trail Making Test B²²  for patients ≥ 8 years of age and NEPSY Visual Attention²³  for 5-7 years of age); (2) attention and processing speed (Digit Symbol Coding Test²⁴  for ≥ 17 years, WISC-IV coding B²⁵  for 8-16 years, and WPPSI-Coding²⁶  for 5-7 years); (3) concentration (Stroop color/words²⁷  for ≥ 8 years and NEPSY Speeded Naming²³  for 5-7 years); (4) verbal fluency (Controlled Oral Word Association Test²⁸  for ≥ 8 years and NEPSY Verbal Fluency²³  for 5-7 years); (5) visual memory (Faces WMS III²⁹  for ≥ 17 years and Faces CMS³⁰  for 5-16 years); (6) verbal memory (WRAML 2³¹  for ≥ 5 years); (7) visuospatial perception (Judgment of Line Orientation²⁸  for ≥ 8 years and NEPSY Arrows²³  for 5-7 years); and (8) fine motor speed and coordination (Grooved Peg Board³²  for ≥ 5 years).

Overall, 144 (49%) of 291 patients ≥ 5 years old completed both baseline and follow-up neurocognitive testing. Of the 176 missed tests, 83 were the result of severe illness or death and 93 were refused or unable to be scheduled.

Pain intensity was scored on a 0 to 10 Likert scale for adults and older children and a 6-point scale using FACES³³  for children younger than 8 years. Pain was categorized as “severe” (≥ 7 on the adult scale and 4-5 on the FACES), “moderate” (5-6 and 2-3, respectively), “mild” (1-4 and 1, respectively), or “none” (0 for both).

Receipt of glucocorticoids was defined as receipt of ≥ 1 mg/kg of prednisone or prednisolone to distinguish higher doses, which are more frequently associated with side effects, from lower doses or no glucocorticoids. Opioids for acute pain were defined as receipt of intravenous or oral morphine, fentanyl, hydromorphone, and meperidine. Opioids for chronic pain were defined as receipt of oxycodone, methadone, and transdermal fentanyl or morphine. Antivirals were categorized as “low activity” (acyclovir) or “high activity” (foscarnet, ganciclovir, or cidofovir).

Polymerase chain reaction (PCR) analyses were performed by persons blinded to patients' clinical status. DNA was extracted from plasma specimens using the QIAGEN column high-throughput DNA extraction system as previously described.³⁴  For each plasma specimen, 400 μL was used for the extraction and the DNA from each sample was eluted in 100 μL Tris 10mM, of which 10 μL was used for PCR. HHV-6 DNA targets were amplified and detected using a real-time quantitative fluorescent probe PCR assay as previously described.³⁵  Detection of 1 copy of HHV-6 DNA/reaction (25 copies/mL of plasma) was considered a positive specimen. All HHV-6 DNA identified by PCR was typed as HHV-6 A or B.³⁵ 

The primary endpoint of the study was delirium episode as a longitudinally measured binary outcome as defined in “Neuropsychiatric and neurocognitive assessments.” Delirium was modeled using longitudinal logistic regression with generalized estimating equations and robust variances based on an independent working correlation matrix to account for multiple observations across time per person.³⁶  HHV-6, the predictor of interest, was modeled as a time-dependent dichotomous variable. At each time point delirium was assessed, HHV-6 was coded as positive using 3 definitions, modeled separately: (1) subject had detectable HHV-6 at the current or any prior time point; (2) subject had HHV-6 more than or equal to 1000 copies/mL at the current or any prior time point; and (3) the time point was within a window spanning 2 days before through 2 weeks after any HHV-6 detection. Covariates evaluated included baseline demographic and clinical variables (Table 1) as well as baseline neurocognitive performance (Trail Making Test B/Visual Attention NEPSY for children 5-7 years of age), baseline laboratory values previously associated with delirium,¹⁴  and receipt of antithymocyte globulin, pain, and medications known to cause delirium (as in “Clinical data and definitions”). Pain and medications known to cause delirium were considered time-dependent. Pain severity was allowed to vary daily. Medications known to cause delirium were considered “in use” if they were administered on the day of or the day before the delirium assessment. As an initial variable selection step, each factor was evaluated in a bivariable model with HHV-6 and considered eligible for the multivariable model if the 2-sided P value was < .20 or its inclusion modified the effect of HHV-6 by more than 10%. Once included in a full multivariable model, each factor was excluded and included in step-down and step-up fashion and retained if its P value was < .10 or its inclusion modified the effect of HHV-6 by > 10%. Potential interactions were examined between HHV-6 and 3 other variables: underlying disease risk, comorbidity score, and conditioning regimen.

Clinically meaningful neurocognitive decline was defined by a 0.75 z-score drop between baseline and 12-week follow-up examinations.³⁷  This definition was used to compare those with versus without HHV-6 reactivation across all 8 neurocognitive domains. Multivariate analyses were performed using results from Trail Making Test B (Visual Attention NEPSY for children 5-7 years of age) as the dependent variable, as it is considered a highly sensitive measure for cognitive dysfunction. Covariates included baseline clinical and demographic variables.

Multivariable Cox regression models evaluating risk of HHV-6 reactivation were constructed using time to first positive (and separately, first HHV-6 ≥ 1000 copies/mL) and included the baseline clinical and demographic variables. In addition, antivirals (acyclovir, ganciclovir, and foscarnet) were included as time-dependent covariates considered “in use” for the day of and day after their administration. Cidofovir was considered “in use” for 7 days after administration.

Cumulative incidence curves for delirium episode and HHV-6 reactivation were evaluated using time from HCT to the primary event of interest, as defined above, censoring at day of last contact and treating death as a competing risk event.³⁸ 

All measures of association are accompanied by 95% confidence intervals, and all P values are 2-sided with significance level at α = .05. Statistical analyses were carried out using SAS statistical software, Version 9.1.³⁹ 

Of the 315 patients included in analyses, 63 (20%) were younger than 21 years, 169 (54%) underwent myeloablative transplants, and 218 (69%) received cells from mismatched or unrelated donors (Table 1). The cell source was peripheral blood stem cells for 233 (74%), bone marrow for 61 (19%), and cord blood for 21 (7%).

HHV-6 was detected in 111 (35%) of the 315 patients by day 84 after HCT (Figure 2A). The median time to first detection among those with reactivation was 20 days after transplantation (interquartile range [IQR], 15-28 days after transplantation), and the median duration of detection was 4 days (IQR, 1-11 days). The median maximum DNA level was 873 copies/mL (IQR, 175-4580 copies/mL), and the HHV-6 DNA level was more than 1000 copies/mL in 53 (17%). All detected HHV-6 was type B.

A positive delirium assessment was obtained for 89 (30%) of 292 patients. More prolonged delirium, a “delirium episode” as defined in “Neuropsychiatric and neurocognitive assessments,” was detected in 59 (20%). The median time of onset of the delirium episode was 15 days (IQR, 12-33 days) after transplantation, and the median duration was 6 days (IQR, 3-15 days; Figure 2B).

HHV-6 reaction best predicted a delirium episode when evaluated using a window that required HHV-6 and delirium to be temporally associated as described in “Statistical analyses”; the P values from univariate models for “any HHV-6,” “any HHV-6 more than 1000 copies/mL,” and “HHV-6 window” were .08, .01, and < .001, respectively. The association between “HHV-6 window” and delirium episode remained statistically significant after controlling for other strong predictors of delirium (Table 2). The other definitions of HHV-6 exposure were not significantly associated with delirium in the final multivariate models; “any HHV-6” had a P value of .42 and “HHV-6 more than 1000 copies/mL” had a P value of .07. No multiple testing adjustments were performed for the different ways of coding HHV-6 exposure.

HHV-6 reactivation appeared to be a stronger predictor of delirium in patients with more advanced underlying disease versus those with less advanced underlying disease: adjusted odds ratio 3.7, 95% confidence interval, 1.6, 8.4, P = .002 versus adjusted odds ratio 0.71, 95% confidence interval, 0.17, 3.0, P = .64, respectively. Although suggestive of a potential interaction between HHV-6 and underlying disease, the interaction term was not statistically significant (P = .10). However, comparing the adjusted odds ratios for HHV-6 from the separately fit models within each disease risk group yielded a P value of .05.

Of the 19 patients with temporally associated HHV-6 and delirium, only 4 had CSF obtained clinically during the course of delirium; 2 of these samples had HHV-6 DNA detected. The 2 patients with HHV-6 detected in their CSF each had 3 positive plasma samples with HHV-6 DNA levels ranging between 108 and 1059 copies/mL for one and 270 and 5068 copies/mL for the other. In contrast, the 2 patients with negative CSF each had low-level HHV-6 DNA detected in only one plasma specimen (60 and 41 HHV-6 DNA copies/mL). All 4 patients had only one CSF specimen obtained during the delirium episode. Brain imaging was obtained in 9 (47%) of the 19 patients with temporally associated HHV-6 and delirium; 6 of these were normal, whereas 3 had abnormalities attributed to radiation or immunosuppressive neurotoxicity.

Among patients completing neurocognitive assessments, clinically significant neurocognitive decline was more common in patients with HHV-6 reactivation in domains involving executive functioning (Table 3). The differences were statistically significant in the domains of attention/processing speed and concentration. HHV-6 was not associated with declines in domains involving memory (visual memory, verbal memory) or other functions (visuospatial perception, fine motor speed/coordination). Multivariate analysis was performed on the outcome cognitive flexibility/divided attention where HHV-6 reactivation was a statistically significant predictor of neurocognitive decline after controlling for multiple baseline factors (Table 4).

Cord blood, reduced intensity conditioning, and unrelated transplant were associated with HHV-6 reactivation at any level of DNA (Table 5). Receipt of antivirals was not associated with HHV-6 reactivation. However, few patients received HHV-6–active antivirals early after transplantation.

In this prospective study of 315 HCT recipients, we found HHV-6 reactivation to be a strong predictor of CNS dysfunction as measured by delirium and neurocognitive decline, even after controlling for known risk factors. These data provide supporting evidence that HHV-6 has a direct effect on the CNS and that HHV-6 DNA detection in the blood after HCT is clinically meaningful.

There is little dispute that HHV-6 causes encephalitis after HCT when detection of the virus in the CSF is associated with characteristic signs of encephalopathy and abnormalities on brain imaging without another etiology identified.^4-7  The clinical significance of HHV-6 reactivation (detection of viral nucleic acids in plasma or whole blood) is less certain. Results from several studies suggest that HHV-6 reactivation increases risk of encephalitis or CNS dysfunction.^3,40-42  Past studies varied in size and lacked systematic assessment of CNS function, leaving questions regarding the extent of the effect of HHV-6. We used detection of delirium as an indicator of acute CNS dysfunction and decline of neurocognitive function as a marker of more persistent dysfunction. Delirium is clinically meaningful in HCT recipients as its occurrence in the first month after transplantation has been associated with impaired neurocognitive functioning and persistent distress at 80 days²⁰  and persistent distress, decreased health-related quality of life, and neurocognitive dysfunction at 1 year after transplantation.²¹  The observed temporal associations between HHV-6 reactivation and delirium and neurocognitive decline provide new evidence that HHV-6 directly affects the CNS and that HHV-6 reactivation after HCT is clinically important. Evidence for a causal relationship between HHV-6 and CNS dysfunction is extended by the finding that higher levels of HHV-6 strengthened the association with neurocognitive decline.

Unfortunately, only a small subset (4 of 19) of our patients with temporally related HHV-6 and delirium had CSF obtained around the time of the delirium episode, and each of these 4 patients had only one CSF sample obtained. Only 2 of these 4 patients had HHV-6 DNA detected. Although HHV-6 DNA has been shown to persist longer in CSF compared with plasma or serum,³⁵  it has also been shown that, when both CSF and plasma have reverted to undetectable levels of HHV-6, active infection may still be ongoing in brain tissue.⁴³  More complete virologic assessment of patients with HHV-6 and delirium would improve understanding of the viral dynamics of HHV-6–related CNS disease.

Consistent with previous studies,^1,41,42  we demonstrated that recipients of cord blood or unrelated transplants were at increased risk for HHV-6 reactivation. We also found a stronger effect of HHV-6 on delirium in patients with more advanced underlying disease. More advanced disease may be a marker for more vulnerable neurologic status, or a patient at higher risk for delirium with HHV-6 reactivation. These findings highlight the potential importance of targeting cord blood transplant recipients or patients with more advanced underlying disease for an antiviral intervention study.

This study has limitations that should be recognized. CSF testing and brain imaging were not available on all patients with HHV-6–associated delirium, making it difficult to determine the proportion of cases with HHV-6 encephalitis. In addition, we could not obtain complete neurocognitive testing on many patients. However, the present study has several strengths not previously available in a single study: a prospective design, large sample size of representative HCT patients, frequent assessments of neuropsychiatric and neurocognitive function blinded to HHV-6 reactivation status, use of frequent quantitative HHV-6 assessments, and use of comprehensive statistical analyses.

In conclusion, we have demonstrated an independent and quantitative association between HHV-6 reactivation and CNS dysfunction after HCT. A randomized antiviral trial is warranted to definitively establish causality and to provide rational approaches to treatment and/or prevention of HHV-6.

The authors thank the patients and their families who participated in this study, without whose patience and dedication this study could not have been undertaken. We also thank Dr Mohamed Sorror for sharing data regarding the HCT comorbidity index, nurse Karin Rogers for her professionalism and sensitivity in approaching patients, and Anna Rashevsky for her excellent laboratory work.

This work was supported by the National Institutes of Health (grants R01AI057639 and CA18029).

The funding source had no involvement in the study design, collection, analysis, and interpretation of data. All authors had full access to the data and responsibility for the decision to submit for publication.

National Institutes of Health

Contribution: D.M.Z., J.R.F., D.B., M.B., A.L.A., C.D., M.-L.H., L.C., and W.M.L. participated in the study design, data analysis and interpretation, and critical review of the report; D.M.Z. wrote the first draft and all revisions of the report; and H.X. and W.M.L. performed the statistical analyses and participated in revision of the report.

Conflict-of-interest disclosure: M.B. received research funding for clinical trials and consulting fees from Roche/Genentech and Chimerix Inc. The remaining authors declare no competing financial interests.

Correspondence: Danielle M. Zerr, Seattle Children's Hospital, Mail Stop R5441, 4800 Sand Point Way NE, Seattle, WA 98115; e-mail: danielle.zerr@seattlechildrens.org.