From 2014 to 2017
For HHV6, in-house PCR used primers and Taqman probe (Applied
Biosystems, Foster City, CA, USA) targeted at the U67 gene of HHV6, and
the viral load threshold of detection was 233 copies/mL plasma at 95%
confidence level. For EBV, in-house PCR was targeted at the highly
conserved, repetitive region of the EBV genome designated the BamHI-W
region, and the viral load threshold of detection was 265 copies/mL
plasma at 95% confidence level. One copy of EBV DNA is equivalent to
0.26 International Units (IU) based on the 1st WHO International
Standard for Human EBV for Nucleic Acid Amplification Techniques (NIBSC
09/260). For CMV, primers and probe targeted at UL54 gene of human
cytomegalovirus were used (Roche COBAS AmpliPrep/COBAS TaqMan CMV Test),
and the viral load threshold of detection was 56 IU/mL plasma at 95%
confidence level.
2.3 Statistical analysis
Outcome variables including ICU admission, the need for dialysis,
in-hospital mortality, the presence of flares, hospital readmission
within 1 year and mortality within 1 year were all treated as binary
variables while outcome variable length of hospital stay (LOS) was
treated as count data i.e. non-negative integer values. Exposersviral reactivation and antiviral treatment were both
treated as binary data with categories ”yes ” or “no ”.
All categorical and continuous outcomes were summarized as frequency
with corresponding percentages and mean (standard deviation) or median
(interquartile range), whichever appropriate, respectively based onviral reactivation and antiviral treatment status.
Categorical outcome variables were compared using Chi-Square test while
continuous variables were compared using two-sample t-test or
Mann–Whitney U test, whichever appropriate.
Univariate logistic regression models were fit to find associated risk
factors for binary outcomes. Association from logistic regression was
expressed as odds ratio (OR) with 95% confidence interval (95% CI).
Outcome variable LOS was fit using the Poisson regression model and
association between LOS and exposers were expressed as incidence rate
ratio (IRR) with 95% CI. Percentage of reactivation from onset date was
analyzed for individual herpesviruses. In case of repeat PCR in this
analysis, first positive episode was analyzed and the corresponding
number of days from onset of rash to the 1st positive test was
calculated. A p-value <0.05 was regarded as statistically
significant. For statistical analysis, we used the software SAS (version
9.4, Cary, NC, USA).
This study was approved by Singhealth IRB: 2014/2011.