Design and setting
We conducted this study at the Institute for Clinical Evaluative Sciences (ICES) according to a pre-specified protocol that was approved by the research ethics board at Sunnybrook Health Sciences Centre (Toronto, Canada). Participant informed consent was not required for this study. We conducted a retrospective, population-based, matched cohort study of older adults using linked healthcare databases in Ontario, Canada. Ontario has approximately 13 million residents, 2 million of whom are aged 65 years or older [21]. Residents have universal access to hospital care and physician services and those aged 65 or older have universal prescription drug coverage. The reporting of this study follows guidelines set out for observational studies (Additional file 1: Table S1) [22].
Data sources
We ascertained patient characteristics, drug use, covariate information, and outcome data using records from six databases. We obtained vital statistics from the Registered Persons Database (RPDB), which contains demographic information on all Ontario residents ever issued a health card. We used the Ontario Drug Benefit (ODB) database to identify prescription drug use, including dispensing date, quantity of pills, dose, and number of days supplied. This database contains highly accurate records of all outpatient prescriptions dispensed to patients aged 65 years or older, with an error rate of less than 1% [23]. We identified diagnostic and procedural information on all hospitalizations and emergency room visits from the Canadian Institute for Health Information Discharge Abstract Database (CIHI-DAD) and the National Ambulatory Care Reporting System (NACRS), respectively. We obtained covariate information from the Ontario Health Insurance Plan (OHIP) database, which includes health claims for inpatient and outpatient physician services. We used the ICES Physician Database (IPDB) to ascertain antiviral drug prescriber information. Previously, we have used these databases to research health adverse drug events and health outcomes, including acyclovir-induced acute kidney injury [24–26]. With the exception of antiviral prescriber specialty and income quintile (missing in 14.5% and 0.4% of patients, respectively), the databases were complete for all variables used in this study. Given the ability of our databases to capture healthcare activity province-wide, the only loss to follow-up would be if patients emigrated from Ontario (a rate estimated to be less than 1% per year) [27]. The database codes used in the analysis are defined in Additional file 1: Table S2.
Patients
We established a cohort of residents aged ≥66 years in Ontario, Canada who filled a new outpatient prescription with ≥7-day supply for oral acyclovir, valacyclovir, or famciclovir from April 2002 to December 2011, a period spanning 9 years. We restricted our analysis to those who had evidence of a herpes zoster diagnosis in the 90 days prior to or 30 days following the time of the prescription (database diagnosis codes presented in Additional file 1: Table S2). The date of the first eligible prescription for a study antiviral served as the index date for that patient and marked the start date of follow-up. We excluded the following patients from the analysis: i) those in their first year of eligibility for prescription drug coverage (age 65) to avoid incomplete medication records, ii) those living in long-term care facilities since residents may have frequent episodes of confusion or delirium for many reasons, iii) those with end-stage renal disease, iv) those who had a prescription for any antiviral in the prior 180 days in order to capture new usage, v) those who had a prescription for more than one type of antiviral on the index date in order to compare mutually exclusive groups, and vi) those who had a hospital admission or discharge on their index date or a hospital discharge in the prior two days to ensure these were new outpatient antiviral prescriptions (as patients continuing an antiviral treatment initiated in hospital would have their outpatient antiviral prescription dispensed on the same or next day of hospital discharge).
To select two groups of antiviral users that were well-balanced on the baseline characteristics we measured in this study, we matched each low-dose user with a high-dose user on a 1:6 basis using the following variables: age (within two years), sex, presence of chronic kidney disease, and type of antiviral prescribed (acyclovir, valacyclovir, or famciclovir). In Ontario, the validated algorithm for chronic kidney disease identifies older adults with a median estimated glomerular filtration rate (eGFR) of 38 mL/min per 1.73 m2 (interquartile range 27 to 52), whereas its absence identifies those with a median eGFR of 69 mL/min per 1.73 m2 (interquartile range 56 to 82) [28].
Antiviral dose
To align with recommendations in drug prescribing references, a higher dose of antiviral therapy was defined as at least 4,000 mg/day for acyclovir, 3,000 mg/day for valacyclovir, and 1,500 mg/day for famciclovir. A lower dose of antiviral therapy was defined as 3,200 mg/day, 2,400 mg/day, 1,600 mg/day or 800 mg/day for acyclovir, 2,000 mg/day, 1,500 mg/day, 1,000 mg/day, or 500 mg/day for valacyclovir, and 1,000 mg/day, 500 mg/day, or 350 mg/day for famciclovir (Table 1).
Outcomes
We followed all patients for 30 days after the index date for the assessment of two pre-specified outcomes. The primary outcome was hospital admission with evidence of an urgent computed tomography (CT) scan of the head within the first five days of admission (inclusive of any scans performed in the emergency room preceding an admission). Based on prospective studies of common clinical practice, neuroimaging is frequently used in the routine evaluation of patients who present to hospital acutely confused, even among those without focal neurological findings or head trauma [29–31]. Unlike diagnostic codes for acute delirium, the receipt of a CT scan of the head is well coded in our data sources (these codes have high sensitivity and specificity for receipt of the imaging as they are associated with physician reimbursement) [32]. We also expected urgent CT scans of the head conducted for reasons unrelated to antiviral dosing to occur at a similar frequency in higher and lower dose groups; therefore, not impacting estimates of difference in risk. We have used this outcome of urgent CT scans of the head in other population-based drug safety studies to characterize the risk of drug-induced delirium [33]. Our secondary outcome was all-cause mortality. Death is accurately coded in our data sources (sensitivity 94%, positive predictive value 100%) [34].
Statistical analysis
We compared baseline characteristics between those prescribed a higher or lower antiviral dose using standardized differences [35]. This metric describes differences between group means relative to the pooled standard deviation and is considered a meaningful difference if greater than 10%. We estimated the odds ratio and 95% confidence intervals for inpatient CT scan of the head with higher antiviral dose compared to lower antiviral dose using conditional logistic regression analyses (accounting for matched sets). We interpreted odds ratios as relative risks which was appropriate given the low incidence of observed events. We examined the relative risk between higher dose and lower dose (referent dose) antiviral and each outcome first in the entire matched cohort and then in four pre-defined subgroups based on: age, sex, presence of chronic kidney disease, and antiviral type. We examined whether relative risks differed among subgroups using tests for interaction. We conducted all analysis with Statistical Analysis Software (SAS) version 9.2 (SAS Institute Incorporated, Cary, North Carolina, USA, 2008).
