Initial Vancomycin Taper for the Prevention of Recurrent Clostridium Difficile Infection

STUDY POPULATION

This is a multi-centre study involving institutions in British Columbia, Ontario, Quebec and Newfoundland. The study population will be drawn from patients cared for as inpatients or outpatients at the participating hospitals. Such patients will have a test positive for C. difficile and will be receiving treatment. The trial will involve only adult patients 18 years of age and older.

Criteria for Recruitment

The microbiology laboratory will notify the study team about a positive CDI test via telephone, email, or fax. The nature of recruitment will then depend on the inpatient status of the patient at the time of the test.

Inpatients:

Pre-existing approval for approaching patients for this study will be obtained from the relevant department heads. The study team will speak with a member of the inpatient treating team (resident physician or faculty physician as appropriate) to determine if the patient is appropriate for recruitment. If this seems to be the case, the patient's file will be rapidly screened to determine eligibility and if the patient is eligible they will be approached for consent.

Outpatients:

The physician who ordered the C. difficile test will be contracted to determine if the patient is appropriate for recruitment. At the invitation of this physician, the investigators will then contact the patient via telephone to evaluate suitability for inclusion and arrange an intake visit.

RANDOMIZATION

For patients who have enrolled in the study, randomization will occur centrally at McGill via an existing internet application and will be performed by permuted block with randomized block sizes. This randomization will be stratified for first episode or first recurrence at study entry to ensure these factors are properly balanced.

TRIAL SCHEDULE

• Day 1: Patient diagnosed with C. difficile and started on standard of care oral vancomycin treatment -> Determine eligibility and obtain permission for approach

• Day 7-10 (Patient's C. difficile has improved and meets eligibility): Consent obtained; randomization; distribution of study drug for day 15 start -> Collection of demographics

• Day 15-28 -> Receipt of study therapy

• Day 28: In person or remote visit

• Day 56: In person or remote visit -> Primary outcome determined, quality of life questionnaire

• Day 90: Study ends for the patient -> Secondary outcomes can be determined

• weekly until Day 56: Brief questionnaire -> By email/text/phone

• biweekly after Day 56: Brief questionnaire -> By email/text/phone

• Ad hoc: If patient has symptoms of recurrence of C. difficile -> Review by ID physician in clinic if possible, otherwise usual doctors or emergency room

Patients will be able to come be assessed for potential relapse by infectious diseases physicians at each site (who may or may not be a part of the study) or could see their usual doctors.

SAMPLE SIZE AND STATISTICAL METHODS

The initial estimates were a risk of recurrence in the control group of 25% and in the intervention group of 15%. With 80% power and an overall alpha of 0.05, it would require 496 patients (rounded to 500) to demonstrate superiority.

After the recruitment of 50 patients, the analytic plan was modified to use a Bayesian framework on April 3rd, 2023, prior to completion of study enrolment and before data were locked and blinded for analysis. This was done to allow for earlier stopping while preserving the overall type 1 error rate and maximal sample size in response to an update to the Infectious Diseases Society of America guidelines for the treatment of C. difficile, which were changed to recommend fidaxomicin as first line therapy in place of vancomycin.

We used adaptR to simulate 50,000 2-armed trials under both the null and alternative hypotheses using the funded 500 patient sample size and interim analyses at 125, 250 and 500 patients. The thresholds for superiority (relative risk <1) were 99% at the first analysis and 97.5% at subsequent analyses. A futility analysis was included whereby the trial would be stopped if there was less than a 15% probability of at least a 4% absolute risk reduction (number needed to treat, NNT ≤ 25). These simulations estimated an overall type 1 error rate of 5.1% with 72.6% power and a median expected sample size of 250 patients.

All binary outcomes were analysed using a Bayesian Generalised Linear Model (binomial family, log link) which yields log relative risk (RR) with 95% Bayesian credible intervals (CrI). These were exponentiated to give risk ratios. Priors were minimally-informative [N~ (0,100)]. Stratification was included in the model and we ran 4 chains of 50,000 Markov Chain Monte Carlo simulations.