Exploring the effect of transient exposure on the risk of acute events by means of time-window designs: an application to fluoroquinolone antibacterials and arrhythmia.

Purpose: To compare different strategies in allocating time-window at risk and in choosing optimal comparator in the setting of proarrhythmic effects of antibacterial fluoroquinolones.
Design: A population-based cohort study.
Setting: Resident population of a Northern Italian Province.
Participants: The study cohort comprised 75 226 patients who received at least one prescription for fluoroquinolone from 1997 to 1999. Cohort members who experienced at least an antiarrhythmic drug prescription and/or hospitalisation with diagnosis of arrhythmia during the follow-up were identified.
Analysis: Trends in incidence rates of arrhythmia according to different strategies in allocating time-windows at risk were calculated. Standardised incidence ratios were obtained by comparing arrhythmic events experienced by the study cohort during the time-windows at risk with those expected from both external (unexposed target population) and internal (study cohort during reference time-windows) comparators.
Results: The rate of arrhythmia was 21.0 per 100 000 person-weeks during the in-treatment period, remained stable during the following week, and progressively decreased until to converge to 7.7 per 100 000 person-weeks from 16th weeks after stop of exposure. Standardised incidence ratios were 2.35 (95%CI: 1.82, 2.99) and 1.86 (1.44, 2.37) according to external and internal comparators respectively. Evidence of linear increasing of proarrhythmic effect with the length of the antibacterial therapy appeared from every approaches, but it was stronger by using internal comparator.
Conclusions: Time-windows designs should be performed by allocating the time-windows at risk and comparing outcome events observed in the cohort during these periods with those experienced by the same cohort during an adequate reference period.
((c) 2005 John Wiley & Sons, Ltd.)