Bias in observational study of the effectiveness of nasal corticosteroids in asthma - 18/08/11

Doi : 10.1016/j.jaci.2004.12.1118

Samy Suissa, PhD ^⁎ , Pierre Ernst, MD
From the Division of Clinical Epidemiology, Royal Victoria Hospital, McGill University Health Centre, and the Departments of Epidemiology and Biostatistics and of Medicine, McGill University

Reprint requests: Samy Suissa, PhD, Division of Clinical Epidemiology, Royal Victoria Hospital, 687 Pine Avenue West, Ross 4.29, Montreal, Quebec, Canada H3A 1A1.

Montreal, Quebec, Canada

Abstract

Background

A recent observational study suggests that intranasal corticosteroids used to treat allergic rhinitis are effective at preventing asthma outcomes, such as emergency visits. The approach to data analysis may have led to biased results because of misclassification of immortal time.

Objective

To illustrate the bias in the cohort approach and to present the proper time-dependent analysis by replicating the recent study using data from another source.

Methods

From an existing cohort of 30,569 patients with asthma age 5 to 44 years and identified from the Saskatchewan Health databases (1975-1997), we formed the cohort of all subjects who were in the source population between January 1, 1989, and December 31, 1991. Subjects were followed to the first asthma hospitalization. All prescriptions dispensed during follow-up were identified. We replicated the time-fixed approach to data analysis used in the recent study and compared it with time-dependent approaches.

Results

The cohort included 20,173 subjects, of whom 1849 were hospitalized for asthma between January 1, 1989, and December 31, 1991. The time-fixed approach misclassified more than 5000 person-years of follow-up, corresponding to 44% of the exposed person-time. As a result, the rate ratio of asthma hospitalization after any use of nasal corticosteroids (NCSs) was 0.57 by the biased time-fixed approach compared with 1.13 by the proper time-dependent approach. The time-fixed approach produced a paradoxical protective effect of NCS with 1 or less canisters dispensed per year (odds ratio, 0.47), which was further exaggerated when the cohort was extended to 5 years (odds ratio, 0.33). Adjusted time-dependent analyses found no protective effect, even when NCSs were dispensed regularly (rate ratio, 1.10; 95% CI, 0.54-2.21).

Conclusion

The time-fixed approach to the analysis of the effectiveness of NCSs on asthma outcomes leads, by its inherent misclassification of immortal time, to a considerable exaggeration of the protective effect of these medications in preventing severe asthma exacerbations.

Le texte complet de cet article est disponible en PDF.

Key words : Cohort studies, biases, drug effectiveness, databases, nasal corticosteroids

Abbreviations used : ED, NCS, RR

Plan

Supported by grants from the Canadian Institutes of Health Research and Fonds de la recherche en santé du Québec. The database used for illustration was acquired thanks to grants from AstraZeneca, Boehringer-Ingelheim, and GlaxoSmithKline. Dr Suissa is the recipient of the Distinguished Investigator award from the Canadian Institutes of Health Research.
Disclaimer: This study is based in part on nonidentifiable data provided by the Saskatchewan Department of Health. The interpretations and conclusions contained herein do not necessarily represent those of the Government of Saskatchewan or the Saskatchewan Department of Health.
Disclosure of potential conflict of interest: S. Suissa has received grants/research support from and is on the Speakers' Bureau of Schering-Plough, AstraZeneca, and GlaxoSmithKline. P. Ernst has consultant arrangements with and is on the Speakers' Bureau of Merck, GlaxoSmithKline, and AstraZeneca and has received grants/research support from GlaxoSmithKline.