To the Editor:
We read with interest the recent paper by Gerrits et al. 1 reporting on the effectiveness of N‐acetylcysteine in reducing re-hospitalisations for chronic obstructive pulmonary disease (COPD). We are concerned, however, that the apparent benefit is actually the result of bias in study design and analysis.
The paper mentions that patients were considered users if they received N‐acetylcysteine “immediately” after discharge. However, we are not given the definition of “immediately” norare we provided with the average and range of the length of time from hospital discharge to dispensing of the first prescription for N‐acetylcysteine. Since follow-up time starts at discharge from hospital for all subjects, irrespective of whether they will or will not be dispensed N‐acetylcysteine, there is a strong possibility of immortal time bias 2. Indeed, the time period from discharge to the first dispensing is “immortal” by definition since re-admissions could not occur during this period; if they did occur before the first prescription, the subject would have been classified as “unexposed”. Moreover, this immortal time period is in fact unexposed, since the subject has not yet received the drug, but is misclassified as exposed by this approach to data analysis.
Consequently, such misclassified immortal time will cause the rate of re-admission in the “unexposed” to be artificially inflated resulting in an impression of a protective rate ratio for exposure. This bias will also accentuate the apparent dose-response effect, since the daily dose calculation uses the immortal unexposed time and also considers it as exposed. Itwould be useful to redo these analyses, after properly classifying exposures.
- © ERS Journals Ltd