Reply

Abstract

We thank Drs Azoulay and Suissa for their interest in our article and we would like to respond to their comment regarding immortal time bias.1Azoulay L. Suissa S. Gastroenterology. 2012; 142: e20Abstract Full Text Full Text PDF PubMed Scopus (3) Google Scholar, 2Kastelein F. et al.Gastroenterology. 2011; 141: 2000-2008Abstract Full Text Full Text PDF PubMed Scopus (107) Google Scholar In our study, patients were classified as user of nonsteroidal anti-inflammatory drugs (NSAIDs) or statins irrespective of the time patients started with the use of these drugs, since most patients were either in the exposed or nonexposed group for their entire follow-up period. For patients who used NSAIDs and statins only after inclusion in the study, the follow-up time between inclusion in the study and the first drug prescription can be considered as “immortal,” because patients had to survive this period without experiencing an event to be classified as a user of these drugs. Because misclassification of the exposure time is avoided in a time-dependent Cox regression model, we performed such analysis with updated follow-up data. Neoplastic progression was defined as the development of high-grade dysplasia or esophageal adenocarcinoma ≥9 months after inclusion in the study and follow-up time was defined as the time from 9 months after inclusion in the study to the most recent surveillance endoscopy or the endoscopy that resulted in a diagnosis of high-grade dysplasia or esophageal adenocarcinoma. Medication use was defined as a time-dependent variable, with patients classified as nonuser of NSAIDs and statins before the first drug prescription and as user afterward. Five hundred seventy patients were followed in the study for a median duration of 5.4 years (interquartile range, 4.4–6.2) and a total follow-up period of 2912 patient-years. During follow-up, 40 patients developed high-grade dysplasia or esophageal adenocarcinoma. NSAIDs were used by 318 (56%) patients with an exposure time of 1538 (53%) patient-years, statins were used by 209 (37%) patients with an exposure time of 864 (30%) patient-years, and 107 (19%) patients used both NSAIDs and statins with an exposure time of 404 (14%) patient-years. In a time-dependent Cox regression model, NSAID use was associated with a significantly lower risk of neoplastic progression (hazard ratio, 0.49; 95% confidence interval 0.25–0.97) and statin use was associated with a trend toward a lower risk of neoplastic progression (hazard ratio, 0.55; 95% confidence interval 0.23–1.29), after adjusting for age, gender, Barrett length, baseline histology, and use of other medications. Use of both NSAIDs and statins was associated with an additional protective effect (hazard ratio, 0.26; 95% confidence interval, 0.07–0.98). Although time-dependent models do not take into account that patients have no risk of experiencing the event during the immortal time period and therefore may bias against therapy, the results of the time-dependent analyses are not very different from our initial risk estimates.3Levesque L.E. et al.BMJ (Clin Res Ed). 2010; 340: b5087Crossref PubMed Scopus (715) Google Scholar, 4Yang X.L. et al.Diabetes Obes Metab. 2011 Dec 15; ([Epub ahead of print])Google Scholar Thus, our initial conclusion that NSAID and statin use are each associated with approximately 50% reduction in the risk of neoplastic progression, also holds in time-dependent analyses. Immortal Person-Time Bias in Relation to the Use of Nonsteroidal Anti-inflammatory Drugs and Statins in the Prevention of Esophageal Cancer in Patients With Barrett's EsophagusGastroenterologyVol. 142Issue 5PreviewWe read with interest the study published by Kastelein et al1 in a recent issue of Gastroenterology. The authors investigated the effects of nonsteroidal anti-inflammatory drugs (NSAIDs) and statins on the risk of neoplastic progression from Barrett's esophagus (BE). The use of NSAIDs and statins was associated with a 53% and 54% risk reduction, respectively (hazard ratio [HR], 0.47 [95% confidence interval (CI), 0.24–0.93] and HR, 0.46 [95% CI, 0.21–0.99]). Furthermore, the use of a combination of NSAIDs and statins was associated with an even stronger risk reduction (HR, 0.22; 95% CI, 0.06–0.85). Full-Text PDF