Response

Abstract

We appreciate the comments raised by Dr Hamada and his colleagues regarding the use of “sensitivity analysis” in our article. The primary objective of our article was to understand the impact of nasogastric lavage (NGL) on important clinical outcomes. In using a propensity-score matched analysis, we attempted to balance biases related to patients who have undergone NGL compared with those who have not. As astutely pointed out by the authors, important hidden biases can still be missing from the propensity-regression model. However, one must understand that such hidden biases exist in all observational studies.1Rosenbaum P.R. Discussing hidden bias in observational studies.Ann Intern Med. 1991; 115: 901-905Crossref PubMed Scopus (157) Google Scholar This is in itself not unique to the analytic technique applied in this article, but to the actual study design similar to that of other observation studies. Certainly, conducting a randomized controlled trial (RCT) would help to eliminate these hidden biases. But even an RCT is not necessarily immune to all types of biases (ie, selection, attrition bias).2Dumville J.C. Torgerson D.J. Hewitt C.E. Reporting attrition in randomised controlled trials.BMJ. 2006; 332: 969-971Crossref PubMed Scopus (245) Google Scholar, 3Berger V.W. Exner D.V. Detecting selection bias in randomized clinical trials.Control Clin Trials. 1999; 20: 319-327Abstract Full Text Full Text PDF PubMed Scopus (120) Google Scholar Performing a sensitivity analysis may be helpful to estimate the magnitude of these hidden biases. However, the intent of the sensitivity and specificity presented in our article was not to estimate these hidden biases, as incorrectly pointed by the authors. Rather, it was an indirect way to validate our study with previous studies.4Aljebreen A.M. Fallone C.A. Barkun A.N. Nasogastric aspirate predicts high-risk endoscopic lesions in patients with acute upper-GI bleeding.Gastrointest Endosc. 2004; 59: 172-178Abstract Full Text Full Text PDF PubMed Scopus (110) Google Scholar Prior studies have conducted analysis only on the predictive value of bloody aspirates in nasogastric lavage on endoscopic findings, not the actual performance of nasogastric lavage on patient outcomes. Thus, we performed a sensitivity and specificity analysis on the subset of patients who underwent NGL to determine the validity of our study population with prior studies. In this case, we found that our study was consistent with prior studies. On the contrary, we performed a “sensitivity analysis” as defined by Hamada and colleagues to estimate the magnitude of this hidden bias and found that unmeasured covariates would have to increase the probability of having NGL by more than approximately twofold (Ã = 2.0) in order to explain away our finding of faster time to endoscopy. Sensitivity analysis, not a calculation of sensitivity, is essential in a propensity score analysisGastrointestinal EndoscopyVol. 75Issue 6PreviewWe read with great interest the article by Huang et al,1 a retrospective analysis of patients suspected of having acute upper GI bleeding. In their analysis, propensity-matched groups were successfully generated and were compared to evaluate the usefulness of nasogastric lavage in such patients. Full-Text PDF