Written by Girish Mishra, MD, MS Associate Professor of Internal Medicine at Wake Forest University School of Medicine in Winston-Salem, North Carolina, USA.
When designing clinical trials, the inclusion and exclusion criteria that are set form the basis of the study population. Sampling and error form the crux of statistical principles. Defining both the inclusion and exclusion criteria defines the study population. Thus, similar to randomized clinical trials, results from endoscopic studies can vary greatly depending on who and how the patients were chosen or excluded.
Tip 1: Equal, if not more, attention should be devoted to the exclusion criteria as is to the inclusion criteria.
Comments: Without randomization, the exclusion criteria that are set a priori can affect the generalizability of the study and also potentially introduce confounders. By excluding certain patients, will the outcome be biased toward or away from the null hypothesis? In other words, by excluding a subset of patients (age, alcohol intake, inpatient vs. clinic setting), could there be either alpha or beta error?
Example: A study that investigated early EUS vs ERCP for the management of acute biliary pancreatitis clearly defined the study population with stringent inclusion and exclusion criteria. They included all patients with first attack of acute pancreatitis with abnormal liver functions. They excluded patients with recurrent pancreatitis, severe cholangitis, post-ERCP, hyperlipidemia, and a few other subsets of patients. Could the authors have seen even a more positive benefit of EUS if they included cholangitis patients since they were more likely to have a CBD stone detected?
Solution: Always include how the patients were selected. For example, if a study evaluates the sensitivity of EUS-FNA in pancreatic lesions, there should be mention of all consecutive patients that were seen at the institution so as not to introduce selection bias.