The level of evidence (eg, meta-analysis, randomized clinical trial, case report, expert opinion or experience) in a study indicates how prone it is to bias.6 The quality of the study, which is determined by an appraisal process, can also be an indicator of bias. Appraisal should include investigating factors, such as the type, age, sex, and number of study participants (ie, sample size); enrollment criteria; definitions of conditions; which and how examinations were performed; how participants were allocated to different groups; and whether the study was blinded (ie, humans involved in the study did not know which treatment group participants were allocated to). Assessment for potential bias should also include whether clear inclusion and exclusion criteria are given and whether outcome measures are reasonable and relevant. Questions asked during the appraisal process depend on the study design and research question of interest (Tables 2-4).
A common flaw in veterinary publications is not including a sufficient number of subjects or samples to draw robust conclusions.7 A sample size calculation or reasoning regarding the number of participants included in the study should be included in the methods, results, and discussion sections.
In some studies, patients with a specific disease are enrolled without a clear case definition or a documented diagnostic procedure (including which test results indicated physiologic/not physiologic conditions) that can be used to confidently identify the presence of disease, making it difficult to draw a conclusion. How participants are allocated to different groups is important because unequal distribution (eg, including more severely ill patients in one group compared with another) can distort outcomes.8 In addition, a lack of blinding can influence recorded outcomes, as researchers may be (consciously or unconsciously) biased in the conduction or interpretation of a study when they know which patients received specific treatments.