Statistics in Medicine: Understanding Data and Study Design to Improve Interpretation of Results
In a presentation delivered at SDPA’s annual conference in LA, John Barbieri, MD, MBA, examined statistics in medicine cautioning providers to avoid being “deceived by p-values”. While p-values inform a provider that findings of a study are statistically significant, they fail to give the provider any information about effect size, said Dr. Barbieri, director of the advanced acne therapeutics clinic at Brigham and Women’s Hospital. A p-value does not mean the results of the study are clinically meaningful. When reviewing medical literature, he encouraged providers to beware of multiple hypothesis testing.
Coverage: SDPA 19th Annual Fall Dermatology Conference in Los Angeles Nov. 4-7, 2021
Next, Dr. Barbieri used a study example to highlight his recommended steps for interpretation of randomized controlled trials. First, he recommends providers review the study design and ask whether the study is placebo controlled or active comparator. Additionally, he communicated providers should review the population of the participants in the study, found in Table 1 of published articles. It is important to examine the age, gender, race-ethnicity, disease severity and comorbidities. Are these participants representative of your patient population? After these questions have been answered, Dr. Barbieri recommends reviewing the key findings through examination of effect size and confidence intervals. Dr. Barbieri encourages providers examine the subgroup analyses—are there findings that a treatment would work better in specific populations?
Finally, Dr. Barbieri explained observational studies through review of a published example. He reports observational studies tend to answer different questions than a randomized controlled trial. Observational studies can be helpful to gain information regarding a rare disease or an uncommon risk or adverse event to treatment and include cohort studies, case control studies and cross-sectional studies. Again, reviewing the study design is key. A provider needs to examine if the dataset is appropriate to determine if the study is performed on the right population and whether the findings are broadly generalizable. Additionally, the provider will need to identify exposures and outcomes of interest. Dr. Barbieri communicated that when reviewing a study, it is always important to determine if an observed effect is clinically meaningful. Lastly, providers need to consider any limitations of the study through identification of any confounding variables or biases.
Pictured: John Barbieri, MD, MBA, Director of the Advanced Acne Therapeutics Clinic at Brigham and Women’s Hospital
Byline: Sarah B.W. Patton, PA-C