Discussion: Why the p Value Alone Is Not Enough: The Need for Confidence Intervals in Plastic Surgery Research

    loading  Checking for direct PDF access through Ovid

Excerpt

Over the past 20 years, there has been increasing attention on evidence-based medicine, with a resultant emphasis on stringent epidemiologic standards for patient-oriented research.1 Plastic surgery has embraced this trend in an effort to improve clinical practice and patient outcomes.1–3 In their work, Samargandi et al. evaluated the variability in reporting measures of statistical significance and precision within the plastic surgery literature, namely, the p value and confidence intervals as reported in clinical trials.4 They evaluated a randomly selected group of 135 clinical trials from the seven highest impact journals in plastic surgery to determine the frequency of reporting confidence interval with measures of association. Their findings suggest that confidence intervals are underreported; factors associated with increased likelihood of reporting the confidence interval were a coauthor with formal research training and higher methodology scores.
Although plastic surgery as a specialty has embraced evidence-based practices for establishing treatment guidelines and protocols, much work remains to be done in this area. Various organizations, including the Enhancing the Quality and Transparency of Research network, have established reporting guidelines for patient-oriented research.5 Their subgroups, based on study design, may be familiar to some researchers (e.g., Consolidated Standards of Reporting Trials, Strengthening the Reporting of Observational Studies in Epidemiology, Preferred Reporting Items for Systematic Reviews and Meta-analyses), but have not yet been fully embraced by the plastic surgery community.6 The Consolidated Standards of Reporting Trials checklist for reporting a randomized clinical trial includes 25 items that authors need to consider when preparing a report for this study type.7 The checklist specifies that outcomes should include an estimate of effect size (e.g., relative risk, absolute risk reduction, odds ratio) and a measure of precision, such as the confidence interval.
Knowledge and competency with basic epidemiologic concepts is a must for the surgeon who wishes to practice evidence-based medicine. Surgeons today face patient populations that have immense resources of widely varying quality at their disposal. Clinical decision-making may be questioned by patients on the basis of their anecdotal knowledge or Web-based searches. It is imperative for surgeons to have the tools to critically assess the validity of scientific work and provide clear recommendations to patients on the basis of the highest quality data available. An understanding of concepts such as the p value and confidence interval, and their strengths and limitations, is paramount to this goal. Indeed, recognition of the importance of these concepts by the American Society of Plastic Surgeons has led to the integration of epidemiology and biostatistics into the annual in-service examination.
Although the authors’ recommendation for reporting the 95 percent confidence intervals as standard practice should be strongly considered, there are some limitations with this work that merit mention. The 95 percent confidence interval, although arguably the most ubiquitous measure of precision, is not the only measure of precision. Standard deviations and standard errors, although related to the calculation of confidence intervals, are also independent measures of precision. Although the purpose of the study was to evaluate the specific use of confidence intervals, it should be recognized that by ignoring measures other than the confidence interval, their analysis has the potential to underestimate the reporting of measures of precision within the plastic surgery literature overall. Second, although the authors were justified in using only selected journals related to plastic surgery (i.e., those with the highest impact factors), we cannot assume that the propensity to report the confidence interval would be directly correlated with the journal impact factor.
    loading  Loading Related Articles