Beyond P Values

To estimate the minimal important difference (MID) of the Comprehensive Complication Index (CCI®) in patients undergoing abdominal surgery. The CCI® is a validated metric that quantifies cumulative surgical morbidity. While the CCI® is a sensitive endpoint to detect treatment effects, a statistically significant effect does not necessarily translate into clinical relevance. Relevant differences from the patients’ perspective are best captured by the MID. Individual patient data were extracted from surgical studies reporting CCI® at 30 days and using patient-reported outcome measures with established MIDs at baseline and 30 days. To determine the MID for the CCI®, we used an anchor-based approach as recommended by methods guidelines. A patient-reported outcome measure was selected as an anchor only if the Spearman correlation coefficient between its change in score (baseline to 30 days postoperative) and the CCI® was ≥|0.30|. We used linear regression to estimate the MID of the CCI® across different anchors, and triangulation to determine a single MID. Data were extracted from 3 published randomized controlled trials and 1 prospective observational study (n = 1583 patients) in major abdominal surgery. In colorectal surgery cohorts, 2 subscores of the Short Form-36, 2 subscores of the Multidimensional Fatigue Inventory-20, the EuroQol-5-Dimension Index Score, and the EuroQol Visual Analog Scale showed a correlation with the CCI® of ≥|0.30|. This resulted in MID estimates for the CCI® ranging from 6.1 to 22.2. In hepato-pancreato-biliary surgery, 1 subscore of the Short Form-36, and 2 subscores of the Patient Reported Outcome Measure Information System-29 questionnaire qualified as anchors providing MID estimates ranging from 6.2 to 13.8. We propose a mean difference of 12 points in the CCI® between treatment groups as a relevant difference in patients undergoing abdominal surgery. This MID provides an important foundation for sample size calculations and interpretation of randomized controlled trials and large real-world observational studies.