What Injury Mechanism and Patterns of Ligament Status Are Associated With Isolated Coronoid, Isolated Radial Head, and Combined Fractures?

To the Editor,

We read the study conducted by Rhyou and colleagues [2] with great interest and would like to congratulate the authors for their valuable work.

Although overall the study was well done, we found some methodological issues that should be pointed out in order to avoid misinterpretation. We are concerned about the estimated odds ratios (ORs) and confidence intervals (CIs) for the parameters in Tables 1 and 2 in the study. We believe the ORs and CIs were not estimated correctly. We wonder how the authors came up with an OR (95% CI) of 34.5 (3.8–333.3) for medial collateral ligament complete rupture in Table 1 or 46.7 (4.9–415.2) for anteromedial facet in Table 2. We request the authors consider reanalyzing their data.

Even if we assume the estimated ORs and CIs in the study are correct, they are still questionable. In the study, the estimated ORs for the parameters in Tables 1 and 2 were large and the estimated CIs were wide, which is likely due to sparse-data bias [1]. In other words, there are not enough cases in each combination of exposure and outcome to estimate a valid OR [1]. A major limitation of the study is that the authors did not attempt to conduct a sensitivity analysis to test how much the results estimated were the result of a small-sample problem. Greenland and colleagues [1] showed that penalization via data augmentation can limit sparse-data bias effectively. Therefore, we suggest Rhyou and colleagues reanalyze their data to minimize sparse-data bias using the penalization approach as suggested by Greenland and colleagues [1]. With this approach the authors also can estimate valid and robust ORs and CIs for medial bone contusion, which was found to be not applicable in the study.