Distal radius fracture (DRF) is one of the most common fractures worldwide. In Korea, the annual incidence of DRF is reported to be about 130,000 [1]. Although the incidence of conservative treatment compared to surgery is decreasing, conservative treatment remains an important option, especially for elderly patients [2,3]. However, a major issue encountered after closed reduction for conservative management is the redisplacement of the fracture. Redisplacement after a closed reduction of DRF has been reported in 58% to 89% of cases [4]. Even in the absence of early reduction loss, late reduction loss is commonly reported in elderly patients [5]. When the clinical treatment plan changes to surgery due to loss of reduction, patients often express disappointment due to the increased overall cost and treatment duration. Therefore, predicting reduction loss is critical for the success of conservative treatment. Previous studies using plain radiography have shown that factors such as dorsal comminution, associated ulnar fracture, volar cortical integrity, radial shortening, and intra-articular fracture are related to the redisplacement of DRF [5-8]. Among these, previous meta-analyses have indicated that only dorsal comminution and age over 60 years are significant risk factors for secondary displacement [9,10]. However, we believe there are challenges with measuring dorsal comminution on plain radiographs. First, there is no consensus on the definition of dorsal comminution. Second, the interpretation of plain radiography can be vague and subjective. Third, there has been a lack of reliable research regarding each predictive factor.

Therefore, in this study, we aimed to evaluate and compare the reliability of dorsal metaphyseal comminution as a predictive factor of DRF in plain radiography.

We found the incidence of dorsal comminution to be 198 of 440 (45.0%) in plain radiographs analyzed by four raters. When applying the three criteria mentioned above (DC, DC1, and DC2), the frequency of dorsal comminution showed a relatively wide difference among the four raters, ranging from 25.0% to 55.0% (Table 1).

Based on Fleiss’s or Cohen’s kappa values, all predictors of intra-rater reliability were higher than inter-rater reliability. The DCs criterion showed fair to good reliability, while DC1 was found to have fair reliability in both inter-rater (kappa value, 0.265) and intra-rater (kappa value, 0.346) assessments. The intra-rater reliability of DC2 showed the highest kappa value of 0.737, but the inter-rater reliability of DC2 was the lowest (0.175) (Table 2).

The inter-rater kappa values in Table 2 were calculated based on data from all raters. We further compared these values according to rater experience (experienced surgeons vs. novice doctors). Notably, the inter-rater kappa values between the two experienced surgeons (range, 0.316–0.386) were significantly higher than those between the two novice surgeons (range, 0.049–0.192). Overall, all inter-rater kappa values for experienced surgeons exceeded those for novice doctors (Table 3).

In this study, we aimed to assess the reliability of post-reduction predictors of dorsal comminution based on plain radiography. For a predictor to be widely used in clinical decision-making, reproducibility is a critical factor, and the results should not vary across observers. However, we found that the inter-rater reliability of dorsal comminution predictors (kappa values, 0.175 to 0.386) was much lower than each intra-rater reliability (kappa values, 0.346 to 0.737) (Table 2). The low inter-rater reliability can lead to varying outcomes among researchers and confusion in interpreting research results. Moreover, discrepancies among doctors can cause patients to distrust their clinicians.

We focused particularly on dorsal comminution because it is considered one of the most important radiologic factors for predicting reduction loss. This is why the Edinburgh Wrist Calculator (EWC) includes the presence of dorsal comminution as a predictive factor, alongside age and ulnar variance. However, many previous studies have reported that the EWC is a poor predictor of fracture redisplacement [13,14]. Furthermore, some papers argue that dorsal comminution is not a risk factor for instability [10,15]. We believe that the low validity of the EWC and conflicting opinions regarding dorsal comminution as a prognostic factor can be attributed to the low reliability of the measurement index, as there are no objective tools to measure dorsal comminution [6].

Additionally, we found that clinical experience significantly influences the reliability of radiographic interpretation (Table 3). It is supported by a previous study, which found that novice students had accuracy rates ranging from 35% to 85%, whereas graduates exhibited a narrower range of 60% to 90% [16]. This difference can be attributed to the structured organization of knowledge regarding fracture pattern and surgical experience [17]. Moreover, when accounting for the surgeon’s experience, the reliability of dorsal comminution assessment decreases even further.

To overcome this limitation, Rhee and Kim [18] previously devised a quantifying tool called the metaphyseal collapse ratio (MCR). The MCR value is calculated by dividing the maximal radiolucent width by the intercortical distance. This index showed good inter- and intra-observer agreement. However, it did not provide a prognostic analysis for reduction loss. In this regard, LaMartina et al. [6] found that volar cortical integrity can be a strong predictive factor for maintaining alignment after closed reduction. However, these studies did not perform inter- or intra-observer reliability measurements and there are controversies surrounding its predictive ability in previous studies [19,20].

We believe it is difficult to accurately predict reduction loss based solely on plain radiography before conservative treatment, as many factors—including age, osteoporosis, socioeconomic status, muscle strength, and fracture pattern—affect reduction loss [21]. For a more accurate prediction, we think a new type of post-reduction predictor is needed. Computed tomography (CT) scans could serve as a valuable auxiliary tool for predicting reduction loss to address the limitations of plain radiography [22-24]. In the future, methods based on artificial intelligence (AI) and machine learning could help solve this problem [25].

This study has some limitations. First, the small number of patients could lead to statistical errors. 40 subjects were also not enough to perform power analysis. Second, this study did not include a prognostic factor analysis regarding closed reduction loss; therefore, the definitions in our study are not suitable to be used as primary predictors of instability. Third, the three definitions for dorsal comminution were assessed without intervals, which may have introduced bias in the evaluation of the X-rays. Especially, the DCs index is subjective and can vary between raters. While this approach might reflect real-world clinical practice and insight, it lacks objectivity and reproducibility.

The radiological predictors of dorsal comminution evaluated in this study exhibited overall low reliability, limiting the usefulness in predicting reduction loss in DRFs. A more precise definition of dorsal comminution is necessary, and additional methods, such as CT and AI, should be considered to enhance reliability.