Intrarater and Interrater Reliability of Three Published Versions of Eaton Classification for Trapeziometacarpal Joint Osteoarthritis.

PURPOSE: This study aimed to identify the formal differences between the three published versions of the Eaton classification system for trapeziometacarpal joint osteoarthritis and to assess their intra- and interrater reliability using true lateral thumb radiographs.

METHODS

100 lateral thumb radiographs from consecutive patients (72% women; median age, 63 years; interquartile range, 20 years) were assessed independently. Eight observers (4 hand surgeons, three residents, and one medical student) classified each radiograph according to the original textual descriptions of the Eaton-Littler (1973), Eaton et al. (1984), and Eaton-Glickel (1987) classifications. All observers completed a second round of assessment after two weeks. Intrarater and interrater reliability was calculated using Cohen's and Fleiss' κ and a subgroup analysis of stage-specific reliability was performed. A linear mixed-effects model evaluated the effects of classification version and rater training level on intrarater reliability.

RESULTS

The κ values (95% confidence intervals) for intrarater reliability ranged from 0.51 (0.41; 0.61) to 0.98 (0.95; 1.00). The mixed-effects analysis showed that intrarater reliability of the Eaton-Littler classification was significantly lower than that of Eaton et al. (P = .040). The Eaton et al. and Eaton-Glickel classifications yielded nearly identical intrarater results. Training level had no significant effect on intrarater reliability. The interrater reliability (95% confidence interval) for the Eaton-Littler classification demonstrated fair agreement (0.39 [0.37; 0.42]), whereas the Eaton et al. and Eaton-Glickel classifications showed moderate agreement (0.45 [0.43; 0.47]).

CONCLUSIONS

Although observers may achieve acceptable intrarater reproducibility according to their own interpretation of the classification system, all three Eaton classifications involve limited interrater reliability, especially due to low reliability for intermediate stages II and III. Therefore, reliance on these classifications to guide treatment decisions is not supported based on the findings of the present study.

TYPE OF STUDY/LEVEL OF EVIDENCE: Diagnostic III.