Novel Mri Classification for Osteochondritis Dissecans of the Knee: A Validation Study

Abstract

Background:

A simpler three-category magnetic resonance imaging (MRI) classification system for osteochondritis dissecans (OCD) of the knee was recently shown to have comparable reliability to the existing five-category Hefti system. The validity and clinical utility of this simpler system as an alternative to the Hefti system has yet to be established.

Hypothesis/Purpose:

The purpose of this study was to assess whether a novel, simpler three-category MRI classification system for OCD of the knee demonstrates equal or better validity and correlates more strongly with treatment than the five-category Hefti classification.

Methods:

Demographic data and arthroscopic findings were collected from the medical and surgical records of 144 knees of children and adolescents with arthroscopically diagnosed knee OCD. MRI assessment of OCD lesions was conducted by two independent raters. Inter-rater reliability for novel and Hefti classifications was assessed by estimating weighted kappa (k_w ). Agreement between MRI classification and arthroscopic findings was assessed by estimating k_w coefficients. Classifications were dichotomized into 1,2 versus 3 for the novel system, and the sensitivity and specificity of classification was estimated as compared to arthroscopic findings. Correlation between arthroscopic classification and treatment type was assessed by estimating Spearman’s coefficient.

Results:

There was substantial interrater agreement with the proposed system’s ratings on MRI (k_w =0.66; 955 CI=0.56-0.75) and moderate agreement for Hefti system’s ratings on MRI (k_w =0.57; 955 CI=0.47-0.67) (Table 1). There was no difference detected in the agreement statistics for the proposed versus Hefti classifications (p=0.89). Binary agreement of the novel classification using dichotomous categories was slightly worse than the three-category classification. When 1&2s were combined, the agreement was moderate with k=0.41 (95% CI=0.25-0.58) and when 2&3s were combined, the agreement was fair with k=0.34 (95% CI=0.21-0.48). There was strong correlation between novel classification on arthroscopy with treatment type (r=0.85; 95% CI = 0.80-0.89) and between Hefti classification and treatment type (r=0.82; 95% CI = 0.75-0.86) (Table 2).

Conclusion:

The validity and clinical utility of the proposed three-category MRI classification system for knee OCD is comparable to that of the five-category Hefti system. This simpler classification system provides a foundation for the subsequent establishment of an operative treatment algorithm for knee OCD.

Table 1.

Summary of MRI and arthroscopic findings based on one random measurement per patient.

Table 2.

Agreement between arthroscopic classification and treatment.