Reliability of radiographic criteria for trochlear dysplasia: intra- and interobserver analysis on 68 knees

Purpose of the study The relation between patello-femoral instability and trochlear dysplasia wa s identified by Dejour. Trochlear dysplasia, diagnosed on knee lateral Xray when the trochlear groove crosses both femoral condyles (the so-called "cr ossing sign"), must be corrected to improve patello-femoral stability. Howe ver surgery should be related to the severity and the shape of trochlear dy splasia, underlining the importance of a reproducible classification. The a im of this study was to establish intra and inter- observer reliability of Dejour's radiographic criteria. Material 68 preoperative exact knee profile radiographs were harvested from clinical records of 64 patients who underwent trochleoplasty because of patello-fem oral instability and trochlear dysplasia. On these 68 views, the crossing s ign was identified by the senior surgeon (F.G.) who performed or supervised surgery. Method The 68 radiographs were examined independantly by 7 observers (2 juniors, 5 seniors) in order to assess interobserver agreement. Two juniors repeated the observation to test intraobserver agreement. Reproducibility for catego rical data (7 shapes of trochlea according to Dejour (3 for dysplasia)) was evaluated by Kappa statistics, and for numerical data (depth and anterior projection of the trochlear groove with respect to anterior femoral cortex) we used the interclass correlation analysis. Results Two out of the 7 observers rated all the 68 trochleas as dysplastics. The 5 others rated as normal 1 to 6 trochleas out of the 68. None of the 68 troc hleas were recognized with the same shape by the 7 examiners. At best, 6 ob servers agreed on the same shape and for only 12 trochleas. Disagreement wa s mostly related to mistakes between type I and type II of dysplasia. For t rochlear morphology interobserver agreement was slight (Kappa = 0.17) and i ntraobserver agreement was fair (Kappa = 0.3). The mean prominence of the t rochlea was 3 +/- 2.1 mm [-6 to 10], and the mean trochlea depth was 1 +/- 1.9 mm [0 to 11]. These measurements were more reliable since the interclas s correlation coefficients were respectively 0.62 and 0.38. The level of ex perience of the observers had no influence for categorical or numerical dat a. Discussion Our results indicated a low interobserver agreement for trochlear shape ide ntification according to Dejour. The most reliable criteria was measurement of the trochlear prominence which was mostly pathological in our series. T he "crossing sign" was reliable to diagnose dysplasia since the probability to rate as normal a true dysplastic trochlea was only 3.1 per cent, Howeve r, once the dysplasia diagnosed, this classification gave inconsistent resu lts to select the trochlear shape, particularly for type ii. To improve rep roducibility we propose to diagnose a type II only when 5 millimeters separ ate the crossings between the medial and lateral condyles. Conclusion We recommand to use anterior projection of the trochlear groove to rate tro chlear dysplasia and to determine the adequate type of trochleoplasty: elev ating of the lateral facet if non prominent or deepening of the groove when prominent.