RAD represents one of the most frequent causes of secondary osteoarthritis of the hip and is not rare in DDH patients. More than one-third of DDH patients treated with CR have an outcome of RAD, and there is a higher prevalence of RAD with increasing age at reduction. In our study the incidence of RAD was 73.6% (92/125), higher than reported in previous studies [6, 9, 23]. This might be the result of the shorter follow-up time (44.2 ± 14.5 months) and older age (18.3 ± 3.9 months) of the cohort at CR.
RAD is often asymptomatic and the diagnosis relies on radiography. Several radiographic parameters such as AI, CEA, centre-head distance discrepancy (CHDD) and Reimer’s index (RI) have been investigated for predicting RAD. However, there are some disadvantages to those parameters: the measurements of CEA, CHDD and RI are influenced by the shape of the femoral head and the center of the femoral head must be identified precisely. If the ossific nucleus of the femoral head has not appeared, is irregular or is very small, the measurement is unreliable. In addition, CEA is validated for patients aged ≥ 4 years and CHDD is useless in bilateral cases [8, 9, 14].
AI is one of the most popular parameters for predicting RAD [6, 15], and a multi-center study suggested that AI would be the best predictor . The measurement of AI is based on a horizontal Hilgenreiner line that runs through the triradiate cartilage and a line extending from the superolateral margin of the triradiate cartilage to the most lateral ossified margin of the acetabulum. However, the Hilgenreiner line is difficult to measure after ossification of the triradiate cartilage when the patient is older than 8 years, and it is difficult to mark the exact lateral bony margin of the acetabulum because of irregular and indistinct bony shadows on radiographs of dysplastic hips. Furthermore, the value of AI can be affected by the position in which the radiograph was taken, and it is difficult to keep an uncooperative child in the correct position for the radiograph. With increasing lumbar lordosis, which corresponds to the extent of pelvic extension, the AI increases, and with decreasing lordosis or pelvic flexion, the AI decreases. Similarly, if the pelvis is rotated about the longitudinal body axis, the AI toward the side of the rotation decreases and the AI on the opposite side increases . It is often necessary to follow up the AI for a long time to decide when to initiate the following procedure because the development of the acetabulum is a long and slow process. The surgeon needs an earlier and more definitive index to help determine when to carry out the next step of treatment to promote the development of the hip.
Considering the limitations of singular prognosticating factors, Kim et al.  tried to combine the orientation of the sourcil and CHDD to predict the RAD and created a risk scale. However, a limitation to this approach has been raised in that CHDD is only useful in unilateral cases .
The acetabular teardrop is a landmark seen in the inferior medial acetabulum on the AP X-ray of the pelvis. The timing of its appearance and configuration of the teardrop have been correlated with various pediatric hip disorders and could predict the outcome of DDH after CR [16–19]. The sourcil is a curved area of dense bone on the weight-bearing surface of the acetabulum and represents a stress distribution within the hip joint indirectly. In a normal hip, the sourcil is uniformly thick and semilunar in shape, with a horizontal or downward orientation. However, in the dysplastic hip, the orientation of the sourcil is upward, suggesting an uneven distribution of stress within the hip joint. The shape and orientation of the sourcil have also been used to predict RAD [8, 20].
In this study, we tried to find a more reliable and easier parameter to measure. TSL is the curve that makes up the lateral margin of the teardrop and the curve of the sourcil. The results showed that a continuous TSL can be a predictor of RAD (p = 0.001). Compared to CHDD, CEA, RI and AI, TSL is not influenced by the shape of femoral head, does not require identification of the exact point at the center of the femoral head and the lateral bony margin of the acetabulum, and can be used in bilateral cases. The improvement of AI post-reduction by a certain age is a reliable predictive factor of the need for later acetabuloplasty [7, 8], and many studies have focused on the cutoff values of AI at different time points post-reduction to predict RAD [6, 9, 15]. In this study, there was no significant difference in the initial AI between the TSL continuous group and discontinuous group, but the level of AI after CR was lower in the TSL continuous group. This meant AI was more likely to become normal in the TSL continuous group and TSL could predict the potential of AI improvement after CR, so that TSL can be an earlier factor than AI for predicting RAD. However, in the TSL continuous group, there was no significant difference between RAD and non-RAD hips at the time point when TSL became continuous after CR (19.5 ± 8.3 vs. 22.4 ± 14.6 months, p = 0.485). Unlike other parameters, there is not a cutoff value of the time when TSL becomes continuous after CR that determines progression to RAD.
The present study had limitations. It was a single center retrospective study, the number of patients was small and the follow-up time was short. Theoretically, there is an upper time limit for TSL becoming continuous after CR that would predict whether RAD occurs or not, but we did not find such a cutoff value in this study. More patients and longer follow-up will be needed in order to further analyze the effectiveness of TSL as a predictive factor.