When and how to discontinue bracing treatment in adolescent idiopathic scoliosis: results of a survey

Deep learning-based identification of spine growth potential on EOS radiographs

OBJECTIVES

To develop an automatic computer-based method that can help clinicians in assessing spine growth potential based on EOS radiographs.

METHODS

We developed a deep learning-based (DL) algorithm that can mimic the human judgment process to automatically determine spine growth potential and the Risser sign based on full-length spine EOS radiographs. A total of 3383 EOS cases were collected and used for the training and test of the algorithm. Subsequently, the completed DL algorithm underwent clinical validation on an additional 440 cases and was compared to the evaluations of four clinicians.

RESULTS

Regarding the Risser sign, the weighted kappa value of our DL algorithm was 0.933, while that of the four clinicians ranged from 0.909 to 0.930. In the assessment of spine growth potential, the kappa value of our DL algorithm was 0.944, while the kappa values of the four clinicians were 0.916, 0.934, 0.911, and 0.920, respectively. Furthermore, our DL algorithm obtained a slightly higher accuracy (0.973) and Youden index (0.952) compared to the best values achieved by the four clinicians. In addition, the speed of our DL algorithm was 15.2 ± 0.3 s/40 cases, much faster than the inference speeds of the clinicians, ranging from 177.2 ± 28.0 s/40 cases to 241.2 ± 64.1 s/40 cases.

CONCLUSIONS

Our algorithm demonstrated comparable or even better performance compared to clinicians in assessing spine growth potential. This stable, efficient, and convenient algorithm seems to be a promising approach to assist doctors in clinical practice and deserves further study.

CLINICAL RELEVANCE STATEMENT

This method has the ability to quickly ascertain the spine growth potential based on EOS radiographs, and it holds promise to provide assistance to busy doctors in certain clinical scenarios.

KEY POINTS

• In the clinic, there is no available computer-based method that can automatically assess spine growth potential. • We developed a deep learning-based method that could automatically ascertain spine growth potential. • Compared with the results of the clinicians, our algorithm got comparable results.

Does the Use of Sanders Staging and Distal Radius and Ulna Classification Avoid Mismatches in Growth Assessment with Risser Staging Alone?

BACKGROUND

Although Risser stages are visible on the same radiograph of the spine, Risser staging is criticized for its insensitivity in estimating the remaining growth potential and its weak correlation with curve progression in patients with adolescent idiopathic scoliosis. Risser staging is frequently accompanied by other skeletal maturity indices to increase its precision for assessing pubertal growth. However, it remains unknown whether there is any discrepancy between various maturity parameters and the extent of this discrepancy when these indices are used concurrently to assess pubertal growth landmarks, which are important for the timing of brace initiation and weaning.

QUESTIONS/PURPOSES

(1) What is the chronologic order of skeletal maturity grades based on the growth rate and curve progression rate in patients with adolescent idiopathic scoliosis? (2) What are the discrepancies among the grades of each maturity index for indicating the peak growth and start of the growth plateau, and how do these indices correspond to each other? (3) What is the effectiveness of Risser staging, Sanders staging, and the distal radius and ulna classification in assessing peak growth and the beginning of the growth plateau?

METHODS

Between 2014 and 2017, a total of 13,536 patients diagnosed with adolescent idiopathic scoliosis were treated at our tertiary clinic. Of those, 3864 patients with a radiograph of the left hand and wrist and a posteroanterior radiograph of the spine at the same visits including initial presentation were considered potentially eligible for this study. Minimum follow-up was defined as 6 months from the first visit, and the follow-up duration was defined as 2 years since initial consultation. In all, 48% (1867 of 3864) of patients were eligible, of which 26% (485 of 1867) were excluded because they were prescribed bracing at the first consultation. These patients visited the subsequent clinics wearing the brace, which might have affected body height measurement. Six percent (117 of 1867) of eligible patients were also excluded as their major coronal Cobb angle reached the surgical threshold of 50° and had undergone surgery before skeletal maturity. Another 21% (387 of 1867) of patients were lost before minimum follow-up or had incomplete data, leaving 47% (878) for analysis. These 878 patients with 1139 skeletal maturity assessments were studied; 74% (648 of 878) were girls. Standing body height was measured in a standardized manner by a wall-mounted stadiometer. Several surgeons measured curve magnitude as per routine clinical consultation, skeletal maturity was measured according to the distal radius and ulna classification, and two raters measured Risser and Sanders stages. Reliability tests were performed with satisfaction. Data were collected for the included patients at multiple points when skeletal maturity was assessed, and only up to when brace wear started for those who eventually had bracing. The growth rate and curve progression rate were calculated by the change of body height and major coronal Cobb angle over the number of months elapsed between the initial visit and next follow-up. At each skeletal maturity grading, we examined the growth rate (in centimeters per month) and curve progression rate (in degrees per month) since the skeletal maturity assessment, as well as the mean age at which this maturity grading occurred. Each patient was then individually assessed for whether he or she was experiencing peak growth and the beginning of growth plateau at each timepoint by comparing the calculated growth rate with the previously defined peak growth rate of ≥ 0.7 cm per month and the beginning of growth plateau rate of ≤ 0.15 cm per month in this adolescent idiopathic scoliosis population. Among the timepoints at which the peak growth and the beginning of growth plateau occurred, the median maturity grade of each maturity index was identified as the benchmark grade for comparison between indices. We used the McNemar test to investigate whether pubertal growth landmarks were identified by specific maturity grades concurrently. We assessed the effectiveness of these skeletal maturity indices by the difference in proportions (%) between two benchmark grades in indicating peak growth and the growth plateau.

RESULTS

For girls, the chronological order of maturity grades that indicated peak growth was the radius grade, ulna grade, Sanders stage, and Risser stage. Curve progression peaked between the age of 11.6 and 12.1 years at a similar timing by all maturity indices for girls but was inconsistent for boys. For both sexes, radius (R) grade 6, ulna (U) grade 5, Sanders stage (SS) 3, and Risser stage 0+ were the median grades for peak growth, whereas Risser stage 4, R8/9, U7/8, and SS6/7 indicated the beginning of the growth plateau. The largest discrepancy between maturity indices was represented by Risser stage 0+, which corresponded to six grades of the Sanders staging system (SS2 to SS7) and to R6 in only 41% (62 of 152) of girls in the whole cohort. Despite Risser stage 0+ corresponding to the wide range of Sanders and distal radius and ulna grades, none of the R6, U5, SS3, and Risser stage 0+ was found more effective than another grade in indicating the peak growth in girls. R6 most effectively indicated the peak growth in boys, and Risser stage 0+ was the least effective. For the beginning of the growth plateau in girls, SS6/7 was the most effective indicator, followed by U7/8. Risser stage 4 was the least effective because it indicated 29% (95% CI 21% to 36%; p < 0.001) fewer patients who reached the beginning of the growth plateau than did those with R8/9. Risser stage 4 also indicated 36% (95% CI 28% to 43%; p < 0.001) fewer patients who reached the beginning of the growth plateau than those indicated by U7/8, and it identified 39% fewer patients than SS6/7 (95% CI 32% to 47%; p < 0.001). For boys, similarly, R8/9, U7/8, and SS6/7 were all more effective than Risser stage 4 in identifying when the growth plateau began.

CONCLUSION

Risser stage 0+ corresponds to a wide range of Sanders and distal radius and ulna grades. Risser stage 0+ is least effective in indicating the peak growth in boys, and Risser stage 4 is the least effective maturity grade for indicating when the growth plateau starts in both sexes. The concurrent use of R6 and SS3 can be useful for detecting the peak growth, and SS6/7 in conjunction with U7/8 is most effective in indicating the beginning of the growth plateau. Using a combination of specific grades of Sanders staging and the distal radius and ulna classification can indicate pubertal growth landmarks with reduced risk of underestimating or overestimating skeletal maturity. These findings may aid in refining clinical decision-making of brace initiation and weaning at a more precise timing. Among Risser stage 0, the appearance of R6, U5, and SS3 provide the most effective assessment of peak growth that can indicate the most effective bracing period within which curve progression occurs. For initiation of the growth plateau, Risser 4 is not useful, and SS6/7, R8/9 and U7/8 should be used instead.

LEVEL OF EVIDENCE

Level III, diagnostic study.

Bracing In The Treatment Of Adolescent Idiopathic Scoliosis: Evidence To Date

Brace effectiveness for adolescent idiopathic scoliosis was controversial until recent studies provided high quality of evidence that bracing can decrease likelihood of progression and need for operative treatment. Very low evidence exists regarding bracing over 40^ο and adult degenerative scoliosis. Initial in-brace correction and compliance seem to be the most important predictive factors for successful treatment outcome. However, the amount of correction and adherence to wearing hours have not been established yet. Moderate evidence suggests that thoracic and double curves, and curves over 30^ο at an early growth stage have more risk for failure. High and low body mass index scores are also associated with low successful rates. CAD/CAM braces have shown better initial correction and are more comfortable than conventional plaster cast braces. For a curve at high risk of progression, rigid and day-time braces are significantly more effective than soft or night-time braces. No safe conclusion on effectiveness can be drawn while comparing symmetrical and asymmetrical brace designs. The addition of physiotherapeutic scoliosis-specific exercises in brace treatment can provide better outcomes and is recommended, when possible. Despite the growing evidence for brace effectiveness, there is still an imperative need for future high methodological quality studies to be conducted.