and 4cp curve types did not differ significantly in function (Table 2 and Additional file 4).SRS-22r totalThe covariates retained included height, SEQ and age, but only age had a significant main effect, such that for every 1-year increase in age, the SRS-22r total dropped by 0.08 (p = 0.047) (Table 2 and Additional file 4).SAQ prominenceTo meet the normality assumption, the SAQ prominence was transformed to its square root. Only classification was retained as covariate, with 3cp having a significant main effect. Best scores were observed in patients classified as 3c. Those with a 3cp curve type had the worst scores of all and were the only type significantly worse than the 3c classifica- tion (0.25, p = 0.00) (Table 2 and Additional file 4).SAQ waistSAQ waist was transformed to its power of −0.3 to meet-0.3The patients who scored ≥35 of the SEQ had signifi- cantly worse SAQ waist scores than the ones with lower self-efficacy (0.10, p = 0.01). Likewise, patients who wore brace had worse waist scores than those without (0.08, p = 0.03).SAQ trunk shiftAge, height and curve type were retained as covariates. Patients aged 10 and 11 behaved differently. To address this difference, age was divided into covariates Age and Age 10–11, but only Age 10–11 had a significant main effect. Ten and 11 years old patients had better scores on average by 0.83 points than their older counterparts (p=0.00). Taller patients also had better scores (−1.89, p = 0.02) where for every 1 cm increase in height, patients had better score by 0.02. Patients with 3c curve types had significantly better score compared to patients with 3cp and 4cp curve patterns (by 0.49, p = 0.01 and 0.36, p = 0.047, respectively) (Table 2 and Additional file 4).SAQ generalHeight and brace wear were retained as covariates. Patients who wore a brace had significantly better scores by 0.72 (p = 0.00) (Table 2 and Additional file 4).SAQ curveCovariates included brace wear and classification. The model predicted that persons classified into 3cp havethe normality assumption—(SAQ waist)retained self-efficacy and brace-wear covariates. Self- efficacy was divided into covariates SEQ (overall effect of SEQ) and SEQ2 (the effect of SEQ scores when ≥35) (Table 2 and Additional file 4).. The modelSchreiber et al. Scoliosis (2015) 10:24Page 8 of 12about 90 % higher odds of having score of >3, indicating worse outcomes (p = 0.01) (Table 2 and Additional file 4).DiscussionThis is the first RCT investigating the effect of Schroth exercises on SRS-22r, SAQ questionnaires’ scores and back muscle endurance. Schroth exercises added to standard of care improved the SRS-22r pain scores and back muscle endurance after 3 months, and the self- image scores after 6 months of intervention. The Schroth intervention did not have significant effect on other outcomes.In the only prospective study on Schroth exercises that examined the back muscle properties, strength, rather than endurance, was assessed using manual muscle test- ing scores ranging from 1 to 5 [31]. Otman et al. found that muscle strength increased significantly after 1 year. In two other studies, supervised resistive rotational exer- cises significantly increased strength after 4 months [30, 63]. To our knowledge, for the first time, Schroth exer- cises combined with the standard of care have been demonstrated to increase the back extension endurance.Only one other study retrospectively investigating Schroth exercises and spinal stabilization compared to stabilization alone tested their effect on the SRS-22r, but not the SAQ questionnaire [33]. Noh et al. reported bet- ter SRS-22r results at 4 months for both groups but the experimental group demonstrated greater benefits, but significantly only for self-image (from 3.3 ± 1.2 to 4.2 ± 1.0) and the total score (from 3.8 ± 1.8 to 4.5 ± 0.4). Monticone et al’s recent RCT found positive effects of scoliosis-specific active self-correction and task-oriented exercises on changes on the Cobb angles and SRS-22r scores at skeletal maturity in 110 patients with AIS and curves