A retrospective case-series of adolescent idiopathic scoliosis patients with 5 to 15 years follow-up after segmental pedicle screw instrumentation Authors: Ida Bromell Kindberg, Efe Levent Aras, Jon Kaspersen, Sasa Randelovics, Simon Toftgaard Skov, Søren Eiskjær Published: 2026 ABSTRACT AbstractBackgroundThe long-term outcome of surgery for adolescent idiopathic scoliosis (AIS) using segmental pedicle screw constructs is sparsely described in the scientific literature.MethodsRetrospective case series of 120 consecutive patients operated for AIS using segmental pedicle screw constructs. Age between 11 and 21 years of age at the time of surgery (25 males) and followed median 9 years after surgery. Outcome measures consisted of radiological variables, the SRS-22 and the EQ-5D-5L questionnaires, questions concerning work status and the most advanced educational level at last follow-up. An electronic questionnaire was answered in 120 of 148 cases (81%) and 112 (76%) showed up for the long-term radiology control. Preoperative PROMS were available from the national research register for spine surgery. Attributable to the incompleteness of the register only a third of the PROM data were paired.ResultsThe frequency of complications was 12.5%. Mean curve correction was 78% immediately preoperative with a 4° increase at last follow-up. SRS-22 total/self-image and the EQ-5D utility scores were significantly improved at last follow-up. At last follow-up 84% were satisfied or very satisfied with the results of the surgical intervention. Socioeconomic factors, Lenke type 2, number of levels instrumented and curve correction were important predictors for satisfaction in a logistic regression model (AUC 85%).ConclusionsOnly 16% were not satisfied with the results of AIS surgery after 5 to 15 years. Total PROM scores and self-image scores were significantly improved at last follow-up. Socioeconomic factors are important when analyzing the results of AIS surgery.Keywords: Adolescent idiopathic scoliosis, Pedicle screw instrumentation, Long-term outcome, Patient reported outcome, Socioeconomic factors, Logistic regression analysis Introduction Surgery for adolescent idiopathic scoliosis (AIS) has evolved considerably during the last decades. The evolution process has comprised segmental pedicle screw instrumentation, Ponte osteotomies, navigation, robotics, intraoperative monitoring, minimal invasive surgery, enhanced recovery after surgery, etc [1]. Unfortunately, the scientific literature concerning the long-term prognosis after AIS surgery is sparse. The literature often focuses on outdated techniques, radiographic detail analysis, do not report any PROMs and includes selective patient population or mixed instrumentation techniques. Most of the long-term studies are retrospective in nature, comprise relatively few patients, lack some data, the long-term follow-up rate is low and the PROMs have undergone some change over time [[2], [3], [4]]. Consequently, it is very difficult to answer any questions the young AIS patient or her/his parents may have regarding the prognosis 5 to 10 to 15 years after surgery for AIS with modern surgical techniques. Is the prognosis stable or can we expect an inevitable deterioration over time? Based on the general uncertainty mentioned above we really do not know if some subgroups fare worse than others. In the era of personalized medicine and shared decision making this can hardly be considered sufficient [5]. We hypothesize that the mid-to-long term outcome of AIS surgery will be stable or improved in comparison with the preoperative status and that we will be able to demonstrate a significant correlation between some of the variables and the final patient reported outcome. The primary aim of this study was to report the mid-to-long term follow-up radiographic and PROM outcome after posterior spinal fusion with segmental pedicle screw instrumentation in a cohort of AIS patients. The secondary aim was to correlate the PROM outcome to a number of variables in the data set using multivariable logistic regression analysis. Methods The current study was a retrospective case-series with a follow-up time between 5 and 15 years. All the surgeries were performed at the same tertiary university hospital between January 2009 and September 2020. A search was conducted in the national patient register using the International Classification of Diseases (ICD) codes for AIS (M41.12*) and the Nomesco Classification of Surgical Procedures (NCSP) 2011 codes for instrumented posterior spinal fusion (NAG7*) [6]. Individual-level data linkage was done using the unique personal identification number. The search identified 148 consecutive patients, which all were eligible for inclusion according to our inclusion criteria: Age ≥11 and ≤21 years at the time of surgery, minimum 5 years of follow-up and a confirmed diagnosis of idiopathic scoliosis. The patients had a median age of 25 years at last follow-up, range [18–34]. Ninety-five of the patients were female. The median age at the time of surgery was 15 years of age, range [12–21]. The time to the last follow-up was 9.0 years median, range [5–16]. The surgical procedure consisted of posterior spinal fusion with bilateral segmental pedicle screw instrumentation. The posterior release included Ponte osteotomies and facet joint resections. To reduce the curve we used a combination of translation, rod rotation, in situ rod bending and distraction/compression. 3D-scanning with cone beam CT and subsequent navigation was used routinely in most cases. Allograft bone transplant (femoral heads) was used in all cases. All procedures were performed by 1 or the other of the 2 senior orthopedic surgeons. The PROM data were acquired by automated survey invitations using Digital Post/e-Boks in combination with a customized REDCap database [7]. Digital Post is a public solution for safe communication between the national authorities and the citizens. The unique personal identification number is the key to any communication via Digital Post. Reminders were sent twice every fortnight using triggered e-mail. The patients answered the PROM questions by self-entry into the electronic survey system (the PROM part of the REDCap database) and gave consent to the anonymized use of their data by electronic signature. The radiological data were collected from the Picture Archiving and Communications System (PACS) and the electronic patient records (EPR) and were also registered in the REDCap database. An unbiased observer collected all the radiological data and the data from the EPR. 120 of the 148 patients (81%) answered the PROM questionnaire. Eight of the 120 patients declined to participate in the last radiology follow-up (7%). Fig. 1 demonstrates the flow of the study population. Flowchart. At the last follow-up the following PROMs were collected: The SRS-22 questionnaire [8], the EQ-5D-5L questionnaire [9], the Spinal Appearance Questionnaire (SAQ) [10], questions concerning the current work situation and questions concerning the most advanced educational level completed at last follow-up. Preoperative PROM data were collected from a national research registry for patients undergoing spine surgery [11]. Preoperative PROM data were available for the SRS-22 and the EQ-5D-3L. For 40 of the 120 patients the data from the mid-to-long term follow-up could be paired with preoperative SRS-22 and EQ-5D-3L DaneSpine data. For comparison between preoperative and mid to-long term PROMS we also used preoperative DaneSpine data from 57 patients (SRS-22) and 51 patients (EQ-5D-3L). The 57 and 51 unpaired patients supplying the preoperative data sets underwent surgery for AIS at our department after September 2020 and were not yet eligible for mid-to-long term follow up. The Spinal Appearance Questionnaire (SAQ) was not a part of DaneSpine. The preoperative EQ5D data were collected using the EQ-5D-3L questionnaire as a measure on the general health-related quality of life (HRQL) and the data at last follow-up were collected using the EQ-5D-5L questionnaire. The EQ-5D-5L enables a more nuanced description of the general health status (5 vs. 3 health states), decreases ceiling problems and the utilities are normally distributed [12]. Thus, for comparision it is necessary to translate the EQ-5D-3L utilities to EQ-5D-5L utility values. This was done using the validated reverse cross walk score (spread sheet available on the web site) [13]. The 2 last questions of the SRS-22 questionnaire concerns the satisfaction with the results of the surgical procedure. The preoperative values of SRS-22 question 21 and 22 are open to considerable interpretation uncertainty so we chose to compare the sum of the first 20 questions from before surgery to the concordant sum at last follow-up. Radiological data for 112 patients were available preoperatively, immediately postoperatively, 2 years after scoliosis surgery and at last follow-up. The collected data included: Lenke type, image modality, Cobb angles, thoracic kyphosis, sagittal vertical axis (SVA), instrumented levels, instrumentation system, number of screws, number of Ponte osteotomies, the use of navigation with the O-arm, adding-on, proximal, and distal junctional failure. From the EPR we acquired data on length of stay (LOS), complications, reoperations, blood loss and the use of intraoperative monitoring. All radiological and clinical data were extracted by an unbiased observer who had not had any part in the surgical interventions. For the statistical analysis we used R version 4.5.1 (R Foundation for Statistical Computing). Besides R base the R packages EQ5D, StepReg, robustrank, and gtsummary were used. Data are presented as mean and SD, medians and range/upper (Q3) and lower (Q1) quartiles or counts/frequencies as appropriate. We only used nonparametric test: Fisher’s exact test or the chi-square test, Wilcoxon rank sum test and the partially matched Wilcoxon rank based test which is used to compare 2 groups when some of the data pairs are missing or unmatched [14]. For the multivariate logistic regression analysis, we used the R package StepReg and bidirectional step wise regression (entry/removal threshold p = .15). For all other analyses the significance level was 2-tailed and set at 0.05. According to national legislation, retrospective quality improvement studies based on anonymized data do not require approval from a regional research ethics committee. It is only required that such studies are registered with the regional authorities. However, in this case all patients gave consent to the use of their data with an electronic signature on the REDCap survey instrument. The REDCap database was hosted by the regional health authorities and all data were handled in compliance with the General Data Protection Regulation (GDPR) and the national data protection laws. Results For the demographics see Table 1. Biplanar slot scanning was used in 67.5% of cases and digital standing radiographs in 32.5% of cases preoperatively. At last follow-up the corresponding numbers were 92% and 8%. The instrumentation systems used were Solera, Legacy, Expedium, Mesa and Mont Blanc in 80% of cases. Pedicle screws were the only anchors used. Table 1Demographics.Table 1 dummy alt textVariableN = 120*Sex Female95 (79%) Male25 (21%)Age at surgery15.00 (14.00, 17.00)Age at last FU25.0 (22.0, 27.0)Follow-up time in years9.02 (6.50, 11.21)Lenke class 167 (56%) 212 (10%) 315 (13%) 44 (3.3%) 522 (18%)Number of Ponte osteotomies4.00 (0.00, 6.00)Number of screws22.0 (18.0, 22.0)Number of instrumented levels11.00 (9.00, 12.00)⁎n (%); Median (Q1, Q3). Demographics. n (%); Median (Q1, Q3). Navigation using cone beam 3D-scanning was used in 85% of cases. Freehand insertion of the pedicle screws with fluoroscopy control or with intraoperative 3D control scanning was used in 10% and 5% of the cases respectively. Intraoperative monitoring using somatosensory and motor evoked potentials was utilized in all cases. Intraoperative blood loss was median 1,000 mL with Q1 to Q3 range [713–1,500]. LOS was median 7 days with Q1 to Q3 range [6–8]. Table 2 shows the number and type of complications. A total of 15 complications were registered (14 patients). The patient with a transient neurodeficit ultimately developed adding-on which required an extension of the fusion distally. The frequency of complications was 12.5% (15/120). Table 2Complications and reoperations.Table 2 dummy alt textComplicationsNumbersReoperationsTimingPain/irritation66After 2 yDeep infection44After 2 yDistal add-on22After 2 yIntercostal nerve root injury11In hospitalTransient neurodeficit10PeroperativeDiscogenic pain11After 2 yJunctional kyphosis11After 2 y Complications and reoperations. Six patients had their implants removed due to pain/discomfort. There was no obvious reason for the pain in these 6 patients—the radiographs, the CT scans and the intraoperative findings were all normal. Tissue cultures obtained intraoperatively did not identify pathogens in any of the 6 cases with pain/discomfort. Not counting the 6 patients with pain/irritation the frequency of complications was 7.5% (9/120). All 4 cases of deep infection was late infections with cutibacterium acnes. Table 3 shows the evolution of the thoracic and thoracolumbar/lumbar Cobb angles over time and stratified according to Lenke class 1 to 4 and 5 (no Lenke 6 curves). Table 3Cobb angle changes by Lenke class and time.Table 3 dummy alt textCobb anglesLenke 1–4N = 98*Lenke 5N = 22*Thoracic preop51.38 (13.37)19.36 (8.47)Thoracic immediately postop11.02 (6.84)5.82 (5.21)Thoracic 2 y FU13.43 (6.55)7.18 (4.60)Thoracic last FU14.27 (8.53)7.67 (4.90)Lumbar preop33.29 (13.71)41.73 (6.70)Lumbar immediately postop9.58 (6.18)8.68 (6.12)Lumbar 2 y FU10.73 (7.54)9.95 (5.48)Lumbar last FU10.85 (7.22)12.10 (8.46)⁎Mean (SD). Cobb angle changes by Lenke class and time. Mean (SD). The mean major (thoracic) curve reduction for the Lenke 1 to 4 curves was 78.8% (SD 11.4) from preoperative to immediately postoperative and from preoperative to last follow-up 73.6% (SD 11.4). The mean major (lumbar) curve correction for the Lenke 5 curves was 79.6% (SD 13.3) from preoperative to immediately postoperative and 70.8% from preoperative to last follow-up. Preoperatively the proximal thoracic curve measured mean 18.7° (SD 10.9) versus mean 10.6° (SD 6.3) at last follow-up. See Fig. 2 for an example of the preoperative and postoperative radiographs—13 year old female with 8 years of follow-up. Preoperative and postoperative radiographs at last follow-up (8 years after the surgery). The preoperative thoracic kyphosis measured mean 37° (SD 11.6), range [2–70]. The thoracic kyphosis measured mean 31° (SD 9.9) and range [12–60] at the last follow-up. At the last follow-up the sagittal vertical axis (SVA) touched or was posterior to the posterosuperior corner (negative values) of the S1 vertebra with a mean distance of −34 mm (SD 23.3) and range [−98–0]. We observed 1 case with proximal junctional kyphosis (18°) which was symptomatic and required a reoperation with cranial extension of the instrumentation (listed in Table 2). We did not observe any cases with distal junctional failure or listhesis. Likewise, we did not observe any cases with a distal junctional disc angle >10° (angle between the lower endplate of the most distally instrumented vertebra and the upper endplate of the adjacent noninstrumented vertebra) at last follow-up. In 14 patients we observed a distance of more than 20 mm between the T1 plumbline and the centre sacral vertical line. None of the 14 cases (11.7%) with coronal imbalance were reoperated. To add data concerning intra and inter-rater agreement we randomly selected 25 patients and the unbiased observer and one of the coauthors repeated the 300 Cobb angle measurements (25% were digital radiographs and 75% EOS). Intrarater reliability measured as Cohen’s kappa was 0.97 and inter-rater reliability was 0.92. The mean differences and sd for the 300 measurements were 0.39 ± 3.8 and 0.58 ± 5.2 respectively. For the 25 measurement of the cobb angles of the 3 curves at the 4 points in time the intrarater kappa values ranged between 0.65 to 0.98 and the inter-rater kappa values ranged between 0.45 to 0.97. The values of Cohen’s Kappa indicate moderate to almost perfect agreement. At last follow-up 84% were satisfied or very satisfied with the results of the surgical intervention (SRS-22 question 21—are you satisfied with the results of your back management?). At last follow-up 88% answered they would definitely or probably have the same management again (SRS-22 question 22). Table 4, Table 5 shows the detailed scores of the PROMs at last follow-up. Univariate statistics (chi-square test and Wilcoxon rank-sum test) were used to compare PROMs at last follow-up stratified by work status and highest educational level. Table 4Disease-specific PROMs at last follow-up.Table 4 dummy alt textVariableN = 120*SRS22 function4.27 (0.83)SRS22 pain3.89 (0.89)SRS22 self image4.03 (0.76)SRS22 mental health3.93 (0.66)SRS22 total score4.06 (0.66)SAQ appearance score17.8 (4.9)SAQ expectance score11.0 (4.8)SAQ total score29 (9)⁎Mean (SD). Disease-specific PROMs at last follow-up. Mean (SD). Table 5EQ5D-5L values at last follow-up.Table 5 dummy alt textVariableN = 120*EQ5D-5L mobility 189 (74%) 224 (20%) 36 (5.0%) 41 (0.8%)EQ5D-5L self care 1104 (87%) 213 (11%) 33 (2.5%)EQ5D-5L usual activities 176 (63%) 226 (22%) 311 (9.2%) 46 (5.0%) 51 (0.8%)EQ5D-5L pain/discomfort 128 (23%) 253 (44%) 327 (23%) 410 (8.3%) 52 (1.7%)EQ5D-5L anxiety/depression 174 (62%) 227 (23%) 313 (11%) 45 (4.2%) 51 (0.8%)EQ5D-5L utility score0.82 (0.22)EQ5D-5L general health VAS75 (18)⁎n (%); Mean (SD). EQ5D-5L values at last follow-up. n (%); Mean (SD). The variable highest educational level achieved was stratified into 2 groups—cases with only an elementary school education and cases with a higher educational level. Work status was divided into those who worked full time and those who did not. For those who did not answer very satisfied or satisfied to the SRS-22 question 21 significantly fewer worked full time or had an educational level above elementary school. The SR22 domain score were significantly or highly significantly improved for those with a higher education or working full time for the domains function, pain, mental health and subtotal (Total20) and total score, but not for self-image (work). Almost the same pattern was evident for the EQ-5D-5L utility score, EQ5D VAS and dimensions scores with significantly (p < .05) or highly significantly (p < .001) improved values for those who worked full time or had a higher education than elementary school. The only nonsignificant EQ-5D-5L dimension was self-care (p = .70 and p = .14 respectively). The outcome variable in the logistic regression analysis was the dichotomized answer to the SRS-22 question 21. The predictor variables were sex, age at surgery, follow-up time in years, the dichotomized work status, the dichotomized educational level, Lenke class coded as dummy variables and the decrease in the proximal, thoracic and lumbar Cobb angles (coded as positive or negative integers), number of levels instrumented, thoracic kyphosis at last follow up, LOS, coronal malalignment dichotomized (>20 mm or ≤20 mm), most caudal instrumented level dichotomized (L1 or below). For 112 patients a full data set was available for the outcome and predictor variables. The variance inflation factor for the variables in the model were all below 2 indicating minimal collinearity. The area under the curve (AUC) was 0.85 for the final model. The final model is summarised in Table 6. Table 6Logistic regression model.Table 6 dummy alt textVariableOdds ratioStatisticp-valueOdds ratio CIIntercept4.360.74.46[0.08, 434.84]Not full time work0.14−2.82.00[0.04, 0.55]Higher education12.972.79.01[2.14, 78.73]Decrease in proximal thoracic Cobb angle1.112.06.04[1.01, 1.22]Lenke class 20.10−2.05.04[0.01, 0.89]Decrease in thoracic Cobb angle1.062.03.04[1.00, 1.13]Number of instrumented levels0.72−1.46.14[0.46, 1.13] Logistic regression model. Table 7 shows the changes in SRS-22 and EQ5D values. Concerning the SRS-22 domains the scores for both the Self-image and Total score (20 questions) increased significantly or highly significantly from preoperative to last follow-up for both the unpaired and paired data. The EQ5D utility score increased significantly from preoperative to last follow-up—though not for the paired data. The change in the general health VAS was nonsignificant. Table 7SRS22 and EQ5D from preoperative to last follow-up.Table 7 dummy alt textVariablePairedDaneSpinePreoperative*N = 40Pairedlast FUN = 40UnpairedDaneSpinePreoperativeN = 51/57Unpairedlast FUN = 80Wilcoxon partially matched data p-valueSRS22 function4.26 (0.76)4.35 (0.83)4.10 (0.64)4.24 (0.83).08SRS22 pain3.81 (0.93)3.91 (0.91)3.65 (0.85)3.87 (0.89).08SRS22 self image3.08 (0.73)4.21 (0.53)‡3.08 (0.76)3.94 (0.85)‡<.001‡SRS22 mental Health3.97 (0.73)4.09 (0.60)3.80 (0.70)4.00 (0.67).53SRS22 total 20 score3.78 (0.61)4.14 (0.60)‡3.69 (0.57)3.98 (0.71)‡<.001‡EQ5D-5L utility score0.81 (0.18)0.83 (0.23)0.81 (0.14)0.82 (0.22)†.04†General health VAS80.8 (18.1)78.2 (16.9)73.1 (18.9)72.8 (18.7).28⁎Mean (SD).†<0.05.‡<0.001. SRS22 and EQ5D from preoperative to last follow-up. Mean (SD). <0.05. <0.001. Any differences between the 2 unpaired data sets were nonsignificant for the following variables: Age, sex, Lenke class, proximal thoracic, preoperative thoracic and lumbar Cobb angles, number of instrumented levels, and cranial and caudal instrumented levels. To further qualify the use of both unpaired and paired data sets we have compared the 2 preoperative datasets and found no significant difference in the SRS22 values (dimension scores and total 20 scores) and no significant difference in the EQ-5D-3L values (domain scores and utility index) between the paired and unpaired data sets. To evaluate any response bias, we performed a no-response analysis comparing demographic, radiographic and data on reoperations for the 28 patients not responding to the PROM questionnaires with the 120 patients included in the study. The most important findings are summarized in Table 8. We could not demonstrate any significant differences in any of the variables but noted a slightly increased reoperation rate in the no-response group (1 case with deep infection and 4 cases with distal adding-on required reoperations). Table 8No response analysis.Table 8 dummy alt textVariableNonrespondentN = 28*RespondentN = 120*p-value†Age at surgery15.8 (2.6)15.7 (2.1)>.9Sex.4 Female20 (71%)95 (79%) Male8 (29%)25 (21%)Lenke Class.15 110 (36%)67 (56%) 24 (14%)12 (10%) 35 (18%)15 (13%) 42 (7.1%)4 (3.3%) 56 (21%)22 (18%) 61 (3.6%)0 (0%)High thoracic Cobb preop18.5 (15.2)18.7 (10.9).4Thoracic Cobb preop48.1 (17.5)45.5 (17.7).6Lumbar Cobb preop36.9 (11.6)34.8 (13.1).4Thoracic kyphosis preop31.1 (14.6)31.0 (11.6).8Number of screws19.8 (3.1)20.6 (3.4).2Implant density1.0 (0.1)1.0 (0.0).4Number of reoperations0.5 Reoperation5 (18%)15 (13%) No reoperation23 (82%)105 (88%)⁎Mean (SD); n (%).†Wilcoxon rank sum test; Pearson's Chi-squared test; Fisher's exact test. No response analysis. Mean (SD); n (%). Wilcoxon rank sum test; Pearson's Chi-squared test; Fisher's exact test. Furthermore, a logistic regression analysis (stepwise regression) with outcome variable response status and p-entrance .15 resulted in a model with only the intercept. Discussion The surgical techniques used in all cases in the current study are modern techniques which reflect a well-established paradigm for the surgical treatment of AIS today. Even though several instrumentation systems were used the basis for all the instrumentation systems was the pedicle screw. The implant density was 0.98 clearly indicating segmental instrumentation. Navigation was used in almost all the cases. We had only 1 misplaced screw requiring reoperation in 120 cases (0.04%). Intraoperative neuromonitoring enabled us to discover and quickly handle the case with transient spinal cord injury resulting in full recovery for the patient in question. The use of biplanar slot scanning for the radiographic controls and digital radiography as a whole made possible and improved the detailed analysis of the radiographs. The number of complications in the current study is on par with other studies with long term follow-up. According to Newton et al. [3] the revision rate is 8% to 14.5% in long term follow-up studies after AIS surgery. However, it is questionable if all implant-related complications require revision surgery [15]. It is possible we could have evaded implant removal in some of the patients with discomfort/pain around the implants with no obvious reason for the pain. The 4 cases with late infection with cutibacterium acnes were only diagnosed after extended culture incubation (9 days) and these infections can easily go unnoticed. Implants were simply removed in most of the cases with late complications. However, the case with discogenic pain and the cases with adding-on and junctional failure all required an extension of the instrumentation. Pseudoarthroses are relatively rarely seen after AIS surgery. This complication was not registered in the current study. The patients with pain/irritation all underwent CT scan before removal of the implant to eliminate nonfusion complications. Our well-functioning bone bank and the fact we always reserve a femoral head weighing at least 100 g for the AIS patients potentially contributed to the zero number of pseudoarthrosis cases. Lenke 3 curves can easily be misclassified as Lenke 6 curves possibly contributing to the absence of Lenke 6 curves. Side bending radiographs in the EOS box is hindered by the reduced space inside the box—another contributing factor. However, after reviewing the radiographs of all patients with a Lenke 3 classification we found no compelling reason to change the classification. Besides, the surgical strategy is the same for the 2 curve types. After reviewing the radiographs of the 28 patients initially excluded, we classified one of these as Lenke 6. However, as mentioned above we could not demonstrate any significant differences in any of the variables listed in Table 8 (this includes the Lenke classification). The curve correction in the current study is comparable to or better than demonstrated by the few other studies with long-term follow-up using contemporary surgical techniques [2,3,[16], [17], [18]]. Percentagewise the correction was not maintained at last follow-up decreasing with 5% to 10% from immediately postoperatively to last follow-up. However, in absolute numbers the decrease was on average 3° to 4° which equals the measuring uncertainty. The clinical implications of the decrease in curve correction are probably negligible as the self-image scores of the SRS-22 at last follow-up remained significantly increased compared to the preoperative scores. Thoracic kyphosis improved with values which must be considered within normal limits. The SVA distances at last follow-up did not indicate a problematic sagittal balance in any of the 112 cases. Whether the sagittal balance and the coronal imbalance (14 cases) will become a problem with increasing age remains to be seen. The logistic regression analysis did not indicate that neither thoracic kyphosis nor coronal imbalance influenced the satisfaction with the surgical intervention at last follow-up. A high percentage of patients declared they were satisfied or very satisfied with the results of the AIS surgeries. However, 12% to 16% were uncertain or dissatisfied. The reoperated patients were as expected significantly more dissatisfied/uncertain. The mean curve correction at last follow-up was greater for the reoperated patients though not significantly increased compared to the nonreoperated patients. This had no influence on the SRS-22 dimension scores which were significantly decreased for all dimensions including self-image but except function. Evidently, expectations were not fulfilled for the reoperated patients. Not surprisingly, educational level and work status had significant influence on the disease specific as well as the general HRQL scores. These variables can be considered proxies for socioeconomic status and usually they are not considered when the results of AIS surgery are analysed. In the current study both univariate and multivariate statistics indicated a significant correlation with the PROMs underlining the importance of incorporating socioeconomic factors in AIS research. A large population based study has to some extent corroborated our results [19] and our results are also corroborated by Bago et al. [20]—a mixed study with patients 12 to 40 years of age and scheduled for both surgical and nonsurgical treatment. The patients with only elementary school education or without work were younger and some are students and will improve their educational level. The ages from the mid-teens to the third decade is an interval with many substantial life changes and it is to be expected that these changes will influence the HRQL scores to some degree. Besides, it has been demonstrated that external factors eg, the COVID-19 pandemia can have a relatively long-lasting influence on HRQL [21]. Still, our results indicate a highly significant improvement from preoperative to last follow-up for the SRS-22 scores for self-image and the subtotal score (Total 20)—apparent for both the paired and unpaired data. The SRS-22 scores for function and pain improved but not significantly and the EQ5D general health VAS decreased. These findings are corroborated by several contemporary studies [[2], [3], [4],22]. Newton et al. [3] demonstrated a decline to preoperative levels for all SRS-22 domains except for self-image and the total score. The follow-up rate in the case-series study by Newton et al. [3] was only 19% and included 102 with pedicle screw constructs. Ahonen et al. [2] using the SRS-24 questionnaire corroborated the findings in the current study concerning self-image, but in fact the total score declined significantly from preoperative to last follow-up for the surgically treated patients. The follow-up rate in the cohort study by Ahonen et al. [2] was the same as the current study but the number of patients was less than half the number of patients in the current study. The meta-analysis by Hevia et al. [4] do corroborate our results regarding no significant change from preoperative to last follow-up for the SRS-22 domains pain and function and a significant improvement in self-image. Contrary to the results of the current study Hevia et al. [4] demonstrated a significantly worsened activity score, a significantly improved mental health score and a nonsignificant change in total SRS score. The lasting improvement in self-image is seen in all long-term studies without any exception. Activity generally must be expected to decline with increasing age or a change to a lifestyle dictated by a sedentary job situation. Besides, the SR22 questions regarding activity are not unambiguously formulated. For 1 in 8 in the current study the answer to question 5 of the SRS-22 was inconsistent with the answer to the EQ-5D-5L activity domain. It is difficult to explain why mental health should improve/decline 10 years after surgery and the 2 contemporary studies mentioned above show no significant change in this SRSS22 dimension. The significant improvement in the SRS-22 total score depends to some extent on the improvement in the self-image score. The EQ5D is one of the most widely used tools to measure HRQL [23]. Data from large population based studies are available for both the domain scores and especially for the utility score [24]. Data for the EQ5D are available for both healthy individuals and for a large number of chronic diseases [25] and this enables us to compare our data to healthy individuals and to patients with other chronic diseases and consequently to place AIS somewhere on a disease severity scale. The EQ-5D-5L mean utility score was 0.82 at last follow-up significantly less than the mean Danish population score (0.93) for the same age group [25]. However, the EQ-5D-5L utility score was greater than or equal to 0.93 for one third of our patients. The mean EQ-5D-5L utility score is comparable to the mean score for a skilled worker or a person with a short duration higher education in the same age group [25]. Using the NICE Decision Support Unit EEPRU models [26] the EQ-5D-5L mean utility calculated in the current study can be compared to a catalogue of utility scores for 199 chronic conditions [24]. The EQ-5D-5L mean (0.82) translates to 0.76 after mapping. This is comparable to the utility score for irritable bowel syndrome (0.77) and slightly improved compared to the score for undefined scoliosis (0.73) and arthrosis of the hip and knee (0.72 and 0.74 respectively) [24]. Incidentally, this is the first time an EQ-5D-5L utility score have been reported for adolescent idiopathic scoliosis patient with a mid-to-long term follow-up after scoliosis surgery. The mapping from EQ-5D-3L to 5L is problematic as the EQ-5D-3L suffers from relatively severe ceiling problems—many patients will report perfect health (11111) simply because the instrument is not sensible enough to detect their mild problems [26]. These ceiling problems can contribute to any nonsignificant differences between the utility scores. Anyway, the comparison between preoperative and EQ-5D-3L values and postoperative EQ-5D-5L values is fraught with difficulties and our result can be questioned. The logistic regression model obviously does not indicate causality. We chose to include the model to demonstrate the corelation with socioeconomic factors—having an educational level higher than elementary school or working full time you were approximately 13 and 7 times more likely to be satisfied with the surgical treatment. This indicates that surgical implants and surgical techniques and curve correction should not be regarded as the only factors responsible for a satisfactory result after AIS surgery. Indeed, the correlation to curve correction was relatively moderate with at most an odds ratio of 1.1 per degree correction. The correlation between number of instrumented levels and satisfaction is as expected—decrease in satisfaction with increased number of levels instrumented. The reason for the reciprocal correlation between Lenke class 2 curves and a satisfactory result of surgery is not obvious. However, it is often difficult to correct the relatively stiff proximal curve, the instrumentation often reaches the mobile cervicothoracic junction and shoulder balance can be problematic. The limitations of the logistic regression analysis per se are the relatively low number of cases, calibration was not analysed and external validation was not undertaken. The AUC value of 0.85 is acceptable. The limitations of the current study were the relatively few patients and the absence of a control group. However, the number of patients included in the current study is among the largest compared to other long-term studies. Concerning the lack of a control group we already have sufficient evidence to conclude that healthy individuals PROM scores are significantly improved compared to AIS patients [2,4]. We can corroborate these findings when comparing with the catalogued EQ-5D-5L utility scores for the Danish population. Comparative studies between nonsurgically treated and surgically treated patients suffers from severe cross-over between treatments and consequently it is difficult to make any conclusions [27]. The reason for the insufficient completeness of the DaneSpine register is lack of funding. Only recently have we been able to engage an academic employee who is responsible for securing that all PROMs are entered in the register. Choosing to include the patients with missing preoperative PROMs in the statistical analysis meant that valuable data could be used in a sensible way. Besides, the 2 unpaired data sets and the preoperative paired and unpaired data sets were comparable as mentioned earlier. The statistics behind the partially matched Wilcoxon rank sum test are sound as long as data are missing completely at random (MCAR) which test for MCAR indicated. Besides, the 2 unpaired data sets and the preoperative paired and unpaired data sets were compatible as mentioned earlier. The follow-up times differed making it possible meaningfully to include this variable in the logistic regression analysis—we do not consider this a weakness. The strength of the current study is the high follow-up rate and the consecutive enrolment of patients. The imaging techniques, surgical techniques, implants etc. were all contemporary and uniform for all of the patients. The digitalized radiographs allowed detailed radiographic analyses and 93% participated in the radiographic follow-up. The PROMS used were well validated and are generally recognized as suitable in the current patient population. The patients filled in the questionnaires electronically without any meddling from health personnel. The radiographic measurements were undertaken by an unbiased observer. The inclusion of socioeconomic factors allowed us to include these in our data analysis. The use of the EQ-5D-5L questionnaire enabled comparison with a large population pool of both healthy individuals and individuals with chronic diseases. Conclusion In the current quasi-longitudinal study (unpaired PROMs in two third of cases) PROMs consisting of the SRS-22 and EQ5D had returned to preoperative levels 5 to 15 years after AIS surgery with the exception of the SRS22 self-image score and the subtotal score (Total 20) which both were significantly improved. The EQ-5D-5L utility score (paired data) did not differ significantly from preoperative to last follow-up. The EQ-5D-5L mean utility score at last follow-up was 0.81. This value is significantly lower than the population mean for the same age group (0.93). The curve correction at last follow-up was 71% to 74%. In absolute numbers the correction was reduced with approximately 4° from immediately preoperative to last follow-up. At last follow-up 84% of cases were satisfied with the results of the treatment. Logistic regression analysis demonstrated the dependency of satisfaction on socioeconomic factors but also on the degree of curve correction and number of levels instrumented (less is better) and Lenke class (Lenke class 2 fares worse). The study confirms that some HRQL value is added and conserved 5 to 15 years after AIS surgery as well as a maintained curve correction. CRediT authorship statement Ida Bromell Kindberg: Data curation, Methodology, Formal analysis, Writing—original draft, Writing—review & editing. Efe Levent Aras: Conceptualization, Methodology, Supervision, Writing—original draft, Writing—review & editing. Jon Kaspersen: Conceptualization, Data curation, Supervision, Writing—review & editing. Sasa Randelovics: Conceptualization, Data curation, Writing—review & editing. Simon Toftgaard Skov: Conceptualization, Methodology, Writing—review & editing. Søren Eiskjær: Conceptualization, Methodology, Formal analysis, Supervision, Writing—original draft, Writing—review & editing. Funding No funding of any sort was received. Declaration of competing interests The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper. Footnotes FDA device/drug status: Not applicable. Author disclosures: IBK: Nothing to disclose. ELA: Nothing to disclose. JK: Nothing to disclose. SR: Nothing to disclose. STS: Nothing to disclose. SE: Nothing to disclose. References