Abstract
Study Design
Cross-sectional cohort study.
Objective
To classify spinal morphology using the “current” and “theoretical” Roussouly systems and assess sagittal alignment in an asymptomatic cohort.
Methods
467 asymptomatic volunteers were recruited from 5 countries. Radiographic parameters were measured via the EOS imaging system. “Current” and “theoretical” Roussouly classification was assigned with sagittal whole spine imaging using sacral slope (SS), pelvic incidence (PI), and the lumbar apex. One-way analysis of variance (ANOVA) was performed to compare subject characteristics across Roussouly types, followed by post hoc Bonferroni correction.
Results
Volunteers were categorized into 4 groups (Types 1-4) and 1 subgroup (Type 3 AP) using the “current” and “theoretical” Roussouly systems. The mean PI in “current” Roussouly groups was 40.8° (Type 1), 43.6° (Type 2), 52.4° (Type 3), 62.4° (Type 4), and 43.7° (Type 3AP). The mean PI in “theoretical” Roussouly groups was 36.5° (Type 1), 39.1°(Type 2), 52.5° (Type 3), 67.3° (Type 4), and 51.0° (Type 3AP). The difference in PI between “current” and “theoretical” Roussouly types was significant for Type 1 (P = .02), Type 2 (P < .001), Type 4 (P < .001), and Type 3AP (P < .001). 34.7% of subjects had a “current” Roussouly type different from the “theoretical” type. Type 3 theoretical shape had the most frequent mismatch, constituting 61.1% of the mismatched subjects. 51.5% of mismatched Type 3 become “current” Type 4.
Conclusion
The distribution of Roussouly types differs depending on whether the “current” or “theoretical” classification are employed. A sizeable proportion of volunteers exhibited current and theoretical type mismatch, highlighting the need to interpret sagittal alignment cautiously when utilizing the Roussouly system.
Keywords
Introduction
Adult spinal deformity (ASD) is defined as deviations of spinal alignment from established normal limits in patients >18 years old in the coronal, axial, or sagittal planes. 1 It also affects a significant proportion of the population, with a reported prevalence of up to 68% in patients over 60 years of age, 2 harming the quality of life comparable to other known chronic diseases. 3 Imbalance in the sagittal plane was shown to have an association with lower quality of life.4-6 Thus, a key goal of ASD surgery is to restore the sagittal balance.
However, the goals of sagittal alignment remain under investigation. The Roussouly classification system categorizes spine morphology into discrete types according to sagittal parameters to aid surgical planning. The “current” (original) system classifies patients into four distinct types based on their sacral slope (SS). 7 Later literature proposed the “theoretical” classification, categorizing patients into four types based on pelvic incidence (PI). 8 The rationale behind the distinction between the “current” and “theoretical” classification schemes is that degenerative changes and pathologies can alter the SS,9,10 but the PI stays constant. 8 The mismatch between the current and theoretical might indicate spinal pathology. An increase in mechanical complications was observed when ASD patients were not restored according to their “theoretical” type.11,12 However, other literature also reported that patients restored solely by Roussouly classification were not associated with better clinical outcomes at two years postoperatively. 13
The Roussouly classification may be limited due to the imposition of discrete types onto the continuous spectrum of spine shapes based on cutoffs in the SS or PI. 14 To address this shortcoming, the numerical Global Alignment and Proportion (GAP) score was proposed to classify sagittal alignment to predict mechanical complications of ASD surgery. 15 However, some literature found no significant differences in postoperative complications in patients corrected according to GAP score, while those that were Roussouly-restored experienced significantly reduced postoperative complications.16,17 The best guideline for restoring sagittal balance in ASD patients remains unclear.
This study aims to classify spinal morphology using the “current” and “theoretical” Roussouly classifications and the applicability of the Roussouly system to assess sagittal alignment in a large, multiethnic, asymptomatic cohort to establish a reference for future studies of the Roussouly system. Additionally, this study investigates and compares the distributions of sagittal parameters in subjects according to the current and theoretical Roussouly definitions.
Materials and Methods
Study Design
The asymptomatic Multi-Ethnic Alignment Normative Study (MEANS) cohort is a prospectively enrolled, cross-sectional group of adult volunteers from multiple institutions in five countries (France, Japan, Singapore, Tunisia, and the United States). MEANS aims to investigate skeletal alignment within a cohort of asymptomatic adult volunteers. Volunteers included in MEANS had a Visual Analog Scale ≤2, with neither known spinal disorder, nor any history of surgical or non-surgical treatment for any spine-related disorder. Exclusion criteria in MEANS were any volunteers with a Cobb angle >20°, congenital abnormalities such as abnormal vertebral number or transitional anatomy, an Oswestry Disability Index (ODI) score above 20, history of hip or knee arthroplasty, or any lower extremity realignment surgery of any kind. Overall, 467 adult volunteers were included in the MEANS cohort for this study.
Radiographic and Clinical Variables
Distribution of Variables and Differences Between the Current Roussouly Types.
Abbreviations: TK indicates Thoracic Kyphosis (T1-T12); LL, Lumbar Lordosis (L1-S1); dLL, Distal Lumbar Lordosis (L4-S1); pLL, Proximal Lumbar Lordosis (L1-L4); LDI, Lordosis Distribution Index; Max TK, Maximum Thoracic Kyphosis; Max LL, Maximum Lumbar Lordosis; PI, Pelvic Incidence; PT, Pelvic Tilt; SS, Sacral Slope; GT, Global Tilt.
an (%); mean (SD).
bPearson’s Chi-squared test; Kruskal-Wallis rank sum test.
Statistical Analysis
Statistical analysis was performed in RStudio (version 1.4.1103, Posit Software, PBC, Boston, MA) and SAS (SAS 9.4, SAS Institute Inc, Cary, NC). Tables were produced using Microsoft Word. Mean, median, standard deviation (SD), or interquartile range (IQR) were calculated for continuous variables. N (%) was calculated for categorical variables. Shapiro-Wilk’s test was used to assess normality. Kruskal-Wallis rank sum test was used to compare continuous variables by group, followed by post-hoc Dunn’s test with Bonferroni correction. Chi-squared test was used to compare categorical variables. Wilcoxon rank-sum test was used to compare two samples of continuous variables. The significance level was set to P < .05. When appropriate, a 95% confidence interval (CI) is reported.
Results
Demographic Variables
In the asymptomatic MEANS cohort (n = 467), the mean age was 40.4 ± 14.8, 61% (284/467) were female, BMI was 24.5 ± 5.3 kg/m2 (the BMI of 110 volunteers was not recorded), and ODI was 2.3 ± 4.1. Of the entire cohort, France comprised 21% (98/247), Japan 25% (119/467), Singapore 17% (79/467), Tunisia 17% (80/467), and the U.S.A 19% (91/467).
Distribution of “Current” Roussouly Spine Types and Their Relationships With Sagittal Parameters
“Current” Type 1 constituted 7.5% (35/467), Type 2 21.4% (100/467), Type 3 47.5% (222/467), Type 4 23.6% (110/467) (Table 1). A subgroup of Type 3 with anteverted pelvis (Type 3AP) constituted 11.1% (52/467). The mean PI in “current” Type 1 was 40.8° ± 9.2, Type 2 43.6° ± 7.6, Type 3 52.4° ± 7.6, Type 4 62.4° ± 8.6, Type 3AP 43.7° ± 5.0. Current Type 1 had a lower arc of 26.7° ± 9.5, Type 2 32.3° ± 9.2, Type 3 37.0° ± 7.3, and Type 4 42.9° ± 9.5 (P < .001). Current Type 1 had an upper arc of 18.8° ± 7.2, Type 2 15.4° ± 6.0, Type 3 16.9° ± 6.3, and Type 4 16.2° ± 6.5 (P = .039). All current Type 1 subjects (35/35) had their lumbar apex at L5, 77% (77/100) of Type 2 at L4 and 21% (21/100) at L3, 69% (153/222) of Type 3 subjects at L4 and 25% (56/222) at L3, and 47% of Type 4 (52/110) subjects at L4, 44% (48/110) at L3. 83% (29/35) of current Type 1 subjects had an inflection point between L1-L3, 75% (75/100) of current Type 2 subjects between T12-L2, 80.1% (177/222) of current Type 3 subjects between T12-L2, and 89.9% (99/109) of current Type 4 subjects between T11-L2. The LDI of Type 1 was higher than all other types (P < .001), Type 2 was higher than Type 4 (P = .009), and Type 3 was not different from Type 4 (P = .063). Type 3 TK, Max TK, PT, GT, and PI-LL mismatch did not differ significantly between current types (P > .05) (Table 1).
Distribution of “Theoretical” Roussouly Spine Types and Their Relationships With Sagittal Parameters
Distribution of Variables and Differences Between the Theoretical Roussouly Types.
Abbreviations: TK indicates thoracic kyphosis (T1-T12); LL, lumbar lordosis (L1-S1); dLL, distal lumbar lordosis (L4-S1); pLL, proximal lumbar lordosis (L1-L4); Max TK, maximum thoracic kyphosis; Max LL, maximum lumbar lordosis; PI, pelvic incidence; PT, pelvic tilt; SS, sacral slope.
an (%); Mean (SD).
bKruskal-Wallis rank sum test; Pearson’s Chi-squared test.
Distribution of Variables and Differences Between Current and Theoretical Roussouly Types.
Abbreviations: TK indicates thoracic kyphosis (T1-T12); LL, lumbar lordosis (L1-S1); dLL, distal lumbar lordosis (L4-S1); pLL, proximal lumbar lordosis (L1-L4); Max TK, maximum thoracic kyphosis; Max LL, maximum lumbar lordosis; PI, pelvic incidence; PT, pelvic tilt; SS, sacral slope. Inflection Point and Lumbar Apex Values Indicate the Most Frequently Appearing. Significant P-values are Bolded.
Comparison of “Current” vs “Theoretical” Roussouly Spine Types
“Current” Types 1 and 2 constituted larger portions of the MEANS cohort, at 7.5% (35/467) and 21.4% (100/467), respectively, than “theoretical” Types 1 and 2 did, at 5.8% (27/467) and 18.6% (87/467). The difference in PI between “current” and “theoretical” Roussouly types was significant for Type 2 (P < .001), Type 4 (P < .001), and Type 3AP (P < .001) but not Type 1 (P = .1) and Type 3 (P = .85) (Table 3). PI-LL mismatch differed between current and theoretical Type 2, with theoretical Type 2 having higher magnitude mismatch (−10.9°) (P < .001), as well as Type 4, with theoretical Type 4 having more positive mismatch (2.2°) (P < .001). The LL differed between current and theoretical Types 2 and 4: theoretical Type 2 exhibited higher magnitude lordosis (−50.0°) than current Type 2 (−46.8°), while theoretical Type 4 exhibited lower magnitude lordosis (−65.2°) than current Type 4 (−68.5°). Additionally, theoretical Type 2 subjects were younger, while Type 4 subjects were older than the current Type 2 and Type 4 subjects. GT was approximately three times higher in current Type 2 than theoretical Type 2 (P < .001), similar between current and theoretical Type 3 (P = .7), and was lower in current Type 4 than theoretical Type 4 (P < .001) (Table 3).
Frequency of Mismatched Roussouly Types and Mismatch Rates.
Distribution of Variables and Differences Between “Consistent” and “Mismatched” Roussouly Types. Consistent Roussouly Types are Those Where the Theoretical Match the Current Type, Whereas Mismatched Roussouly Types Indicate Subjects Whose Current Type is Different From the Theoretical Type.
Abbreviations: TK indicates thoracic kyphosis (T1-T12); LL, lumbar lordosis (L1-S1); dLL, distal lumbar lordosis (L4-S1); pLL, proximal lumbar lordosis (L1-L4); Max TK, maximum thoracic kyphosis; Max LL, maximum lumbar lordosis; PI, pelvic incidence; PT, pelvic tilt; SS, sacral slope. Inflection Point and Lumbar Apex Values Indicate the Most Frequently Appearing. Significant P-values are Bolded. P-Value is Not Applicable for Type 1 as There is N = 1 for the Mismatched Subject.
Comparison of Multi-Ethnic Alignment Normative Study (MEANS) Roussouly Type Sagittal Parameters With the Ideal Roussouly Sagittal Parameters
Based on the “current” Roussouly classification, 34.7% (162/467) subjects had a PI that differed from the ideal reference PI per Roussouly type, 21.8% (102/467) had a lumbar apex different from the ideal reference lumbar apex, and 55.2% (257/467) had an inflection point distinct from the ideal reference inflection point. The mean difference between each subject’s LL and the ideal reference LL per Roussouly type was 6.6° ± 5.3, LDI and the ideal reference LDI was 11.8% ± 9.1. Based on the “theoretical” Roussouly classification, 34.7% (162/467) of subjects had a SS different from the ideal reference SS, 24.4% (114/467) had a lumbar apex different from the ideal reference lumbar apex, and 51.2% (239/467) had an inflection point different from the ideal reference inflection point. The mean difference between each subject’s LL per theoretical Roussouly type and the ideal reference LL per Roussouly type was 7.9° ± 6.2, LDI and the ideal reference LDI was 10.5% ± 8.5. No significant difference was observed between the current and theoretical Roussouly systems of the rate of lumbar apex (P = .393) and the inflection point (P = .250) mismatch with the ideal reference apex and inflection point positions. Under the current Roussouly system, the difference of LL and the ideal reference LL per Roussouly type was 1.3° less than that under the theoretical system (P = .005). In contrast, under the current system, the difference of LDI and the ideal reference LDI was 1.3% more than that under the theoretical system (P = .022).
Discussion
The Roussouly classification system originally defined four sagittal spine shapes in the asymptomatic population using the sacral slope (SS). 7 Recent literature supported the use of PI-based “theoretical” Roussouly classification, as PI is considered constant despite changes in age or pathologies.9,10 A widely accepted tenet is that restoring sagittal alignment to match the ideal sagittal balance would enhance the patient’s quality of life and lower postoperative complications.12-14 Strategies such as the Roussouly system and the GAP score have been created to restore ideal sagittal balance. The principle behind using the Roussouly system in deformity surgery is to obtain the theoretical/ideal sagittal alignment and restore the current, pathologic spine shape to the theoretical. Recent literature has explored this principle in adult scoliosis (AS): the Roussouly system applies to AS patients and AS modified the theoretical type in one in every three patients. 14 Furthermore, AS patients without postoperative restoration to the theoretical Roussouly type suffered increased rates of mechanical complications; current-theoretical mismatch was a risk factor for mechanical complications. 12 However, others reported that restoration to the theoretical Roussouly type did not offer superior outcomes. 13
Building on recent literature, this study investigated the application of the “current” and “theoretical” Roussouly classifications to a large, multiethnic, asymptomatic cohort. The distributions of the Roussouly types, the current-theoretical mismatch, and the relationship between sagittal parameters and Roussouly types were explored. In the current types, the LL increases in magnitude as the type becomes numerically higher, with pLL contributing to a large proportion of the increase (Table 1). As the current type increases numerically, the inflection point and lumbar apex also move cranially. This pattern is also reflected in the increasing lower arc of lordosis, wherein a more cranial lumbar apex would correlate with a larger lower arc. PI increases as the current type increases numerically, with SS contributing to this increase, while PT does not change (Table 1). LL increases in magnitude in the theoretical types as the type becomes numerically higher, with pLL contributing to a large proportion of the increase (Table 2), like current types. As the theoretical type increases numerically, the inflection point and lumbar apex move cranially, reflected in the increasing lower arc of lordosis. Like current types, PI increases as the theoretical type increase numerically, with SS contributing to this increase (Table 2). Unlike current types, PT generally increases as the theoretical type increases, except for Type 2, whose PT is lower than Type 1’s. Additionally, the PI-LL transitions from negative to positive, indicating that PI becomes greater in magnitude than LL as the theoretical types increase, a finding not observed in the current types (Table 2). Thus, a similar pattern exists in the asymptomatic cohort for both theoretical and current Roussouly types: as the Roussouly type increases numerically, the LL, dLL, and lower arc of lordosis also increase, while the lumbar apex and inflection moves cranially, and the PI and SS increase, as compensation for the increasingly lordotic spine. Furthermore, the LDI of both current and theoretical types decrease as the type numerically increases. This pattern, corroborating with past literature,7,14 supports the applicability of Roussouly classification in a large, multiethnic, asymptomatic setting.
A comparison of the theoretical and current Roussouly types yielded interesting findings (Table 3). Current Type 2 subjects were older, had less LL and dLL, and higher PI, PT, but lower SS and PI-LL mismatch than theoretical Type 2 subjects. No significant differences in sagittal parameters between current and theoretical Type 3 subjects were found. Current Type 4 subjects were younger, had more LL, dLL, and lower PI, PT, but higher SS and PI-LL mismatch than their theoretical counterparts. Importantly, in this cohort, 34.7%, or approximately one in three, of subjects had a “current” Roussouly type different from the “theoretical” type, with Type 3 theoretical shape having the most frequent mismatch. Subjects with consistent types—where the current matches the theoretical—exhibited notable differences from those with mismatched types—where the assigned current type does not match the theoretical. Consistent Type 2 subjects were older, had lower LL, dLL, lower arc, and PI, and higher PT, but lower SS and PI-LL mismatch than the mismatched Type 2s (Table 5) (Figure 1). Consistent Type 3 subjects had similar sagittal parameters to mismatched Type 3 subjects. Consistent Type 4 subjects were younger, had higher LL, dLL, lower arc, and higher PI, SS, but lower PT and PI-LL mismatch than the mismatched Type 4s. A French volunteer with a current-theoretical type mismatch. SS = 28.12 deg, PI = 60.4 deg, lumbar apex = L4, and inflection point = T12. This volunteer is current Type 2 but theoretical Type 4.
Considering recent literature, the sagittal balance of subjects with current-theoretical mismatch is worthy of investigation. In this asymptomatic cohort, the mismatch rate approximates that reported in symptomatic patients, about one in every three. 14 As a mismatch is conceptualized to indicate malalignment, the existence of this sizeable proportion of current-theoretical mismatched subjects in the asymptomatic cohort raises a few possibilities: (1) the mismatched subjects are already undergoing a degenerative/pathologic process yet remain asymptomatic, and/or (2) current-theoretical mismatch is a radiographic “pathology” but does not always indicate clinical pathology. Previous literature found that degenerative processes are associated with Types 1 and 2, and Types 3 and 4 tend to degenerate into Types 1 and 2.14,21 Current Types 1 and 2 constituted larger portions of the MEANS cohort than theoretical Types 1 and 2. Overall, the mismatch direction is from Types 3 and 4 towards Types 1 and 2, which parallels the degenerative process described in the literature. Additionally, mismatched Types 2 and 4 subjects had significantly larger PI-LL mismatch than their consistent counterparts, another clue to an underlying pathologic process. However, mismatched Type 3 subjects did not exhibit significant sagittal balance deviations from their consistent Type 3 counterparts. Moreover, radiographic malalignment may not always result in clinical pathology due to the body’s compensatory mechanism. That the current-theoretical mismatch is a radiographic pathology but not a true clinical pathology is still plausible. Additionally, it should be noted that the MEANS cohort differs in age from the Roussouly study’s cohort; the mean age of the MEANS cohort is 40.4 years, whereas the mean age of the Roussouly study’s cohort is 27 years. 7 Thus, the original Roussouly system was developed in relatively young subjects with expected minimal spinal degeneration; one would expect more spinal degeneration in subjects in the older MEANS cohort. The pelvic parameters PI, SS, and PT change according to aging 22 and thus may contribute to the current-theoretical mismatch and the modification of the Roussouly shapes.
An effort was also made to determine whether the “current” or “theoretical” Roussouly system corresponds better to the ideal reference sagittal parameters per Roussouly type, as reported by previous literature. 12 Significant variations existed between the sagittal parameters of the MEANS subjects per Roussouly type and the ideal reference sagittal parameters. For instance, 34.7% of subjects had a PI or SS that differed from the ideal reference PI or SS per Roussouly type; approximately half of the subjects had an inflection point that did not fit the ideal reference inflection point. Though under the current system, the LL varies less from the ideal reference LL per Roussouly type that under the theoretical system, the variation is marginal by approximately one degree. Thus, no significant difference was observed between the variations of sagittal parameters depending on whether the “current” and “theoretical” Roussouly system was employed. Based on the SS and PI, the “current” and “theoretical” Roussouly systems are similar in describing the asymptomatic spine shape. More importantly, the variations in the sagittal plane in the MEANS cohort against previously reported ideal reference values 12 highlight the human spine’s varied nature and emphasize the importance of considering the sagittal shape and parameters as a continuous spectrum, instead of discretizing the spine into distinct categories like the Roussouly system and defining a single ideal reference value for each category.
Some limitations exist for the present study. Firstly, the study is limited by its cross-sectional nature. To truly investigate the complexity of the current-theoretical mismatch, the study must be prospective: tracking subjects longitudinally to determine whether current-theoretical type mismatch prognosticate pathology. Secondly, the study is limited by the concept of discrete Roussouly types itself. Sagittal parameters such as PI and SS are continuous parameters, and it appears arbitrary to discretize a continuous parameter into four or five categories. Thirdly, stemming from the categorization issue, the findings from this study are limited by statistical methodology. Aggregating unique human subjects for analysis can obscure individual differences, which may explain why mismatched Type 3 subjects did not exhibit significant sagittal shape variation from consistent Type 3 subjects. Fourthly, lumbar apex was selected based on previously published literature,7,8,13 however, degenerative changes can alter the apex of a curve. Another shortcoming of the Roussouly classification systems is that it does not take into account how degenerative changes alter the aging spine.
Conclusions
This study of a large, multiethnic, asymptomatic cohort demonstrated the applicability of the “current” and “theoretical” Roussouly classification systems. As the Roussouly type increases numerically, the lumbar lordosis increases, and the lumbar apex and inflection point migrate cranially. One in three asymptomatic subjects exhibited a mismatch between the current and theoretical spine types, with mismatched subjects having larger PI-LL mismatch. The Roussouly classification system, when applied in deformity surgery, must be correlated with clinical pathology. The human spine shape should be considered more as a continuous spectrum rather than discrete types.
Supplemental Material
Supplemental Material - Characteristics of Spinal Morphology According to the “Current” and “Theoretical” Roussouly Classification Systems in a Diverse, Asymptomatic Cohort: Multi-Ethnic Alignment Normative Study (MEANS)
Supplemental Material for Characteristics of Spinal Morphology According to the “Current” and “Theoretical” Roussouly Classification Systems in a Diverse, Asymptomatic Cohort: Multi-Ethnic Alignment Normative Study (MEANS) by Yong Shen, BA, Zeeshan M. Sardar, MD, MSc, Matan Malka, BA, Prerana Katiyar, BS, Gabriella Greisberg, BS, Fthimnir Hassan, MPH, Justin L. Reyes, MS, Jean-Charles Le Huec, MD, Stephane Bourret, PhD, Kazuhiro Hasegawa, MD, PhD, Hee Kit Wong, MBBS, Gabriel Liu, MBBCh, Hwee Weng Dennis Hey, MBBS, Hend Riahi, MD, Michael Kelly, MD, Joseph M. Lombardi, MD, Lawrence G. Lenke, MD, and Multi-Ethnic Alignment Normative Study Group in Global Spine Journal.
Footnotes
Acknowledgments
EOS imaging supported the multicentric collaborative group by providing a web-based solution for image review. Authors did not receive any funding from EOS imaging relative to this work.
Declaration of Conflicting Interests
The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.
Funding
The author(s) received no financial support for the research, authorship, and/or publication of this article.
Ethical Statement
Device Status/Drug Statement
The manuscript submitted does not contain information about medical device(s)/drug(s).
Supplemental Material
Supplemental material for this article is available online.
References
Supplementary Material
Please find the following supplemental material available below.
For Open Access articles published under a Creative Commons License, all supplemental material carries the same license as the article it is associated with.
For non-Open Access articles published, all supplemental material carries a non-exclusive license, and permission requests for re-use of supplemental material or any part of supplemental material shall be sent directly to the copyright owner as specified in the copyright notice associated with the article.
