Lumbar Lordosis Distribution in Asymptomatic Adult Volunteers: A Systematic Review

Abstract

Background:

Restoring lumbar lordosis is important for adult spinal deformity surgery. Several reports have suggested that lumbar lordosis distribution has a significant impact on the outcome of surgery, including lumbar distribution index (LDI), proximal lumbar lordosis (PLL), and distal lumbar lordosis (DLL). The features of lumbar lordosis distribution are inconclusive in asymptomatic adults.

Questions/Purposes:

We sought to evaluate the variation of lumbar lordosis distribution (LDI, PLL, and DLL) and to identify associated factors in asymptomatic adult volunteers.

Methods:

We performed a systematic review of the Embase and Medline databases to identify studies in asymptomatic adult volunteers to evaluate lumbar lordosis distribution including LDI, PLL, and DLL.

Results:

Twelve articles met eligibility criteria and were included in our review. The respective pooled estimates of mean and variance, respectively, were 65.10% (95% confidence interval [CI]: 62.61–67.58) and 13.70% in LDI, 16.51° (95% CI: 5.54–27.49) and 11.46° in PLL, and 35.47° (95% CI: 32.79–38.18) and 9.10° in DLL. Lumbar lordosis distribution was associated with race, age, sex, body mass index, pelvic incidence, and Roussouly classification.

Conclusions:

This systematic review found that despite a wide variation in LDI and PLL, DLL is maintained in a narrower range in asymptomatic adult volunteers, especially in white populations. Distal lumbar lordosis may be a more reliable radiographic parameter to restore the lumbar lordosis distribution in preoperative planning.

Introduction

Mechanical complications after operative treatment for adult spinal deformity (ASD) are very common, including proximal junctional kyphosis (PJK) or proximal junctional failure (PJF), distal junctional kyphosis or failure, rod breakage, or implant-related complications [30]. The incidence of mechanical complications is in approximately 20% to 50% of patients after ASD surgery, which can result in various functional and clinical consequences, often requiring revision surgery [18]. Appropriate correction of the sagittal spinal alignment through ASD surgery can reduce mechanical complications and improve clinical outcomes [23,30].

There is an increasing interest in the optimal lumbar lordosis for restoring normal lumbar sagittal alignment in preoperative planning, which includes the magnitude and distribution of lordosis. Lumbar lordosis distribution can be reflected by the proximal and distal arc lordosis value and ratio, including lumbar distribution index (LDI), proximal lumbar lordosis (PLL), and distal lumbar lordosis (DLL). The appropriate magnitude of global lumbar lordosis has been established in the Scoliosis Research Society (SRS)-Schwab classification and the Global Alignment and Proportion (GAP) score [23,30]. Lumbar lordosis distribution can further provide an accurate description of the lumbar lordosis morphology, which was confirmed to be associated with the outcome of the ASD surgery [30].

Currently, there are 2 main methods describing the lumbar lordosis distribution: functional and anatomical approaches (Fig. 1). In the former, lumbar lordosis is divided into an upper and lower arc, separated by a horizontal line [19]. In the latter, lumbar lordosis is divided into proximal lordosis and distal lordosis, separated by the upper endplate of L4 [30]. Lumbar distribution index serves as a ratio of lordosis distribution, which is defined as distal lordosis/global lordosis × 100% [30]. Functional lumbar lordosis distribution can be successfully applied in normal people but would be unsuitable in degenerative disease, as the compensatory changes of pelvic tilt can lead to the changes of the horizontal reference in spinal diseases [16]. Thus, the use of this approach may produce misleading inferences in the setting of spinal pathology. On the contrary, use of anatomical landmarks is more convenient and makes it easier to define normative values of lumbar lordosis distribution. This approach can help surgeons to determine the malalignment levels accurately during preoperative planning.

Fig. 1.

Functional and anatomical lumbar lordosis. Functional lumbar lordosis is divided into upper and lower arc, separated by a horizontal line. Anatomical lumbar lordosis is divided into proximal lordosis and distal lordosis, separated by the upper endplate of L4.

At present, there is limited literature exploring the normal anatomical lordosis distribution. Therefore, to have a better understanding of lumbar alignment, the objective of this study was to systematically review the literature to evaluate the variation of lumbar lordosis distribution (LDI, PLL, and DLL) and to identify associated factors in asymptomatic adult volunteers.

Methods

A literature research was performed on the Embase and Medline online databases, in accordance with the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guidelines (https://www.prisma-statement.org/). The following search terms were used: ((‘asymptomatic disease’/exp OR asymptomatic:ti,ab OR ‘normal human’/exp OR volunteers:ti,ab) AND ‘lumbar lordo*’:ti,ab AND [english]/lim). The final search was performed on June 20, 2022. All the articles considered significant were collected for inclusion, and their bibliographies were checked for additional references.

Cross-sectional and case-control studies were included in this review. Studies included in the analysis had to meet the following inclusion criteria: (1) asymptomatic adult volunteers (age ≥18 years) without history of a spinal disorder or spinal surgery; (2) without radiographic spine abnormality; (3) without hip, knee, and ankle abnormalities; (4) radiographic data were collected by lateral radiographs in standing position using Cobb’s angle method; and (5) studies had to analyze at least one of the primary parameters. Primary parameters included LDI, PLL, and DLL.

The initial literature search was performed by 2 authors (T.H.G. and J.T.A.) to identify citations with potential relevance separately. First, titles and abstracts were screened. If abstracts did not provide sufficient data, the entire article was acquired for further evaluation. Afterward, the full texts of the potential eligible articles were assessed for final inclusion. A senior author (Y.Q.S.) was consulted in cases of disagreement on the inclusion of articles.

Data Collection and Analysis

The following variables were collected: (1) study characteristics (author name, study year, study type, level of evidence, and country); (2) patient demographics (race, number, age, and sex); and (3) primary parameter measures (LDI, PLL, and DLL). As for the parameters studied, the PLL was the angle between L1 upper endplate and L4 upper endplate. The DLL was the angle between L4 upper endplate and S1 upper endplate. The DLI was defined as the ratio of DLL to lumbar lordosis. Data (the mean and variance) were extracted from graphs if data were not presented in tables or text.

The data analysis was performed using a random-effects model. The pooled estimates of mean values were calculated as weighted averages of study-specific means by the random-effects model in Stata 15 (Stata Corporation). The pooled estimates of variance values were calculated by Cochrane’s formula in StatsToDo (http://www.statstodo.com/CombineMeansSDs.php). Statistical heterogeneity across the studies was evaluated using the I² test.

Results

A total of 923 titles and abstracts were identified in the initial search. After duplicates were eliminated, 577 records remained. Through the screening of titles and abstracts, a total of 50 potentially suitable articles were analyzed in full; 38 articles were excluded as these articles did not meet the inclusion criteria. Thus, a total of 12 studies [1,4,6,8,9,11,13,14,16,26,28,32] were finally included in this systematic review (Fig. 2, Table 1).

Fig. 2.

Preferred Reporting Items for Systematic Reviews and Meta-Analyses flowchart.

Table 1.

Characteristics of included studies.

Study	Study type	Level of evidence	Country	Population	Number	Age	Male-Female	Radiographic measurement
								LDI	PLL	DLL	LL	PI	PT	SVA	PI-LL	TK
Ma et al [14]	Cross-sectional study	III	China, Nanjing	Asian	490	47.8 ± 16.2 (20–79)	210/280	×		×	×	×	×	×	×	× T5-T12
Li et al [13]	Cross-sectional study	III	China, Jinan	Asian	87	37.3 ± 7.2 (19–45)	54/33		×	×	×	×	×			× T5-T12
Zhou et al [32]	Case-control study	III	China, Nanjing	Asian	158	47.1 ± 11.3	62/96	×		×	×	×	×	×		× T5-T12
Wegner et al [28]	Cross-sectional study	III	USA, New York	White	85	47.9 ± 16.2 (22–77)	20/65	×		×	×	×
Chung et al [4]	Cross-sectional study	IV	South Korea, Suwon	Asian	160	32.8 ± 8.9 (21–50)	75/85			×	×	×	×	×		× T4-T12
Ando et al [1]	Cross-sectional study	III	Japan, Nagoya	Asian	286	64.5 ± 10.2	109/177			×	×	×	×	×	×	× T4-T12
Hyun et al [6]	Cross-sectional study	III	South Korea, Seoul	Asian	150	64.1 ± 6.4	150/0			×	×	×	×	×		× T5-T12
Pesenti et al [16]	Cross-sectional study	III	USA, New York	White	119	50.7 ± 17 (22–78)	38/81		×	×	×	×	×		×
Sohn et al [26]	Cross-sectional study	III	South Korea, Seoul	Asian	128	54.1 ± 10.4 (30–79)	63/65			×	×	×	×
Kim et al [8]	Cross-sectional study	III	South Korea, Seoul	Asian	342	40.9 (19–79)	342/0	× TLL		×	× TLL	×	×			× T5-T12
Lee et al [11]	Cross-sectional study	III	Korea, Seoul	Asian	86	28.2 (19–39)	54/32			×	× TLL	×	×			× T3-JL
Korovessis [9]	Case-control study	III	Greece, Patras	White	100	49 ± 18	100/0			×	× TLL

LDI lumbar distribution index, PLL proximal lumbar lordosis, DLL distal lumbar lordosis, LL lumbar lordosis, PI pelvic incidence, PT pelvic tilt, SVA sagittal vertical axis,TK thoracic kyphosis, TLL Total lumbar lordosis was the angle between the T12 inferior endplate and the S1 superior endplate

Ten of the included studies were cross-sectional, and 2 were case-control studies. Five studies were performed in South Korea, 3 in China, 2 in the United States, 1 in Japan, and 1 in Greece. Three of the studies were conducted in an asymptomatic population under 50 years of age, and 9 were conducted in a population of mixed age. Three of the studies were conducted in male volunteers and 9 were conducted in male and female volunteers. As for the primary outcomes, 4 studies reported LDI, 2 studies described PLL, and all studies reported DLL.

Four studies (1 with white participants and 3 with Asian participants) reported the mean and standard deviation of LDI in 1075 asymptomatic adults. The pooled estimate of the mean was 65.10% (95% confidence interval [CI]: 62.61–67.58, I² = 88.3%) across studies (Fig. 3). The overall pooled estimate of the variance was 13.70%.

Fig. 3.

Forest plot of the lordosis distribution index in all asymptomatic populations. ES estimate, CI confidence interval.

Lumbar distribution index is affected by pelvic incidence (PI), Roussouly classification, sex, and age. It has been found that the value of the LDI decreases as the PI increases. Pesenti et al [16] reported that the proximal lordosis increased with PI, whereas the distal part remained constant to show that the LDI decreased throughout PI groups. In addition, LDI was affected by Roussouly classification. Chung et al [4] reported that a significant difference was found in distal lordosis among the Roussouly types and LDI was 83.4% in Roussouly type 1, 65.2% in type 2, 64.7% in type 3, and 61.5% in type 4. For sex-related difference of LDI, Ma et al [14] reported that male volunteers had significantly higher LDI than female volunteers. However, the relationship between age and LDI is controversial. Kim et al [8] demonstrated that the old men group had a significant increase in LDI, compared with the young men group. In contrast, Wegner et al [28] found that the old group had relatively less value and larger range of LDI than the young group.

Two studies (1 with white participants and 1 with Asian participants) reported the mean and standard deviation of PLL in 206 asymptomatic adults. The pooled estimate of the mean was 16.51° (95% CI: 5.54–27.49, I² = 98.4%) across studies (Fig. 4). The overall pooled estimate of the variance was 11.46°.

Fig. 4.

Forest plot of the proximal lordosis in all asymptomatic populations. ES estimate, CI confidence interval.

Proximal lumbar lordosis is affected by age and PI. Pesenti et al [16] reported that age was negatively strongly correlated with PLL and PI was positively strongly correlated with proximal lordosis. Li et al [13] revealed that PI had a great correlation with proximal lordosis and the muscularity of erector spinae and multifidus were not correlated with proximal lordosis.

Twelve studies (3 with white participants and 9 with Asian participants) reported the mean and standard deviation of DLL in 2191 asymptomatic adults. The pooled estimate of the mean was 35.47° (95% CI: 32.79–38.18, I² = 98.5%) across studies, 36.22° (95% CI: 35.31–37.13, I² = 10.9%) in white participants, and 35.24° (95% CI: 31.85–38.64, I² = 98.9%) in Asian participants (Fig. 5). The overall pooled estimate of the variance was 9.10°.

Fig. 5.

Forest plot of the distal lordosis in all asymptomatic populations. ES estimate, CI confidence interval.

The commonly held opinion is that there is an association between age and DLL. Wegner et al [28] found that the old group (≥60 years) had less DLL than the young group (<60 years). Ma et al [14] also reported that old female volunteers had less DLL than the young female volunteers. Conversely, Kim et al [8] demonstrated that the old men group had a significant increase in DLL compared with the young men group. However, Pesenti et al [16] reported that age was not correlated with DLL. Sohn et al [26] suggested that DLL did not differ between the group above 70 years of age compared with the other age groups, and the magnitude of DLL gradually increased with age groups.

Two researchers agree that the DLL does not differ by sex. Lee et al [11] reported that DLL was similar in male and female young adult participants. Sohn et al [26] suggested that DLL was not found to differ significantly between male and female participants in asymptomatic patients aged above 30 years in the Korean population. However, Ma et al [14] reported that female volunteers had significantly lower DLL than male volunteers in the less than 60 years age group.

One study Ando et al [1] evaluated the relationship between obesity and spinal sagittal alignment to show that obese group (body mass index ≥25 kg/m²) had significantly lower DLL than non-obese group (body mass index <25 kg/m²). There were no differences in radiographic parameters between obese and non-obese males, but obese females had significantly lower DLL compared with non-obese females.

It is debatable whether DLL was affected by PI. Pesenti et al [16] reported that PI was not correlated with DLL. They performed stratification by PI to show that the distal part remained constant. Li et al [13] revealed that PI had a correlation distal lordosis and distal lordosis was not found to differ significantly between low PI group (30°–45°) and moderate PI group (45°–60°). Conversely, Hyun et al [6] also found that PI was an important predictive factor for DLL and the magnitude of DLL increases with the PI.

One study evaluated the relationship between Roussouly classification and DLL. Chung et al [4] reported that although statistically significant, DLL in all types was approximately 30° to 40°.

One study evaluated the relationship between paraspinal muscle and DLL. Li et al [13] reported the muscularity of erector spinae, and multifidus were not correlated with DLL.

Discussion

It is necessary to understand the lordosis distribution of the lumbar spine, as spinal fusion surgery aims to restore the natural spinal alignment. Correction of LL proportional to PI (PI-LL <10°) is critical as described in the SRS-Schwab surgical strategies, which is associated with good health-related quality of life [21]. However, one study reported that there was a 37.9% incidence of mechanical complications in the optimal surgical target of PI-LL <10° [31]. Although PI-LL criteria provided a successful predictor for determining total amount of LL restoration, it did not help to guide short-level spine fusion surgeries in the lower lumbar area or the multilevel spine fusion to achieve harmonious lumbar shape. Yilgor et al [30] proposed LDI as a novel radiographic parameter in the GAP score to predict mechanical complications for surgically treated ASD patients. Lumbar distribution index <50% suggests hypolordotic maldistribution, LDI of 50% to 80% suggests aligned distribution, and LDI >80% suggests hyperlordotic maldistribution [30]. Jacobs et al [7] reported that the GAP score was more capable of predicting mechanical complications than SRS-Schwab classification. Therefore, it seems crucial to restore not only the total amount of LL but also the lumbar lordosis distribution according to PLL and DLL. However, Xu et al [29] reported that LDI was not related to PJK when the distribution of the lordosis was considered to be aligned if the LDI was within the range of 50% to 80%. Tobert et al [27] reported that previously developed categories of LDI did not show differences in PJF. The possible reason is that previously developed category was not accurate enough, as this was derived from ASD patients after surgery rather than from normal people.

Race-related difference should be considered in the assessment of lumbar lordosis distribution [3,14]. For evaluation of LDI and PLL, we could not draw affirmative conclusions due to the small number of studies with white participants. Meanwhile, heterogeneity was low for the included studies with white participants (I² = 10.9%) in DLL analysis, which indicated high agreement of DLL value with a little bias and narrow ranges in white participants. On the contrary, the heterogeneity was significantly high among the included studies with Asian participants (I² = 98.9%). Diebo et al [5] analyzed compensatory patterns among the ASD patients from Japan, Korea, and the United States, reporting that Asian populations had the greater ability to “tolerate” PI-LL mismatch than white populations, which supported our finding that Asian populations may have a wider variation of DLL than white populations in healthy subjects.

The LDI displayed a wide variation, and the normal range of LDI in the literature did not completely coincide with the range of the aligned distribution in GAP. Lumbar distribution index is affected by many factors, such as PI, Roussouly classification, sex, and age. It is found that the value of the LDI decreases as the PI increases in this systematic review. Anwar et al [2] reported that the proportion of total LL contributed at L4-L5 and L5-S1 reduced as PI increased. However, the relationship between age and LDI is controversial. The possible reason is that the aging of lumbar spine differs among the Roussouly types [25]. In type 1, LDI increases with age due to compensatory increase in DLL caused by decreased PLL in the junctional zone (L2-L3). In type 2 or type 3, LDI remained nearly unchanged with age, as a type 2 or 3 spine generally loses few LL with few increasing PT. In type 4, LDI decreases with age due to decreased DLL in the degenerative evolution.

The PLL varies over a wide range and increases as the PI increases in the literature. Pesenti et al [16] report that the lordosis within the proximal portion of the lumbar spine increased throughout the increased PI. Proximal lordosis was responsible for almost 30% of the total lordosis in patients with low PI and 50% of the total lordosis in patients with high PI. Anwar et al [2] reported that PI correlated strongly with the L1 and L2 motion segments. Hence, restoration of PLL is also important in surgeries to correct ASD based on PI parameters.

The DLL has relatively small variations in the literature, especially for white participants. Predictive formula of DLL was determined by individual PI in asymptomatic population in 4 studies [2,6,12,16] (Fig. 6). Three studies showed that poorly positive or no associations were observed for DLL and PI except the Lee et al [12] study. The Anwar et al [2] study and Lee et al [12] study were limited to a young, healthy population and cannot be generalized to the entire population. The difference of DLL value among 3 predictive formulas was less than 5° except for the Lee et al [12] study, which showed that PI produced a small effect on DLL and DLL value was approximately 36° when PI value was 30° to 60°.

Fig. 6.

Relationships between PI and DLL in the 4 predictive formulas when PI value was 30° to 60°. PI pelvic incidence, DLL distal lumbar lordosis.

Although LDI is a useful predictor, this approach has limitations. The LDI appeared highly variable between individuals and displayed a wide range of values. It is not clear to restore lumbar lordosis distribution in individualized treatment. In addition, LDI presents the cumbersome and paradoxical nature [27]. As lordosis is very small in Roussouly type 1, the maximal thoracic kyphosis (TK) is large and extends caudally beyond L2, which may result in that LDI more than 100% [19]. Moreover, the concept of using LDI solely as an absolute numeric value may be misleading, particularly when compensatory mechanisms occur in adjacent segments. Distal lumbar lordosis is considered as a more reliable radiographic parameter to restore the lumbar shapes in preoperative planning. The lower lumbar spine is the most important part of determining the lumbar lordosis, and the value of DLL was nearly constant [13,16]. Although there was a statistical difference in LDI among the Roussouly types, DLL in all types was approximately 30° to 40° [4]. Pesenti et al proposed that DLL should be preserved or restored around 36° to reduce the risk of PJF [17].

A simplified approach to DLL restoration may be easy to remember and useful in functional and anatomical restoration. In anatomical lumbar restoration, Lafage et al redefined age-adjusted targets for ideal LL based on the Schwab formula of PI minus LL (PI − LL), and Michael et al further proposed an age-adjusted formula to identify the amount of ideal LL (LL = PI − 0.3TK − 0.5Age + 10) [10,15]. The ideal DLL value is 36°, and the ideal PLL formula is as follows: Ideal PLL = Ideal LL − 36°, which approach can provide a more precise and individualized interpretation of the lordosis distribution. Interestingly, the 35° also was the cutoff value of sacral slope between Roussouly type 2 and type 3 in functional restoration [19]. The ideal DLL value is 36°, and ideal apex location is based on PI value (PI < 50°, ideal apex is located at the lower L4 vertebral body; PI > 50°, ideal apex is located at the upper L4 vertebral body or the L3-L4 disk) [24]. Thus, lumbar shapes with low PI were reconstructed into Roussouly type 2 and lumbar shapes with high PI were reconstructed into Roussouly type 3, as types 2 and 3 were well-balanced harmonious spines and the inflection point was at the thoracolumbar junction (T12-L1 disk) [24]. Scemama et al [20] reported that reducing the kyphosis and transforming in type 2 could work in ASD patients with thoracolumbar kyphosis (TLK) and low PI. In patients with large PI, a lower correction of lumbar lordosis may be a better option and surgical target of type 3 is enough [22]. Restoring the sagittal spinal contour to the normal shapes of Roussouly according to the PI could reduce mechanical complications [24].

To the best of our knowledge, this is the first systematic review assessing lumbar lordosis distribution in asymptomatic adult volunteers. However, the present review has several limitations. First, we cannot rule out that the analysis of the studies found is to some extent subjective. Second, this review is based on a small number of studies. Third, there is a lack of comparative studies on lumbar lordosis distribution according to race, which needs to be further studied. Fourth, future studies are needed to investigate the correlation between DLL and PI to better understand the lumbar shape.

In conclusion, despite a wide variation in LDI and PLL, DLL is maintained in a narrower range in asymptomatic adult volunteers, especially for white populations. Distal lumbar lordosis may be a more reliable radiographic parameter to restore the lumbar lordosis distribution in preoperative planning.