Key words
lumbar lordosis - sagittal balance - standing radiographs - supine MRI - sacral slope
- lumbar spine
Objectives
Analysis of sagittal balance has gained in importance, and measurement of radiographic
spinopelvic parameters is now a routine part of many surgical interventions. Indeed,
surgical correction of lumbar lordosis must be proportional to pelvic incidence (PI)
[1]
[2]. Reference data of asymptomatic patients may be helpful to understand individual
disease progression, to achieve goals of movement therapies, and to improve the quality
of monitoring therapies [3].
Sagittal spine geometry is position-dependent. Radiologic measurements cannot be automatically
transferred from one position to another [4]
[5]
[6]
[7]
[8]. Only few studies have examined position-dependent measurement or the interchangeability
of radiologic technologies in lumbar spine imaging [4]
[6]. Regarding the interchangeability of CT measurements in supine position with upright
conventional sagittal radiographs, CT measurements have been shown to be an inappropriate
tool for baseline assessment of sagittal lordosis of the injured thoracolumbar spine
if upright conventional sagittal plane radiographs are used for follow-up measurements
[4]. One investigator using MRI measurements in supine position showed that lumbar lordosis
visible in upright position can be reproduced by placing the patient in supine position
with straightened legs [9]. Other studies have shown good correlations for Cobb angles depending on the particular
curve severity in adolescent idiopathic scoliosis [10]. To our knowledge, the present study is the first to investigate the possible interchangeability
of measurements of sagittal lumbar lordosis obtained by MRI in supine position and
upright conventional radiographs.
The main purpose of the current study was to analyze the correlation between global
lumbar lordosis and single level lordosis measured by means of both MRI in supine
position and radiographs in standing position in patients without any severe back
problems at the moment. We wanted to find out whether MR images in supine position
could also be an alternative to radiographs in standing position as a first assessment
of sagittal balance.
Materials and methods
Subjects
This retrospective study was approved by the local Ethics Committee and follows the
guidelines of this journal. 276 patients were reviewed who had simultaneously undergone
an MRI examination of the lumbar spine in supine position and a lateral standing radiograph
of the lumbar spine with viewable femoral heads between April 2014 and March 2016
([Table 1]). The indication for imaging in our internal assessment was suspicion of rheumatological
disease and history of an episode of low back pain. The following inclusion criteria
applied: (1) patients with both lateral radiographs of the lumbar spine with femoral
heads in standing position and lumbar spine MRI in supine position (2) aged between
30 and 85 years (patients under 30 years of age had not received a lateral radiograph).
Exclusion criteria were (1) previous spinal surgery, (2) a Cobb angle of > 20°, (3)
fracture of a vertebra, (4) acute disc herniation, (5) history of total hip replacement,
(6) severe radiological pathologies of the lumbar spine, and (7) flexion contraction
of the hip. Only 63 of the 276 patients reviewed could be included. Because many patients
showed radiological pathologies in the MRI, the exclusion rate was rather high.
Table 1
Demographic data of the 63 patients included.
Tab. 1 Demografische Daten der 63 eingeschlossenen Patienten.
|
mean ± SD
|
range
|
age (yr)
|
62.3 ± 7.3
|
37 – 78
|
gender (m:f)
|
25:38
|
|
lumbar L1-S1 lordosis on radiographs (°)
|
44.99 ± 10 754
|
29 – 59
|
lumbar L1-S1 lordosis on MRI (°)
|
47.91 ± 9170
|
36 – 74
|
Radiographic imaging
The MRI protocol included a T2 weighted turbo spin echo sequence in sagittal orientation.
All lumbar images were acquired with a 1.5 T system (Ingenia, Philips N.V., Amsterdam,
The Netherlands) using a digital posterior coil. The following scan parameters were
used: slice thickness, 4 mm; repetition time (TR), 3000 ms; echo time (TE), 100 ms;
matrix, 704 × 704; reconstructed voxel size, 0.43 × 0.43 × 4 mm; acquired voxel size,
0.60 × 0.86 × 4 mm; scan duration, 4:12 min; patients in supine position with knees
60° flexed on a pillow.
All lateral radiographs of the lumbar spine were acquired in standing position with
a digital flat panel detector system (Siemens Multix Vertix, Siemens Healthcare, Erlangen,
Germany). The following scan parameters were used: tube voltage, 90 kV; tube current,
dose modulated; focus-to-detector distance of 130 cm with the left body side to the
detector and arms crossed over the body, barefooted; central beam through the iliac
crest.
Radiographic measurements
All radiographic and MR images were assessed by means of the Picture Archiving and
Communications System workstation of our hospital (IMPAX EE, Agfa Healthcare, Bonn,
Germany). All measurements were done with SurgiMap Version 1.2.1.70 [11]. On standing lateral radiographs, all single level angles were measured from the
superior endplate of the upper vertebra to the inferior endplate of the lower vertebra
(L1 / L2, L2 / L3, L3 / L4, L4 / L5). L5 / S1 was measured from the superior endplate
L5 to the superior endplate S1. Global lumbar lordosis (L1-S1) was measured from the
superior endplate L1 to the superior endplate S1. In addition, we measured the pelvic
parameters pelvic incidence (PI), pelvic tilt (PT), and sacral slope (SS) [12]. On MR images in supine position, angles were measured according to the standing
radiographs. We only measured one midline-copied image to make sure to always measure
the same layer of an MR image. ([Fig. 1a, b])
Fig. 1 a, b Example of measuring lateral radiographs and MRI using Surgimap 2016.
Abb. 1 a, b Beispiel der Vermessung von lateralen Röntgenbildern und MRT mit Surgimap 2016.
Each parameter was measured by two independent spinal surgeons, and the mean values
were adapted for analysis. Radiographs and MRI scans were evaluated in random order
on different days within two weeks.
Statistical methods
Statistical analysis was carried out with SPSS (IBM SPSS Statistics, Version 23.0.,
Armonk, NY: IBM Corp.). Metric variables were descriptively reported using means (standard
deviation). We used non-parametric tests for statistical analyzes, because histograms
showed that data were not normally distributed. Therefore, group comparisons were
done by means of the Wilcoxon test. Accordingly, the spearman test was used for correlations.
Statistical significance was set at P< 0.05.
Roussouly groups
In 2005, Roussouly analyzed anteroposterior and lateral radiographs of 160 volunteers
in standardized standing position. “The volunteers were mainly medical students, physical
therapists, nurses, and other allied health professionals affiliated with the medical
center where the study was performed. Those patients who were free of current or historical
symptoms suggestive of spinal or orthopedic disease were included in his study” [12] ([Table 1]). He then analyzed the alignment of the spine and pelvis on the lateral radiographs
and used a four-part classification scheme of sagittal morphology to classify each
patient. In this classification, types 1 and 2 have a sacral slope < 35°. In type
1, the apex of lumbar lordosis is with a short lordosis curve in L5, whilst the apex
in type 2 is in the center of L4 with more vertebral bodies included in the lordosis
curve. Type 3 has the sacral slope between 35° and 45°, and the sacral slope in type
4 is more than 45°. Using Roussouly’s classification, the patients in our study were
divided into 4 groups depending on the sacral slope (SS): Group 1 and 2 had SS< 35°,
group 3 showed 35 <SS< 45, and group 4 SS> 45 ([Fig. 2]).
Fig. 2 4 Roussouly types of sagittal alignment (adapted from Roussouly et al., 2005 [12]).
Abb. 2 Die 4 Typen des sagittalen Alignements nach Roussouly (angelehnt an Roussouly et
al., 2005 [12]).
Results
Total cohort
The mean sagittal parameters of our cohort were close to that of the original Roussouly
cohort of 2005 [12] ([Table 2]).
Table 2
Comparison of the mean sagittal parameters between the original Roussouly [Roussouly
2005] cohort and our cohort.
Tab. 2 Vergleich der durchschnittlichen sagittalen Parameter zwischen der originalen Roussouly
Kohorte [Roussouly 2005] und unsere Kohorte.
parameter
|
roussouly cohort n = 160
(mean ° ± SD)
|
our cohort n = 63
(mean ° ± SD)
|
pelvic incidence (PI)
|
51.91 ± 10.71
|
55.90 ± 11.05
|
sacral slope (SS)
|
39.92 ± 8.17
|
34.29 ± 8.81
|
pelvic tilt (PT)
|
11.99 ± 6.46
|
21.00 ± 6.35
|
On standing radiographs, global lumbar lordosis (L1-S1) was 44.99° (± 10 754) across
the four Roussouly groups. On MR images in supine position, global lumbar lordosis
was 47.91° (± 9170). ([Table 1]) Despite this significant difference (p < 0.05), the clinically relevant correlation
was weak: r = 0.611, p < 0.01. The mean intra-individual difference between L1-S1
on standing radiographs and on MRI in supine position was –2.9° (± 9080). Angles on
the standing radiographs ranged between 29° and 59° and from 36° to 78° on the MRI.
Analyses of each single level showed increases in mean lordosis from L1 / L2 to L4 / L5
on both technologies (radiographs 5.25°-22.54°, MRI 3.98°–21.30°). Level L5 / S1 showed
highly interesting values: angles on the radiographs had decreased to 14.83° (± 7600)
but increased to the highest lumbar value of 25.04° (± 3925) on the MRI. Therefore,
the differences between radiographic and MR image were slightly positive in the upper
lumbar spine but had decreased to negative values in the lower lumbar spine ([Fig. 3]). The correlation R was 0.611.
Fig. 3 Difference of lumbar lordosis between radiographic and MR images at every single
lumbar level; in the upper lumbar spine, segmental lordosis is increased in the standing
x-ray; in the two lower levels, especially at L5 / S1, segmental lordosis is much
more pronounced on the MRI in supine position than on the x-ray.
Abb. 3 Unterschied der lumbalen Lordose zwischen Röntgenbildern und MRT in jedem lumbalen
Wirbelsäulensegment; die lumbale Segmentlordose in der oberen LWS ist in den stehenden
Röntgenaufnahmen erhöht; in den zwei unteren Segmenten, ist die Lordose besonders
in L5 / S1 im MRT deutlich höher als im Röntgen.
Findings in Roussouly group 1
18 patients were assigned to group 1. Lumbar L1-S1 lordosis on the radiographs was
39.14° (± 9465) and 42.53° (± 6326) on the MRI, thus somewhat less than in the total
cohort for both technologies ([Fig. 4]). These values presented the best and clinically most significant correlation of
all groups R = 0.707, (p < 0.01) that was even higher than the correlation in the
total cohort. Values of the single level analysis also corresponded with the findings
of the total cohort ([Fig. 5]). The mean intra-individual difference between L1-S1 on standing radiographs and
MRI in supine position was -3.39° (± 6.11) ([Fig. 6]). In this group, L5 / S1 lordosis measured by means of MRI was so high that even
global lordosis was higher on MRI than on the radiograph ([Fig. 6]).
Fig. 4 Comparison of lumbar L5 / S1 lordosis measured by radiographic and MR images for
each Roussouly group.
Abb. 4 Vergleich der Lordose im Segment L5 / S1 im Röntgen und MRT bezogen auf die einzelnen
Roussouly Gruppen.
Fig. 5 Mean single level lordosis of the four Roussouly types measured by means of radiographs
(black) and MRI (red).
Abb. 5 Durchschnittliche Segmentlordose für die vier Roussouly Gruppen im Röntgen (schwarz)
und im MRT (rot).
Fig. 6 Difference in lumbar L1-S1 lordosis measured by radiographic and MR images according
to Roussouly types; in type 1, 2 and 4, lordosis in x-ray is higher than that on MRI;
in type 3, lordosis is higher in standing position. The largest mean difference with
–8.3° is found in type 3.
Abb. 6 Unterschiede der Lordose L1-S1 im Röntgen und MRT bezogen auf die Roussouly Typen;
bei Typ 1, 2 und 4 ist die Lordose im Röntgen mehr als im MRT; bei Typ 3 ist die Lordose
im stehenden Röntgen höher. Den größten durchschnittlichen Unterschied mit –8,3° sieht
man bei Typ 3.
Findings in Roussouly group 2
14 patients were assigned to group 2. Lumbar L1-S1 lordosis on the radiographs was
36.96° (± 4826) and 45.29° (± 5539) on the MRI ([Fig. 6]). This finding was the largest intra-individual difference between L1-S1 on standing
radiographs and the MRI in supine position of all groups, –83 214° (± 5276). For level
L5 / S1, group 2 was the only one showing a significant difference (p < 0.001) ([Fig. 6]); level L4 / L5 showed the nearly the same values on both technologies (radiograph
19.82° and MRI 19.61°). All other groups showed higher angles for L4 / L5 on the radiographic
than on the MR images.
Findings in Roussouly group 3
Group 3 included 19 patients. Lumbar L1-S1 lordosis on the radiographs was 53.87°
(± 7308) and 51.63° (± 11 040) on the MRI. This value was the highest of all groups
and higher than the radiographic value for the total cohort ([Fig. 4]). Although the mean difference was only 2237° (± 11 029), there was no correlation
(R = 0.167) ([Fig. 6]). Group 3 was also the only group with a higher value for L1-S1 lordosis on radiographs
than on MRI. Values for single level lordosis on radiographs were higher in the upper
levels than those in other groups ([Fig. 5]).
Findings in Roussouly group 4
12 patients were assigned to group 4. Lumbar L1-S1 lordosis on the radiographs was
49.08° (± 10 297) and 53.17° (± 8133) on the MRI, which was the highest value of all
MRI measurements ([Fig. 4]). The mean intra-individual difference between L1-S1 on standing radiographs and
MRI in supine position was -40 833° (± 9449) ([Fig. 6]). This finding was not significant and had no correlation. In group 4, L5 / S1 lordosis
(25.71°) on MRI was decreased in comparison to L4 / L5 lordosis (27.63°). In the other
groups, this finding was only seen on the radiographs but not on the MRI. With regard
to the other levels, the findings corresponded to those of the other groups ([Fig. 5]).
Discussion
Evaluating sagittal balance has become more and more important in the treatment of
low back pain and in planning surgical procedures [1]
[2]
[13]
[14]
[15]. No study has yet really addressed the use of MRI with regard to sagittal profile
measurement of the lumbar spine with regard to differences in single levels in consideration
of the Roussouly classification of different types of sacral slope [12]. In this study, we investigated the influence of sacral slope on the correlation
between measurements of lumbar lordosis obtained by standing radiographs and MRI in
supine position.
Previous studies have investigated thoracic kyphosis and − as control − lumbar lordosis
using MR images in supine position to evaluate the coronal Cobb angle [16]
[17]
[18]
[19]
[20].
Inter-observer reliability was higher in MRI than in X-ray measurements, which may
lead to the conclusion that MR images can be measured more precisely [19]. However, we only measured one midline-copied image to make sure to always measure
the same layer of an MR image. The main problem of MRI examinations is the supine
position, which eliminates the effect of gravity. All patients undergoing MRI in supine
position had a 60°-angle pillow under their knees to make lying on the back more comfortable.
A recent study has shown a higher degree of slippage in symptomatic patients with
spondylolisthesis in supine magnetic resonance imaging with straightened lower extremities
than in conventional MRI with flexed lower extremities[21]. Changes in lumbar lordosis due to knee flexion have been controversially discussed
in the literature. The best comparison we have found are changes in lumbar lordosis
when wearing high-heels, which increases knee flexion. For such cases, some authors
described decreased lumbar lordosis in comparison to the normal standing position
[22]
[23]
[24]; others found increased lumbar lordosis, and several authors did not find any differences
[25]
[26]. However, none of these authors had considered the different Roussouly types or
single level lordosis. Therefore, our finding that lumbar lordosis slightly increased
in groups 1, 2, and 4 and decreased in group 3 corresponded with the literature [16]
[17]
[18]
[19]. We also found these effects at each single level, except for level L5 / S1, which
showed a reversed effect. This effect was so great that it also influenced the measurement
of global lumbar L1-S1 lordosis. Thus, we found a slightly greater level of lordosis
on the MRI than on the radiograph. We have no direct anatomic explanation for this
fact, but 60° knee flexion may have this reverse effect on L5 / S1.
In clinical practice, healthy patients could undergo an MRI in supine position to
avoid unnecessary exposure to radiation [27]
[28], if only the level of global lumbar lordosis needs to be assessed as a first planning
of conservative treatment of low back pain. If further information should be required,
particularly single level angles, for instance when planning a surgical procedure
such as spinal fusion [29]
[30], the two scanning technologies should be used simultaneously because of their increasing
differences between upper and lower spinal levels.
Our data analysis according to the four Roussouly groups [12] also showed some differences in the single levels. In group 1, “the sacral slope
is less than 35°, and the apex of lumbar lordosis is the center of the L5 body.” Roussouly
described the lower arc of lordosis as minimal [12]. In comparison, group 2 “has the apex in body L4.” The radiographs in our study
showed the same findings: i. e., the angle L4/5 was higher in group 1 than in group
2. Group 1 had the best correlation between MRI and radiograph measurements.
In group 3, “the apex of lumbar lordosis is in the center of the L4 vertebral body.
The lower arc of lordosis becomes more prominent.” [12] In group 4, “the apex of the lumbar lordosis is located at the base of the L3 vertebral
body or higher.” Our radiographs had the highest angles of all groups at L3/4 and
accordingly in group 3 that also showed a higher angle at L2/3. Because group 3 showed
such high levels of lordosis at L2 / L3 and L3/4 on the radiographs so that also L1-S1
lordosis was still higher than that measured by means of MRI in supine position, we
obtained the presumed result with a smaller angle for L1-S1 on MRI.
The major limitation of the present study is the uncertainty of distinguishing between
the effects of body position within one radiologic technology or between the two different
technologies using the same body position. Another limitation is the retrospective
approach. Although our scanning modalities are usually standardized, correct positioning
of the patient cannot be guaranteed retrospectively. Additionally, our patients had
no severe back problems but only minor disc degeneration. In the case of more degenerative
discs, the stability of the spine may be influenced, which could result in higher
differences. Therefore, further studies with the same design but using degenerated
discs or degenerated spondylolisthesis are recommended.
Conclusion
Although the two scanning technologies investigated showed a significant difference
in global lumbar lordosis, the mean difference was just 2.9°. Thus, there was only
a clinically weak correlation for global lumbar lordosis. That means MRI in supine
position may be used for estimating global lumbar lordosis as required, for instance
to indicate conservative treatment options for low back pain as a first approach.
Analysis of single level lordosis, for example for planning surgical fusion, necessitates
radiographs in standing position, particularly of the lower lumbar spine.