Lumbar spinal stenosis is a highly genetic condition partly mediated by disc degeneration

Abstract

Objectives

Lumbar spinal stenosis is one of the most commonly diagnosed spinal disorders in older adults. Although the pathophysiology of the clinical syndrome is not well understood, a narrow central canal or intervertebral foramen is an essential or defining feature. The aim of the present study was to estimate the magnitude of genetic versus environmental influences on central lumbar spinal stenosis, and investigate disc degeneration and stature or bone development as possible genetic pathways.

Methods

A classic twin study with multivariate analyses considering lumbar level and other covariates was conducted. The study sample comprised 598 male twins (147 monozygotic and 152 dizygotic pairs), 35-70 years of age, from the population-based Finnish Twin Cohort. Primary phenotypes were central lumbar stenosis assessed qualitatively on MRI and quantitatively measured dural sac cross-sectional area. Additional phenotypes to examine possible genetic pathways included disc bulging and standing height, as an indicator of overall skeletal size or development.

Results

The heritability estimate (h²) for qualitatively assessed central lumbar spinal stenosis on MRI was 67% (95%CI: 56.8-74.5). The broad sense heritability estimate for dural sac cross-sectional area was 81.2% (95%CI: 74.5 – 86.1%), with a similar magnitude of genetic influences across lumbar levels (h²=72.4-75.6). The additive genetic correlation of quantitatively assessed stenosis and disc bulging was extremely high. There was no indication of shared genetic influences between stenosis and stature.

Conclusion

Central lumbar spinal stenosis and associated dural sac dimensions are highly genetic, and disc degeneration (bulging) appears to be one pathway through which genes influence spinal stenosis.

The clinical syndrome of lumbar spinal stenosis (LSS) is a commonly diagnosed spinal disorder in older adults seeking care for back related symptoms. Although the pathophysiology of symptomatic lumbar spinal stenosis is not well understood, a narrow central canal, lateral recess or neuroforaminal canal is an essential or defining feature. In the case of degenerative lumbar spinal stenosis, the most common form of lumbar stenosis, disc degeneration, thickening and buckling of the ligamentum flavum and facet hypertrophy contribute to canal narrowing, which is generally greatest at the level of the disc. While disc degeneration is recognized as having a substantial genetic component (1), less is known about the degree to which genes influence other contributors to degenerative stenosis or associated central spinal canal dimensions.

The aim of the present study was to estimate the magnitude of genetic influences on qualitatively assessed central canal lumbar spinal stenosis and quantitatively assessed central canal dimensions in a general population sample of men. We were also interested in estimating the proportion of genetic influences on canal capacity attributable to shared genetic influences with disc degeneration (bulging), as well as on bone size or development, as indicated through stature.

Subjects and Methods

We conducted an extended classic twin study using multivariate analyses with consideration of lumbar level and other covariates. Such a design allows estimating the relative importance of genetic and environmental influences on qualitatively-assessed central lumbar spinal stenosis and relevant canal dimensions, and evaluating the degree to which the same genetic and environmental influences may be responsible for more than one phenotype.

Subjects

The study sample comprised 598 male twins (147 monozygotic and 152 dizygotic pairs), 35-70 years of age, from the population-based Finnish Twin Cohort. They were recruited in 3 steps. Initially, 116 pairs of MZ twins were identified based solely on discordance between co-twins for a specific common behavioral or environmental factor, as has been described elsewhere (2). This initial subsample of MZ twins was compared to the larger twin cohort that has been shown to be representative of the Finnish population, and was found to be highly representative on a multitude of factors (2). Subsequently, the same number of pairs of DZ twins was selected using the same criteria. Finally, the study sample was further increased by adding randomly selected MZ and DZ pairs from the Finnish Twin Cohort, with zygosity determined using a questionnaire with an estimated probability of misclassification of 1.7% (2).

Study protocols were approved by the Ethical Committees of the Faculty of Rehabilitation Medicine of the University of Alberta and the Department of Public Health at the University of Helsinki.

Data Acquisition

All study subjects travelled to an imaging center in mid Finland for lumbar spine magnetic resonance imaging (MRI) between 1991 and 1997, where images were obtained using one of two 1.5 Tesla imagers, a Siemens Magnetom SP4000 scanner or Magnetom Vision scanner (Siemens AG Erlangen, Germany) using previously described protocols (3). The angle of the slice for the axial images was set at mid disc for each of the lumbar disc levels imaged.

Primary phenotypes: Central canal stenosis and relevant canal dimensions

The primary phenotypes of interest included three measures of anatomical lumbar central canal stenosis, one was a qualitative assessment of stenosis and two involved quantitative measurements of relevant dimensions of canal capacity. Qualitative assessments of all subjects' MR images were performed by an experienced orthopaedic spine surgeon blinded to twinship, environmental exposures and symptom history. The central canal at each lumbar disc level from L2-3 through L5-S1 was rated on a scale from 0-3, with 0 representing no stenosis, 1=mild, 2=moderate and 3=severe. The intra-class correlation coefficient for the intra-observer reliability of the measurements, as determined using blinded, repeated measurements of a sample of 30 subjects' images by the spine surgeon, was 0. 85 (95%CI: 0.77-0.90) for L2-L4 and 0.70 (0.57-0.80) for L4-S1.

Quantitative measurements were acquired of central canal dimensions of the L2 vertebra through the L5-S1 disc region. One measure that was used as an indicator of developmental LSS consisted of the mean A-P diameter of the dural sac along the length of each L2, L3, L4 and L5 vertebral body on the mid sagittal MR image. Cross-sectional area measurements of the dural sac at mid disc level at the L2-3, L3-4, L4-5 and L5-S1 discs from axial images were obtained to reflect degenerative or combined LSS. The segmentation or tracing of all images to acquire the quantitative measurements was conducted using mid-sagittal images by one observer, as was the case for the axial images. The intra-class correlation coefficients for the intra-rater reliability determined from blinded, repeated segmentation of a randomly selected set of 30 subjects' images were 0.98 for dural sac cross-sectional area and 0.83 for the mean A-P diameter of the dural sac along the length of the vertebral body.

Other phenotypes for multivariate analyses

To examine possible pathways through which genes or environment influence LSS, we included qualitative assessments of disc bulging using a scale from 0-3 (intra-rater reliability ICC=0.75-0.76, depending on lumbar region), as described earlier (3). Standing height (cm) was examined as an indicator of skeletal size or development.

Data Analysis

Initially, sample description and data normality tests for quantitative measures (Kolmogorov-Smirnov) were computed. Quantitative measures that did not follow a normal distribution were rank-transformed. Phenotype variance was assumed to result from a combination of four different sources: additive genetic influences (A), reflecting the sum of additive allelic effects of many segregating genes, dominance genetic effects (D), reflecting interactions between alleles at the same locus, common environmental influences (C), reflecting effects of environmental factors shared by twins in a pair, and unique environmental influences to each person (E). Biometric modeling enabled estimating A, D, C and E for different traits utilizing information on the twin and co-twin covariance structure and comparing observed and expected variance-covariance matrices.

Preceded by univariate modeling of all variables, different multivariate models were computed by utilizing a “common pathway” approach. As measures of the same trait at different spinal levels are usually correlated, we built a common pathway model with a latent factor, allowing the estimation of influences that are common to all the spinal levels considered. This procedure offers an overall heritability estimate across spinal levels. Similarly, it is possible to estimate to what extent genetic and environmental influences may underlie the relation of disc bulging and dural sac size by estimating a common pathway model including two latent factors corresponding to the common variability across the spinal area of interest for both traits.

Models were fit in Mx software (4) utilizing the raw data option, and including age as definition variable to regress out age-effects on estimates. For ordinal measures a threshold modeling technique was implemented. The common pathway models including quantitative and ordinal measures (e.g. dural sac cross-sectional area and disc bulging) were fit to polyserial correlation matrices.

Statistically non-significant path estimates were tested sequentially using likelihood-ratio tests (5) in both univariate and multivariate models, in order to find the most parsimonious yet biologically acceptable model providing the best fit to the data. We supplemented this fitting process with Akaike's Information Criterion (AIC), which balances model fit with model parsimony. Smaller AIC values indicate a better fit.

Results

MZ and DZ twins were of similar mean age (49.6 vs. 50.1 years, respectively), weight, height and BMI. Some differences in means and within-pair resemblances between the zygosity groups were detected for qualitatively assessed stenosis and quantitative canal capacity measures at different spinal levels, with smaller measurements for the DZ, as compared to MZ twins (Table 1). Thus, statistical models were also tested allowing heterogeneity in means for MZ and DZ pairs. Although results were virtually the same (estimates differed <2%), the broad sense (overall) heritability estimates are reported from the analyses allowing heterogeneity in means.

Table 1. Descriptive characteristics of the MZ and DZ twins and within-pair similarities.

Subject Characteristics	Mean (95%CI)		Diff *	Within-pair correlations (95%CI)#
	MZ	DZ	P-value	MZ (n=147 pairs)	DZ (n=152 pairs)
Age	49.6 (48.3 – 50.9)	50.1 (48.9 – 51.3)	0.56	1.00 (--)	1.00 (--)
Body weight (kg)	79.3 (77.6 – 81.1)	79.8 (78.3 – 81.3)	0.66	0.69 (0.60 – 0.77)	0.40 (0.25 – 0.53)
Standing height (cm)	175.0 (173.9 – 176.1)	176.1 (175.3 – 177.0)	0.10	0.92 (0.90 - 0.95)	0.51 (0.38 - 0.63)
BMI (kg/m2)	25.9 (25.4 – 26.4)	25.7 (25.3 – 26.1)	0.58	0.65 (0.55 – 0.74)	0.30 (0.14 – 0.44)
Dural sac cross-sectional area (mm2)
L2-L3	177 (171 – 184)	165 (160 – 171)	<0.01	0.73 (0.64 – 0.80)	0.23 (0.07 – 0.38)
L3-L4	154 (147 – 161)	143 (137 – 148)	0.01	0.77 (0.69 – 0.83)	0.25 (0.09 – 0.40)
L4-L5	157 (149 – 164)	140 (132 – 147)	<0.01	0.71 (0.62 – 0.79)	0.31 (0.15 – 0.45)
L5-S1	164 (154 – 174)	136 (128 – 144)	<0.01	0.69 (0.60 – 0.77)	0.32 (0.16 – 0.46)
Mean A-P diameter of dural sac (mm)
L2	13.3 (13.0 – 13.5)	13.0 (12.8 – 13.2)	0.11	0.64 (0.53 – 0.72)	0.13 (-0.03 – 0.29)
L3	12.4 (12.1 – 12.7)	12.0 (11.8 – 12.2)	0.04	0.59 (0.47 – 0.68)	0.12 (-0.05 – 0.28)
L4	12.9 (12.5 – 13.2)	12.1 (11.8 – 12.3)	<0.01	0.48 (0.34 – 0.59)	0.22 (0.06 – 0.37)
L5	15.6 (15.1 – 16.0)	14.8 (14.5 – 15.2)	0.01	0.49 (0.36 – 0.60)	0.14 (-0.03 – 0.29)
Qualitative LSS index (0-3)
L2-L3	0.16 (0.10 – 0.23)	0.18 (0.12 – 0.23)	0.80	0.68 (0.58 – 0.76)	0.40 (0.25 – 0.53)
L3-L4	0.34 (0.25 – 0.44)	0.36 (0.27 – 0.44).	0.90	0.72 (0.64 – 0.80)	0.20 (0.04 – 0.36)
L4-L5	0.56 (0.47 – 0.65)	0.69 (0.58 – 0.79)	0.08	0.43 (0.29 – 0.55)	0.41 (0.26 – 0.54)
L5-S1	0.11 (0.07 – 0.16)	0.14 (0.10 – 0.18)	0.51	-0.02 (-0.19–0.14)	0.07 (-0.10 – 0.23)
Disc bulging (0-3)
L1-L2	0.37 (0.28 – 0.67)	0.42 (0.34 – 0.50)	0.447	0.53 (0.40 – 0.64)	0.25 (0.08 – 0.40)
L2-L3	0.51 (0.41 – 0.59)	0.64 (0.55 – 0.74)	0.029	0.42 (0.28 – 0.55)	0.18 (0.01 – 0.33)
L3-L4	0.34 (-0.27 – 0.95)	0.77 (0.68 – 0.85)	0.172	0.42 (0.27 – 0.55)	0.02 (-0.15 – 0.19)
L4-L5	0.97 (0.88 – 1.05)	1.04 (0.95 – 1.12)	0.243	0.27 (0.11 – 0.41)	0.09 (-0.08 – 0.25)
L5-S1	0.82 (0.74 – 0.90)	0.86 (0.77 – 0.94)	0.527	0.36 (0.21 – 0.49)	0.08 (-0.09 – 0.24)

^*“Diff” for MZ to DZ differences in mean values (quantitative observations)

^#Within-pair correlations are based on Intra-class correlation coefficients for quantitative variables and polychoric correlation coefficients for categorical variables.

Intra-class correlation coefficients demonstrated clearly higher degrees of concordance between MZ twin siblings than between DZ twins for all spinal stenosis phenotypes, supporting a substantial genetic influence (Table 1). The best fitting multivariate model explaining qualitatively assessed LSS included additive genetic and specific environmental influences at all levels, with minor shared environmental influences detected at some of the higher lumbar levels (Figure 1). This model yielded a heritability estimate of 66.9% for genetic influences shared across all spinal levels. Environmental influences increased consistently with decreasing spinal level, from 40% at L2-L3 to 81.6% at L5-S1.

Figure 1.

Graphical representation of best fitting model the qualitative stenosis score variable, including % variance explained, path coefficients and heritability estimates (h²). The As, Cs and Es notations with subscripts represent additive genetic, common environmental or specific environmental influences unique to particular lumbar levels.

The broad sense heritability estimate for the mean A-P diameter of the dural sac along the length of the vertebral body, across lumbar levels, was 73.1% (95%CI: 63.5-80.3%). While the broad sense heritability estimates (and 95%CIs) are best for gauging overall genetic vs. environmental influences on the stenosis phenotypes of interest, the complex multivariate twin models allow comparisons of estimates for individual spinal levels. For the mean A-P diameter a multivariate model including additive and dominance genetic and unique environmental factors fit best and showed a strong dominance genetic component (72.6%). Again, environmental influences increased consistently with decreasing spinal level, from 35% at L2 to 50.5% at L5.

The broad sense heritability estimate for dural sac cross-sectional area at the mid disc level was 81.2% (95%CI: 74.5–86.1%) for influences shared across spinal levels. The best fitting model explaining dural sac cross-sectional area at the mid disc level included additive as well as dominance genetic effects and unique environmental effects. Heritability estimates specific to each spinal level were stable across the lumbar region, ranging from 72.6 to 76.3%, with the largest part of these estimates accounted for by dominance genetic effects (Figure 2).

Figure 2.

Graphical representation of best fitting multivariate model (ADE) for mean dural sac CSA and intervertebral disc bulging. A genetic correlation (rg) for the two major phenotypes of -1.00, % variance explained, path coefficients and heritability estimates (h²) are depicted. The Ds, As and Es notations with subscripts at the bottom of the figure represent specific genetic additive, genetic dominance or environmental influences unique to particular lumbar levels.

The multivariate model for qualitatively assessed disc bulging showed the best fit when including additive genetic, dominance genetic and specific environmental effects (Figure 2). Heritability estimates specific to each level were moderate (∼24% to 44%), and mostly due to dominance genetic influences. The genetic influences affecting disc bulging at the lowest level, L5-S1, were highly specific to this particular site.

The multivariate analysis including dural sac cross-sectional area and disc bulging demonstrated an extremely high additive genetic correlation of rg= -1.00 (95%CI: -1.00 – -0.46) between the two phenotypes (Figure 2), which suggests that only the additive genetic influences on dural sac cross-sectional area across multiple spinal levels are shared with those of disc bulging. A moderate specific environmental correlation of re= -0.44 (95%CI: -0.60 – -0.26) was also detected, suggesting that part of the environmental influences on dural sac cross-sectional area were shared with those of disc bulging (Figure 2). Conversely, the multivariate analysis examining shared genetic and environmental influences for standing height (estimated to be approximately 90% heritable) and dural sac CSA revealed no genetic correlation.

Discussion

Lumbar spinal stenosis is considered primarily a degenerative condition, but the current study demonstrates that it is also highly genetic, and disc degeneration (bulging) appears to be one pathway through which genes influence spinal stenosis. Heritability estimates were 66.9% for lumbar spinal stenosis assessed qualitatively from the clinical perspective of an experienced spine surgeon and 81.2% for dural sac cross-sectional area measured quantitatively at the narrowest point at the disc level.

There was an unexpected difference in the mean measures of canal size between MZ and DZ twins despite the twins coming from the same base population of Finns and being selected in an analogous way. Whether this is due to developmental differences in utero that may be related to the splitting of the zygote, imaging factors or chance is unclear. While the contribution of genetic influences to the variance in qualitatively assessed lumbar spinal stenosis was significantly less at the L5-S1 level, followed by L4-5, as compared to the higher lumbar levels of L2-3 and L3-4, heritability estimates were similarly high for dural sac cross-sectional area at all lumbar levels. The differences between heritability estimates by spinal level as assessed qualitatively suggest that the spine surgeon's assessments may be taking other factors into account in determining stenosis than purely dural sac cross-sectional area, which are affected more by environmental influences in the lower than upper lumbar levels.

The best fitting model explaining the narrowest point at the disc level included additive and dominance genetic influences, as well as unique environmental effects. Whereas the mean anterior-posterior diameter of the dural sac along the length of the vertebral body (bony canal) had a particularly strong dominance genetic component, suggesting that there may be one or more major genes or important gene-gene interactions. If this is the case and the apparent dominance component is not due to chance fluctuation, this may have implications for the search for influential genes with substantial effects. However, while 598 twins is a relatively large sample, the study is underpowered to reliably distinguish additive and dominance genetic effects simultaneously, which leads to relatively wide confidence intervals for some estimates (Figure 2).

The present study also explored possible pathways through which this association could be explained by investigating shared genetic influences affecting both disc degeneration and dural sac cross-sectional area. Our analyses provided evidence that only additive genetic influences on dural sac cross-sectional area are fully shared with those of disc bulging, while dominance genetic influences are completely independent. This suggests the presence of two sets of genes: a group which effects are additive to each other and usually small; and a second group of gene variants with specific effects that tend to interact with each other with larger effects. The fact that skeletal size or development as depicted through standing height had no genetic association with dural sac CSA is consistent with earlier research on the size of the bony lumbar vertebral canal and vertebral body size or stature, outside of achondroplasia.

Lastly, while a narrow canal, either central or foraminal, is an essential aspect of the clinical diagnosis of spinal stenosis and has been linked to symptoms, as with other musculoskeletal conditions, findings on imaging are poorly correlated with symptoms and disability. There may be neurovascular or inflammatory factors or other mediators that cause symptoms to manifest in association with a narrow canal, which may have different genetic and environmental influences. Thus, it cannot be assumed that the heritability estimates from the present study will generalize to the clinical syndrome of spinal stenosis. Nonetheless, the present study substantially advances knowledge of the etiology of an important risk factor or prerequisite of an increasingly common clinical syndrome responsible for chronic pain and disability in older adults.

Supplementary Material

Supplemental Figure

Supplemental Figure: The upper, axial MR image displays the segmentation conducted to obtain the dural sac cross-sectional area. The lower, sagittal MR image displays the segmentation conducted to obtain the mean A-P diameter of the dural sac along the length of each L2, L3, L4 and L5 vertebral body.