Introduction

Low back pain (LBP) is one of the largest causes of disability worldwide. It is a leading cause of work absenteeism and medical consultation [1]. This is reflected by a huge medical and economic social burden. The prevalence of LBP is higher in females, and a 2019 estimation of 568.4 million LBP cases overall makes it the leading morbidity worldwide [2]. One of the main causes of LBP is spine degeneration, in particular the age-related changes of the intervertebral discs (IVD) [3]. The IVD is the largest avascular organ in the body, and it composed of three anatomical regions: nucleus pulposus (NP), a gelatinous proteoglycan-rich central structure, annulus fibrosus (AF): a ring of ligamental fibres surrounding the NP and cartilaginous endplates (CEP) that enclose the disc and separate it from vertebral bodies. Vertebral CEPs play an important role in the homeostasis of IVDs serving as an interface between discs and bones and providing nutrients to IVD cells. Another important function of CEP is to prevent contact between host immune cells and the IVD acting as a physical barrier [4]. With age, IVD faces biochemical and morphological changes and starts to degenerate. Intervertebral disc degeneration (IDD) is a pathological process that leads to the loss of function of this anatomical structure favouring the onset of a plethora of disorders that eventually lead to LBP [5].

One of the IDDs is a bone marrow lesion within the vertebra and change to adjacent bony endplates detectable with magnetic resonance imaging (MRI) named Modic change (MC) [6]. Evidence suggests that these manifestations are part of the same pathological process influencing the onset of LBP [7]. MC was first reported by Ross in 1987 [8] and subsequently described by Modic et al. Risk factors for MC are similar to those for IDD and include increased age, high body mass index (BMI) and smoking; however, the condition also has a heritable component estimated at 30% [9].

Despite known risks and heritability, it is not clear how the onset of MC is triggered, and why this only occurs in a subset of people affected by IDD. One possible cause could be the subjective ability of bone marrow to respond to inflammation coupled with the damage of CEP (ED) [10]. When bone marrow and protruded disc tissue get in contact, the immune-privileged disc tissue could trigger an autoimmune response which could enhance the onset of MC [11, 12]. One characteristic of IDD is a compromised endplate, which, when damaged, along with the ingrowth of blood vessels leads to the infiltration of activated immunocytes and inflammatory cytokines in the disc space.

If MC has an autoimmune component, then people with autoimmune conditions would be expected to manifest higher prevalence and severity of MC. An autoimmune diagnosis may be associated with altered immunoreactivity [13], including a more severe response if NP tissue is exposed to immune cells. The aim of the current investigation was to assess prevalence and severity of MC and LBP in participants with an autoimmune disease diagnosis in a well-documented cohort of adult twin volunteers having prospectively gathered data over many years.

Methods

This cross-sectional retrospective study determined whether adults diagnosed with an autoimmune condition demonstrate a higher prevalence or severity of MC. The cohort is composed of twins enrolled in the TwinsUK adult twin registry based on King’s College London [14]. Currently, the registry comprises over 15,000 twins, mostly female, aged 18–88 making it one of the largest twin registries in the world. Participants are sent regular questionnaires that include questions regarding lifetime diagnosis of an autoimmune disease. All twins have signed informed consent forms for research approved by St. Thomas’ Hospital Ethics Committee and Liverpool East Research Ethics Committee (REC reference 19/NW/0187), IRAS ID 258513. The TwinsUK spine study was started in 1996.

T2-weighted images (T2WI) sagittal spine MRIs were obtained using a Siemens scanner (Munich, Germany) with 1.0-T superconducting magnet, and supporting data were collected from twins who were included in the study. Variables such as age, sex and BMI were obtained, as well as lifestyle and clinical information.

Only participants with availability of MRI and autoimmune questionnaires were included in the study. Data were collected from self-reported questionnaires and health assessment visits of twins previously invited to participate in various studies examining a wide range of traits and common medical conditions. Questionnaires included doctor’s diagnosis of inflammatory bowel disease, rheumatoid arthritis, coeliac disease, vitiligo, autoimmune thyroid disease (hypothyroidism and hyperthyroidism), lupus, multiple sclerosis and type I diabetes were included as autoimmune conditions of interest (Fig. 1). Circulating anti-thyroid peroxidase antibodies (TPOAb) were measured merging the results coming from two different assays: Roche assay (TPOAb titre of > 34 kU/l considered positive) and Abbott assay (TPOAb titre of > 6 kU/l considered positive). Patients with missing data were assumed to be negatives for that specific condition.

Fig. 1
figure 1

Autoimmune conditions prevalence. The pie chart shows the prevalence in the sample (n = 764) of the autoimmune conditions included in the study

Full size image

An individual was considered autoimmune positive if they had ever received a diagnosis of at least one of the conditions above or had circulating antibodies over the positivity threshold. Coding of the largest autoimmune subgroup, thyroid related was cross-referenced and integrated with prescription data that listed medications to treat autoimmune disorders affecting the thyroid.

Coding MRI scans

Lumbar MC was identified on MRI by the presence of altered imaging signal at vertebral body levels as previously [9]. MC was coded as a positive binary variable if MC was present at any one of the lumbar segments [15]. A quantitative variable was derived from the size of both superior and inferior bone marrow lesions evaluated on a scale ranging from 0 to 5 as previously described [16], and values were summed to create a MC lesion size score for each lumbar level [17].

IDD had been coded as the sum of disc bulge, disc imaging signal intensity, disc height and osteophytes formation at each spine level previously [18]. Briefly, each measure was assigned a score (0–3), grading the severity of the phenotype, and a total degeneration score, summing scores for each of the 5 lumbar discs, was assigned for each participant (total IDD range = 0–60). Back pain was evaluated through several methods. First, the Medical Research Council Back and Neck Pain Questionnaire was administered at the MRI scan visit. In addition, self-reported episodes of disabling LBP lasting more than 1 month at any period during lifetime were collected on subsequent questionnaires [19].

Statistical analysis

Statistical parametric and non-parametric tests were used for comparing the two groups. T-test was used for comparing age and BMI, Chi-squared for MC prevalence and Mann–Whitney U for MC width. Multivariable generalized mixed linear models (GLMM) were fitted to the data to calculate odds ratios for risk of developing MC and LBP in participants with an autoimmune disorder. Moreover, it was investigated associations between diagnosis of an autoimmune disorder and the size of MC lesion. Multivariable analysis was adjusted for covariates including age, sex, smoking, BMI and family structure, i.e. twin relatedness as well as for number of autoimmune conditions and IDD. An individual was considered as smoker if reported to smoke more than 10 cigarette packets per year. Data processing and analysis were planned and executed in Python version 3.10.5 using “statsmodels” and “scipy” packages and RStudio version 2022.07.1 using lme4 library. A post hoc power analysis was carried out using G*Power software [20].

Results

Data were obtained for 764 twins having MRI coded for MC and completed autoimmune self-report questionnaires (Table 1). Twins were predominantly female (n = 737 (96%)) with mean age of 54 years (range 34–73 years) and mean BMI = 25 kg/m2 (range 16.23–51.40 kg/m2). Participants reported smoking more than 10 packets a year or more (n = 213 (28%)), and 164 (22%) participants had had an episode of disabling LBP over their lifetime lasting more than 1 month (Table 1). BMI was significantly different between participants diagnosed with an autoimmune disease and those with no autoimmune diagnosis (t-test p value = 0.001). This is because included autoimmune disorders were affecting thyroid and is well grounded that these conditions influence BMI [21].

Table 1 Baseline and outcome characteristics of the study participants
Full size table

Modic change

MC was defined as any individual having at least one MC affected endplate. A prevalence of 31% and 33% was detected in the autoimmune-positive and -negative participants, respectively, there was no significant difference between the two proportions (χ2 p value = 1). Both incidence and size of MC were higher in L4-5 and L5-S1 than other lumbar levels (Fig. 2). Table 2 summarizes the results of generalized mixed logistic regression model corrected for age, BMI, smoking and family structure. Neither having a diagnosis of autoimmune disease nor the number of morbidities was associated with the presence of MC (p value > 0.05). MC was found to be associated independently to IDD (OR 1.22 CI 1.16–1.30, p value = \(6.7\times {10}^{-13}\)). MC size and autoimmune diagnosis (r2 = 0.4) were not associated, results are summarized in Table 3. No association between autoimmune diagnosis and MC size was detected. Interestingly, BMI was associated with MC size even after correcting for IDD, providing evidence of an independent association between BMI and MC size.

Fig. 2
figure 2

Prevalence and size of MC in the sample. Prevalence of MC (Panel A) is measured as total prevalence of MC at specific level among MC-positive participants. Size of MC (Panel B) is measured as mean MC width at specific level among all MC-positive participants

Full size image
Table 2 Risk factors for MC
Full size table
Table 3 Risk factors for MC size
Full size table

Finally, we examined the association between autoimmune disorders and episodes of LBP in the twin participants while correcting for BMI, age and smoking (r2 = 0.4, Table 4). No association was found between autoimmune phenotype and LBP (p value > 0.05).

Table 4 Risk factors for LBP
Full size table

Discussion

This is the first study using a large population sample having MRI scans coded for MC and recorded autoimmune diagnosis. Autoimmune diagnosis does not appear to influence the development of MC suggesting that the development of MC is not autoimmune in aetiology. We explored MC both as a binary variable and as continuous measure of size. None of the models showed that autoimmune diagnosis was correlated to MC size, width or measures of back pain but our study does reveal that MC is related to raised BMI.

Related to our hypothesis that more severe autoimmune diathesis following an autoimmune diagnosis could result in larger MC lesions, we investigated the correlation between autoimmune diagnosis and MC width but found no association. BMI and MC were associated when correcting for disc degeneration which suggests an independent association, as we [22] and others have reported [23, 24]. Clarity around how BMI influences MC prevalence or size is needed [25]. Whether high BMI places a physiological burden upon endplates, subjecting them to increased microtrauma or whether systemic inflammation by lipid-induced endo- or paracrine responses promotes MC is not clear. Our findings that BMI is independently associated with size but not prevalence of MC could suggest that the association is driven by adipose-derived inflammation. In support to this, Teichtahl et al. showed that increased spinal adiposity is correlated with MC [26]. Conversely, high BMI increasing spine workload and influencing MC prevalence was not supported by our findings.

Autoimmunity has been suggested to play a role in pain perception, a dysregulated interplay between nervous and immune systems, especially the interaction of nociceptor and immune cells [27]. Pathological changes promoted by altered metabolite transport due to ED and the subsequent recruitment of inflammatory cytokines such as interleukin-1 (IL-1), tumour necrosis factor-alpha (TNF-α) and IL-6 have been demonstrated to stimulate pain receptors in tissues and play a causal role in LBP [28]. For this reason, we examined whether diagnosis of an autoimmune disorder increased the reporting of back pain. This question fell within our general investigation that autoimmune diagnosis and MC may correlate; thus, MC symptomology, for example LBP, may also correlate with autoimmune diagnosis. We found no evidence to support either of these relationships. We found previously in TwinsUK that smoking and high BMI were associated to LBP [29]. High BMI and smoking likely have several influences on back pain from reducing blood flow and disc nutrition, to socio-economic or physical workload factors [30, 31]. The endocrine responses of adipose tissue chronically elevate inflammatory markers [32] with may increase pain susceptibility. Even if autoimmunity has a preponderant component in chronic pain perception, LBP is directly influenced by IDD, smoking and BMI.

The aetiology of autoimmune disease depends, in part, from genetic susceptibilities, along with several other risk factors; lifestyle, environment and epigenetics have been proposed and demonstrated to contribute [33]. Autoimmunity is a diverse phenomenon classified into “systemic” and “organ specific” which share risk alleles at HLA locus on chromosome 6 [34]. In contrast, alleles associated with cartilage production and vertebral health have been associated with MC. Both candidate and genome-wide gene association studies fail to provide evidence of a shared genetic risk for autoimmune diagnosis and MC [35,36,37]. Autoimmune disorders are highly heterogeneous, and future investigations would likely benefit from disease-specific foci when examining links with other conditions [38]. MC does not appear to be an autoimmune disorder, the autoimmune-like response of cells may not relate to autoimmune diagnosis.

Finally, alternative hypotheses cannot be discounted; mounting evidence suggest a role for occult infection in the IVD [12, 39,40,41]. The resulting low-level or occult infection could promote inflammatory processes, leading to MC [42]. This theory is supported by studies investigating the presence of bacteria in degenerate and MC adjacent discs[43], Propionibacterium acnes (among others) has been cultured and whole bacterial genome-sequencing studies have reported an array of findings in disc tissue [44, 45] as well its injection in animal models disc caused the development of MC [46, 47]. Moreover, it was shown that patients with LBP and type I MC treated with antibiotics improved pain symptoms and MRI Modic grading [48]. There has, however, been relatively little scale, scientifically robust research investigating the presence of bacterial infection in the disc especially in relation to chronic LBP and MC [49], and bacterial contaminants are often posited to explain findings of disc microbes [50]. Whether dysregulated bacteria and occult infection or pain-generating cytokines are the impetus for an inflammatory stimulus, leading to MC is not clear.

A study may represent a false negative, or lack of detectable association may arise where studies are of insufficient size. According to the power estimation of our study, we had 80% power to detect an association between the occurrence of MC and autoimmune disease of effect size of 0.1. Despite the high power, we did not detect any increased MC prevalence. This suggests that if MC does occur more frequently in participants with autoimmune diagnoses, it must be at a rate less than 10%. Our best conclusion is that there is no relationship between autoimmune diagnosis and MC development. Exposure of normally immune-privileged NP cells to the external circulus may be necessary for this response; however, we did not find evidence, this reaction is promoted by carrying an autoimmune disorder but instead could depend on a specific adaptive immunity alteration. We cannot, therefore, exclude that an autoimmune response is ongoing during MC onset, but this is not related to a susceptibility to autoimmunity proper of autoimmune disorders.

We acknowledge several limitations to this study. We considered autoimmunity as a single phenotype obtained by merging very different diseases. For this reason, further investigations should focus on studying relationships between autoimmunity and MC in specific autoimmune disorders by including biological samples and genetic data in order to obtain a deeper understanding of the complex mechanisms underlying. Self-report was used to classify diagnosis of autoimmune conditions and reliance on participant self-report may bias results, although the hypothesis being tested was obscure to participants. The cohort is predominantly female, which may have favoured our study as autoimmune diseases are twice as prevalent in females [51], allowing us to include conditions rare in the general population. It does, however, prevent us drawing conclusions about autoimmune diagnosis and MC risk in males. Lastly, our MC categorization was based on T2WI images only. To distinguish MC types (I, II or III), both T1WI and T2WI are required—but not available in this cohort for funding reasons.

Conclusions

This study is the first to examine autoimmune disorders and MC prevalence in a large, population-based cohort. The study was well powered to detect a small effect. No association was found between having a diagnosis of one or more autoimmune conditions and the prevalence, width or severity of MC. Interestingly, MC extension was associated to increased BMI after correcting for IDD, sign of an independent association. These results are applicable to females since the results presented were replicated excluding male participants.