Exploring the Links Between Ankylosing Spondylitis and Amyotrophic Lateral Sclerosis: A Bidirectional Mendelian Randomization Study

Abstract

Background: Very few case reports have explored a potential link between ankylosing spondylitis (AS), an autoimmune disorder, and amyotrophic lateral sclerosis (ALS), a progressive neurodegenerative condition. We aimed to investigate whether genetic liability to AS causally influences the risk of ALS, and vice versa, using a bidirectional two-sample Mendelian randomization (MR) framework. Methods: We performed a two-sample MR study to evaluate the bidirectional causal relationship between genetic liability to ankylosing spondylitis and ALS risk. We used 6 valid single nucleotide polymorphisms (SNPs) from genome-wide association study (GWAS) data (AS: 1462 cases and 164,682 controls; ALS: 27,205 cases and 110,881 controls). We used the inverse-variance weighted (IVW) approach as the primary statistical method for causal estimation, with sensitivity analyses (including MR-Egger, weighted median, weighted mode, Mendelian Randomization Pleiotropy Residual Sum and Outlier (MR-PRESSO), leave-one-out, and single SNP analysis) to assess pleiotropy and heterogeneity. Results: There was no evidence of a causal association between genetic predispositions to ankylosing spondylitis (AS) and amyotrophic lateral sclerosis (ALS) (IVW OR = 1.01; 95% CI: 0.99–1.02; p = 0.10). The results from the weighted median, weighted mode, MR-Egger, and simple mode methods were consistent and nonsignificant. In the reverse analysis, genetic liability to ALS showed no causal effect on AS (IVW OR = 0.88; 95% CI: 0.70–1.12; p = 0.33), with similar null findings across all sensitivity methods. Conclusions: Overall, our bidirectional two-sample MR analyses provided no evidence supporting a causal relationship between AS and ALS.

1. Introduction

Amyotrophic lateral sclerosis (ALS), a devastating and fatal neurodegenerative disorder, is caused by the chronic erosion of motor neurons in the brain and spinal cord. It is often characterized by the gradual paralysis of skeletal and respiratory muscles [1,2]. With an estimated lifetime risk of approximately one in 400 individuals, ALS has a median survival of less than five years post diagnosis, and fewer than 10% of patients live beyond a decade [3,4]. The global prevalence of ALS is expected to increase from roughly 3.1 to 5.2 instances per 100,000 individuals between 2015 and 2040, with an average annual increase of 2.12% within that time period [2,5,6,7,8].

ALS occurs across all populations but is more common in individuals of European and non-Hispanic ancestry [7,9]. Both familial and sporadic cases are recognized, and may thus suggest the interplay of genetic susceptibility and environmental exposure. About 10% of patients report a family history, and more than half of these cases carry a pathogenic mutation [10]. Sporadic ALS, however, is a more complex form with heritability estimates of 40–50% [11,12]. Superoxide dismutase 1 (SOD1) and chromosome 9 open reading frame 72 (C9orf72) mutations are perhaps the most common genetic contributors to both familial and sporadic forms [4,10]. In addition, environmental risk factors such as heavy metal exposure, industrial dust, air pollution, smoking, and high levels of physical activity have also been implicated [2,6]. An age of between 50 and 70 years and male sex further increase disease risk [2]. Although the etiology of ALS is heterogeneous, a few studies have suggested that immune dysregulation and autoimmune disease may contribute to its pathogenesis [13,14]. Current therapies provide only modest benefits, and no treatment has yet been shown to alter the course of the disease [15].

Ankylosing spondylitis (AS), a severe autoimmune condition, mainly affects the spine and sacroiliac joints [16]. Very few reports have explored a potential link between ankylosing spondylitis (AS) and amyotrophic lateral sclerosis (ALS). These isolated case studies describe patients who were initially diagnosed with AS and subsequently developed ALS, and thereby suggesting a possible connection between the two conditions [17,18]. These observations raise important questions about potential shared mechanisms (e.g., immune-mediated or inflammatory pathways) that may predispose individuals with AS to ALS. A large nested case–control study conducted in Sweden investigated the relationship between autoimmune diseases and amyotrophic lateral sclerosis (ALS) [19]. The study included 3561 patients diagnosed with ALS between 1990 and 2013 and 35,610 matched controls. Individuals with a prior history of autoimmune conditions, particularly ankylosing spondylitis (AS), were found to have an increased risk of developing ALS [19]. However, the authors noted that diagnostic overlap and the long latency period before ALS diagnosis could have introduced misclassification bias, as early ALS symptoms may have been mistaken for AS.

More broadly, observational studies (e.g., case-control, case reports, cohort studies) face inherent limitations, including measurement error, unmeasured confounding, and temporality (i.e., reverse causation). For example, while co-occurrence between autoimmune disorders such as AS and neurodegenerative diseases like ALS has been observed [17], the directionality and causality of this relationship remain uncertain. To address these limitations, we applied a Mendelian randomization (MR) approach. MR is a causal inference technique which uses genetic variants, such as single-nucleotide polymorphisms (SNPs), as genetic instruments to assess the causal relationship between an exposure and an outcome [20]. The underlying mechanism of MR is grounded in the second law of Mendel (Law of independent assortment of genes), which ensures the random allocation of alleles during gamete formation [20]. This natural randomization mimics the principles of a randomized controlled trial, and thereby minimizes bias and measurement error inherent in traditional observational designs [21,22].

Here, we employed a bidirectional two-sample MR framework to investigate whether a genetic liability to ankylosing spondylitis causally influences the risk of developing amyotrophic lateral sclerosis, and vice versa. The bidirectional two-sample MR approach strengthens causal inference by reducing the likelihood of reverse causation and clarifying potential bidirectional biological mechanisms between the exposure and outcome [20].

2. Methods

2.1. Study Design

To assess the causal link between AS and ALS, we implemented a two-sample Mendelian randomization (MR) technique. Unlike a single-sample MR, this two-sample MR design uses two independent summary-level genome-wide association studies (GWASs) of the same underlying population for the exposure and the outcome [23]. Causal inference in MR relies on three core assumptions, as summarized in Figure 1. By using SNPs as genetic instruments, MR strengthens causal inference, reduces bias, and increases confidence in the robustness of the findings. Modern MR approaches apply multiple complementary estimators to improve causal inference and account for pleiotropy. Consistent with current standards, we used IVW as the primary method, supported by MR-Egger, simple mode, and weighted median/mode. We also applied MR-PRESSO for pleiotropy correction and to ensure reliable causal estimates.

2.2. Data Source

Summary-level genetic association data for the AS (exposure) and ALS (outcome) datasets were obtained from publicly available genome-wide association study (GWAS) datasets. Genetic variants (SNPs) associated with ankylosing spondylitis (OpenGWAS ID: finn-b-M13_ANKYLOSPON) were retrieved from the MRC IEU OpenGWAS project (https://gwas.mrcieu.ac.uk/) (Accessed 10 September 2025); this dataset comprised 1462 cases and 164,682 controls, with 16,380,022 SNPs. Summary statistics for amyotrophic lateral sclerosis (OpenGWAS ID: ebi-a-GCST90027164) were also retrieved from the same source. The ALS dataset represents a large-scale meta-analysis including 27,205 cases and 110,881 controls, identifying 15 loci associated with ALS [24].

Table 1. GWAS Summary statistics used in the Two-Sample Mendelian Randomization of AS and ALS.

Phenotype/Trait	Data Source	OpenGWAS ID	Sample/Cohort	Ancestry	Sample Sizes
Ankylosing Spondylitis	FinnGen Consortium	finn-b-M13_ANKYLOSPON	FinnGen R9	(Finnish) (100%)	1462 (cases) 164,682 (controls)
Amyotrophic lateral sclerosis	van Rheenen et al., 2021 [24]	ebi-a-GCST90027164	SLAP, MND Bank, SLALOM, BePS, SLAGEN, PARALS, QSkin	Pan-European (100%)	27,205 (cases) 110,881 (controls)

provides a summary of the datasets used in this study.

This study made exclusive use of publicly accessible datasets, all of which were obtained from open genomic repositories; thus, no ethical clearance or participant consent was needed.

2.3. Genetic Variant Selection

We selected independent single-nucleotide polymorphisms (SNPs) associated with ankylosing spondylitis at the genome-wide significance threshold of p < 5 × 10⁻⁸. Linkage disequilibrium (LD) clumping was performed using an r² < 0.01 within a 10,000 kb window based on the 1000 Genomes Project European reference panel, implemented through PLINK v.1.90b7.7 software. From the exposure GWAS, we retrieved the effect estimates (β) and corresponding standard errors for the SNPs significantly associated with AS, and then extracted the same SNPs from the outcome (ALS) GWAS. The exposure and outcome summary statistics were harmonized to align effect alleles and ensure that all genetic associations corresponded to the same allele orientation. Palindromic SNPs with ambiguous strand alignment were also removed. Details of the selected SNPs and harmonization process are provided in the Supplementary Material. Power calculations for this bidirectional MR study were performed on the mRnd website (http://cnsgenomics.com/shiny/mRnd) (Accessed 20 October 2025) [25].

2.4. Primary MR Analyses

We examined the causal effect of genetic predisposition to ankylosing spondylitis (AS) on amyotrophic lateral sclerosis (ALS) using the inverse variance weighted (IVW) method as the main analytical approach. The IVW method estimates the causal effect by performing a weighted regression of SNP-outcome associations on SNP-exposure associations, with the assumption of 100% SNPs validity, and the intercept constrained to zero. All analyses were conducted in R (v4.4.1) using the MRC IEU TwoSampleMR package.

2.5. Bidirectional MR Analyses

To assess reverse causality, we performed bidirectional two-sample MR to examine the causal effect of genetic liability to ALS on AS. The analytical procedures mirrored those described above and followed established MR guidelines.

2.6. Sensitivity Analyses

The genetic instrument strength was evaluated by computing the F-statistics for each variant. In this case, values >10 indicate sufficiently strong instruments, minimal possibility of weak instrument bias, and potential violations of the first MR assumption [22,23].

To evaluate the possible presence of directional horizontal pleiotropy on the derived causal estimates, we also employed the MR-Egger regression, the weighted median, and the weighted mode methods. MR-Egger uses an intercept term to detect and adjust for pleiotropy bias, even when some genetic instruments are invalid, while the Inverse Variance Weight method assumes no horizontal pleiotropy [21,26]. Weighted median methods yield reliable estimates even if less than 50% of the instruments are valid, whereas the weighted mode method can remain robust even when more than half of the instruments are invalid [21,25]. Additionally, we applied the Mendelian Randomization Pleiotropy Residual Sum and Outlier (MR-PRESSO) method, which identifies and corrects for pleiotropic outliers through 3 main steps: a global test, outlier detection, and then a distortion test [27].

Finally, we assessed heterogeneity among the SNP-specific estimates using Cochran’s Q test, where p > 0.05 indicates no significant heterogeneity [22].

3. Results

3.1. Genetic Instrument Strength

To investigate the strength of our genetic instruments, we used F-statistics in our MR analysis. We observed a high F-statistics (commonly > 10) in this study. We observed high F-statistic values across all instrumental variables (range: 32.7–1408), well above the accepted threshold of 10. This indicates that the selected SNPs are strong instruments, minimizing the likelihood of weak instrument bias. These findings support the first core assumption of Mendelian randomization that the genetic variants are robustly associated with the exposure of interest, ankylosing spondylitis (AS). Power analysis indicated that the study had limited power (<1%) to detect modest causal effects between ALS and AS in both directions (Supplementary Material).

3.1.1. Causal Estimates of Genetic Liability to Ankylosing Spondylitis (AS) to Amyotrophic Lateral Sclerosis (ALS)

We found no evidence that genetic susceptibility to ankylosing spondylitis causally influences the risk of amyotrophic lateral sclerosis across all the statistical methods employed. Using six valid SNPs, the inverse variance weighted (IVW) analysis yielded an odds ratio (OR) of 1.01 (95% CI: 0.99–1.02; p = 0.10). The weighted median and weighted mode estimates were directionally consistent but not statistically significant (weighted median OR = 1.01; 95% CI: 0.99–1.02; p = 0.15; weighted mode OR = 1.00; 95% CI: 0.98–1.01; p = 0.95). Similarly, the MR-Egger (OR = 1.00; 95% CI: 0.98–1.02; p = 0.81) and simple mode (OR = 1.02; 95% CI: 0.99–1.05; p = 0.21) analyses showed no significant causal relationship. These findings are summarized in Figure 2a.

Each panel displays a table listing the causal estimates generated using multiple MR methods, including inverse-variance weighted (IVW), MR-Egger, weighted median, and weighted mode. The table reports the odds ratios (ORs), 95% confidence intervals (CIs), and corresponding p-values for each method. The forest plots visually represent the effect estimates from each MR approach. Points represent the estimated causal effect sizes (expressed as ORs), while the horizontal bars denote their 95% CIs. The vertical line indicates the null value (OR = 1). Estimates to the right of the line suggest a positive association, values to the left suggest a negative association, and CIs crossing the null line indicate insufficient evidence for a causal effect.

3.1.2. Causal Estimates of Effect of Genetic Susceptibility to Amyotrophic Lateral Sclerosis on Ankylosing Spondylitis Risk

In the bidirectional analysis, we evaluated the potential causal effect of genetic susceptibility to ALS on AS risk using 14 valid SNPs. The inverse variance weighted (IVW) method showed no evidence of a causal association (OR = 0.88; 95% CI: 0.70–1.12; p = 0.33). Similar null associations were observed across other MR methods, including the weighted median (OR = 0.88; 95% CI: 0.67–1.16; p = 0.38), weighted mode (OR = 0.74; 95% CI: 0.51–1.07; p = 0.14), MR-Egger (OR = 0.86; 95% CI: 0.56–1.31; p = 0.50), and simple mode (OR = 1.13; 95% CI: 0.70–1.84; p = 0.61) methods. These findings are summarized in Figure 2b.

3.2. Sensitivity Analyses

We found no evidence of heterogeneity or directional pleiotropy in either direction of analysis. The Cochran’s Q test and MR-Egger intercepts indicated no significant heterogeneity or horizontal pleiotropy, respectively. Similarly, the MR-PRESSO global test showed no indication of unbalanced pleiotropy, and no outlier variants were detected. These findings suggest that our causal estimates are unlikely to be biased by pleiotropic effects or heterogeneity (

Table 2. Summary of heterogeneity and pleiotropy tests in primary and bidirectional Mendelian randomization analyses.

Exposure	Outcome	Sensitivity Test
		Cochran’s Q
		MR-Egger			IVW
		Q	df	p value	Q	df	p value
AS	ALS	3.28	4	0.51	5.13	5	0.40
ALS	AS	18.9	12	0.09	19	13	0.12
		MR-Egger intercept
		Egger intercept			p-value
AS	ALS	0.012			0.25
ALS	AS	0.005			0.86
		MR-PRESSO (Global test)
		RSS obs			p-value
AS	ALS	16.35			0.39
ALS	AS	21.87			0.15

).

Single SNP analyses revealed no individual variant exerting a disproportionate influence on the overall causal estimate. The leave-one-out analysis showed that omitting any individual SNP had a negligible impact on the direction or magnitude of the causal association, indicating that no single variant was driving the results. The MR scatter plots further confirmed the absence of a consistent directional effect between ankylosing spondylitis and amyotrophic lateral sclerosis, aligning with the null findings observed in both directions of the bidirectional MR analyses (Figure 3, Figure 4 and Figure 5).

4. Discussion

In this study, we followed a bidirectional two-sample MR framework to explore the effect of genetic susceptibility to ankylosing spondylitis (AS) on amyotrophic lateral sclerosis (ALS) risk. Our analysis yielded null evidence supporting a causal relationship between genetic susceptibility to AS and ALS in either direction, across multiple statistical methods.

Studies exploring the association between ankylosing spondylitis (AS) and amyotrophic lateral sclerosis (ALS) remain scarce. To date, only a few isolated case reports have described individuals who developed ALS following a diagnosis of AS [17,18]. However, evidence from these observational studies does not necessarily imply a direct causal relationship. The observed co-occurrence may instead reflect residual confounding, reverse causation or shared environmental and lifestyle determinants. Furthermore, the retrospective design of these studies limits the ability to establish temporality, making it difficult to determine whether AS preceded ALS onset or vice versa. Our findings are in contrast with a large case–control study, where individuals with AS were found to have an increased risk of ALS when the diagnoses occurred within five years (OR 3.30, 95% CI 1.17–9.30). However, this association was not significant when the interval exceeded six years (OR 1.30, 95% CI 0.79–2.13) [19]. This short-term association may perhaps be a result of reverse causation where early symptoms of ALS mimic AS and subsequently lead to misdiagnosis as AS. Time-varying confounders such as comorbidities, physical activity and surveillance intensity may also have contributed to the observed association, as these factors can influence the likelihood of ALS diagnosis or progression independently of AS.

A few clinical reports have described patients treated with TNF-α inhibitors for ankylosing spondylitis who subsequently developed amyotrophic lateral sclerosis [28,29,30]. Tumor necrosis factor-alpha (TNF-α) plays an important role in inflammatory regulation; its inhibition has been associated with adverse neurological effects in rare cases [19,31]. Although the underlying mechanisms remain unclear, it is possible that these agents alter TNF-α neuroprotective functions within the central nervous system or trigger secondary neurotoxic side effects [32]. Therefore, future studies should explore whether TNF-α signaling mediates a biological link between AS and ALS, and identify the molecular pathways through which such effects may occur. In addition, metabolomic signatures such as glycolysis and arachidonic acid metabolism have been identified in association with AS [33]. These metabolic pathways have also been implicated in the pathogenesis of ALS, and suggest potential shared metabolic mechanisms between the two diseases [34]. Therefore, pathway enrichment and integrative multi-omics studies should be explored to elucidate the functional relevance of these metabolic processes and clarify whether they represent causal mechanisms or secondary disease responses.

This study has several notable strengths. Notably, the Mendelian randomization (MR) approach strengthens causal inference by reducing the impact of confounding and reverse causation inherent in observational research. Also, we used independent, non-overlapping GWAS samples of the same European ancestry for the exposure and outcome, and thus minimized potential bias from sample overlap. Third, the genetic instruments were robust, as indicated by F-statistics exceeding conventional thresholds (>10), and our findings were consistently supported across multiple sensitivity analyses, including MR-Egger, weighted median, weighted mode, leave-one-out, and MR-PRESSO, which together showed no evidence of heterogeneity or pleiotropy. Fourth, our bidirectional analysis allowed for evaluation of reverse causation, providing a more comprehensive understanding of the relationship between AS and ALS. Nevertheless, there are several limitations. First, our analyses were restricted to individuals of European ancestry, and therefore, our findings may not apply to individuals of non-European ancestries. Secondly, the limited number of independent SNPs and the small variance explained by the exposure instruments likely reduced the study’s power to detect modest causal effects. Third, we were unable to account for potential gene–environment interactions (e.g., smoking, physical activity, diet) that could modify the effect of AS on ALS. Thus, future MR studies could employ the multivariable MR technique to incorporate these factors and better assess their combined influence on the causal AS-ALS relationship. Although our study found limited evidence for a causal link between AS and ALS, a relationship cannot be completely ruled out. The association may be non-linear, which would not be captured by standard MR analyses that assume linear effects, thereby highlighting an exciting avenue for future research.

To the best of our knowledge, this is the first study to investigate the causal link between genetic predisposition to AS and ALS. Our findings contribute novel knowledge to a scarcely explored domain of neuroepidemiology and provides a background for further exploration on the genetic link between AS and ALS. Hence, future studies should aim to replicate these findings in larger well-powered AS and ALS GWASs to increase the number of robust genetic instruments that could improve the causal inference analysis. Also, the integration of larger omics datasets could help provide insight into the molecular and biological pathways linking AS and ALS. Additionally, applying other advanced statistical genetic methods to understand shared genetic loci, and revisiting the reverse-direction analysis with larger ALS datasets will be important to clarify any bidirectional relationship.

5. Conclusions

Our Mendelian randomization analyses revealed null evidence of a causal association between ankylosing spondylitis (AS) and amyotrophic lateral sclerosis (ALS) in either direction. While previous observational reports have noted cases of co-occurrence between both conditions, our findings suggest that this overlap is not explained by shared genetic liability. These findings underscore the need for further investigation that will explore alternative explanations, including environmental influences, neurotoxic treatment effects, or neuroimmune mechanisms that might contribute to their previously observed coexistence.