Background
Amyotrophic lateral sclerosis (ALS) is an adult-onset neurodegenerative disease characterized by rapid motor neuron degeneration and subsequent respiratory failure [
1]. ALS is relatively rare worldwide: the standardized disease incidence is about 1.89 and 0.83 per 100,000 person-years of follow-up in the European and East Asian populations, respectively [
2]. However, the burden of ALS is substantial. Clinically, ALS patients often suffer from loss of independence due to progressive functional impairments of upper and lower motor neurons. In addition, ALS patients only have a median survival time of 2–4 years starting from disease onset and less than 10% patients can survive beyond 10 years [
3]. Economically, the annual cost per ALS patient is considerable: it ranges from ~$11,000 in Denmark to ~$70,000 in the USA [
4]. Subsequently, the population-wide national economic burden of ALS is estimated to be ~$1.023 billion in the USA, far greater than that of the other two common neuromuscular diseases (Duchenne muscular dystrophy and myotonic dystrophy) [
5]. Importantly, ALS incidence is expected to increase due to population aging worldwide, further aggravating the socioeconomic burden associated with ALS in the coming years [
6]. Therefore, it becomes critically important to understand the etiology of ALS and identify risk factors that can causally influence ALS. Identifying causal risk factors for ALS can potentially lead to the discovery of new pathogenic pathways underlying ALS, guide the development of effective medical treatment and patient care, and facilitate healthcare policy making and healthcare resource allocation.
While the causes and pathogenesis of ALS remain largely unknown, several genetic and environmental risk factors have been identified to be associated with the development of ALS [
7‐
9]. Among these identified risk factors, the association between antecedent diseases and ALS is of particular interest. Specifically, previous observational epidemiological studies have found that ALS patients tend to have less antecedent diseases (e.g., liver/lung/thyroid disease, diabetes, hypertension, hyperlipidemia, and arthritis) as compared to the general age/gender/geography-matched control patients [
10‐
13]. In addition, the presence of antecedent diseases appears to lead to a substantial delay in the onset age of ALS patients [
10‐
13]. The evidence suggests that antecedent diseases may be markers of causal risk factors for ALS or may themselves be involved in the pathogenesis of ALS. Importantly, several pieces of evidence have recently emerged to support a potentially pathological role of one antecedent metabolic disease, type 2 diabetes (T2D), in ALS. The association between T2D and ALS was first discovered a decade ago [
12]. Follow-up observational studies and hospital disease registries have provided additional empirical evidence supporting the association between T2D and ALS (Table
1). This association evidence, when further paired with the observations that high energy consumption often accompanies ALS progression [
23,
24] and the observations that glucose intolerance and insulin resistance have been linked to ALS [
25,
26], leads to a hypothesis that T2D may mechanically and causally affect ALS.
Table 1
Estimated effect sizes of T2D on ALS in previous observational studies
DOvidio | 2018 | Italy | 0.30 (0.19–0.45) | |
Visser | 2017 | Netherlands | 0.77 (0.33–1.21) | |
Hollinger | 2016 | USA | 0.80 (0.53–1.21) | |
Mitchell | 2015 | USA | 0.47 (0.38–0.58) | |
Mariosa | 2015 | Sweden | 0.79 (0.68–0.91) | |
Seelen | 2014 | Netherlands | 0.72 (0.51–1.01) | |
Turner | 2013 | England | 0.98 (0.85–1.13) | |
Kioumourtzoglou | 2015 | Denmark | 0.61 (0.46–0.80) | |
Moglia | 2017 | Italy | 1.05 (0.78–1.42) | |
Korner | 2012 | Germany | 1.11 (0.76–1.60) | |
Armon | 1991 | USA | 1.00 (0.29–3.50) | |
Sun | 2015 | China | 1.35 (1.10–1.67) | |
Pool 1 | | | 0.73 (0.59–0.90) | |
Pool 2 | | | 0.77 (0.62–0.96) | |
Unfortunately, establishing a causal relationship between T2D and ALS has been challenging so far. Almost all previous studies on T2D and ALS are observational in nature, and observational studies have inherent drawbacks that make it hard for them to reach a causal conclusion using standard statistical tools. In particular, while these previous observational studies have attempted to adjust for the effects of many confounding factors (e.g., income, education, marital status, age, gender, and race) (Additional file
1: Table S1), it is not possible to control for all confounding factors there. Unadjusted confounding factors can potentially bias the association evidence between T2D and ALS. Moving beyond observational studies is not straightforward either. For example, because ALS is rare in the population, it becomes difficult to collect large samples to carry out longitudinal studies for examining the influence of T2D on ALS [
27]. In addition, due to ethical considerations, it is almost impossible to validate the causal association between T2D and ALS directly by performing randomized controlled trails. Therefore, it remains unclear whether the association between T2D and ALS observed in previous studies is causal or not: is T2D protective against ALS or is the absence of T2D just an early manifestation of ALS [
10,
11,
14,
19,
28,
29]?
Besides a lack of evidence on the causal relationship between T2D and ALS, likely also due to the observational nature of previous studies, there is a lack of consensus on whether T2D is protective for ALS in all human populations. For example, in the European population, despite minor conflicting results, most observational studies found that T2D is associated with decreased susceptibility to ALS (Additional file
1: Figure S1), suggesting a possible neuroprotection role of T2D on ALS. In contrast, however, in the East Asian population, it was found that T2D can increase the risk of ALS [
22].
Mendelian randomization (MR) is an advanced statistical method that can help establish a causal relationship between an exposure of interest (e.g., T2D in the present study) and an outcome of interest in observational studies by employing single-nucleotide polymorphisms (SNPs) as instrumental variables for the exposure [
30‐
36]. MR relies on the idea that SNPs associated with T2D would also be associated with the risk of ALS through the path of T2D, if T2D is causally associated with ALS. Therefore, even though SNPs that are selected as instruments are not causal for T2D but are only associated with T2D, MR can still help establish the causal association between T2D and ALS [
37]. Large-scale genome-wide association studies (GWASs) on T2D performed in the recent years have identified many SNPs associated with T2D, making it feasible to choose appropriate SNPs to serve as valid instruments for T2D [
38].
To ensure the validity of the causal conclusion from MR, each selected instrumental variable needs to satisfy three MR modeling assumptions (Additional file
1: Figure S2A and Additional file
2) [
39‐
41]: (i) it should be strongly associated with T2D; this is referred to as the relevance assumption; (ii) it should not be associated with any other confounders that may be associated with both T2D and ALS; this is referred to as the independence assumption; (iii) it influences ALS only through the path of T2D and does not have horizontal pleiotropic effects; this is referred to as the exclusion restriction assumption. Note that the first assumption (i.e., the relevance assumption) can be directly tested based on the observed data while the last two assumptions (i.e., the independence and exclusion restriction assumptions) are difficult to validate in practice. We will later examine the validity of the last two assumptions through various sensitivity analyses.
In the present study, our main objective is to investigate the causal relationship between T2D and ALS in both the European and East Asian populations. To achieve this objective, we conducted the largest two-sample MR analysis to date based on summary statistics publicly available from large-scale GWASs with ~63,000 cases for T2D and ~42,000 cases for ALS in the European population, and with ~191,000 individuals for T2D and ~4100 individuals for ALS in the East Asian population.
Publisher’s Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.