Background

Amyotrophic lateral sclerosis (ALS) is a neurodegenerative disease, featuring progressive loss of motor neurons in the brain and spinal cord. Globally, around 4.5 per 100,000 people are living with this disease and 1.9 per 100,000 people are newly diagnosed with ALS each year [1]. As no effective treatment is currently available, patients with ALS usually die within 3–5 years after symptom onset [2]. The cause of ALS remains unknown. Although various risk factors have been proposed, older age, male gender, family history, and specific genetic mutations are the only established risk factors [3]. A vivid area of research in ALS concerns medication use, with the aim to improve understanding of disease etiology as well as identify novel therapeutic targets (e.g., through drug repurposing). To this end, we performed a systematic review of the existing literature aiming to provide a comprehensive view of the current knowledge on the roles of different medication use on the subsequent risk of ALS.

Methods

This systematic review was conducted according to the Preferred Reporting of Systematic Reviews and Meta Analyses (PRISMA) guidelines [4]. A unique searching strategy (online material) was applied in the databases Medline (1947-), Embase (1974-), and Web of Science (1900-) from the dates of database establishment to December 10, 2021. References of identified articles were further searched for additional relevant articles.

Inclusion and exclusion criteria

All original articles in English which used medication use as exposure and development of ALS/ motor neuron disease (MND) as outcome among human subjects were included. Clinical trials, systematic reviews and meta-analyses, comments, editorials, and case reports were excluded.

Ascertainment of exposure and outcome

Exposure was defined as any medication use according to questionnaire data, medical records, or drug prescription registers. Alternative medicine which lacks biological plausibility and is untested or dietary supplements were not considered as exposure. The diagnosis of ALS was derived from patient register, medical records, or other sources.

Study selection and quality assessment

The screening process was divided into two stages (stage 1: title and abstract; stage 2: full text). Two researchers independently evaluated the articles and reached consensus at each stage. The screening tools Rayyan and Endnote were used. A pre-validated tool, the Newcastle–Ottawa Assessment Scale (NOS) [5], was used at the final stage for quality assessment of included articles. The NOS scales the quality of each article based on the selection of study groups, the comparability of the groups, and the ascertainment of the exposure and outcome of interest (from 0 to 9; 9 being the highest quality).

Results

Out of a total 3360 unique studies that were screened by title and abstract, 33 articles were selected for a full-text screening. After consensus by the two researchers who performed the screening independently, 25 articles were included in the review (Fig. 1). Many articles were excluded because of wrong article type (review, clinical trial, post-marketing surveillance, etc.), exposure (dietary supplements, chemicals, etc.), outcome (prognosis, comorbidities, etc.), or study subject (mice, cell cultures, etc.). Finally, a total of seven categories of medications, including cholesterol lowering drugs, anti-inflammatory drugs, immunosuppressants, antibiotics, oral contraceptives (OCs) or hormone replacement therapy (HRT), antihypertensive drugs, antidiabetics, and drugs for psychiatric and neurological disorders, were assessed for an association with risk of ALS.

Fig. 1
figure 1

Flowchart of the study selection. Steps and number of articles included and excluded per step during the study selection process

Full size image

Table 1 summarizes the characteristics of the 25 articles, including publication information, study type, demographics of the participants, data source, exposure, and outcome ascertainment, adjusted covariates, main findings, consideration of reverse causation, and quality score. Among these studies, there were 22 distinct study populations including 171,407 persons with ALS. Nineteen of these studies considered to some extent reverse causation and 23 studies had a quality score ≥ 5.

Table 1 Literature summaries of association between medication use and risk of amyotrophic lateral sclerosis (25 articles including 7 drug categories)
Full size table

Cholesterol lowering drugs (mainly statins)

Eight studies have explored the associations of cholesterol lowering drugs with the subsequent risk of ALS [7,8,9,10,11,12,13, 22], among which seven studies focused on statins [7, 9,10,11,12,13, 22] while one focused on a collection of cholesterol lowering drugs [8]. Four studies demonstrated a protective role of cholesterol lowering drugs in ALS development [8,9,10,11], while another four reported a lack of or near-null association between statins and risk of ALS [7, 12, 13, 22].

A case–control study from the Netherlands included 334 patients with ALS and 538 age- and sex- matched controls and found an inverse association between cholesterol lowering agent use and ALS risk (OR = 0.6; 95% CI 0.4–0.9) [8]. A US population-based case–control study included 10,450 cases and 104,500 age-, sex-, and calendar year-matched controls and found a 13% lower ALS risk in relation to statin use, but not for non-statin anti-cholesterol medication use, after adjusting for age, sex, calendar year, race, socioeconomic status, Medicare use, indications for statin prescription, obesity, chronic obstructive pulmonary disease (COPD), and average number of physician visits [10]. In this study, lipophilic statins showed greater inverse association with ALS, compared with other statins [10]. Another Dutch study also found a 55% lower risk of ALS in relation to statin use in a case–control study of 722 patients with sporadic ALS and 2268 age- and gender-matched controls, adjusted for gender, age, education, current smoking, and current alcohol consumption [9]. Excluding smoking and alcohol consumption or adding BMI in the multivariate analyses or excluding patients with C9orf72 repeat expansion did not change the result [9]. Finally, a US retrospective cohort study, including 144,214 individuals with statin use and 144,301 individuals without statin use, who were individually matched by a propensity score, found statin use to be associated with a lower risk of ALS (RR = 0.46; 95% CI 0.30–0.69) [11].

In contrast, four studies showed a null or near-null association between statin use and risk of ALS, including a Danish population-based case–control study with 556 cases and 5560 controls [7], a Swedish nested case–control study with 2475 cases and 12,375 controls [12], and a German case–control study with 200 cases and 197 controls [22], demonstrating a null association; and a Danish population-based cohort study including 974,304 individuals with statin use and 1,948,606 individuals without statin use [13], showing a weak association between statin use and ALS risk.

Anti-inflammatory drugs, immunosuppressants, or antibiotics

Two studies examined steroids [22, 24] and five have evaluated the relationship of non-steroidal anti-inflammatory drugs (NSAIDs) [14,15,16, 22, 24] in relation to risk of ALS, among which four studies examined aspirin [14,15,16, 24]. In addition, three studies assessed the association between immunosuppressant use as a broad category and ALS risk [9, 17, 22] whereas two studies explored antibiotic use in relation to ALS risk [6, 18], focusing on a broad category of drugs or sub-class or individual drugs.

The findings on the association of NSAIDs with ALS risk are not conclusive. A German case–control study found a twofold risk increase of ALS in individuals with use of NSAIDs (OR = 2.3, 95% CI 1.2–4.4) [22]. However, two US studies, including one case–control study (111 cases and 258 controls) and one prospective cohort study (708 cases of ALS during a follow-up of 9,727,583 person-years) found that neither non-aspirin NSAID nor aspirin use was associated with the risk of ALS [14, 15]. A population-based case–control study in Taiwan including 729 ALS cases and 14,580 controls also failed to observe an association between NSAID use and risk of ALS [24]. However, together with another population-based case–control study in Taiwan (including the same 729 cases as the previously mentioned study in Taiwan and 7290 controls) [16], both studies showed a lower ALS risk with aspirin use, including some evidence for a dose–response relationship [16]. When it comes to steroids, no association with ALS was found in either the German or Taiwanese study [22, 24].

Regarding immunosuppressants, a Dutch population-based case–control study found a 74% reduced ALS risk in relation to the use of immunosuppressive drugs (OR = 0.26, 95% CI 0.08–0.86) [9]. A US case–control study [17] and a German case–control study [22] failed however to observe an association between immunosuppressant use and ALS. The US study did report a lower ALS risk among individuals with a history of chemotherapy (OR = 0.46, 95% CI 0.22–0.89) [17].

A nested case–control study in Sweden, including 2484 newly diagnosed patients with ALS and 12,420 sex-, birth year-, and area of residence-matched population controls, found that antibiotic use was associated with a higher risk of ALS (OR = 1.11, 95% CI 1.01–1.22) with a dose–response relationship (P for trend = 0.0069), while no association was noted for macrolide-type antibiotics [18]. A US study found azithromycin, a macrolide-type antibiotic, to be associated with a lower risk of ALS, on the other hand [6].

OCs or HRT

Five studies have assessed the roles of OCs or hormone use as a broad category on the risk of ALS [19,20,21,22,23]. No association was found between OC or HRT use and ALS risk in either a US case–control study with 62 cases and 131 controls [19] or a British prospective cohort study including 1,319,360 women among whom 752 developed MND [20]. However, a case–control study conducted in Ireland, Italy, and the Netherlands found a lower risk of ALS in relation to OC use (OR = 0.65, 95% CI 0.51–0.84) with a dose–response relationship in all three countries, whereas a lower risk in relation to HRT use was only shown in the Netherlands (OR = 0.57, 95% CI 0.37–0.85) [21]. In addition, OC use was found to be associated with a lower risk of ALS (OR = 0.4, 95% CI 0.2–0.7) in a German case–control study [22], and HRT use was found to be associated with a lower ALS risk in a US cohort study (RR = 0.42, 95% CI 0.28–0.63) [23].

Antihypertensive drugs

Four studies have evaluated the association of antihypertensive use with risk of ALS [6, 8, 16, 25]. A case–control study in the Netherlands did not find an association between any antihypertensive use and ALS risk [8]. A US population-based case–control study, on the other hand, screened 685 individual prescription drugs and found that, among antihypertensive drugs, the use of amlodipine (calcium channel blocker), furosemide (loop diuretic), lisinopril (angiotensin-converting enzyme inhibitor/ACEI), or metoprolol (beta blocker) was associated with a lower risk of ALS [6]. Two other studies examined the roles of ACEIs or angiotensin II receptor blockers (ARBs) on ALS risk. The first is an Italian nested case–control study, with 1200 cases and 120,000 controls, showing no clear association between ACEIs or ARBs and the risk of MND [25]. In contrast, the other study, a case–control study in Taiwan with 729 cases and 14,580 controls found use of ACEIs to be associated with a lower risk of ALS, including a dose–response relationship when studying ACEIs by cumulative defined daily dose (cDDD) [16]. However, no association was found for other antihypertensives [16]. Among the individual compounds of ACEIs, captopril and enalapril were shown to be associated with a lower ALS risk [16].

Antidiabetics

Four studies have estimated the association of antidiabetic use with the risk of ALS [6, 8, 12, 22]. Among these, two case–control studies (one from the Netherlands and one from Germany) did not observe an association between antidiabetic use and risk of ALS [8, 22]. Contrary to these findings, a nested case–control study in Sweden reported a lower risk of ALS among persons exposed to antidiabetics, among which an inverse association was noted for insulin, metformin and sulfonylureas [12]. A US study also found a lower ALS risk among individuals who used human recombinant Glucagon using either a 1- or 3-year lag time, although a lower risk for metformin was only noted using a 1-year lag time, compared with controls [6].

Proton pump inhibitor (PPI)

Only one study has evaluated the association between PPI use and ALS risk [26]. This study included 2484 cases with ALS and 24,840 controls in Sweden and failed to find an association between PPI use and ALS risk [26].

Drugs for psychiatric and neurological disorders

An Italian prospective cohort study with 687,324 participants examined the association of multiple nervous system drugs (opioids, antiepileptic drugs, anti-parkinsonian drugs, antipsychotics, and antidepressants) with ALS risk [30]. In this study, a 41% lower ALS risk was found in relation to the use of opioids (HR = 0.59, 95% CI 0.35–0.97) and a higher ALS risk was suggested in relation to the use of antiepileptics (HR 1.35, 95% CI 0.92–2.00), whereas no association was noted for use of anti-parkinsonian drugs, antipsychotics, or antidepressants [30]. A nested case–control study in Sweden found however a higher prevalence of antidepressant use before ALS diagnosis, compared with controls [28]. A German case–control study found a threefold risk increase of ALS in individuals with use of antidepressants (OR = 3.2, 95% CI 1.6–6.6) [22]. Lithium, used for bipolar and some depressive disorders, was found to be associated with a lower ALS risk in a US cohort study [29]. The central nervous system (CNS) stimulant amphetamine, used in the treatment of attention deficit hyperactivity disorder (ADHD) and narcolepsy, was not found to be associated with ALS risk in a US case–control study [27]. Skeletal muscle relaxant, baclofen, was found to be associated with a higher risk of ALS in a US case–control study [6].

Discussion

This systematic review summarized all available studies assessing the link between the use of seven classes of medications, including cholesterol lowering drugs, anti-inflammatory drugs, immunosuppressants, and antibiotics, oral contraceptives (OCs) or hormone replacement therapy (HRT), antihypertensive drugs, antidiabetics, proton pump inhibitors, and drugs for psychiatric and neurological disorders and risk of ALS. In brief, we found that there was generally a lack of agreement on the associations of these medications with the subsequent risk of ALS between studies. However, it appeared that statins, aspirin, OCs/HRT, antihypertensives, and antidiabetics were unlikely related to a higher risk of ALS. The positive associations suggested for the use of antibiotics, antidepressants and skeletal muscle relaxants might be attributable to prodromal symptoms of ALS.

An inverse association of cholesterol lowering drugs with ALS risk was found in four studies [8,9,10,11], while a null or marginal association was observed in four [7, 12, 13, 22]. The marginal association was likely due to the higher risk of ALS during the first year after statin use [13]. The differing results between studies might be attributable to diverse study designs, sample sizes, variable adjustments, or biological difference between populations. Although still controversial, some studies have suggested that hyperlipidemia may be a risk factor ALS [31, 32]. If this is the case, the null or inverse association noted between cholesterol lowering drugs, especially statins, and ALS risk might indicate a potentially protective role of statins on ALS. After considering indications for statin use, Freedman et al. indeed found a lower ALS risk in relation to statin use, but not other cholesterol-lowering drugs [10]. The potentially protective role of statins may be attributed to their antioxidative and anti-inflammatory properties [33].

Steroids were not associated with the risk of ALS in the present study. Some experimental studies have shown that steroids may exhibit a neuroprotective role on neurons, glial cells, and blood vessels via steroid receptor signaling [34, 35], whereas others found a detrimental effect of steroids on motor neurons [36,37,38]. In terms of NSAIDs, most previous studies failed to show an association with ALS risk [14,15,16], except for one [22]. The discrepancy between studies might be partly due to sample size, recall bias with self-reported data, not considering reverse causation, and lack of correction for multiple testing. Real population differences might also exist. For instance, two US studies did not find an association between aspirin use and ALS [14, 15], whereas two Taiwanese studies found aspirin use to be associated with a lower risk of ALS [16]. A potentially protective role of aspirin on ALS has been identified in an experimental study which suggested that a derivative of aspirin “AAD-2004” might decrease motor neuron degeneration and improve motor function and life span by blocking free radical production, prostaglandin E [2] formation, and microglial activation in the spinal cord of ALS mice [39].

The findings on immunosuppressants are inconsistent [9, 17, 22]. Further, immunosuppressive treatment was not shown to be related to disease progression and survival of ALS patients [40, 41]. Emerging evidence suggests that gut microbiota dysbiosis might influence the development and disease progression of ALS through the microbiota-gut-brain axis [42, 43]. The relationship between antibiotics use and ALS risk, however, is less studied in human. One study showed a positive association, which was mainly attributable to the use of antibiotics during the years before ALS diagnosis [18].

There is similarly no agreement on the association of OCs or HRT use with the risk of ALS, but findings point to a lowered or null risk. OCs use differs greatly between populations due to cultural and socioeconomic differences [21] and is associated with many factors such as age, marital status, and health status [44, 45]. Such differences may have contributed to the difference in findings of different studies. In animal and cell studies, however, estrogen was found to exert neuroprotective effects on motor neurons through binding estrogen receptor α [46, 47] and hinder the progression of the female ovariectomized SOD1G93A mice. The prognostic role of estrogen has also been studied in ALS patients, however, with no considerable influence on disease progression [22].

In general, antihypertensive drugs were not shown to be associated with the risk of ALS. An Italian study demonstrated null associations of ACEIs and ARBs with the risk of MND [25], which differed from the finding of a protective role of ACEIs on ALS risk in a Taiwanese study [16]. The ethnic and age differences between the studies might explain the discrepancy to some extent. Further, the disparity in vitamin E exposure between the two populations may also contribute as ACEIs may function through increasing the levels of vitamin E whereas a protective role of ACEIs might disappear at the presence of relatively high vitamin E levels [25].

Antidiabetics use was shown to be associated with a lower risk of ALS, although two studies did not find statistically significant results [8, 22] which might be partly attributed to their relatively small sample sizes. It is difficult to conclude a neuroprotective role of antidiabetics on development of ALS due to the difficulty of discerning antidiabetics use from underlying diabetes. Previous studies have indeed found type 2 diabetes to be associated with a lower risk of ALS [48, 49]. A recent study has further suggested that the relationship might be causal in the European population [49]. Experimental studies have also supported a protective role of antidiabetics on ALS development as ALS mice treated with pioglitazone exhibited delayed onset and prolonged survival, in parallel with morphological and functional improvement of motor neurons [50].

Antidepressants and a skeletal muscle relaxant, baclofen, were found to be associated with a higher risk of ALS. Reverse causation might be the main explanation for these associations as depression and muscle problems are also symptoms of ALS and the risk is decreasing with longer time before diagnosis [6, 22, 28]. For instance, patients with ALS were more likely to receive a diagnosis of depression before diagnosis of ALS, especially during the year before ALS diagnosis [28]. Although a higher ALS risk was suggested in relation to antiepileptics use in one study [30], the association was not statistically significant. No genetic correlation was observed between ALS and epilepsy either [51]. Opioids and lithium, on the other hand, were found to be associated with a lower risk of ALS. Opioids exert neuroprotective effect through stimulating opioid receptors and upregulating excitatory amino acid transporters [52]. The protective effect of lithium in ALS may occur through an activation of autophagy in CNS [53]. Besides, findings that ALS patients with the UNC13A C/C genotype might benefit from lithium treatment suggesting a role of lithium in the promotion of synaptogenesis and neuronal outgrowth [54]. However, as these medications have been relatively rarely studied so far, a conclusion cannot be drawn before more studies are available.

To our knowledge, this is the first systematic review to evaluate the association between medication use and subsequent risk of ALS. The strengths of this review include the systematic search strategy using three large databases and the large numbers of articles included without restricting the publishing time. Further, the quality of the included studies was moderate to high according to the results of the quality assessment tool NOS.

Nonetheless, this review has several limitations. First, variation exists between studies regarding genetic background, ethnicity, geographic location, sample size, study design, consideration of reverse causation, and multivariable adjustment, making it difficult to make direct comparison between studies. Specifically, as ALS likely has a long pre-clinical stage with various prodromal symptoms and there is a known diagnostic delay of about 1 year for ALS, due to patient and healthcare-related delays [55], studies not considering potential reverse causation due to such might reach artificial conclusions regarding beneficial or harmful medications. Regardless, similar findings across different genetic background, ethnicity, and geographic location may on the other hand provide a good opportunity to find a robust risk factor for ALS. Second, some medications were only studied as a broad category, leaving the drug-specific effect unraveled. Third, some studies were based on patients with other diseases such as psychiatric disorders [29] and non-melanoma skin cancer [56], which may lead to low generalizability of the results to the general population. Finally, we focused on articles in English in three large databases (Medline, Embase, and Web of Science), studies from other databases or in other languages were not included. The number of the latter is however presumably small.

The increased availability of medication data obtained from medical records and administrative databases has greatly promoted the development of pharmaco-epidemiological research [57]. Although there are inherent challenges in pharmaco-epidemiological studies, namely that it is difficult to disentangle the effect of the studied medication from its underlying disease indication, so-called confounding by indication or indication bias, pharmaco-epidemiological studies are currently the main research tool to evaluate the long-term benefits and adverse events of medication use in the real-world practice [58]. Indication bias might be partially addressed by specific study designs, such as restricting the analysis to a target population (e.g., patients with specific disease) with or without medication use and active comparator design (i.e., comparing the effect of the target medication with another active drug) [59]. These designs also have limitations. For instance, they often rely on the assumption that there is little difference between the comparison groups other than medication use; however, it is almost always true that treated and untreated individuals as well as patients with different treatments for the same disease differ in disease severity, comorbidity profiles, etc.

As previous reviews have summarized the existing literature on the roles of specific diseases, including cardiometabolic and inflammatory diseases, on the risk of ALS [60, 61], in the present review, we focused on medication use as potential risk factor for ALS. However, because among the studies included in the present review, few have made efforts to address the concern of indication bias, it remains largely unknown whether it is the use of a specific medication or its underlying disease that is the real relevant factor for ALS. Although indication bias is unlikely to be completely ruled out in any pharmaco-epidemiological studies to date, new efforts should regardless be encouraged to use state-of-the-art study designs, such as analysis of target population [62], use of active comparator [63], application of propensity score to match comparison groups [64], and target trial emulation [65], to estimate the potential influence of indication bias. It would also be useful to provide more information, if available, on medication use, including indication for use, specific types of medication instead of using only a broad category of the medication, and dose and quantity of medication use.

In addition to indication bias, as important potential confounders such as smoking are rarely available in previous studies, residual confounding due to factors other than indication can also be a common source of systemic error. Further, the low incidence of ALS makes some studies underpowered to disclose real associations with relatively small magnitude from noise. Finally, the heterogeneity of ALS as a disease means that an overall null association for ALS in general does not necessarily preclude an association for a specific type of ALS and this has rarely been addressed in the literature.

Finally, the present review focused on studies where different medications were studied as potential risk factors for ALS. Additional efforts should be made to summarize the existing literature where medications are studied as potential factors influencing prognosis of ALS. For instance, one study suggested an association between status use and functional decline among ALS patients [66], whereas another study failed to do so [67]. In another study, aspirin or other NSAIDs use was found to be associated with a shorter survival among ALS patients; however, the association disappeared after correcting for multiple testing [68]. Similarly, although animal and cell studies have found estrogen to exert neuroprotective effect on motor neurons [46, 47], human studies failed to show an influence of estrogen on disease progression of ALS [22]. Finally, in terms of antidiabetics, a randomized trial evaluated pioglitazone as an add-on therapy in ALS patients without however showing beneficial effects on survival [69].

Conclusions

Although there is currently no strong evidence to link any medications with a higher risk of ALS, it appears that statins, aspirin, OCs/HRT, antihypertensives, and antidiabetics were unlikely related to a higher risk of ALS. Although this might provide evidence for the continued use of these medications among individuals at high-risk of ALS, interpretation of these results should be done with caution as most of the studies examined broad categories of these medications, leaving the potentially harmful role of specific drug classes unraveled. The positive association noted for antibiotics, antidepressants, and skeletal muscle relaxants might on the other hand be attributable to prodromal ALS symptoms or reverse causation due to diagnostic delay.