Amyotrophic Lateral Sclerosis as a Multistep Process in the United States: A Population‐Based Study

ABSTRACT

Background

Amyotrophic lateral sclerosis (ALS) is a fatal, progressive neurodegenerative disease that typically results in death within 3–5 years from symptom onset. However, little is known about the environmental exposures, clinical aspects, or social determinants of health factors that may be associated with the disease. Multistep modeling has been previously applied to cancer research, demonstrating a linear relationship between logs of incidence and age. This method may help to understand the mechanisms involved in the development of ALS in the United States (e.g., environmental exposures, genetic mutations). We aim to assess whether ALS is a multistep process among patients enrolled in the largest ALS registry in the world—the United States' National ALS Registry.

Methods

Incident ALS cases, defined as confirmed and likely, cases between 2012 and 2019 were obtained from the National ALS Registry. Age‐standardized incidence was calculated for all cases and by sex. The log incidence of ALS was regressed against the log of age (years) at case determination, on average, for each year and by sex.

Findings

Between 2012 and 2019, there was a mean of 5253 incident ALS cases (confirmed or likely) per year. We identified a linear relationship between the log of the average incidence and log age overall (r ² = 0.99), for men (r ² = 0.99), and for women (r ² = 0.98). The incidence slope estimates were 4.8 (95% CI: 4.4–5.1) overall, 4.7 (95% CI: 4.4–5.1) for men, and 5.0 (95% CI: 4.5–5.5) for women.

Interpretation

The linear relationships observed overall, for men, and for women are consistent with a multi‐step process. The slope estimates, on average, are approximately 5.0, which suggests that the development of ALS is a six‐step process. Further investigation of these steps can elucidate potential risk factors and treatments for ALS.

1. Introduction

Amyotrophic lateral sclerosis (ALS) is a rare, neurodegenerative disorder that is characterized by progressive deterioration in muscular and respiratory function that results in respiratory failure and muscular atrophy and weakness [1]. The pathophysiological mechanisms and potential causes that lead to the development of ALS are not completely understood. ALS is attributed to genetic mutations, such as SOD1 and C9orf72, in approximately 10% of people with a family history of ALS, and also in up to 15% of people with non‐familial ALS, but for most people, there appears to be no specific cause or origin [2, 3]. While environmental exposures, occupational exposures, and genetic predispositions have been implicated in the development of the disease, it is unknown if external exposures can trigger ALS from isolated, singular events or from accumulated, sequential episodes [4, 5]. The Agency for Toxic Substances and Disease Registry (ATSDR) is tasked with identifying exposures to communities that could result in adverse health effects and disease to aid in protection and prevention [6, 7]. ATSDR's ALS work meets this mandate [7].

The Armitage‐Doll model is a mathematical model originally applied to cancer research to assess the development of cancer as a multi‐step process [8]. This model demonstrates a proportional relationship between incidence and age in a given year, with the assumption that the disease develops in a single step [8]. Previous research has adapted this approach to establish multi‐step models among ALS populations and determine the number of steps involved in the development of the disease [9, 10, 11, 12]. These studies have reported that ALS cohorts in Australia, Europe, and Japan required six steps for ALS development, while the South Korean cohort required five steps [9, 10, 11, 12]. This indicates that there may be a minimum number of risk factor encounters that contribute to the development of the disease.

To date, this model has not been applied to ALS populations in North America. In the United States (US), ALS is a non‐notifiable disease, making it difficult to track and research the potential risk factors that can lead to the disease. Establishing a multi‐step model in a US‐based population could provide further insights to mechanisms of the disease and potential connections with similar cohorts in other countries. Therefore, the objective of this analysis was to apply the Armitage‐Doll model to an ALS population in the United States enrolled and identified in the National ALS Registry at the Centers for Disease Control and Prevention (CDC)/Agency for Toxic Substances and Disease Registries (ATSDR) [6, 7].

2. Methods

2.1. Analytic Population

In 2008, the ALS Registry Act was signed by Congress, establishing the National ALS Registry (Registry) via CDC/ATSDR [7]. After 2 years of pilot testing, the Registry was launched in 2010 and began collecting data from national administrative databases (e.g., Centers for Medicare Services, Veterans Benefits Affairs, Veterans Health Affairs) and also directly enrolling people with ALS into its web portal. The mission of the Registry is to estimate incidence, prevalence, and mortality of ALS in the United States, identify epidemiologic trends, and examine environmental exposures and social determinants of health that may be potential risk factors. Additional details on the Registry's objectives and methods are described in previous publications [13].

The Registry identifies people with ALS from national administrative databases from the Centers for Medicare and Medicaid Services (CMS), Veterans Health Administration (VHA), and the Veterans Benefits Administration (VBA), which provide data on inpatient, outpatient, patient pharmacy records, and pensions and compensation for disabilities for veterans. To categorize ALS cases, an algorithm is applied to each database utilizing International Classification of Diseases (ICD)‐9 and ICD‐10 codes, number of neurologist visits, and the use of Food and Drug Administration (FDA) approved ALS drugs Riluzole or Edaravone [14]. Only people categorized as having “confirmed ALS” or “likely ALS” are counted as a case for analysis [14]. People who are categorized as “confirmed ALS” have met multiple of the aforementioned criteria, while those categorized as “likely ALS” only have a prescription for Riluzole in their records. People with ALS were also identified from our secure web portal located on the CDC ALS website, in which patients can self‐register as a case. A screening process was used during the registration process, where patient responses to six validation questions determine if the person is a valid ALS case [15]. Patient information from the web portal and national administrative databases are merged and then deduplicated for analyses.

2.2. Model

We adapted the modeling techniques described by Armitage and Doll in their cancer research for ALS [8]. This model assumes that ALS is caused by a single step, in which the incidence (i) of ALS in a specific year will be proportional to the risk (u) of undergoing the specific step in that year. While the risk depends on the level of exposure to the disease‐causing event, the average background risk (u) of a step will be proportional to the incidence (i) when there is no information on the level of exposure. Thus, each step is associated with the risk u _i . If ALS is caused by more than one step, then the probability of undergoing the nth step at t years of age is u _n t. The plot of the log of ALS incidence against the log of age results in a linear relationship. This model follows the assumption that there is a logarithmic increase in incidence as age increases, which follows the power law that n − 1 is associated with the rate of increase. Thus, the number of steps required for ALS to develop is represented by the slope, n − 1, with n being the number of steps. However, since the model in other studies was less linear at older age groups (85+ years), these age groups were excluded from the model [8, 16].

2.3. Statistical Analysis

We calculated age‐standardized incidence per 100,000 persons for age groups 20–84 years during 2012–2019, overall, by year, by sex, and by race from the United States Census data. Only Black and White races were included in the analysis by race because other races were grouped together in the data and population size could not be accurately determined. We excluded case data from 2011 because all case data in 2011 were considered incident. The youngest (18–19 years) and oldest (85+ years) cases were excluded to reduce the chance of measurement errors, cohort effects, and the effect of under‐ascertainment of cases for those age groups. The remaining cases were divided into 5‐year age groups. We determined the midpoint of each age group and conducted unweighted linear least squares regression to plot the log of ALS incidence against the log of age. Unweighted linear regression was used to prevent the slope from being biased towards the oldest age groups. We conducted regressions overall and by sex. The 95% confidence intervals (CIs) were reported for the slope estimate of all models. All analyses were done in SAS v9.4 (SAS Institute Inc., Cary, NC) [17].

2.4. Role of the Funding Source

This work was funded by CDC/ATSDR. The findings and conclusions in this report are those of the author(s) and do not necessarily represent the official position of CDC/ATSDR.

3. Results

There was a mean of 5253 confirmed and likely incident cases per year between 2012 and 2019 (Table 1). The mean age at diagnosis ranged from 64.9 to 68.0 years. Each year, one half to two‐thirds of participants were men. Approximately 80% of patients identified as White. The highest number of incident cases was observed in 2015 (6838), while the lowest was observed in 2018 (3941). The age‐standardized incident rates were highest overall in 2015 (1.56), for men in 2012 (1.86), for women in 2015 (1.32), for White patients in 2015 (1.59), and for Black patients in 2015 (0.99). The rates were lowest in 2018 overall (0.71), for men (0.99) and women (0.47), White patients (0.72), and Black patients (0.48). Across all years, men had higher incident rates compared to women and the overall cohort. Additionally, White patients generally had higher incident rates compared to Black patients.

TABLE 1.

Basic characteristics of participants in the National ALS Registry 2012–2019, by year.

Year	Number of confirmed and likely cases	Population estimates	Age‐standardized incidences (men, women, both)	Age‐standardized incidences (White, Black, both)	Mean (std) age (years) at case	Number of men (%)	Number of White individuals (%)
2012	5004	240,134,326	1.86, 1.07, 1.44	1.42, 0.87, 1.29	64.9 (11.6)	3010 (60.2)	4140 (82.7)
2013	5087	242,425,013	1.82, 1.05, 1.41	1.42, 0.78, 1.27	65.3 (11.1)	3049 (59.9)	4274 (84.0)
2014	5191	244,737,285	1.78, 1.03, 1.41	1.44, 0.79, 1.29	65.1 (11.1)	3034 (58.4)	4419 (85.2)
2015	6838	247,017,112	1.83, 1.32, 1.56	1.59, 0.99, 1.44	66.9 (10.4)	3729 (54.5)	5681 (83.1)
2016	5229	249,291,898	1.62, 0.80, 1.19	1.21, 0.84, 1.09	65.8 (10.8)	3276 (62.7)	3866 (73.9)
2017	4802	251,400,193	1.33, 0.84, 1.07	1.09, 0.73, 0.99	65.5 (10.7)	2774 (57.8)	3617 (75.3)
2018	3941	253,368,356	0.99, 0.47, 0.71	0.72, 0.48, 0.65	68.0 (9.9)	2444 (62.0)	2613 (66.3)
2019	5931	255,200,373	1.63, 0.90, 1.24	1.25, 0.86, 1.12	65.2 (10.8)	3595 (60.6)	4329 (73.0)

The relationship between the log of the average incidence and the log of age was linear overall (r ² = 0.99) (Figure 1), with a slope estimate of 4.8 (95% CI: 4.4–5.1) (Table 2, Figure 2), consistent with an Armitage and Doll model and a six‐step process. Across 2012–2019, the range for the slope estimates was 4.3–5.3. By individual year, the slope estimate was highest in 2018 (5.3; 95% CI: 4.8–5.7) and lowest in 2019 (4.3; 95% CI: 3.9–4.8).

FIGURE 1.

Log (incidence) versus log (age) ALS incidence, 2012–2019; US National ALS Registry.

TABLE 2.

Slope estimates (95% CI) for ALS incidence in the period 2012–2019, overall and by sex and race; US National ALS registry.

Models	(Average) 2012–2019	2012	2013	2014	2015	2016	2017	2018	2019
Total	4.8 (4.4–5.1)	4.6 (4.1–5.1)	4.6 (4.2–4.9)	5.0 (4.5–5.6)	5.1 (4.7–5.6)	4.8 (4.3–5.2)	5.2 (4.6–5.8)	5.3 (4.8–5.7)	4.3 (3.9–4.8)
Men	4.7 (4.4–5.1)	4.7 (4.2–5.2)	4.5 (4.2–4.8)	4.8 (4.4–5.2)	5.1 (4.6–5.6)	4.7 (4.2–5.1)	4.6 (4.0–5.2)	5.4 (4.8–5.9)	4.4 (4.0–4.8)
Women	5.0 (4.5–5.5)	4.6 (4.1–5.1)	4.8 (4.5–5.2)	5.1 (4.3–6.0)	5.3 (4.8–5.8)	4.8 (4.2–5.4)	5.0 (4.5–5.4)	5.0 (4.1–6.0)	4.2 (3.6–4.7)
Black	4.2 (3.8–4.6)	4.4 (3.8–5.0)	4.6 (3.9–5.2)	4.9 (4.0–5.7)	4.7 (3.9–5.6)	4.8 (4.1–5.4)	3.8 (3.2–4.5)	4.0 (3.5–4.6)	4.0 (3.2–4.7)
White	4.8 (4.3–5.2)	4.9 (4.2–5.6)	4.5 (4.1–4.8)	5.1 (4.6–5.6)	5.3 (4.8–5.9)	4.8 (4.4–5.3)	4.9 (4.3–5.5)	5.8 (5.1–6.4)	4.6 (4.2–5.0)

FIGURE 2.

Forest plot of slope estimates (95% CI) for ALS incidence, 2012–2019, US National ALS Registry.

By sex, the relationship was also linear for men (r ² = 0.99) and women (r ² = 0.98). The slope estimates for both groups were similar to the estimate for the overall cohort, with women having a slightly higher slope estimate of 5.0 (95% CI: 4.5–5.5) compared to the slope estimate of 4.7 (95% CI: 4.4–5.1) for men. The slope estimates for both sexes, on average, suggested a six‐step process. By individual year, women also generally had higher slope estimates compared to men. For men, the highest slope estimate was in 2018 (5.4; 95% CI: 4.8–5.9) and the lowest estimates were in 2019 (4.4; 95% CI: 4.0–4.8). For women, the highest slope estimate was in 2015 (5.3; 95% CI: 4.8–5.8) while the lowest was in 2019 (4.2; 95% CI:3.6–4.7).

By race, the relationship was linear for both White (r ² = 0.98) and Black (r ² = 0.98) patients. The average slope estimates for White patients (4.8; 95% CI: 4.3–5.2) were similar to the overall cohort's average estimates, while the average slope estimates for Black patients (4.2; 95% CI: 3.8–4.6) were lower than the overall cohort's average slope estimates. The slope estimates for Black patients suggest, on average, a five‐step process, while a six‐step process is suggested for White patients. By individual year, Black patients generally had lower slope estimates compared to White patients. For Black patients, the highest slope estimate was in 2014 (4.9; 95% CI: 4.0–5.7) and the lowest estimates were in 2017 (3.8; 95% CI: 3.2–4.5). For White patients, the highest slope estimate was in 2018 (5.8; 95% CI: 5.1–6.4) while the lowest was in 2013 (4.5; 95% CI: 4.1–4.8).

4. Discussion

In this analysis, we observed a linear relationship between log incidence and log age at case confirmation across each year, overall, by sex, and by race, establishing for the first time the multi‐step model in a North American ALS population. On average, the slope estimates for the overall population and, separately, for men and women ranged from 4.7 to 5.0, which suggests that ALS is a six‐step process and that this is consistent by sex. By race, the average slope estimates ranged from 4.2 to 4.8, which suggests a five‐step to six‐step process. By given year, between 2012 and 2018, the slope estimates overall and by sex were also five, consistent with a six‐step process, while in 2019, the slope estimates were four, suggesting a five‐step process. By race, between 2012 and 2019, the slope estimates for White patients were consistent with a six‐step process, with the exception of 2018, which suggested a seven‐step process. For Black patients, between 2013 and 2017, the slope estimates were five, suggesting a six‐step process. In years 2012 and 2017–2019, the slope estimates were four, consistent with a five‐step process. Overall, our results are consistent with previous findings from multi‐step modeling studies conducted outside of the US, further reinforcing that ALS may require five to six pathophysiological steps to develop [9, 10, 11, 12, 18].

A step can be referred to as a single event that contributes to a triggering cascade that leads to the development of ALS, in which the suggested number of steps represents the threshold that has to be crossed for ALS to develop. Establishing the number of steps for ALS development could provide crucial insights into understanding which key risk factors are involved. It is widely believed that a combination of genetic and environmental factors can contribute to the risk of developing the disease [5]. However, the exact mechanisms or pathways have yet to be elucidated. Genetic mutations associated with ALS may contribute one or multiple steps to the process once expressed, while environmental exposures may represent the initial or final steps in the cascade for people with familial ALS. The number of steps a genetic mutation contributes to the process could be dependent on the penetrance of these mutations, as it has been suggested that, compared to patients with no genetic mutation, patients with low penetrance mutations require a higher number of steps for ALS to develop, while patients with high penetrance mutations may require fewer steps for ALS development [10]. Since familial ALS only represents an estimated 10% of cases, environmental factors or somatic mutations may be more influential in the multi‐step process for people with apparently non‐genetic ALS. It has been hypothesized that exposure to toxic environmental insults throughout one's lifetime could contribute to the neuronal degradation associated with ALS. Environmental exposures, such as first and second‐hand smoke, pesticides, heavy metals, cyanobacteria, and occupational exposures have been linked to the risk of developing ALS [19, 20, 21, 22, 23, 24, 25, 26]. Although many of these exposures have been studied in isolation to assess the risk of ALS, a step may be triggered by a combination of environmental exposures that trigger neuron damage or somatic mutations that contribute to developing ALS [27, 28, 29, 30]. The combination of environmental exposures required to trigger a step may also depend on the dose of exposure, length of exposure, and prior history of environmental exposures [31, 32]. Thus, future research could consider investigating combinations of environmental exposures throughout a person's lifetime.

We report that the slope estimates are closer to 4.0 for incident ALS in 2019, across all groups, except for White patients, suggesting a five‐step process, which is slightly lower than the six‐step process observed on average or for each year between 2012 and 2018. This suggests that persons diagnosed in 2019 required fewer steps to develop ALS compared to those diagnosed in previous years. That year had the second highest number of confirmed and likely cases but the second lowest mean age at case confirmation. Participants in 2019 may have had fewer protective biological mechanisms compared to other years, which may have made them more susceptible to neuronal degradation from environmental risk factors [33]. This would result in participants developing ALS slightly earlier compared to other years. Additionally, for 2012 and 2017–2019, Black patients had slope estimates that suggested a five‐step process while White patients had slope estimates that were consistent with a six‐ to seven‐step process. While the reason for this is unclear, it is possible that Black patients experience long‐term exposure to environmental pollutants more often than White patients, which may reduce the number of steps that contribute to developing ALS. Black Americans are disproportionately impacted by environmental hazards as they are more likely to live in communities that have high levels of air pollution, have poor infrastructure that can introduce heavy metals in the drinking water, and are near industrial facilities that release pollutants that contaminate the air, soil, and water compared to White Americans [34, 35, 36, 37]. Living in proximity to environmental hazards has been associated with poorer health outcomes and increased risks of developing cancers, neurological conditions, and somatic mutations [38, 39, 40]. Therefore, due to the higher likelihood of residential proximity to environmental pollutants, Black patients may be more vulnerable to the neuronal damage that leads to developing ALS compared to White patients.

A previous study found that populations with ALS across England, Ireland, Italy, the Netherlands, and Scotland all had slope estimates of 4.5 to 5.1, consistent with a six‐step process [9]. These results were also consistent after stratification by sex. Similarly, a study investigating the multi‐step process among populations with ALS in Australia also reported a linear relationship, with the suggestion that six steps may be required for ALS to develop [12]. A follow‐up study from these researchers incorporated South Korean, Japanese, and Australian populations with ALS and found similar results as well [11]. They established that both Japanese and Australian populations may require six steps to develop ALS, while South Korean populations may require five steps. In another study in an Italian population with ALS, the presence of a genetic mutation was taken into consideration when applying the multi‐step model [10]. While the slope estimates among all patients suggested a six‐step process, when broken down by genetic mutation, the estimations were lower. Analysis of patients with definite or probable or familial ALS demonstrated a four‐step process overall [10]. Patients with the C9orf72 mutation had a slope estimate that suggested a three‐step process. Those with the SOD1 mutation had a slope consistent with a two‐step process. TARDBP mutations aligned with a four‐step process, while those with familial ALS who tested negative for the aforementioned genes had slope estimates that suggested a five‐step process. Collectively, the similar findings across these studies and the present analysis suggest that differences by sex or geography are unlikely and that people with ALS experience similar pathophysiological mechanisms across diverse backgrounds. However, the lower slope estimates observed in those with ALS who have genetic mutations suggest genetic predispositions may play a greater role in the interaction between genetic and environmental triggers that lead to the development of ALS. In a study using the model among a Danish population, they also found overall that slopes were consistent with a five to six‐step process [18]. When stratified by sex, females were found to require half a step more than males, which aligns with the results of our analysis. Prior diagnosis of cardiovascular disease (CVD) was found to require almost one and a half more steps compared to those without the diagnosis, suggesting it may be a modifying factor for ALS. When the competing risk of death was controlled for and the analysis was conducted only among those with CVD in the population, they found that the slope estimates were consistent with a two‐step process, suggesting that within a CVD population, those with a CVD diagnosis require fewer steps to develop ALS. These results highlight that lifestyle factors and exposures can be considered when creating models and that competing causes of death add an additional complexity to the interpretation of the multi‐step model. Risk factors may not represent a single step in the pathway and may be a part of other interactions in the pathophysiological mechanisms that contribute to steps in the model.

The findings from this analysis should be considered in the context of its limitations. Approximately 80% of participants in the Registry identify as White. Therefore, the results may not be generalizable to other race/ethnic groups with ALS in the United States. The Registry also does not contain information on genetic data associated with ALS, which prevented further stratification and analysis by genetic mutation. Participation in the web portal is also voluntary and we are unable to obtain case information from sources such as private insurance. Issues with case ascertainment for certain years can also occur as we receive case data from large national administrative databases that may contain incomplete patient information that is difficult to match with existing patient records. Additionally, delays with receiving data from these databases restrict our ability to report on incident data from recent years. Therefore, the incidence reported by the Registry may be lower than the true incidence, which is difficult to obtain for a non‐notifiable disease.

The results from this analysis establish ALS as a multi‐step process in a US‐based population with ALS, giving support and further evidence to reports from previous multistep models applied to European, East Asian, and Australian populations. These findings suggest that patients with ALS across diverse geographic locations could experience similar pathophysiological mechanisms that contribute to the development of ALS, but that more research is needed to investigate this model by race/ethnicity. Determining these mechanisms and how they contribute to the multi‐step process could provide justification for environmental interventions that target common environmental risk factors.

Author Contributions

J.B., J.R., P.M., T.L., M.H., T.N., and D.K.H. assisted with study design, data collection and analysis, and preparation of the manuscript. A.A.C. assisted with data interpretation and preparation of the manuscript.

Disclosure

The findings and conclusions in this report are those of the author(s) and do not necessarily represent the official position of the Centers for Disease Control and Prevention/the Agency for Toxic Substances and Disease Registry.

Conflicts of Interest

The authors declare no conflicts of interest.

Funding Statement

This work was funded by Centers for Disease Control and Prevention .

Data Availability Statement

The data that support the findings of this study are available from the corresponding author upon reasonable request.