Residential exposure associations with ALS risk, survival, and phenotype: a Michigan-based case-control study

Abstract

Background:

Environmental exposures impact amyotrophic lateral sclerosis (ALS) risk and progression, a fatal and progressive neurodegenerative disease. Better characterization of these exposures is needed to decrease disease burden.

Objective:

To identify exposures in the residential setting that associate with ALS risk, survival, and onset segment.

Methods:

ALS and control participants recruited from University of Michigan completed a survey that ascertained exposure risks in the residential setting. ALS risk was assessed using logistic regression models followed by latent profile analysis to consider exposure profiles. A case-only analysis considered the contribution of the residential exposure variables via a Cox proportional hazards model for survival outcomes and multinomial logistic regression for onset segment, a polytomous outcome.

Results:

This study included 367 ALS and 255 control participants. Twelve residential variables associated with ALS risk after correcting for multiple comparison testing, with storage in an attached garage of chemical products including gasoline or kerosene (odds ratio (OR)=1.14, p_adjusted<0.001), gasoline-powered equipment (OR=1.16, p_adjusted<0.001), and lawn care products (OR=1.15, p_adjusted<0.001) representing the top three risk factors sorted by p_adjusted. Latent profile analysis indicated that storage of these chemical products in both attached and detached garages increased ALS risk. Although residential variables were not associated with poorer ALS survival following multiple testing corrections, storing pesticides, lawn care products, and woodworking supplies in the home were associated with shorter ALS survival using nominal p-values. No exposures were associated with ALS onset segment.

Conclusion:

Residential exposures may be important modifiable components of the ALS susceptibility and prognosis exposome.

Introduction

The fatal neurodegenerative disease amyotrophic lateral sclerosis (ALS) causes primarily progressive impairments in voluntary motor function.¹ Although a large number of genes are found to associate with ALS risk, the vast majority of ALS patients are without disease-causing mutations and are termed “sporadic.”^2,3 It is believed that in these sporadic cases, environmental and occupational exposures greatly contribute to disease risk and possibly to disease progression.⁴ Exposures that have been identified in many ALS cohorts include pesticides and metals.⁵ We recently replicated the impact of persistent organic pollutant mixtures with ALS risk and progression.^6–8 A more thorough understanding of such exposures could be leveraged to reduce disease burden by reducing or mitigating exposures. In our Michigan-based cohort,⁹ we have found that several self-reported exposure types in occupational^10,11 and avocational settings¹² influence ALS risk. These include production occupation work, self-reported occupational exposure to metals, particulate matter (PM), volatile organic compounds (VOCs), combustion and diesel exhaust, wood working, hunting and shooting, and playing golf in years leading up to ALS symptom onset. A comprehensive analysis in residential settings is needed given that many exposures also can occur at home, individuals spend most of their time indoors at home, and residential exposures affect all occupants of the home.

Methods

Participants

ALS participant recruitment drew from patients receiving care at the University of Michigan (UM) Pranger ALS Clinic. All patients with a Gold Coast ALS diagnosis were approached. Controls were identified from population outreach efforts via either the University of Michigan Health Research recruitment tool provided by the Michigan Institute for Clinical and Health Research or random address sampling. Controls were excluded if a first- or second-degree blood relative had ALS. All participants were at least 18 years old and were required to provide consent in English. The study was approved by the University of Michigan Institutional Review Board (HUM28826). Controls received compensation for survey completion and donated blood and urine samples for future research.

Survey development and administration

The survey (Supplemental Table 1), designed to assess exposure in the residential setting, was developed based on the expertise of an exposure scientist (SAB) and was also informed by Agency for Toxic Substances and Disease Registry (ATSDR) instruments.¹³ Participants were also asked to provide the dates or ages lived for each residence to calculate exposure durations. Completion instructions accompanied the survey. Survey staff contacted participants for survey completion reminders and to clarify responses as needed. A close contact was able to assist in the event an ALS participant could not communicate via voice or if they died. Survey responses were entered into a Redcap database. Data validation steps included checking allowable date ranges and continuity of time lived in each residence, needed to calculate exposure durations. For example, ensuring that move-out dates occurred after move-in dates and that dates lived for all residencies occurred after a participant’s birth date and before other key dates such as study entry date, onset date, and death date. The survey focused on exposures at four homes: the current, the birth home, and the other two longest lived homes. For ALS participants, homes lived in exclusively after symptom onset were excluded from the analyses (n=52), as were homes for ALS participants missing an onset date (n=0). For controls, homes lived in after study consent were excluded from analysis (n=3). To help ensure sufficient coverage and representativeness of the survey data, cases (n=34) and controls (n=25) who lived less than half their life in the listed homes and less than 10 of their last 20 years in the listed homes, prior to onset for case and survey consent for controls, were also removed.

Statistical Analysis

Descriptive statistics were tabulated for the study cohort. Missing covariate and residential data were imputed using a two-stage multiple imputation with chained equations (MICE) procedure,¹⁴ stratified by ALS and healthy control status so that ALS specific characteristics could be used. Missing covariate information for controls was imputed conditional on sex, education, age at survey consent, and military service. Missing covariate information for cases was imputed conditional on sex, onset segment, age at symptom onset, El Escorial criteria, family history of ALS, education, age at diagnosis, military service, and the Nelson-Aalen estimator of the cumulative hazard rate. After imputing missing covariate information, missing data for residential variables corresponding to houses listed in the survey were imputed using the imputed covariate data and MICE for multi-level data, where individuals have up to four residences. The total number of imputed datasets was 20. For each imputed dataset, the number of life-years of exposure, the metric of interest in this analysis, was calculated as the duration of time lived in the houses with the exposure. For garage exposures, exposures were separated by whether exposures occurred in an attached or detached garage.

Three types of outcomes were considered: logistic regression models were used to estimate one-at-a-time associations between residential exposures and case/control status; multinomial regression models estimated associations between residential exposures and three onset segments (bulbar, cervical, lumbar); and Cox proportional hazards models estimated associations between residential exposures and post-diagnosis survival. Logistic and multinomial regression models were adjusted for military service, sex, education, life-years covered by self-reported residences, and life-years not covered by self-reported residences. The hazards models were adjusted for military service, sex, age at diagnosis, education, log-transformed time between symptom onset and diagnosis, El Escorial criteria, onset segment, family history of ALS, life-years covered by the self-reported residences, and life-years not covered by the self-reported residences. Note that adjusting for the latter two variables in all outcome models provides an implicit adjustment for participant age and accounts for participants having differing numbers of life-years across their self-reported residences. For exposures pertaining to chemicals stored in garages, two variables were used: the number of life-years for exposures with attached garages; and life-years exposed for detached garages (regardless of items stored). Including both variables in the same regression model separates the more transitory nature of exposure occurring due to detached garages, which occurs only when the participant is in the garage, although exposure to the stored chemical may occur at other times, e.g., gardening or painting. Attached garages used to store chemicals have an additional and potentially important exposure pathway due to migration of contaminated air (and potentially dust) into the house. Each of the non-garage exposures was modeled using an individual regression model. Rubin’s rules for pooled inference across imputed datasets were used to obtain confidence intervals and p-values, and the Benjamini-Hochberg procedure was used to control the false discovery rate.^15,16 Kaplan-Meier survival curves, stratified by whether participants ever or never stored chemicals, were constructed using the residential variables that attained the nominal significance level of 0.05 prior to multiple testing correction.

Lastly, to assess whether residential exposure profiles associate with ALS risk, post-diagnosis survival, and phenotype, unsupervised latent profile analysis was performed, with a goal of minimizing the number of interpretable profiles. Latent profiles were estimated using the tidyLPA package in R, with equal variances across the profiles and zero covariances across the exposures.¹⁷ Then, the estimated latent profiles were associated with ALS risk, post-diagnosis survival, and phenotype, in a manner similar to the individual exposure variables. Descriptive statistics by latent profile assignment were calculated. Because survival and onset segment are ALS specific outcomes and the latent profile analysis was run on both cases and controls, a sensitivity analysis rerunning the latent profile analysis on cases only for the survival and onset segment analyses was performed.

Results

Overview of participants and residential exposures

Participants in this analysis included 367 ALS and 255 control participants. Among ALS participants, the survey participation rate was 55%. Self-reported occupational and avocational exposure data for these participants were previously reported.^10–12 ALS cases were more male (54.5%) compared to controls (45.5%, p=0.033), had lower educational attainment (p<0.001), and were more likely to have a military service history at marginal statistical significance (p=0.067) (Table 1).

Table 1.

Participant Demographics

Covariate	Cases (N = 367)	Controls (N = 255)	P-Value
Age (years)*	63.0 (55.6–70.0)	61.6 (55.0–68.3)	0.093
Sex			0.033
Female	167 (45.5)	139 (54.5)
Male	200 (54.5)	116 (45.5)
Military Service			0.067
No	308 (83.9)	225 (88.2)
Yes	58 (15.8)	26 (10.2)
Missing	1 (0.3)	4 (1.6)
Education			<0.001
High School or less	111 (30.2)	24 (9.4)
Some Postsecondary	115 (31.3)	71 (27.8)
Bachelor’s Degree	84 (22.9)	73 (28.6)
Graduate Degree	57 (15.5)	85 (33.3)
Missing	0 (0.0)	2 (0.8)
El Escorial Criteria
Suspected	13 (3.5)
Possible	36 (9.8)
Probable, Lab Supported	98 (26.7)
Probable	119 (32.4)
Definite	101 (27.5)
Onset Segment
Bulbar	104 (28.3)
Cervical	124 (33.8)
Lumbar	137 (37.3)
Missing	2 (0.5)
Family History of ALS
Yes	31 (8.5)
No	321 (87.5)
Missing	15 (4.1)
C9orf72
Positive	22 (6.0)
Negative	231 (62.9)
Not tested	114 (31.1)
Time Between Symptom Onset and Diagnosis (years)	1.06 (0.65–1.84)

Table of descriptive statistics for the study population. For continuous variables, median (25th – 75th percentile), and for categorical variables, N (%). P-values for continuous and categorical variables correspond to analysis of variance tests and chi-squared tests, respectively.

^*Age is defined at survey consent for controls and at symptom onset date for cases.

Most participants provided self-reported residential exposure histories for 4 homes (current residence, birth residence, and two other longest-lived residencies). ALS and control participants reported a total of 1,236 and 947 residences, respectively. Among the ALS participants, 55 residencies reported after symptom onset were excluded. The total time captured was 27,860 life-years across all participants. Survey question missingness is shown in Supplemental Table S1. Overall missingness was less than 15% for all variables.

ALS risk

Twelve residential characteristics were significantly associated with ALS risk following the multiple comparison correction (Table 2). The top seven (by lowest adjusted p-value) included those that involved storage of various chemicals in an attached garage: gasoline and/or kerosene (odds ratio (OR)=1.14, p_adjusted<0.001); gasoline-powered equipment (OR=1.16, p_adjusted<0.001); lawn care products (OR=1.15, p_adjusted<0.001); pesticides (OR=1.12, p_adjusted=0.011); and paint (OR=1.11, p_adjusted=0.013). Also in the top seven were humidifier in the home (OR=1.09, p_adjusted=0.013) and central heating using oil fuel (OR=1.18, p_adjusted=0.013). Here, the ORs correspond to 5 life-years of additional exposure. Notably, fuels (gasoline and kerosene), paints, pesticides and many lawn care products are volatile and have toxic components. Most participants reported storing multiple items in the attached garage (Supplemental Figure S1). Sensitivity analyses excluding participants with a known ALS family history followed by those with a C9orf72 expansion (using data from our recent publication¹⁸) did not reveal substantial changes in the ORs (Supplemental Table S2).

Table 2.

Residential Exposure Risk Model

Description	Exposed N (%)	OR	95% CI	P-Value	Q-Value (BH)
Storage in attached garage
Gasoline and/or kerosene	402 (64.6%)	1.14	(1.07, 1.22)	0.000	0.000
Gasoline-powered equipment	404 (65.0%)	1.16	(1.08, 1.24)	0.000	0.000
Lawn care products	395 (63.5%)	1.15	(1.07, 1.23)	0.000	0.000
Pesticides	389 (62.5%)	1.12	(1.05, 1.19)	0.001	0.011
Paint	330 (53.1%)	1.11	(1.04, 1.18)	0.002	0.013
Solvents	321 (51.6%)	1.09	(1.02, 1.16)	0.009	0.037
Woodworking supplies	241 (38.7%)	1.09	(1.02, 1.17)	0.012	0.045
Bleach	161 (25.9%)	1.11	(1.02, 1.21)	0.018	0.062
Car Parked	321 (51.6%)	1.04	(0.97, 1.10)	0.261	0.511
Ammonia	118 (19.0%)	1.03	(0.94, 1.12)	0.565	0.726
Storage in detached garage
Gasoline-powered equipment	415 (66.7%)	1.11	(1.03, 1.19)	0.003	0.015
Lawn care products	347 (55.8%)	1.11	(1.04, 1.19)	0.003	0.015
Pesticides	309 (49.7%)	1.08	(1.00, 1.15)	0.039	0.125
Gasoline and/or kerosene	402 (64.6%)	1.07	(1.00, 1.15)	0.044	0.132
Solvents	301 (48.4%)	1.04	(0.98, 1.11)	0.205	0.461
Woodworking supplies	190 (30.5%)	1.05	(0.97, 1.13)	0.218	0.467
Paint	264 (42.4%)	1.03	(0.97, 1.11)	0.325	0.532
Bleach	106 (17.0%)	1.03	(0.93, 1.14)	0.561	0.726
Car Parked	178 (28.6%)	1.02	(0.94, 1.11)	0.600	0.739
Ammonia	86 (13.8%)	1.00	(0.90, 1.12)	0.966	0.978
Others
Humidifier	392 (63.0%)	1.09	(1.03, 1.16)	0.002	0.013
Central heating - oil	201 (32.3%)	1.18	(1.06, 1.30)	0.002	0.013
Pet dog	545 (87.6%)	1.08	(1.02, 1.14)	0.005	0.022
Fireplace / wood stove	437 (70.3%)	1.05	(0.99, 1.11)	0.107	0.301
Solvents	357 (57.4%)	0.96	(0.91, 1.01)	0.114	0.302
Pet cat	389 (62.5%)	0.96	(0.91, 1.01)	0.122	0.305
Farm site nearby	243 (39.1%)	1.05	(0.98, 1.12)	0.143	0.339
Industrial operation site nearby	78 (12.5%)	1.09	(0.94, 1.26)	0.252	0.511
Water leaks - floor	136 (21.9%)	0.96	(0.89, 1.03)	0.274	0.514
Pesticides	431 (69.3%)	0.97	(0.93, 1.02)	0.292	0.526
Stand-alone air purifier (filter)	165 (26.5%)	0.96	(0.89, 1.04)	0.318	0.532
Gasoline and/or kerosene	100 (16.1%)	0.95	(0.85, 1.06)	0.331	0.532
Pet bird	139 (22.3%)	0.96	(0.88, 1.05)	0.356	0.552
Gasoline-powered equipment	75 (12.1%)	0.95	(0.83, 1.08)	0.447	0.652
Clutter or storage of material or chemicals next to home	80 (12.9%)	1.04	(0.94, 1.15)	0.449	0.652
Water leaks - walls	144 (23.2%)	0.97	(0.90, 1.05)	0.496	0.683
Water leaks - ceiling	218 (35.0%)	0.98	(0.91, 1.04)	0.501	0.683
Use wood stove or fireplace	400 (64.3%)	1.02	(0.96, 1.08)	0.608	0.739
Lawn care products	126 (20.3%)	0.98	(0.90, 1.08)	0.726	0.860
Greater than 4 blocks from major road	433 (69.6%)	1.01	(0.96, 1.06)	0.807	0.931
Woodworking supplies	408 (65.6%)	0.99	(0.95, 1.05)	0.839	0.944
Central heating - other	197 (31.7%)	1.01	(0.92, 1.10)	0.877	0.956
Golf course site nearby	78 (12.5%)	0.99	(0.88, 1.11)	0.892	0.956
1 to 4 blocks from major road	450 (72.3%)	1.00	(0.95, 1.05)	0.959	0.978
Gas station site nearby	123 (19.8%)	1.00	(0.91, 1.10)	0.978	0.978

Models are adjusted for military service, sex, education, number of years in the survey, and number of years not in the survey. Interpretation of the odds ratios correspond to five life-years of additional exposure. Variables are ranked by q-values.

BH, Benjamini-Hochberg; CI, confidence interval; N, number; OR, odds ratio

Unsupervised latent profile analysis classified ALS and control participants into three subpopulations (Figure 1; Table 3): first, participants (n=111) who had high storage of chemicals in a detached garage; second, participants (n=230) who had high storage of chemicals in an attached garage; and third, participants (n=281) who had low storage of chemicals in a garage. The low storage profile had a higher proportion of control participants (52.3%) compared to high detached (29.7%) and high attached (32.6%) garage storage with p<0.001. Low storage participants were younger (median age of 59.4 years) than those in high detached (63.7 years) or high attached (65.3 years) garage storage with p<0.001. Using individuals with low storage as a reference, participants who stored products in a detached (OR=2.16, 95%CI 1.26–3.71, p=0.005) and an attached (OR=2.37, 95%CI 1.55–3.62, p<0.001) garage had an elevated ALS risk. Note, these relatively large ORs apply to 5 additional life-years with the exposure noted.

Figure 1. Latent Profile Analysis for ALS and Control Participants.

Latent profile analysis separated ALS and control participants into three profiles: those with high storage of products in their detached garage (n=111, circles, blue in color figure); those with high storage of products in an attached garage (n=230, triangles, red in color figure); and those with low storage of products in a garage (n=281, diamonds, yellow in color figure). With low storage as the reference, participants who had high storage of products in a detached (OR=2.16, 95%CI 1.26–3.71, p=0.005) and high storage in an attached (OR=2.37, 95%CI 1.55–3.62, p<0.001) garage were at an elevated ALS risk.

Table 3.

Participant Demographics by Latent Profile

Covariate	Latent Profile 1 (N = 111)	Latent Profile 2 (N = 230)	Latent Profile 3 (N = 281)	P-Value
Age (years)*	63.7 (57.2–70.8)	65.3 (59.2–70.2)	59.4 (51.9–66.4)	<0.001
Sex				0.289
Female	51 (45.9)	107 (46.5)	148 (52.7)
Male	60 (54.1)	123 (53.5)	133 (47.3)
Military Service				<0.001
No	89 (80.2)	187 (81.3)	257 (91.5)
Yes	22 (19.8)	41 (17.8)	21 (7.5)
Missing	0 (0.0)	2 (0.9)	3 (1.1)
Education				0.002
High School or less	34 (30.6)	40 (17.4)	61 (21.7)
Some Postsecondary	41 (36.9)	64 (27.8)	81 (28.8)
Bachelor’s Degree	15 (13.5)	75 (32.6)	67 (23.8)
Graduate Degree	21 (18.9)	51 (22.2)	70 (24.9)
Missing	0 (0.0)	0 (0.0)	2 (0.7)
Subject Type				<0.001
Case	78 (70.3)	155 (67.4)	134 (47.7)
Control	33 (29.7)	75 (32.6)	147 (52.3)
El Escorial Criteria**				0.168
Suspected	1 (1.3)	7 (4.5)	5 (3.7)
Possible	7 (9.0)	14 (9.0)	15 (11.2)
Probable, Lab Supported	21 (26.9)	50 (32.3)	27 (20.1)
Probable	24 (30.8)	52 (33.5)	43 (32.1)
Definite	25 (32.1)	32 (20.6)	44 (32.8)
Onset Segment				0.544
Bulbar	24 (30.8)	45 (29.0)	35 (26.1)
Cervical	22 (28.2)	58 (37.4)	44 (32.8)
Lumbar	32 (41.0)	51 (32.9)	54 (40.3)
Missing	0 (0.0)	1 (0.6)	1 (0.7)
Family History of ALS				0.455
Yes	9 (11.5)	13 (8.4)	9 (6.7)
No	65 (83.3)	136 (87.7)	120 (89.6)
Missing	4 (5.1)	6 (3.9)	5 (3.7)
Time Between Symptom Onset and Diagnosis (years)	1.02 (0.69–2.04)	1.02 (0.60–1.76)	1.12 (0.73–1.87)	0.291

Table of descriptive statistics for the study population. For continuous variables, median (25th – 75th percentile), and for categorical variables, N (%). P-values for continuous and categorical variables correspond to analysis of variance tests and chi-squared tests, respectively.

^*Age is defined at survey consent for controls and at symptom onset date for cases.

^**P-value for Chi-squared test excludes the suspected category.

ALS survival and onset segment

For the second outcome examined, none of the residential variables were associated with ALS survival after multiple comparisons adjustment (Table 4). Nonetheless, several variables approached statistical significance that suggested poorer ALS survival: storage of pesticides in the home (hazard ratio (HR)=1.04, p=0.011, q=0.405), storage of lawn care products in the home (HR=1.07, p=0.018, q=0.405), and storage of woodworking supplies in the home (HR=1.04, p=0.043, q=0.532) (Supplemental Figure S2). No residential variable was significantly associated with onset segment (Supplemental Table S3).

Table 4.

Residential Exposure Survival Model

Description	Exposed N (%)	HR	95% CI	P-Value	Q-Value (BH)
Storage in attached garage
Lawn care products	245 (66.8%)	0.96	(0.92, 1.00)	0.065	0.532
Ammonia	75 (20.4%)	1.05	(1.00, 1.10)	0.071	0.532
Bleach	108 (29.4%)	1.03	(0.99, 1.08)	0.149	0.671
Paint	202 (55.0%)	0.97	(0.93, 1.01)	0.154	0.671
Woodworking supplies	158 (43.1%)	0.98	(0.94, 1.02)	0.267	0.707
Gasoline and/or kerosene	250 (68.1%)	0.98	(0.94, 1.02)	0.357	0.803
Car Parked	192 (52.3%)	1.02	(0.98, 1.06)	0.432	0.884
Solvents	201 (54.8%)	1.01	(0.97, 1.05)	0.610	0.982
Gasoline-powered equipment	249 (67.8%)	1.00	(0.96, 1.05)	0.951	0.982
Pesticides	240 (65.4%)	1.00	(0.96, 1.04)	0.953	0.982
Storage in detached garage
Car Parked	104 (28.3%)	1.04	(1.00, 1.09)	0.058	0.532
Paint	156 (42.5%)	0.98	(0.94, 1.02)	0.316	0.790
Lawn care products	210 (57.2%)	0.98	(0.94, 1.02)	0.338	0.801
Woodworking supplies	115 (31.3%)	0.99	(0.95, 1.03)	0.671	0.982
Ammonia	48 (13.1%)	1.01	(0.95, 1.08)	0.685	0.982
Gasoline and/or kerosene	235 (64.0%)	0.99	(0.95, 1.04)	0.697	0.982
Gasoline-powered equipment	246 (67.0%)	1.01	(0.96, 1.05)	0.742	0.982
Pesticides	184 (50.1%)	1.00	(0.96, 1.05)	0.842	0.982
Bleach	62 (16.9%)	1.00	(0.95, 1.06)	0.920	0.982
Solvents	184 (50.1%)	1.00	(0.96, 1.04)	0.950	0.982
Others
Pesticides	249 (67.8%)	1.04	(1.01, 1.07)	0.011	0.405
Lawn care products	69 (18.8%)	1.07	(1.01, 1.14)	0.018	0.405
Woodworking supplies	246 (67.0%)	1.04	(1.00, 1.07)	0.043	0.532
Solvents	202 (55.0%)	1.03	(0.99, 1.06)	0.119	0.671
Use wood stove or fireplace	236 (64.3%)	1.03	(0.99, 1.07)	0.153	0.671
Humidifier	246 (67.0%)	0.98	(0.95, 1.01)	0.176	0.671
Central heating - other	125 (34.1%)	0.96	(0.91, 1.02)	0.179	0.671
Gasoline-powered equipment	38 (10.4%)	1.06	(0.97, 1.16)	0.215	0.706
Greater than 4 blocks from major road	249 (67.8%)	1.02	(0.99, 1.05)	0.233	0.706
Golf course site nearby	42 (11.4%)	1.05	(0.97, 1.14)	0.249	0.706
Clutter or storage of material or chemicals next to home	49 (13.4%)	0.97	(0.92, 1.02)	0.251	0.706
Gas station site nearby	76 (20.7%)	1.02	(0.97, 1.09)	0.419	0.884
Pet dog	333 (90.7%)	1.01	(0.98, 1.05)	0.500	0.939
Fireplace / wood stove	265 (72.2%)	1.01	(0.98, 1.05)	0.501	0.939
Water leaks - ceiling	120 (32.7%)	0.99	(0.95, 1.03)	0.525	0.945
Stand-alone air purifier (filter)	92 (25.1%)	0.99	(0.94, 1.04)	0.635	0.982
Farm site nearby	151 (41.1%)	0.99	(0.96, 1.03)	0.699	0.982
Central heating - oil	135 (36.8%)	0.99	(0.94, 1.05)	0.708	0.982
Water leaks - floor	75 (20.4%)	1.01	(0.96, 1.06)	0.767	0.982
Water leaks - walls	79 (21.5%)	1.01	(0.96, 1.06)	0.782	0.982
1 to 4 blocks from major road	264 (71.9%)	1.00	(0.97, 1.03)	0.910	0.982
Pet cat	215 (58.6%)	1.00	(0.97, 1.03)	0.926	0.982
Gasoline and/or kerosene	54 (14.7%)	1.00	(0.92, 1.08)	0.953	0.982
Industrial operation site nearby	52 (14.2%)	1.00	(0.93, 1.08)	0.982	0.982
Pet bird	76 (20.7%)	1.00	(0.94, 1.06)	0.982	0.982

Models are adjusted for military service, sex, age at diagnosis, education, time between symptom onset and diagnosis, El Escorial criteria, onset segment, and family history of ALS. Interpretation of the hazard ratios correspond to five life-years of additional exposure.

BH, Benjamini-Hochberg; CI, confidence interval; HR, hazard ratio; N, number

Latent profile analyses were also conducted on the ALS-only population for survival and onset segment (Supplemental Figure S3A). These case-only latent profiles resembled those shown earlier for the full cohort: those ALS participants who had high storage in a detached garage; those who had high storage in an attached garage; and those who had low storage in a garage. Survival and onset segment did not differ by unsupervised latent profile (Supplemental Figure S3B–C).

Lastly, additional sensitivity analyses including assuming missing survey responses represented no exposure (Supplemental Table S4) and a complete case analysis (Supplemental Table S5) also did not reveal substantial changes in the results for both the risk and survival outcomes.

Discussion

Identifying disease-provoking exposures can inform and motivate interventions to reduce exposure, risk, and ultimately the ALS burden. Residential exposures, the focus of this study, are an important component of the exposome, and the home is a setting where behavior modifications could possibly lessen ALS risk.

A consistent risk factor in our study was garage storage of fuels (gasoline and/or kerosene) and gasoline-powered equipment, and the use of oil for home heating. This factor may represent several types of exposures and exposure pathways. First, fuels and oils contain mixtures of VOCs that include a number of toxicants, e.g., benzene, and inhalation exposure can occur in the garage and house for an attached garage as vapors are released and migrate into the house. VOC releases from gasoline powered equipment can be large, especially for older equipment that has rudimentary emission controls and likely worn seals and gaskets. Second, the presence of these fuels and equipment, as well as an oil furnace, indicates local combustion sources that can cause exposures to many pollutants, e.g., carbon monoxide, nitrogen oxides, and particulate matter.^19,20 Gasoline powered equipment, in particular, can have significant emissions and users can experience considerable exposure. While furnaces should be vented outdoors to limit indoor exposure, the presence of an oil furnace likely indicates older housing or rural neighborhoods where cleaner fuels (natural gas and electricity) are not used for heating, and thus ambient air pollutant levels may be locally elevated. Some fraction of the outdoor pollutants will enter occupied spaces, and older homes may be especially “leaky” and prone to entry of outdoor pollutants. Residential VOC and combustion sources and the risk of ALS have not been investigated to our knowledge. However, recent studies have suggested a link between ambient PM levels and ALS risk,^21,22 but impacts of the various sources contributing to PM levels, and specifically residential combustion sources, have not been characterized or apportioned.

We also show that storage of paint, pesticides, lawn care products, solvents, and bleach in an attached garage was an ALS risk. With the exception of bleach, all of these chemicals have been associated with ALS. For instance, a case-control study in Italy found occupational paint, pesticide, and solvent exposure associated with ALS risk.²³ A New England case-control study found similar results for occupational use of paint thinners, a volatile solvent.²⁴ Another New England cohort also found an association of paint with ALS.²⁵ Although this study investigated the use of chemicals in the occupational and hobby setting, the data were aggregated and the exposure setting cannot be determined.²⁵ Lastly, pesticide exposure is linked to ALS in several cohorts.⁵ While the neurotoxic effects of solvents are not entirely known,²⁶ the neurotoxic effects of pesticides are felt to be due in part to the consequence of ion channel dysregulation.⁸ Overall, our findings are very consistent with the literature, but our results are highly significant in extending the literature by showing that such exposures in residential settings have a meaningful impact on ALS risk.

Storage of woodworking supplies in an attached garage was also an ALS risk. This finding agrees with our recent analysis showing participation in wood working as a hobby increased ALS risk,¹² however, the link between wood working and ALS has not been thoroughly investigated. A University of Washington study did not find an association between wood working and ALS,²⁷ despite a non-significant suggestion of ALS with wood working chemicals. Storage of woodworking supplies could expose individuals to products containing VOCs and formaldehyde,²⁸ and these chemicals can be released from some wood species and wood products, e.g., plywood and particle board.^29,30 In addition to studies linking solvents to ALS noted above, formaldehyde also shows an association.^31,32 We must also consider that storage of woodworking chemicals may be an indicator of wood working as a hobby, and possibly carpentry and home renovations. Sawdust emitted during such activities is a form of PM, and exposure to sawdust from some wood species and woodworking activities is associated with increased risk of cancer, asthma and other respiratory effects in occupational settings.^33,34

An intriguing finding in our study was that one-at-a-time analyses revealed that storage of several chemicals in an attached, but not a detached, garage was associated with ALS. Flows of air and airborne pollutants from an attached garage to the adjoining residence have been demonstrated,³⁵ which is pertinent to both particulate matter and VOCs, e.g., benzene in gasoline.³⁵ Other emissions occurring in garages, e.g., carbon monoxide³⁶ and polybrominated diphenyl ethers,³⁷ can migrate from the garage to the living space, and building codes have been designed and updated to limit such flows and reduce indoor exposure.³⁸

The latent profile analysis also supported the increased ALS risk for participants storing chemicals in garages, and the higher risk for those with attached garages. This finding suggests the exposure pathway discussed above, namely, that polluted air in the garage migrates into the living space, and that this is an independent or additive effect to exposure from engagement in related hobbies or avocations, e.g., painting, gardening, and woodworking. Such findings would be strengthened by indoor air quality studies that measure and apportion exposures and concentrations in avocational, indoor, garage, and outdoor settings.

The two other home characteristics that reached statistical significance for ALS risk included having a humidifier or a pet dog in the home. We can speculate on various exposures that might explain these results, for example, having a dog might expose residents to pesticides used to control fleas, ticks and other vectors, and a humidifier might be associated with microbial contamination or biocides. At the same time, these could be spurious or confounded associations. At present, we consider these associations as speculative findings that require further evidence and replication.

Our study is unique because it focuses on risks for self-reported home characteristics. Prior ALS studies using residential information have used solely residential location and geospatial approaches to examine exposures originating from outside the home, such as traffic related air pollutants,²¹ electromagnetic radiation,³⁹ hazardous air pollutants,⁴⁰ pesticide exposure,⁴¹ cyanobacteria,⁴² and lead.⁴³ Thus, these studies do not include potential risk factors associated with the home and the indoor environment. This information may be obtained by monitoring the indoor environment, conducting walk-through inspections, and using self-reported survey information. Sometimes the necessary information is known only to the occupants, e.g., how frequently gas stoves or kitchen exhaust fans are used. Self-reported residential exposure data have been used to investigate a variety of health studies. As examples, self-reported residential pesticide use shows varying associations with breast cancer risk,^44–46 and the self-reported presence of gas stoves has long been linked to respiratory disease.⁴⁷ These self-reports are able to ascertain specific exposure sources and activities that may increase exposure propensity, beyond what occurs in the broader ambient environment.

No residential exposure variable was consistently and significantly (p<0.05) associated with ALS survival after adjusting for multiple comparisons or using latent profile analysis. However, storage of pesticides in the home was associated with survival with marginal statistical significance, nominal p<0.05, a finding in agreement with our occupational survival analysis¹⁰ and our analysis of the organochlorine pesticides.^6,8 Overall, associations with residential exposure variables and ALS phenotypic features were not significant. Residential characteristics might play only a limited role in ALS risk, the self-reported information might inadequately capture key information (i.e., exposure measurement error), or possibly a larger sample size is needed to provide further insight. Additional analyses of residential exposures, possibly with a reduced set of exposure risks, might help tease out these associations.

This study does have limitations. Recall bias is possible with any retrospective case-control study, and perhaps particularly important in this study where we request information from decades earlier. The survey questions focused on the storage of chemicals and not necessarily their use. However, as noted earlier, the assumption of storage within the residence leading to exposure is supported by the literature. Still, the study design does not allow an assessment of the severity and frequency of exposures to stored chemical products. Residential characteristics over the residential history reported prior to ALS diagnosis were weighted by life-years available in the survey, and exposure periods (windows) that might be crucial to disease development or progression were not considered, although such windows are also not yet known. Additionally, the survey may not capture all residential exposure risks, including those that occur via air, water, and dust contamination. In addition, we did not adjust for home age or neighborhood health, but these features are not available for much of the period covered in this analysis. The risks identified in the present analysis overall centered on storage of chemicals in an attached garage. Storage of these same products in the home were not a risk. This could be due to sample size/power or differences in patterns of where people tend to store chemicals in their home. Lastly, our survey does not ascertain household income or economic stability, so we are unable to evaluate whether this social determinant of health contributes to the presented findings.

In conclusion, we find that storage of chemicals, especially in attached garages, associate with an increased ALS risk and may be a modifiable factor that can help decrease the overall disease risk.

Data availability

Sharing of non-identifiable data will be considered at the reasonable request of a qualified investigator.