Methods
Study Population
The Prospective ALS study The Netherlands (PAN) is a population based case control study performed in The Netherlands during the period 1 January 2006 to 31 December 2010. Complete case ascertainment was ensured by continuous recruitment through multiple sources: neurologists, rehabilitation physicians, the Dutch Neuromuscular Patient Association and our ALS website.
All patients diagnosed with possible, probable (laboratory supported) or definite ALS according to the revised El Escorial criteria were included. Medical records were scrutinised for eligibility of patients, excluding patients with an ALS mimic syndrome or with a first, second or third degree family member with ALS. As exogenous factors—probably—had only a minor role in the development of ALS in patients with the highly penetrant C9ORF72 repeat expansion, these patients, 43 in total, were excluded from our analysis.
To ascertain population based controls, the general practitioner of the participating patient was asked to select individuals from his register in alphabetical order starting at the surname of the patient. The Dutch health care system ensures that every inhabitant is registered with a general practitioner, which makes this roster representative of the population. Controls were matched to patients for gender and age (±5 years). This study, however, did not use individual matching, meaning that some general practitioners delivered several controls while others delivered none. As can be seen in Table 1, our case and control groups were well matched for age and gender. Blood relatives or spouses of patients were not eligible to be controls, to prevent over matching.
Ethics approval was provided by the institutional review board of the University Medical Centre Utrecht. All participants gave written informed consent.
Data Collection
A structured questionnaire was used to collect demographic and clinical characteristics of participants and to obtain data regarding lifetime physical activities. Participants were asked to recollect all of their jobs and to describe the various activities they had to perform during these jobs. They were also asked to list all of their leisure time activities, consisting of sports and hobbies. For each activity, the participant was asked to state the number of years and how many hours per week the activity was performed. Specific questions were asked about vigorous physical activities (eg, marathon, triathlon, etc). This questionnaire was part of a larger questionnaire containing questions on several other exogenous factors. Participants were, therefore, blinded to the hypothesis being tested. In the patient group, only data referring to the period before disease onset were analysed. Survival status of the patients was recorded up to 8 August 2011 and obtained through the municipal personal records database or from the general practitioner. If the questionnaire was not completed in full or if data were found to be inconsistent, participants were approached by telephone to complete or correct the data. To ensure blinding, all questionnaires were coded prior to processing and analysis.
Classification of Physical Activities
To objectively quantify the cumulative lifetime physical activity level of participants, all reported activities were scored and coded based on the Compendium of Physical Activities. The Compendium provides a coding scheme that links specific activities performed in various settings with their respective MET. The definition of an MET is the ratio of work metabolic rate to a standard resting metabolic rate. A MET score of 1.0 (ie, the standard or resting metabolic rate while sitting quietly) is defined as 1 kcal×kg body weight×h. MET levels for specific activities, as reported in the Compendium, were established by reviewing published and unpublished studies that measured the energy cost of human physical activities. The compendium describes 605 specific activities. Assignment of MET scores to the activities enabled us to calculate cumulative scores of all reported physical activities:
where k represents an activity from the lifetime job or leisure time history. Because of the magnitude of the cumulative score, it was divided by 1000. Activities that had a MET score of ≤1.5 (eg, listening to music, reading, playing chess, needlework) were not included in the analysis. Subsequently, military service (not occupation) or periods spent as a homemaker were excluded because of difficulties quantifying these activities. Military service was mandatory for male study participants during a 15–24 month period around the age of 18 years and will therefore have minimal influence on total cumulative physical activity. In our study, 34% of patients compared with 35% of controls joined the military service (χ test: p=0.73), and 12% of both patients and controls reported periods spent as a homemaker (χ test: p=0.77).
Statistical Methods
Univariate and multivariate logistic regression were used to determine the association of physical activity and the risk of ALS. Standard unconditional logistic regression was used as the study did not include individual case control pairs but was frequency matched. The risk of ALS with cumulative scores of physical activity was analysed separately for leisure time activity, occupational activity and total activity (the combined leisure time and occupational activity) as a continuous variable. Furthermore, to determine a dose–response relationship, physical activity was categorised into quartiles based on the data of controls. The first quartile with the lowest intensity in physical activities was defined as the reference category. Multivariate logistic regression was used to determine the association between the four levels of physical activity and ALS. A separate multivariate logistic regression analysis was performed to determine the effect of vigorous physical activity (ever/never) on the risk of ALS. OR and p values were derived from these analyses. In the multivariate model, the ORs were adjusted for gender, age (at onset for patients and at the date the questionnaire was completed for controls), level of education (divided into seven categories ranging from no education to university), premorbid body mass index, current alcohol consumption and current smoking. In patients, current alcohol consumption and current smoking were determined at the time of disease onset, so before diagnosis and before the questionnaire was completed.
To determine a difference in the maximum intensity of the activities performed, the maximum MET scores were calculated (excluding duration in years or hours per week) and analysed using the Mann–Whitney U test.
A Cox regression analysis was performed to determine whether survival of patients was associated with physical activity. Survival was defined as the time from symptom onset to death or to the censoring date of 8 August 2011. The HR derived from these analyses were adjusted for gender, age at onset, site of onset and current smoking. Physical activity was entered into the model as a continuous variable. The same method was used to determine the effect of physical activity on the age at onset of ALS patients, adjusting for gender and site of onset. To adjust appropriately for age, an interaction term of diagnosis and physical activity was introduced into the Cox regression analysis using age at the time of completing the questionnaire for controls.
In the above mentioned models, we performed a complete case analysis, using only those cases without any missing values. A Bonferroni correction for multiple testing was applied adjusting for three tests (leisure time, occupational and total activity); a p value of 0.05/3=0.017 was considered significant.