Risk factors in Swedish young men for amyotrophic lateral sclerosis in adulthood

All Swedes have a unique personal identity number making linkage between different registers possible. Using this number, data from the Swedish Military Service Conscription Register were linked to the Swedish Hospital Discharge and death registries. The Swedish Hospital Discharge Register includes both inpatient (started 1986 and is complete since 1987) and outpatient visits (from 2001). A cohort of young individuals aged 16–25 (mean age 18.3, SD 0.7) who enlisted for military service between 1968 and 2005 (i.e., born between 1950 and 1987) was extracted from the conscription register (n = 1,886,542). At that time, all Swedish males had to enlist according to Swedish law. Exceptions were made regarding those who were imprisoned or had severe chronic somatic or mental conditions or functional disabilities documented by a medical certificate (approximately 2–3% annually). Exclusion criteria were: late allocated personal identity number (i.e., personal identity number reallocated from deceased person), female sex, ALS diagnosis before or at conscription, and missing data regarding conscription test center (Fig. 1). To further reduce baseline misclassification, we excluded incident ALS cases registered within 5 years after conscription (n = 9). The remaining 1,819,817 subjects were included in the present study (Fig. 1).

The enlisted men were examined over 2 days by psychologists and physicians at any of six conscription centers, Southern, Western, Eastern, Central (called Bergslagen), Northern lower, and Northern upper. The standardized procedure included measurement of weight and height.

The statistical calculations were performed with SAS version 9.4 (SAS Institute, Cary, NC, USA). The follow-up period started at the date of conscription (baseline) and subjects were followed until the time of: (1) first ALS hospitalization or hospital-based outpatient clinic contact, (2) death, (3) emigration, or (4) the end of follow-up, December 31, 2014 (minimum 9 years, maximum 46 years). For comparison of the population characteristics among men with and without ALS during follow-up, binary logistic regression was used for non-psychotic psychiatric diagnoses, multinomial logistic regression for conscription test center, ordinal logistic regression for ordered categories, and linear regression for continuous variables. Tests based on scaled Schoenfeld residuals were used to assess the proportional hazards assumption [31]. No significant violation was found for any of the variables in any of the models. Thus, there was no evidence for time-dependent associations, and Cox proportional hazards models were used to assess the influence of presumptive predictors and potential confounders for an ALS diagnosis.

The following presumptive independent predictors for ALS were treated as continuous variables: body mass index, W _max, hand grip, elbow flexion, knee extension, ESR, and EVF. We examined deviations from linearity with restricted cubic splines, and found no significant deviations and no improvement in Akaike Information Criterion. Categorical variables were conscription test center, cardiovascular fitness, overall muscle strength, parental education, and non-psychotic psychiatric disorders. Cardiovascular fitness, overall muscle strength, and parental education were trichotomized as low, medium, and high. As temporal differences, as well as procedural differences among sites could result in bias, conscription year and test center were considered potential confounders and adjusted for in all regression models. Furthermore, age at conscription was adjusted for. Because of the large number of observations, the risk for type I errors is considered low.

For the exposure variables, overall muscle strength and cardiovascular fitness, respectively, which both had more than 10% missing values and over 5% difference between the proportions of missing values among ALS cases compared to non-cases, we performed analyses using multiple imputations by the fully conditional specification approach [32], using 20 imputations and 22 iterations. All variables included in any of the other analyses were included in the imputation model, with the exception of variables that were functions or transformations of other variables, such as grip strength per kg, where only the original variable was included. In addition, we included height, weight, blood pressure, resting heart rate, stress resilience, IQ and command suitability evaluated at the conscription and indicators for diabetes, death and emigration, and the estimated cumulative baseline hazard and event indicator for ALS as recommended in [33]. Convergence was assessed using plots of the mean and standard deviation and found satisfactory.

The Ethics Committee of the University of Gothenburg and Confidentiality Clearance at Statistics Sweden approved the study. Requirement for informed consent was waived for the current study, because it was a secondary analysis of existing data. The investigation conforms to the principles outlined in the Declaration of Helsinki.

To our knowledge, this is the first study to demonstrate a prospective relationship with low overall muscle strength measured decades prior to a clinical diagnosis of ALS. It has recently been shown that respiratory muscle strength is a strong predictor for survival in ALS [34]. In line with Mattsson et al. [9], we found no association between each separate subparameter of muscle strength, i.e., grip strength, biceps strength, and knee extension and later ALS. As in the previous studies [7, 9], we found that low BMI was associated with future ALS. These findings suggest that the previously reported association between BMI and future ALS may actually reflect early muscular parameters.

It has been suggested that there is a risk phenotype for ALS characterized by relatively more type 1 muscle fibers (slow twitch fibers) [9]. Type 1 muscle fibers are the predominant muscle fiber type in endurance sports which may explain the higher incidence of ALS among professional athletes [19, 35]. The premorbid overall lower muscle strength seen in the present study could, perhaps, be explained by a smaller proportion of the larger type 2b fibers and in particular to the lesser force generating capacity of type 1 compared with type 2b fibers [36]. Muscle biopsies demonstrated that type 1 fibers are affected earlier and more severely in ALS [37]. There are so far no data on premorbid muscle fiber pathology in ALS.

Other plausible premorbid mechanisms may be motor nerve hyperexcitability [38], androgen insensitivity [39, 40], and inflammation [41, 42]. In the current study, we did not find evidence of an association between a systemic marker of inflammation (erythrocyte sedimentation rate) and future ALS. However, a significant association was observed for erythrocyte volume fraction (EVF), i.e., each percent increase was associated with a 4% decreased risk of later ALS. Although changes in EVF could predict disease progression in patients diagnosed with ALS [43], there are, to our knowledge, no previous reports indicating lower EVF as an early predictor of ALS. The present observation thus needs further investigation and it is far too early to speculate in mechanisms such as motor neuron and skeletal muscle oxygenation.

Somewhat surprisingly, the previously reported prospective association with high level of physical fitness in Swedish conscripts [8, 9] was not confirmed in our larger cohort. In addition to sample size, some other differences may help to explain the divergent results. The previous Swedish conscription study by Mattsson et al. only studied men born between 1951 and 1965 (based on deaths only, n = 85 cases), whereas we studied men born between 1950 and 1987 which means that the differences could be due to temporal trends. In our study, 42% of the ALS cases were still alive [9]. In addition, a recent Swedish conscription study by Longinetti et al. [8] showed that individuals with physical fitness above the highest tertile tended to have a higher risk of ALS before the age of 45 years, statistically significant associations only at age 41–43 years at ALS onset and only after adjustment for BMI and resting heart rate. Our cohort size is similar to that of the Longinetti et al. study, but our longer follow-up time yielded 87 additional ALS cases. The study of Longinetti et al. compares conscripts with high physical fitness (W _max/kg: ≥ 4.25 W/kg) with conscripts with lower physical fitness in fitted flexible parametric models with attained age as the time scale. We found no evidence for time-dependent associations for cardiovascular fitness or W _max/kg in any of the models.

As expected due to the higher incidence of ALS in the background population [44], men who enlisted in Northern Sweden were more likely to develop ALS than their peers who enlisted at other sites. The most plausible explanation for this is a genetic vulnerability. Familial aggregation and genetic risk factors have been estimated to explain 10–15% of all ALS cases [45]. One of the two major genetic contributions to ALS known to date is the C9ORF72 gene at chromosome 9 which is overrepresented in Northern Sweden [19]. In addition, persons living in northern Sweden might be differentially exposed to (unknown) environmental factors.

We could not detect any associations between non-psychotic mental disorder at conscription and later ALS. The previously demonstrated association with psychiatric disorders occurring more proximate in time [23] might suggest that mental disorders are prodromal symptoms of ALS-related neurodegeneration rather than predictors or risk factors per se.

Strengths of this study include the large sample size (> 1.8 million participants) and the prospective population-based design. The study population is largely representative of the male general population as only a small proportion (approximately 2–3% annually) of Swedish men were exempted from conscription prior to 2005. The follow-up was long, allowing for a considerably larger number of ALS cases compared to the previous studies employing the conscription register. The Swedish Hospital Discharge Register has a high accuracy for chronic diseases [46]. Both the Hospital Discharge Register [47] and the Swedish Cause of Death Register [9] have been shown to provide accurate ALS data.

The data retrieval of non-psychotic disorders from conscription data is a strength of the present study including psychologists and medical doctors for baseline measure of mental health at conscription. The general incidence is likely to be underestimated for non-psychotic disorders in the Hospital Discharge Register, since these disorders are mostly seen in primary care (where no national registers are available). This is less of a concern for estimates for psychotic disorders (diagnostic codes used for exclusion only to avoid misclassification of conscripts with prodromal episodes), as most persons with these disorders will come to the attention of psychiatric services at some point in time. However, psychotic disorders could be underestimated in this study, because outpatient hospital data, which are likely more sensitive than inpatient data, are only available for 2001–2014. We believe that it is unlikely that the associations we examined would be particularly sensitive to this lack of information which should be similar for individuals with and without an ALS diagnosis. Psychiatric diagnosis codes may change over time with each edition of ICD. The diagnostic validity of the schizophrenic disorders in the Swedish National Hospital Discharge Register has been shown to be high [48]. While not all mental diagnoses have been individually validated, the positive predictive values for most ICD diagnoses in the Swedish National Hospital Discharge Register are reported to be 85–95% [46].

There are further limitations to our study. For some exposure variables, for example, overall muscle strength and cardiovascular fitness, the amount of missing values could have biased the results. However, adjusting for the impact of missing values with multiple imputations showed similar associations, although it did not reach statistical significance for low muscle strength. The study population is homogeneous, with a majority of Caucasian males at the age of 18, which confines the generalizability to other populations, and to women. A draw back of the present study is that ALS cases may be under ascertained, since outpatient hospital data, which are likely more sensitive than inpatient data, are only available for 2001–2014. Before 1987, a case had to be diagnosed with ALS, die from it, and have ALS recorded on their death certificate to appear in the Cause of Death Register. There is a lag between diagnosis and death (probably 5–10% of ALS patients survive more than 10 years after diagnosis), so we may not have captured all ALS cases. In addition, a median age of 50.0 years for ALS cases is a little bit younger than what has typically been reported (i.e., 55 or 60 years) [1]. There are two likely explanations for this. First, men who signed for military service in the recent years were not old enough to develop ALS at the end of the observation period, and second, it has been shown that women have a later onset-age compared to men [49]. A median age at death of 43.5 years regarding the 12 cases captured solely in Cause of Death Register is about 20 years younger than what has typically been reported which was probably due to a selection of individuals with an early onset ALS. The median age at death was higher when all cases that died were included.

Our study design did not allow for the examination of youth metabolic factors that may be of importance for BMI and the development of ALS. Obesity and type 2 diabetes (not type 1) have been shown protective for ALS [21, 50]. Recent research shows that alterations in the carbohydrate, lipid, and apolipoprotein metabolisms are associated with ALS risk [51]. High glucose level was associated with a lower incidence, whereas high LDL-C/HDL-C and high apoB/apoA-I ratios were associated with a higher ALS incidence. Higher levels of LDL and apoB might affect the energy supply to the nervous system. Increased apoB has been associated with inflammation and oxidative stress—a common disease mechanism of ALS [52]. In addition, we were not able to study the development of certain exposures that may affect the risk of ALS such as smoking habits, diet, physical activity, exposure to toxins, and occupation during follow-up.