Background Amyotrophic lateral sclerosis (ALS) appears to be a sporadic disorder in 95% of cases. Although few personal characteristics associated with developing ALS are known, identification of those at risk is essential to any vision of early intervention. There is persistent anecdotal observation that those with ALS are premorbidly physically ‘fitter’, although such observations are susceptible to bias. Hospital admission for coronary heart disease (CHD) might serve as an objective marker of reduced cardiovascular fitness.
Methods A record linkage study of two large databases of hospital admissions, the Oxford Record Linkage Study (ORLS) and an English national record linkage dataset of Hospital Episode Statistics was undertaken. The ratio of the rate of ALS in people without a record of CHD to that in those with a record of CHD was calculated, factoring out premature death in both cohorts. Similar analysis for Parkinson's disease (PD) and multiple sclerosis (MS) was undertaken.
Results In the English population, the rate ratio for ALS in the non-CHD cohort, compared with the CHD cohort, was 1.14 (95% CI 1.05 to 1.22); for PD it was 0.95 (95% CI 0.93 to 0.98); and for MS 0.95 (95% CI 0.88 to 1.04). The ORLS data yielded similar findings.
Conclusions Those without a record of CHD were at modestly higher risk of ALS, but not for PD or MS. This lends support to the assertion that ALS arises within a population who may have relatively higher levels of cardiovascular fitness.
Statistics from Altmetric.com
Only 5% of amyotrophic lateral sclerosis (ALS) cases are associated with apparent Mendelian inheritance, although what appears to be an essentially sporadic disease in the other 95% may be explained by a more complex susceptibility profile conferred by multiple gene variants.1 A major challenge for ALS research is to identify patients at a sufficiently early stage of the disease when neuroprotective therapies are more likely to be effective. To this end a greater understanding of the population at risk of ALS would be a major advance.
Relative ‘physical fitness’ in those diagnosed with ALS is a persistent anecdotal observation, noted by both patients and their clinicians. Although unconfirmed in some studies,2 a high level of prior athleticism found in ALS patients specifically raised the possibility that, rather than the disease being a result of physical exercise per se, ‘being slim and athletic might be a phenotypic expression of genetic susceptibility to ALS, mediated by some environmental agent’.3 In this model, the observation of increased risk of ALS among footballers,4 military personnel5 and ‘blue collar’ employment6 might simply reflect a higher proportion of people with athletic physique in such professions. Such a physique might reasonably be expected to be associated with a lower incidence of cardiovascular disease,7 and a beneficial vascular profile, including lower premorbid body mass index, has been demonstrated in ALS patients.8
We have previously used a hospital admission record linkage database to explore (and reject) another aetiological hypothesis in ALS—namely, that head injury and other trauma requiring hospitalisation might be a risk factor for ALS—using record linkage to avoid the issue of recall bias.9 To exploit the power of this resource as a way to study prior physical fitness in those subsequently developing ALS, the same methodology was applied to study the rate ratio of ALS in people with and without a prior record of coronary heart disease (CHD). The prediction was that ALS would be more common in those without a prior record of CHD.
Populations and datasets
Three hospital admission datasets, from two record linkage systems, were used. The first system was the Oxford Record Linkage Study (ORLS),10 comprising admissions in the former Oxford National Health Service (NHS) region from 1963 to 1998 (ORLS1) and from 1999 to 2008 (ORLS2). The patient identifiers in the NHS were different before and after 1999 and the two datasets cannot be linked to one another, hence they were treated separately. The second system was that of English National Linked Hospital Episode Statistics11 spanning 1999–2008. The linked national file was built by staff of the ORLS. The datasets include brief statistical abstracts of records of all hospital admissions, including day cases, in UK NHS hospitals, and data derived from death certificates. All records for the same person are linked together. The datasets and the methods used have been described in detail elsewhere.12 ORLS1 commenced in a relatively small population around Oxford city and gradually expanded until, by 1987, it covered the four counties of the former Oxford NHS health region (population 2.5 million). The population of England is about 50 million.
A cohort without a record of CHD was built consisting of all those admitted to hospital with a wide range of non-cardiovascular conditions comprising mainly minor medical and surgical conditions. This cohort was used as a ‘reference cohort’, presumed to approximate the general population in its risk of major diseases.9 10 12 Referred to as the ‘non-CHD’ cohort, it comprised common conditions that, both individually and in combination, are thought very unlikely to be associated with either an atypically high or low risk of ALS. Each person's first recorded admission for ALS in the cohort without CHD, or in the CHD cohort, was identified and the dataset was then searched for any subsequent hospital admission for, or death from, ALS more than a year after the reference or CHD admission. We omitted cases within a year (a common practice in such record linkage studies) to avoid any biases, such as surveillance bias, that might arise from a second disease (eg, ALS, Parkinson's disease (PD) and multiple sclerosis (MS)) being identified initially as a direct result of care for a first disease (eg, CHD).
Calculation of rates and rate ratios
The rates of subsequent ALS were calculated as follows. The ‘date of entry’ into each cohort was the date of first admission in the reference cohort, or for CHD. The ‘date of exit’ was the date of subsequent admission for ALS (if any occurred), or death, or the end of the data file, whichever was the earliest. Censoring for death meant that, for example, an individual in the CHD cohort with acute myocardial infarction, who died soon after onset, ceased to be included within the denominator as they were no longer ‘at risk’ of future ALS. Each such person was excluded from the analysis from the exact day of death; and, in this way, the analysis adjusted very precisely for differential mortality in the CHD and non-CHD cohorts.
In the ORLS study, in comparing the reference with the CHD cohort, the rates of subsequent ALS were calculated, stratified and standardised by age at entry (in 5 year age groups), sex, calendar year of first recorded admission and district of residence, using the indirect method of standardisation and taking the combined reference and CHD cohorts as the standard population. The stratum-specific rates from the standard population were applied to the reference cohort, and to the CHD cohort, in order to obtain the ‘expected’ number of cases of ALS in each individual cohort based on the stratum-specific rates in the two cohorts combined. The expected numbers were compared with the observed numbers in each cohort, and the rate ratio was calculated as the (observed ÷ expected numbers in the non-CHD cohort) divided by the (observed ÷ expected numbers in the CHD cohort). This follows the methods described in detail by Breslow and Day.13 The same methods were used in the English cohort except that health region (rather than health district) was stratified and standardised, as were quintiles of an index of socioeconomic deprivation (not available in the ORLS), age, sex and year of admission. All those individuals in each dataset with the reference conditions, and all those with CHD, were included to maximise the numbers in the study.
An additional analysis was performed excluding those cases of ALS within 5 years of the diagnosis of CHD, to study a ‘survivor’ population and reduce further any concern that those in the CHD population might die prior to the mean age of onset of ALS.
All procedures were repeated using PD and then MS as the outcome disease.
In the ORLS1 dataset there were 559 536 people in the reference ‘non-CHD’ cohort (205 developed ALS) and 128 899 people in the CHD cohort (53 developed ALS). In ORLS2, the corresponding figures were 308 905 (77 developed ALS) and 115 082 (40 developed ALS); and, in the English national cohort, there were 5 835 276 (1351 developed ALS) and 2 760 677 (1313 developed ALS). The observed and expected number of cases of ALS, PD and MS are shown for the CHD and non-CHD cohorts in table 1.
In ORLS1, the rate ratio for ALS in the non-CHD cohort, compared with ALS in the CHD cohort, was 1.32 (95% CI 0.97 to 1.82), in ORLS2 it was 1.57 (1.06 to 2.38) and in the English national dataset it was 1.14 (1.05 to 1.22). With the subsequent exclusion of cases of ALS <5 years after admission for CHD, the rate ratio for ALS in the non-CHD cohort remained elevated in the English national dataset at 1.23 (1.01 to 1.37, p=0.03) and also, although not significantly so given the much smaller numbers, in the ORLS1 at 1.23 (0.84 to 1.85, p=0.32) and ORLS2 at 1.69 (0.83 to 3.7, p=0.16).
In ORLS1, the rate ratio for PD in the non-CHD cohort, comparing its rate with that in the CHD cohort, was 1.14 (1.03 to 1.25), in ORLS2 it was 0.95 (0.84 to 1.09) and in the English national dataset it was 0.95 (0.93 to 0.98).
In ORLS1, the rate ratio for MS in the non-CHD cohort, comparing its rate with that in the CHD cohort, was 1.19 (0.83 to 1.75), in ORLS2 it was 1.05 (0.74 to 1.54) and in the English national dataset it was 0.95 (0.88 to 1.04).
This study demonstrated an increased rate ratio of ALS in those without a prior hospital record of CHD, within both a local and national dataset of hospital records. The analyses were stratified and standardised by age and other factors (see method) so that the comparisons between the non-CHD and CHD cohorts were equivalent in these respects. The three datasets are almost independent of each other. ORLS1 and the English national set are wholly independent. ORLS2 is the subset, for the former Oxford NHS region, from the English national set. The findings in each dataset corroborate those in the others. Analysis of a longer surviving population, restricted to those who survived at least 5 years after admission for CHD or reference cohort condition, provided similar results of an excess of ALS in people without a prior record of CHD. We undertook this additional analysis to further allay concern that those individuals with CHD might not survive long enough to be at risk of subsequent ALS. The size of the excess risk in the non-CHD cohort was fairly small, indicating that the excess risk associated with cardiovascular fitness is modest.
This elevation of risk in the non-CHD cohort was not consistently found for the outcomes of two ‘disease control’ populations—namely, PD (a neurodegenerative disorder affecting a similar age group to those with ALS) or MS (an autoimmune disorder affecting a significantly younger population). Although the rate ratio for PD was slightly high in the ORLS1 dataset, it was not in the two other datasets. In fact, in the national data the rate ratio (at 0.95) for PD was significantly lower in the non-CHD cohort. However, the numbers of observed and expected cases of PD were very large in this population and even a small difference, like this, is likely to be statistically significant. There was no significant difference in risk of MS comparing the non-CHD and CHD cohorts.
These results suggest that those who develop ALS have a lower premorbid likelihood of CHD which in turn provides evidence, albeit indirect, that ALS arises in a population with relatively greater cardiovascular fitness. More directed studies are needed however. One design might be a case control study in which people with ALS, and appropriate controls, are questioned about past history of physical activity, and also tested for objective markers of physical fitness not likely to have been affected by the onset of ALS. Another design might be a prospective study of cohorts of people with high and low levels of physical fitness, followed for the subsequent development of disease, including ALS. The latter design would, however, be very costly and time consuming, given the rarity of ALS.
Comorbidities in ALS
There is sparse literature concerning comorbidities among ALS patients, and in one case control study no difference was seen across a range of disease groups.13 With reference to vascular diseases, a large study of the natural history of ALS noted a 12% prior incidence of stroke or ischaemic changes on MRI and 9% incidence of concurrent diabetes (with a 42% family history of diabetes).14 Another such study considering vascular risk factors demonstrated that ALS patients had a more favourable lipid profile (with less use of cholesterol lowering medication) and lower body mass index.8 Hypolipidaemia was also observed in the presymptomatic stages of the transgenic superoxide dismutase-1 (SOD1) mouse model of ALS,15 and hyperlipidaemia associated with more favourable prognosis in established human ALS,16 although in another study only body mass index was an independent prognostic factor.17 Such circumstantial evidence adds to the growing view that the vascular profile may be an aspect of pathogenesis in ALS whether or not it is also a surrogate marker for any or all of greater premorbid fitness, physical activity or metabolic rate.
The concept that physical exercise itself might somehow be linked to the aetiology of ALS persists (reviewed by Harwood and colleagues18), and this might in theory be reflected in a reduced cardiovascular risk profile. Physical exercise has been previously associated with anterior horn cell pathology in poliomyelitis.19 More recently, phrenic nerve stimulation in a mouse model of ALS was associated with disease acceleration and spread to the adjacent forelimb.20 Handedness was found to be concordant with laterality in upper limb onset ALS21 although the authors pointed out that this might equally reflect a cortical vulnerability rather than simply use of the limb. It should be noted that physical exercise in established ALS may even be moderately beneficial.22 23 It has been inversely associated with the incidence of Alzheimer's disease,24 and no association between physical activity and the subsequent development of PD was found in a prospective study.25
Caveats to interpretation
Although an inherently indirect approach, the use of record linkage is a resource-efficient study method, compared with personal follow-up of patients in a prospective cohort study design, and it eliminates the issues of selection or recall bias found in case control studies based on interviews with patients. The limitations of using routinely collected administrative data are well recognised nonetheless. They include lack of information about the diagnostic characteristics that underpinned the clinical diagnoses transcribed from clinical case notes. However, any sources of error in diagnosing (or missing) ALS would apply similarly in the non-CHD and CHD cohorts, and should not be a source of bias. Migration of subjects into and out of the area covered by record linkage means that it is not possible to calculate absolute event rates. This is why an ‘internal’ comparison group was used—the reference cohort, drawn from within the record linkage dataset—and why relative risks in the form of the rate ratios are given. Rate ratios, comparing the non-CHD and CHD with the reference cohorts, were adjusted for age and socioeconomic status but not for other potential confounders, which include blood pressure, blood glucose, obesity, serum cholesterol, diet, psychosocial factors and family history. Smoking is also an established risk factor for CHD and potential confound. Analysis of select class II and III studies reported smoking to be a clear risk factor for ALS26 27 although a meta-analysis did not support this.28 The present study cannot take account of risk factors that do not directly result in hospital admission, and so it relies on the assumption that these are equally distributed between the cohorts. With respect to smoking, we also studied the effect of prior admission for chronic obstructive airways disease (also strongly linked to smoking) on subsequent ALS, using the same methodology and found no significant effect (data not shown). The possibility that one or more medications in the routine secondary prevention of further cardiovascular events (eg, antiplatelets or statins) might be somehow protective for ALS is also acknowledged but there is currently no published evidence to support this.
Therapy for ALS has not been promising to date. The failure of all but one therapeutic trial may in part reflect the remoteness of the onset of the pathological cascade initiating disease to the clinical horizon. If, ultimately, primary prevention is to be realised in ALS, then identifying modifiable risk factors is essential. If more certain evidence emerges that factors associated with fitness or athletic body structure are important (or perhaps a surrogate marker of a vulnerable motor system architecture), future focused studies of the genetic, molecular, biochemical and neuronal substrates of physical fitness may be revealing. A population with a genetic profile promoting physical fitness in early life would have an obvious evolutionary advantage but might harbour age related vulnerabilities unmasked by the dramatic increase in life expectancy in the past two centuries, leading to an excess risk of ALS. We conclude from our indirect observations that the broad hypothesis of athleticism as an association in those at risk of ALS is, at the very least, still worthy of pursuit.
This article conforms to STROBE guidelines.
Funding This work was supported by the Medical Research Council/Motor Neurone Disease Association Lady Edith Wolfson Fellowship (grant No G0701923) to MRT.
Competing interests None.
Provenance and peer review Not commissioned; externally peer reviewed.
Data sharing statement The authors adhere to the Medical Research Council policy on data sharing whereby data arising from MRC funded research should be made available to the scientific community with as few restrictions as possible.
If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.