Amyotrophic lateral sclerosis (ALS) is a relatively rare neurodegenerative disease of unknown aetiology characterised by rapidly progressive paralysis and respiratory failure.1 While the typical phenotype of ALS has been well characterised over the years, broad assumptions about the clinical features of ALS have been generally based on relatively small cohorts of patients attending tertiary referral centres. Until recently, there has been a tendency to assume that ALS is a degenerative disease of the motor system that manifests as a number of differing phenotypes. The observation in some familial forms of ALS that the phenotype does not “breed true” has contributed to a view that phenotypic characterisation is of limited value.2

However, with the increasing recognition of phenotypic heterogeneity in ALS, including extramotor manifestations,3 and with the recognition of overlap between ALS and other neurodegenerative diseases in some kindreds,4^–6 the importance of descriptive epidemiology and detailed phenotypic characterisation has re-emerged. It is increasingly evident that accurate phenotype description will form the basis of future genetic association studies designed to identify both susceptibility and protective factors for ALS.7

The low incidence of ALS in the general population (1.5 to 2.5 per 100 000 persons per year) hampers the collection of epidemiological data from large numbers of patients.8 Studies based in tertiary referral centres are inherently biased, and in the case of rare diseases such as ALS, can lead to erroneous conclusions. In ALS, a series of population based registers,9^–12 mostly from European countries, have attempted to resolve the bias generated by cohort selection. These prospective ALS registers have provided detailed epidemiological and phenotypic data on relatively homogenous populations of approximately 4 million people each, and have provided important insights into the clinical progression of the disease.

The working hypotheses is that ALS, like other chronic neurodegenerative diseases such as Parkinson’s disease (PD) and Alzheimer’s disease (AD), is a complex genetic disorder. Identification of subpopulations that share environmental and/or genetic risks is likely to contribute to our understanding of disease pathogenesis, and ultimately lead to the development of new therapies. To achieve this, data on large, well characterised populations must be collected and analysed. However, each register is limited by the relative size of its base population (generating a maximum of 110 new patients per annum each).

The purpose of this paper is to explore the contributions and limitations of existing population based registers in ALS research, and outline a way forward for ALS epidemiology that will seek to progress our understanding of the disease and address the limitations of the previous investigations.

CONTRIBUTION OF EXISTING REGISTERS

Application and limitations of the El Escorial and Airlie House Criteria

The El Escorial Criteria (EEC) were established at an international consensus conference in 1990 to facilitate collaboration in clinical trials and other studies in ALS. According to these criteria, cases are classified into four categories (suspected, possible, probable and definite) based on the presence of upper and/or lower motor neuron signs in four defined CNS regions (bulbar, cervical, thoracic and lumbosacral).13

Diagnostic confirmation required that patients demonstrated symptom progression over a 6 month period and that ALS mimic syndromes (ie, conditions that can phenotypically resemble ALS) were excluded (table 1).14 15 When applied correctly, they have been generally accepted as being both sensitive and specific for ALS. The EEC have been widely employed in epidemiological and genetic research in ALS. However, the use of the EEC remains problematic. Rigid application of the EEC excludes at least 40% of ALS patients from enrolment in clinical trials at presentation, and up to 10% of patients at the time of death.16 This is because the EEC are sensitive to the “extent of disease” rather than to the “burden of disease” and do not discriminate between spinal disease and “bulbar” disease. As bulbar onset disease is associated with a much reduced life expectancy, patients who present with a bulbar phenotype may never fulfil the EEC for “definite ALS”. For this reason, the EEC are not particularly useful as a prognostic device, as evidenced by two separate population based studies,16 17 although it must be acknowledged that this was not the purpose of the classification.

A study of the reliability of the EEC in the clinical setting revealed a low concordance (κ = 18.5) for categorising ALS patients among independent observers. Agreement improved after training (κ = 68.3) and was better when the evaluation was based on direct clinical examination (κ = 70.8) rather than on medical record review (κ = 13.7).18 In a second study based on the Scottish ALS Register, the overall reliability of the EEC criteria was much better (weighted κ = 0.78) but the expert raters were working in only one centre and had extensive experience in conducting clinical and epidemiological research in ALS.19 These studies highlight the need for careful training of personnel, particularly in the setting of multicentre research. Interestingly, no such training has been performed in recent clinical trials.18

The criteria were revised in 1998 (known as the Airlie House Criteria) to include the addition of EMG as a diagnostic tool, the addition of a “laboratory supported” category and the deletion of the “suspected ALS” category.20 Formal validation of the Airlie House Criteria has never been undertaken. However, a number of population based studies have shown that the new criteria are no better than the EEC in selecting patients with a “restricted” form of ALS for clinical trials.16 17 Furthermore, deletion of the “suspected” category has caused difficulties in determining the status of patients with involvement of a single limb, or patients with pure lower motor neuron disease.16 To date, the Airlie House Criteria have not been widely adopted by ALS researchers and most therapeutic trials continue to use the original EEC to determine patient eligibility.

In summary, the unquestioning adoption of the EEC as diagnostic criteria for ALS, in the absence of proper validation studies, may have been an error. Furthermore, population based epidemiological studies have shown that the criteria may be excessively restrictive in capturing all patients with ALS, particularly in the context of clinical trial enrolment, and that their reliability needs definitive clarification. Further refining of clinical criteria of ALS should be undertaken in the context of these considerations, and should utilise population based registers rather than tertiary referral cohorts of patients.

Incidence and prevalence data

Age and gender specific incidence rates published by the European registers show that the incidence of ALS increases after the age of 40 years, reaching a peak in the late sixties or early seventies, followed by a rapid decline.9^–12 The pattern is consistent across the four published European registries (table 2) with only a slight difference in the age peak (fig 1). Only one study from Rochester, Minnesota, USA, described a continuous rise in incidence with age over the period 1925–1977. This report was, however, based on a small number of patients (n = 44)21 and a more recent updated study from the same area, including data up to 1998, indicates a decline in ALS incidence after the age of 60 years.22

• Download figure
• Open in new tab
• Download powerpoint

Figure 1 (A) Age specific incidence of amyotrophic lateral sclerosis for males in four European population based registries. (B) Age specific incidence for females in four European population based registries.

The pattern of incidence among the elderly may have more than one explanation. The apparent rapid decline of ALS incidence after age 75 years may arise from problems with case ascertainment among this age group due to the difficulty in diagnosing ALS among the elderly because of comorbidity, difficult access to specialised care or a more rapid and aggressive disease that cause elderly patients to die before the diagnosis of ALS is firmly established. Alternatively, the pattern of ALS incidence in the very elderly may also be dependent on aetiology. If ALS genuinely decreases in the very elderly, it might imply exposure to an environmental toxin with a long latency between exposure and disease onset, as possibly indicated by a case control study from Washington state.23 An alternative explanation is that people who survive beyond the age of 80 years are protected against motor neuron damage either because that individual is genetically unsusceptible to the causative toxin or the exposure occurred after a critical age.24 25 In other words, the pool of individuals susceptible to developing ALS have already succumbed to the disease by their eighth decade of life and what remains in the population is a pool of unsusceptible individuals. This pattern of variable susceptibility to a neurodegenerative disease has been observed in AD, where about 30% of individuals do not develop AD over a reasonable life expectancy.26

Few studies have comprehensively analysed the temporal trends in ALS incidence. Data from Norway show that ALS mortality has increased substantially among the elderly over the past decades.27 In contrast, national US mortality data show a small overall increase in the death rate between 1979 and 1983, with a subsequent plateau.28 There is only one prospective, population based study that examined ALS incidence over time. This study in Rochester, Minnesota, USA, showed the incidence of ALS to be constant at 1.7 per 100 000 person-years between 1925 and 1998.22 Increasing ALS incidence over time has been reported by other non-population based studies,29 but this is almost certainly because of improvements in case ascertainment and diagnostic methods rather than a genuine increase in incidence.

Many studies on ALS, including clinical trials, are based on prevalent cases. A recent study conducted in Ireland30 suggests that prevalence cohorts have different characteristics compared with incident cohorts, with fewer patients with bulbar onset and more subjects with age younger than 65 years (see page 30). Therefore, prevalent cohorts select subjects with a better prognosis. This issue has important implications both for collection of DNA for genetic studies and for the recruitment of patients for clinical trials.

Gender distribution in ALS

Non-population based studies of ALS published before 1990 consistently report a 2:1 male to female gender ratio.8 29 31^–43 More recent studies from the European population based prospective registers report a change in the gender ratio that is clearly approaching unity (M:F ratio 1.3:1 in Piemonte and Ireland).10 12 Data from Rochester also show a decreasing male to female ratio over a 73 year period (ranging from 1.2 in the period 1925–1984 to 1.0 in 1998).22 A review of 16 published studies looking at male/female ratio confirmed the presence of a significant trend towards one in most recent studies (fig 2).44

• Download figure
• Open in new tab
• Download powerpoint

Figure 2 Male to female ratio of the incidence of amyotrophic lateral sclerosis by mid point in the study period, age adjusted for 1990 US population aged 65–74 years (Spearmann R, p = 0.05).

There are several possible explanations for the increasing incidence of ALS among women. From a methodological perspective, the apparent increase may reflect better ascertainment among women compared with previous time periods. Alternatively, a genuine alteration in incidence may be a consequence of a change in prevalence of risk factors across genders. Socioeconomic changes during the past century mean that women are increasingly exposed to the same occupational and environmental risk factors as men. For example, two case control studies have reported an increased risk for ALS among smokers,45 46 and the increased prevalence of smoking among women in the past decades could partially explain the increase in ALS incidence among women. This hypothesis remains to be verified in a larger case control study.

Prognostic indicators

Only population based prospective studies can adequately determine survival of ALS and assess prognostic factors.16 17 47 Strikingly similar survival and prognostic data have been published from three population based studies, two in Italy and one in Ireland. The cumulative probability of surviving after diagnosis was 78% at 12 months, 56% at 24 months and 32% at 48 months. Population based median survival times in Europe were 19 months from first symptoms (range 16–31) and 30 months from diagnosis (range 27–39). Negative prognostic indicators included site of disease onset, older age and progression rate of respiratory, bulbar and lower limb symptoms.

Respiratory decline has been traditionally measured using forced vital capacity.48 However, this measure is not suitable for patients with bulbar onset disease, and although a reliable indicator of survival, is inferior to sniff nasal inspiratory pressure as a prognostic indicator.49

In summary, prognostic indicators are best identified from population based studies, as the data are least likely to be affected by bias. The observation that early disease progression is indicative of overall prognosis is of specific clinical relevance in discussing outcome with individual patients. In addition, sniff nasal inspiratory pressure should be incorporated as an adjunct standard measure of respiratory dysfunction in ALS.

Finally, the general consensus is that ALS prognosis is improving, but data from two population based studies seem to show a different picture. A recent report from the Scottish Registry described a decline in survival after ALS diagnosis (2.4 months) between 1989 and 199850 despite the introduction of riluzole and percutaneous endoscopic gastrostomy. No increase in suicide or change in practice in the last period of life seemed to emerge from these data. Interestingly, another report from Rochester in the US showed no improvement in survival comparing survival before and after 1990.22 These intriguing data on prognosis need to be confirmed, and possible causes of this trend should be explored.

Benefits of tertiary care management

As ALS is a relatively rare condition, it is most efficiently managed in a tertiary referral centre. Specialist multidisciplinary clinics provide tertiary services to patients with ALS. These clinics comprise a wide range of clinical professional specialists with expertise in ALS. Ideally, such clinics provide both diagnostic and management services, and facilitate continuity of care by liaising with the primary care clinician and community based services.51 52 However, patients who attend multidisciplinary clinics tend to be younger and to have longer duration illness,52 53 and accordingly “clinic cohorts” are not considered representative of the general population of ALS. Additionally, familial ALS tends to be over represented in tertiary clinic populations,52 53 and comparison between clinic based cohorts and population based cohorts of patients have confirmed a referral bias.52 53 Outcome data from tertiary clinics must thus be interpreted with caution.

Two population based studies have demonstrated an independent survival benefit for patients who attend multidisciplinary ALS clinics. This benefit is independent of other prognostic factors including age, disease duration, bulbar onset disease and rate of progression.52 54 Additionally, one study has shown that patients attending a multidisciplinary clinic have fewer hospital admissions and shorter durations of stay than those who attend general clinics.54 Increased use of non-invasive ventilation, attention to nutrition and earlier referral to palliative referral services are likely to contribute to the increased survival of those attending multidisciplinary clinics.52 54 These observations, which are only possible by the interrogation of population based registers, provide robust support for increased investment in multidisciplinary clinics for ALS in countries with nationalised health services.

Benefits of riluzole

Riluzole is the only treatment known to be effective in slowing disease progression in ALS. Outcome data from clinical trials indicated a marginal benefit in selected patients with “probable” or “definite” ALS by the EEC. Inclusion criteria for the riluzole trials were stringent, with a high cut-off rate for respiratory dysfunction. However, as demonstrated by two population based studies, many patients with ALS do not fulfil the criteria for inclusion in clinical trials until late in the course of their disease, and up to 10% of patients die without ever having fulfilled the criteria.16 Two population based, phase IV studies have demonstrated the efficacy of riluzole, although these studies were based on retrospective data and should be interpreted cautiously.55 56 Survival in patients treated with riluzole was similar in both studies, and confirms the outcome of the more stringently controlled clinical trials in all patients, including those with “possible” ALS who were excluded from the original clinical trials. An additional improvement in survival has recently been shown in patients who underwent percutaneous endoscopic gastrostomy and were taking riluzole.57

The use of population based studies to determine the clinical effectiveness of riluzole is particularly relevant when national health authorities debate the funding of relatively costly medications. Despite this, there has been a surprising lack of effort to develop further insights into the clinical use and efficacy of riluzole, either from the pharmaceutical industry or among academicians.

LIMITATIONS OF EXISTING POPULATION BASED STUDIES

Ascertainment problems within populations

A population based registry is designed to detect all cases of a specific disease in a population defined by geographic boundaries, and this methodology has contributed significantly to our clinical knowledge of rare conditions, such as ALS. The key design feature of a registry is its use of multiple sources of information on ALS patients to minimise the possibility that ascertainment will be biased according to the clinical practice of survey centres. The ability of this multifaceted approach to identify all ALS cases in a population has been confirmed by data from Ireland and Scotland with the use of the capture–recapture method.58 However, it remains possible that some groups of patients, such as the extremely old, are missed by the existing surveillance system.

Despite their success, several limitations of population based design are recognised. The reliability of the EEC criteria, which are used to determine eligibility for enrolment on the registers, is a function of the expertise and training of those who collect the data.18 19 Reviewers of medical records must have been adequately trained and validated in category designation. Furthermore, comparison across populations requires a validation process to ensure inter-rater reliability.

However, the greatest concern in patient ascertainment is the introduction of new data protection laws within Europe which threaten to impede data collection. In some countries, written informed consent must now be obtained from all individuals included on the registry, and the exact purpose of the registry must be stated and registered with a national data protection agency. Accessing files from multiple different agencies to permit “capture–recapture” analysis requires ethical submissions to multiple authorities, which is time consuming and expensive.

2.2 Risk identification in ALS: are registers useful?

Environmental and lifestyle risk factors in ALS

A large number of putative environmental and lifestyle risk factors have been suggested in ALS,59 including heavy manual labour, smoking and previous trauma.36 43 46 60^–72 Only one possible protective risk factor, vitamin E, has been identified to date.64 Unfortunately, the majority of analytical epidemiological ALS research is inconclusive because of poor study design, lack of replication studies and the enrolment of insufficient numbers to achieve sufficient power. With the exception of the studies of Italian football players (relative risk of 6.5),63 the relative risk identified in most studies has been less than 2, and large numbers are required to identify a weak effect.

Genetic risk

Studies of familial aggregation and twin studies are powerful techniques to determine whether a disease is genetic or environmental in nature. Such studies have been widely used in other neurodegenerative diseases, such as AD and PD,73^–76 but there are relatively few genetic epidemiological studies in ALS. A non-population based study conducted on a clinical series of 151 mainly sporadic ALS cases examined the association of ALS with PD and dementia.5 The risk of dementia was significantly higher in the relatives of ALS patients than in those of controls (relative risk (RR) 1.9; 95% CI 1.1 to 3.1). The risk of PD was high in relatives of patients with familial ALS (RR 5.6; 95% CI 0.6 to 50.3) and with sporadic ALS (RR 1.8; 95% CI 0.5 to 6.0), but did not reach statistical significance. These findings suggest a shared genetic susceptibility of ALS and dementia and perhaps PD. Similarly, a study in Sweden based on 92 cases and 372 controls found an increased risk for ALS in the presence of an association of positive family history for neurodegenerative disorders, exposure to metals and male gender.37 A large population based study is required to confirm this finding.

Similarly, twin studies are a well established genetic epidemiological technique employed to determine the relative contribution of genetics and environment to the cause of a disease.74 76 The only twin study in ALS (based on death certificates from the UK) found the hereditability of ALS to be between 0.38 and 0.85, supporting the hypothesis that sporadic ALS is genetic in aetiology.77 To date, this study has not been replicated.

Are ALS and frontotemporal dementia related diseases?

Recent clinical, neuropathological and genetic data support the notion that ALS and frontotemporal dementia (FTD) form part of a spectrum of disease. Approximately 5% of patients with ALS have clinically florid dementia78 and at least half of patients with “classical” ALS have subtle frontal and temporal lobe impairment termed “mild cognitive impairment.79^–81 Both ALS and FTD patients have ubiquitin inclusions containing the newly identified TDP43 protein.82 Recently, a new linkage to chromosome 9p has been reported which may replace an earlier reported linkage to chromosome 9q.83

Reliable data on the frequency of ALS, FTD and the overlap between these two syndromes within the same population based setting are missing. A large prospective, population based study will help delineate the nature and frequency of frontal involvement in ALS and may expand our definition of the clinical ALS syndrome.

CONCLUSION

The emergence of population based datasets over the past 10 years has contributed substantially to the clinical knowledge of ALS. However, the limitations of small population based registries are all too apparent. Large, well characterised populations are required to provide adequately powered data on environmental and genetic susceptibility. The EURALS consortium, which was established to provide a mechanism for large scale descriptive epidemiology, draws from a population of over 25 million Europeans. It is the largest ever population based dataset of ALS and will have the capability of conducting large scale case controlled studies of sufficient power on both environmental and genetic risk factors to resolve many of the existing controversies in ALS research.