Despite the large number of studies on the natural history of amyotrophic lateral sclerosis (ALS), most of the data have been based on large clinical series enrolled in ALS referral centres1^–11; these are selected groups of patients and the collection of data is largely retrospective. Recent population based series based on the presence of registries in a defined geographical area12^–18 have shown a broader range of clinical phenotypes and have prospective collection of information both on prognostic indicators and outcome; findings from these studies are more likely to be representative of the entire clinical spectrum of the disease. Most of the studies have been conducted before the introduction of standardised criteria for diagnosis of ALS (El Escorial19 and Airlie House Criteria20) and it is still not clear if the El Escorial categories have any prognostic value.

The aim of this study was to describe the survival of patients from a population based cohort of ALS incident cases from Southern Italy; to evaluate the prognostic role of selected clinical indicators; and, in particular, the assess the role of El Escorial categories at diagnosis.

PATIENTS AND METHODS

Study population

A prospective registry of all newly diagnosed ALS cases with several sources of information was established in Puglia, Southern Italy, in 1997 (reference population 4 025 329 subjects).21

This multisource system allowed us to identify all newly diagnosed ALS cases resident in Puglia in the 2 year period from 1 January 1998 to 31 December 1999.

Diagnostic criteria

El Escorial criteria19 and their Airlie House revised version were the diagnostic criteria used.20 Individuals younger than 18 years were excluded to avoid misclassification of other motoneuron diseases of genetic origin. Subjects with spinal onset were also classified based on the presence of exclusively or predominantly upper motoneuron (UMN) signs at the time of diagnosis (considering the extent of involvement of the UMN and lower motoneurons (LMN) and the number of affected regions).

Data collection and follow-up

All patients gave consent for the study. Ad hoc forms were used by the investigators to register and follow-up all eligible patients. Records were taken of the main demographic and clinical findings (date of onset of symptoms, date of diagnosis, type of onset and presence of UMN and LMN signs in the four regions indicated by the El Escorial criteria). Data were stored on a security protected computerised database, with separated anonymous files.

All patients were routinely followed-up during the course of their illness, on average every 6 months, by direct examination of the patients or by telephone conversations. Therefore, detailed information concerning the progression, medications and treatment provided to patients (eg, riluzole, non-invasive ventilation (NIV), percutaneous endoscopic gastrostomy (PEG) and tracheostomy) were collected. Date and cause of death were recorded.

Vital status was investigated using: (1) the demographic service of the town of residency of each patient; (2) death certificates obtained from the National Death Data Base Registry.

Date of the last follow-up for this study was 30 June 2004. Vital status (alive or dead) was checked at the censoring date for all patients in the study.

Statistical analysis

Univariate comparison of demographic and clinical variables used either the t test or χ2 test. Survival curves were estimated by the Kaplan–Meier method and differences in survival were measured by the log rank test.22 Patients who underwent tracheostomy were coded as deceased on the date of the tracheostomy. In the primary analysis, survival time was measured from time of diagnosis. This was done because classification of patients based on El Escorial criteria (considered one of the main variables of interest) was done on the date of diagnosis. In a secondary analysis, survival time was measured from onset of symptoms. Multivariate analyses of the risk for death associated with selected independent variables were performed using the Cox proportional hazards model.22 The following variables were considered in the analysis: age (both as a continuous and categorical variable: ⩽45, 45–65, 65–75, >75 years), gender, time interval from first symptoms to ALS diagnosis (ODI), localisation of symptoms at onset, El Escorial category at entry, riluzole use, PEG and NIV insertion, presence of predominant UMN signs and management by multidisciplinary clinics for ALS. Riluzole therapy, PEG and NIV were included as time dependent variables in the model. Bulbar onset ALS also included patients with generalised (bulbar and spinal) onset. Significance was tested at the 5% level. No missing values were present for each variable included in the model. All statistical analyses were done using Stata software, version 9.0.23

RESULTS

During the 2 year surveillance period, we identified 130 ALS patients (81 males, 49 females). Sixty-five per cent of cases (n = 84) attended the four multidisciplinary ALS clinics in the area of the study and 35% (n = 46) attended the general neurology clinics. Four of the original 130 patients were lost to follow-up.

Median age at diagnosis was 65.2 years (range 19–80); ALS was diagnosed before the age of 45 years in 11 cases (9%). Seventy-four per cent of cases (n = 93) had spinal onset disease, 19% (n = 24) had bulbar onset and 7% (n = 9) generalised onset ALS. The proportion of bulbar onset progressively increased with advancing age in both men and women and resulted in the most common site of onset among subjects aged >75 years (58%; p = 0.006). At the time of diagnosis, bulbar signs were present in 78% of cases (n = 99).

Twenty-eight cases had definite ALS, 43 probable ALS, 44 possible ALS and 11 suspected ALS at the time of diagnosis. Of 44 possible cases, 29 (18 males and 11 females) were reclassified after EMG as probable laboratory supported ALS. Fourteen cases (11%) presented at diagnosis with clinical signs characterised by predominant involvement of UMN. In this population based series, only two cases (1.5%) had familial ALS.

Median ODI was 9.3 months. This interval was shorter (7.4 months; range 3.8–49.2) for subjects with bulbar onset compared with subjects with limb onset (10 months; range 1–70.7) ALS and was significantly longer in patients aged 65–75 years (12 months) compared with the other age groups (t test 12.5; p = 0.006). No difference was observed in ODI between cases with predominant UMN involvement (9.9 months; range 3.6–68.8) and cases with LMN features (9.4; range 1–70.7).

Median survival time from symptoms onset was 27.5 months overall (range 6.6–97.8). Survival rates at 1, 2 and 4 years from symptom onset were 95% (95% CI 89.9 to 98.2%), 73% (95% CI 64.4 to 80.5%) and 41.3 (95% CI 32.6 to 50.4%).

Median survival time from diagnosis was 15.7 months overall (range 0.3–46.9); survival rates at 1, 2 and 4 years from diagnosis were 71.4% (95% CI 62.7 to 79.1%) 44.4% (95% CI 35.6 to 53.6%) and 27% (95% CI 19.5 to 35.6%).

During the follow-up period, riluzole was prescribed in 73 cases (58%) whereas eight patients (6%) underwent PEG, three (2.5%) NIV and 13 (10%) tracheostomy.

Outcome was significantly influenced by age at diagnosis; survival rate at 4 years was 73% for patients aged 45 years or less whereas no patient aged >75 years survived longer than 36.7 months from diagnosis (log rank 19.4; p = 0.0002) (fig 1). Bulbar and generalised site of symptom onset (4 year survival rates 8% and 0%, compared with 34% of cases with spinal onset; log rank 14.4; p = 0.0007) (fig 1) and the involvement of the bulbar region at the time of diagnosis (18% vs 59% of patients without any bulbar signs; log rank 11.8; p = 0.0006) were also associated with a worse prognosis (table 1).

• Download figure
• Open in new tab
• Download powerpoint

Figure 1 Cumulative survival of amyotrophic lateral sclerosis (ALS) from diagnosis by different prognostic indicators: age (A), presence of bulbar signs at diagnosis (B), site of symptom onset (C) and El Escorial category at diagnosis (D).

No significant differences were found in survival rates and in median survival time from diagnosis after stratifying patients according to the median value of ODI (⩽9 months 31% vs 23% of cases with ODI >9 months; log rank 0.5; p = 0.5) and gender (males 29% vs females 23%; log rank 0.06; p = 0.8) (table 1).

Analysis of survival to death of both El Escorial and Airlie House Criteria categories at diagnosis revealed that clinical features at diagnosis of definite or suspected ALS were associated with poorer survival (log rank 14.4; p = 0.002) (fig 1). In this series, suspected ALS cases were characterised by a significantly higher 4 year mortality rate (91%) and a shorter survival (15.3 months) compared with possible ALS (57%; 18.1 months) and probable ALS (77%; 18.8 months) (table 1) and similar to those of definite ALS (86%; 15.4 months). The results were similar after exclusion of suspect ALS cases that did not developed UMN signs during the course of their illness. During the follow-up, six of 11 suspect ALS cases became probable or definite ALS within 1 year while the remaining patients (5% of total sample) died with only LMN signs within 3 years from the diagnosis. When we looked at Airlie House Criteria categories, the new category of probable, laboratory supported ALS presented with survival rates (38%) similar to those of probable (23%) and possible (53%; log rank 1.1; p = 0.3) ALS (table 1).

We examined the impact on survival of management by multidisciplinary clinics.24 No differences were found in median survival times (15.7 months (range 0.8–46.9) vs 16.5 (range 0.3–45.1)) and 4 year survival rates of patients who received their care at ALS multidisciplinary clinics (26%) compared with patients followed-up by general neurology clinics (29%), regardless of the site of symptom onset (log rank 0.1; p = 0.7). Finally, cases with predominant UMN involvement (11%; n = 14) presented a longer median survival time from onset (45.2 months) and from diagnosis (23.8), with better survival (43% vs 25%) compared with patients with predominant LMN features (27.1 months from onset and 14.1 months from diagnosis), and the difference was borderline significant (log rank 2.8; p = 0.09) (table 1).

Multivariate analysis confirmed that advanced age and bulbar or generalised onset of symptoms were the main factors to influence the prognosis (table 2).

We also built up similar multivariate models after stratifying patients according to site of first symptoms (bulbar vs spinal) to assess if age and site of onset were independently related to outcome. We found that age at diagnosis was significantly related to poor survival outcome among spinal ALS (hazard ratio (HR) for patients aged >75 years compared with patients aged 45 years or less: 11 (95% CI 1.5 to 78.5); p = 0.01) whereas it did not influence survival among bulbar ALS (for patients aged >75 years compared with patients aged 45 years or less: HR 4.5 (95% CI 0.4 to 46.5); p = 0.2). Among cases with spinal onset, clinical features characterised by predominant UMN involvement were associated with a borderline increased survival (HR 0.5 (95% CI 0.2 to 1.3); p = 0.1).

Finally, we repeated the analysis of survival from symptom onset and the results were similar.

DISCUSSION

In this population based prospective study of newly diagnosed ALS cases, we found that advanced age and bulbar onset of symptoms are indicators of poor prognosis and have independent effects on survival. El Escorial diagnostic category at the time of diagnosis did not predict survival. Spinal onset ALS with predominant UMN involvement presented a trend towards better prognosis.

Median survival time was 28 months from first symptoms and 16 months from ALS diagnosis, with a 4 year survival rate after diagnosis of approximately 30%. These results are similar to those reported in other previous population based studies12^–16 and confirm that the median survival times from population based studies are definitively shorter than in recent clinical series.25 Tertiary centres probably select subjects with better prognosis because of age and clinical characteristics.3 15

Onset of disease at a younger age was associated with a better prognosis, as more than 70% of ALS patients aged ⩽45 years survived after 4 years from diagnosis. This finding is consistent with a recent study carried out in a sample of long survival ALS (more than 10 years) from the King’s Database25 and could be related to a different phenotype of the disease in this age group, with a lower prevalence of bulbar ALS. Indeed, no bulbar onset ALS was diagnosed before the age of 45 years in this case series.

Conversely, in keeping with other population based studies, older age at diagnosis,12^–16 bulbar onset of symptoms and presence of bulbar signs at diagnosis12^–15 were the main predictors of shorter survival.

Consistent with other studies,1^–16 advanced age at diagnosis was the most reliable predictor of adverse survival in our case series. No patient aged ⩾75 years survived longer than 36 months. Several factors may possibly coincide to worsen the prognosis in the oldest group of patients, including age related motoneuron depletion,1 7 higher frequency of bulbar ALS,4 contemporary presence of other conditions (such as pulmonary disease) that can further reduce survival14 and a different, less intensive process of care.1 14

Previous population based studies,12^–15 including ours, indicated that bulbar onset of symptoms and any bulbar features at diagnosis were associated with a shorter survival. Because older patients are more likely to have a bulbar onset of disease, some authors1 hypothesised that the prognostic effect of bulbar onset may be explained by the confounding effect of age; however, in this case series, we considered patients separately according to site of first symptoms, to test this hypothesis, and we found no effect of age among bulbar ALS. This indicates that bulbar onset of symptoms was, per se, an independent risk factor for poor survival.

In previous population based studies, some authors have reported worse survival in women3 14 and in cases with a shorter time to diagnosis,14 a surrogate of a more aggressive course of the illness in these subgroups of patients. In the present study, neither gender nor shorter time to diagnosis was a significant predictor of worse survival when all of the other factors were considered.

Three population based studies have evaluated the prognostic role of El Escorial categories and their results are conflicting12 15 16; one study12 found that El Escorial status at diagnosis was directly related to the rate of disease progression, as the category of definite ALS showed a shorter survival compared with other categories, and another16 observed that being classified as definite ALS at diagnosis was a predictor of poor survival. The Irish study15 evidenced no difference in survival among El Escorial and Airlie House categories; clinical features of probable and definite ALS were compatible, with 10 year survival, as observed in a recent clinical based study.26 In our cohort, we found that survival of suspected ALS was similar to that of definite ALS, even after exclusion of suspected cases that did not develop UMN signs during the course of the disease. These results are in agreement with those of the Irish study15 and indicate that El Escorial and Airlie House categories are a poor prognostic indicator and rather reflect the clinical heterogeneity of the disease at presentation. Our data indicate that the susceptibility of separate anatomical systems to degeneration, more than the extent of involvement of different systems in different body regions, may be relevant for prognosis. This is consistent with recent neuropathological data that indicate that there is a clinicopathological continuum ranging from slowly progressive PMA to “typical” ALS.27 However, no firm conclusions on differences among El Escorial categories can be drawn, as our findings are based on small numbers in each subgroup.

Finally, in agreement with a recent report on ALS long term survival,26 we found that in patients with spinal onset of symptoms, the predominance of UMN signs was associated with a trend towards better survival, even if the result was only borderline significant. In another clinical study, UMN dominant ALS, defined by predominantly UMN disease with minor LMN signs, was found to present a disability similar to ALS, but with a slower progression.28 These findings need to be further replicated in larger studies.

The main strength of this study is the completeness of the follow-up; in our study, only 3% of patients were lost to follow-up compared with 22% in the best clinical series.26 The main limitation of our population based study, as in others, is power, because of the relatively limited sample size. However, in contrast with clinical based studies, population based studies have similar survival curves,15 based on a well defined reference population (in our case all of Puglia with approximately 8 million person-years). Therefore, we have a likely model of the natural history of the disease, when both case and outcome ascertainment is complete and loss to follow-up is minimised and not selective.