Survival in frontotemporal dementia

Methods.

Sixty-one cases were selected from two neuropathologic series of patients with dementia in Sydney, Australia (n = 31) and Cambridge, England (n = 30). Both series have been collected as part of multidisciplinary research programs closely linked to specialist tertiary referral dementia clinics in Sydney and Cambridge. In both centers, every attempt was made to enroll all patients with early onset dementia and suspected FTD into the brain donor program with a 90% success rate for obtaining declarations of intent for brain tissue donation in life. Enrollment into the programs began in 1988 (Sydney) and 1990 (Cambridge) with the establishment of special interest clinics. Over a 10-year period (first brains donated in January 1992), similar numbers of cases were collected in the two centers (Sydney 125, Cambridge 105), reflecting the similar population bases. The research programs were approved by the Human Ethics Committees of the Universities of Sydney and New South Wales and the Addenbrooke’s Hospital Local Ethics Committee.

Results.

Of the total group (n = 61), 26 presented clinically with fvFTD, 9 with semantic dementia, 8 with PNFA, 9 with FTD-MND, and 9 with CBD. As shown in table 1, the age at onset of symptoms was not different between clinical subgroups (F[4,56] = 1.6, p > 0.5). For age at diagnosis there was, however, an overall difference (F[4,56] = 2.8, p < 0.05) with post hoc pairwise tests revealing that the PNFA group was older than two of the other groups: FTD-MND and fvFTD (p < 0.01). For age at institutionalization (F[4,45] = 2.6, p < 0.05) and age at death (F[4,56] = 5.3, p < 0.001) there were also intergroup differences with pairwise post hocs again showing differences between the extreme groups: FTD-MND (youngest) vs PNFA (oldest) at p < 0.01.

The total group contained more men (36) than women (25). The sexes were evenly represented in the fvFTD group, but in the FTD-MND, semantic dementia, and CBD subgroups, men predominated. The converse was true for the PNFA group. Overall, 20 of the 61 patients had a positive family history (33%), with a fairly even distribution across the clinical subtypes.

For the whole group (n = 61), the median survival from diagnosis was 3.0 years (95% CI 2.6 to 3.4 years). Seventy-five percent had died by 6.0 years. Taken from symptom onset, the median survival was 6.0 years (95% CI 4.9 to 7.1 years). The median time to institutionalization was 5.0 years (95% CI 4.5 to 5.5 years) from symptom onset but only 1.0 years (95% CI 0.2 to 1.8 years) from diagnosis.

Analysis of the effect of sex, presence of family history, and age at diagnosis (with a median split of above and below 60 years) showed no significant differences. These variables were not considered in subsequent analyses.

Analysis of survival to death according to the clinical diagnosis confirmed differences, both from symptom onset (χ² [df 4] = 29.85, p < 0.0001) and from diagnosis (χ² [df 4] = 20.0, p < 0.0005), between the clinical subtypes. Inspection of the survival curves (see figure 1) indicated the outlying groups were those with FTD-MND, who had the shortest survival, and those with PNFA, who lived the longest. Survival of the fvFTD, semantic dementia, and CBD groups appeared equivalent. We therefore repeated the Kaplan-Meier analyses excluding first the PNFA and then the FTD-MND groups. After removal of the PNFA group there was still an overall difference in survival between the remaining other four groups, both from symptom onset (χ² [df 3] = 23.1, p < 0.0001) and from diagnosis (χ² [df 3] = 15.6, p < 0.001). By contrast, after exclusion of the MND group there was no longer a significant intragroup difference (from symptom onset χ² [df 3] = 3.1; from diagnosis χ² [df 3] = 1.9). Thus, the main difference in survival was attributable in large part to the short survival of the FTD-MND patients. Table 2 shows the 25%, 50%, and 75% and mean (± standard errors) survival from symptom onset and diagnosis to death in the clinical subgroups: note the extremely short survival of the FTD-MND in contrast to the other variants.

• Download figure
• Open in new tab
• Download powerpoint

Figure 1. Kaplan-Meier survival plots for the clinical frontotemporal dementia (FTD) subgroups (▪ = FTD with motor neuron disease, ▴ = behavioral variant FTD, ♦ = semantic dementia, |b = corticobasal degeneration, □ = progressive nonfluent aphasia) from symptom onset (A) and from diagnosis (B).

We were also interested to examine the impact of pathologic diagnosis on survival. All patients with an in vivo diagnosis of FTD-MND had the pathologic hallmarks of this subtype (ubiquitin-positive inclusions in hippocampus or brainstem motor nuclei). In the other clinical subgroups, pathology was less predictable in life, except that the majority of cases with PNFA (7/9) and CBD (7/9) had tau-positive pathology. In patients with fvFTD and semantic dementia, there was an even distribution of tau-positive and tau-negative cases. In view of the poor prognosis of the FTD-MND group, they were excluded from further analysis and we compared the survival of the groups with, and without, tau-positive pathology. As illustrated in figure 2, patients with tau-positive pathology had a better prognosis in terms of survival from symptom onset: median survival 5.0 years (95% CI 3.9 to 6.1 years) in the tau-negative group vs 9.0 years (95% CI 8.1 to 9.9 years) in the tau-positive group (χ² [df = 1] = 7.3, p < 0.01). Interestingly, there was no difference in survival from diagnosis, either in terms of time to institutionalization or to death (3.0 years for tau-positive vs 4.0 years for tau-negative groups). The effect of pathology was confounded by age because the tau-positive group tended to be older at the time of symptom onset, although the difference was not significant (60.3 ± 7.3 vs 56.7 ± 8.3). By the time of diagnosis there was an age difference between the groups (tau-positive 64.3 ± 6.9 vs 59.0 ± 7.3; χ² [df = 1] = 5.5, p 0.02). At death, the difference was even more pronounced: tau-positive 69.4 ± 8.1 vs tau-negative 62.9 ± 7.4 (χ² [df = 1] = 7.9, p < 0.005). Thus, those with tau-positive pathology tend to be older and to have a more indolent disease particularly in the early clinical phase before diagnosis.

• Download figure
• Open in new tab
• Download powerpoint

Figure 2. Kaplan-Meier survival plots for the frontotemporal dementia (FTD) patients split into those with tau-positive (Δγ) and tau-negative (⋄) pathology after excluding those with a clinical diagnosis of FTD with motor neuron disease.

Discussion.

This study provides, for the first time, an accurate estimate of survival in a large group of pathologically verified patients with FTD. The median survival for the entire group was 3.0 years at the time of diagnosis and 75% were dead after 6.0 years, indicating that FTD is a highly malignant disease. This short survival was, in part, attributable to the delay in diagnosis: on average 3.0 years elapsed between symptom onset and diagnosis, but even taken from the time of symptom onset the median survival was merely 6.0 years. The delay in diagnosis replicates previous findings and is similar to that observed in AD.18 One of the most striking findings was that institutionalization occurred on average only 1.0 years after diagnosis. This is probably due to the high prevalence of behavioral changes and greater impairment of activities of daily living compared to patients with AD.18 There were no major demographic differences between the clinical groups, except the finding of a younger age at onset, and especially at death, in the FTD-MND group.

A positive family history taken as a first-degree relative with a history of a disease within the FTD spectrum was present in 20/61 (33%), a figure similar to that reported (38%) in a nationwide survey of familial FTD conducted in the Netherlands19 but lower than that observed in clinic populations (57/101 or 56%).20 In keeping with a recent epidemiologic study from Cambridgeshire,10 our sample contained an excess of men (36 to 25; 1.4 to 1), which was accounted for by their preponderance in the semantic dementia, FTD-MND, and CBD groups. Our study confirms that FTD is predominantly a disease of the presenium with a mean age at diagnosis of 61.5 years, although the range was from 50 to 73 years. Age, sex, and family history had no effect on survival in our sample.

When survival was analyzed according to clinical subgroup, a number of important differences emerged. Compared with the commonest subtype of fvFTD, those with FTD-MND had a predictably poor prognosis with a median survival of 1.0 years from diagnosis and 2 years from symptom onset. In keeping with the published literature,6 all nine patients presented with behavioral changes, often combined with aphasic symptoms, and went on to develop features of predominately bulbar motor neuron disease (ALS) within 12 months of presentation. Features of motor neuron disease were not apparent when the patients were first diagnosed with FTD. Thus, the mode of presentation was identical to patients with fvFTD uncomplicated by motor neuron disease, yet the time from symptom onset to diagnosis was on average 1.1 years, which contrasts with a delay to diagnosis of 3.3 years in the fvFTD group. This finding strongly suggests that the tempo of neurodegenerative process in FTD-MND is fundamentally different from that found in other variants of FTD. Prognosis in the other clinical subgroups was not significantly different, with a median survival from diagnosis of 3 to 4 years and of 6 to 8 years from symptom onset.

Underlying pathology also influenced survival. Patients with tau-positive pathology (Pick bodies, corticobasal bodies, argyrophilic grains) had a significantly better prognosis in terms of survival from symptom onset. This finding has biologic and clinical implications. Biologically, Pick bodies have more three-repeat than four-repeat tau isoforms whereas the reverse is true for corticobasal bodies and argyrophilic grains.17,21,22⇓⇓ The presence of these tau-positive pathologies appears to be associated with a slower rate of progression and presumably less rapid neuronal loss, regardless of the type of tau deposited. Cases with tau-positive pathology were also older, with a mean age at diagnosis of 64.3 years and at death of 69.4 years. Clinically, tau-positive pathology should be suspected in older patients with a slow onset and predominantly nonfluent aphasic symptoms.

The gold standard for the diagnosis of FTD remains neuropathology. Although a number of clinical criteria for diagnosis in life have been proposed,2,3⇓ there are very few studies that have attempted to validate these clinical criteria against pathologic confirmation.23 Survival statistics must therefore be based upon a neuropathologically confirmed series, yet the use of such a series introduces potential biases. Selected patients with rapid progression might be more likely to be enrolled into a brain donor program. Even if all patients were enrolled in a given center, it is conceivable that a subpopulation would progress rapidly and die, whereas another with a long duration of illness would remain alive. We cannot rule out such biases; however, the brain donor programs have been established in both centers for more than a decade, with every effort made to enroll all potential patients with FTD. Success rates for obtaining declarations of intent in vivo and collection of brains after death have been exceptionally high in Sydney and in Cambridge.

Estimates of survival in AD have ranged from 4.0 to 10.0 years, depending on the nature of the sample and whether onset was considered as the time of diagnosis or from the start of significant symptoms.24 Recent studies have suggested that the lower figure is more accurate. For instance, two clinic- and community-based studies found median survival from disease onset of 4.3 (clinic-based) and 4.5 (community-based) years.25,26⇓ Higher estimates may have resulted from the exclusion of those with rapidly progressive disease, as illustrated by the findings of the Canadian Clinical Progression of Dementia Study Group24: the uncorrected median survival of their group was 6.6 years but after adjustment for length bias the survival was merely 3.3 years from symptom onset. It is arguably spurious, however, to compare survival in sporadic AD, where the average age at onset is typically over 75, with that of FTD, which is predominantly a disease of the presenium. Survival in patients with early onset AD has been investigated less extensively, but two different studies from Scotland and the North of England have produced very similar figures with median survival estimates of between 4.0 and 5.0 years from diagnosis.27,28⇓ Based upon the later estimates in AD it appears that FTD is even more aggressive than AD, with a median survival of 3.0 years from diagnosis and 6.0 years from symptom onset.