Frontotemporal dementia progresses to death faster than Alzheimer disease

For patients with dementia, impairment of cognitive function and loss of functional independence are only part of the burden. In recent years, it has been increasingly appreciated that these diseases also cause early mortality. Premature death has been extensively studied in Alzheimer disease (AD), with estimates of median survival ranging from 3.5 to 6 years after diagnosis.^1–5 There is less information regarding survival in other degenerative dementias, and only recently have disorders such as frontotemporal lobar degeneration (FTLD) been addressed in epidemiologic studies.

FTLD is a major cause of non-AD dementia⁶ and encompasses three clinical syndromes with heterogeneous neuropathology: frontotemporal dementia (FTD), semantic dementia (SD), and progressive aphasia (PA).⁷ In comparison with other neurodegenerative diseases, FTLD shows significant overlap with corticobasal degeneration (CBD) and progressive supranuclear palsy (PSP) but is quite distinct from AD both clinically and neuropathologically. In addition, FTLD might be a more rapidly progressive disease. Patients with Pick disease, a neuropathologic subtype of FTLD, decline faster than patients with AD on certain neuropsychological and functional measures.⁸ In addition, there is a trend toward higher rates of cortical atrophy in patients with the “frontal variant” of FTLD than in AD patients.⁹

We sought to characterize survival in FTLD, directly comparing the results with AD and CBD/PSP. We also performed a hazards analysis to determine what demographic, neuropsychiatric, neuropsychological, neurologic, and neuropathologic factors are associated with reduced survival in FTLD.

Methods.

Consecutive patients presenting between 1988 and 2003 to a University of California neurology clinic (Los Angeles until 1998 or San Francisco thereafter) and receiving a clinical diagnosis of FTLD, probable AD, CBD, or PSP met inclusion criteria for the study (table 1). Because of the smaller number of patients with CBD and PSP clinical syndromes and the close relationship between these diseases, these conditions were analyzed together as a single CBD/PSP group. Because of increasing evidence of both clinical and pathologic overlap between FTLD and CBD/PSP, we were interested in comparing survival in these two groups. For this reason, we analyzed survival in FTLD and CBD/PSP separately rather than merging these groups a priori.

Initial diagnoses were established by consensus of a multidisciplinary team. For this study, the entire medical record, with all available data including longitudinal follow-up, was reviewed retrospectively, and patients were classified based on published criteria.^7,10,11 The study was approved by the Committee on Human Research at the University of California, San Francisco.

Among the 80 patients with a clinical diagnosis of FTLD who had died, autopsy results were available for 39 (49%); the remainder did not have autopsies or the results were unavailable. Of those examined, 14 (36%) had tau-positive FTLD-spectrum pathology (8 with Pick disease, 2 with PSP pathology, 1 with CBD pathology, 1 with tau and synuclein pathology, 1 with tau and ubiquitin pathology, and 1 with tangle-only dementia). Another 21 patients (54%) had tau-negative FTLD-spectrum autopsy diagnoses, including 11 with dementia lacking distinctive histology (DLDH), 6 with ubiquitin-positive inclusions (FTLD-U), 1 coroner’s case in which ubiquitin staining was not performed to distinguish between DLDH and FTLD-U, 2 with progressive subcortical gliosis, and 1 with neuropil spheroids.¹² Of the remaining autopsies, three had AD pathology and one had no abnormalities; these cases were included in the survival analyses because they met the inclusion criteria of a clinical diagnosis of FTLD, although the results did not significantly differ when they were excluded. Most cases were studied pathologically at either the University of Southern California or the University of Pennsylvania.

Results.

One hundred seventy-seven patients with FTLD, 395 with AD, and 47 with CBD/PSP were included in the study. Of these, 80 FTLD, 80 AD, and 14 CBD/PSP patients had died. The characteristics of the patients in each cohort are shown in table 1. As expected, patients with AD were older than those in the other cohorts (70.1 ± 9.6 years; p < 0.0001), and patients with FTD had the youngest age at onset (mean 56.6 ± 9.9 years). The only difference in years of education was the higher level in SD and PA compared with AD (p < 0.001); AD was not significantly different from FTD. Most of the conditions were more common in men, with the exception of PA and AD (p < 0.0001). Mini-Mental Status Examination (MMSE) scores at initial evaluation were 21.2 ± 7.5 in FTLD and 20.2 ± 6.2 in AD. The only difference in MMSE was the higher score in the CBD/PSP cohort than in the AD group (p < 0.0001). The delay between first symptom (determined in retrospect) and initial evaluation did not significantly differ between FTLD (4.5 ± 2.9 years) and AD (4.0 ± 2.8 years); the only difference was the longer delay in SD patients than in those with AD or PA (p < 0.05).

Survival in FTLD.

Median survival in FTLD was 11.0 ± 0.9 years from onset of first symptom (figure 1A) and 3.6 ± 0.4 years from presentation to clinic (figure 1B). Among the three FTLD subtypes, survival was shorter in FTD than in SD (8.7 ± 1.2 vs 11.9 ± 0.2 years from onset, p < 0.005 [figure 1C]; 3.0 ± 0.5 vs 5.3 ± 0.4 years from presentation, p < 0.05 [figure 1D]). Median survival in PA was intermediate between the other subtypes (9.4 ± 2.4 years from onset and 4.2 ± 1.1 years from presentation), although fewer patients had progressed to death and the PA survival curves were not significantly different from those for FTD or SD.

• Download figure
• Open in new tab
• Download powerpoint

Figure 1. Kaplan–Meier curves for frontotemporal lobar degeneration (A, B) and its three subtypes (C, D), plotting survival from symptom onset (A, C) and survival from initial presentation (B, D). The dotted lines in A and B indicate median survival. The p value for log-rank comparisons between survival curves is shown. FTLD = frontotemporal lobar degeneration; FTD = frontotemporal dementia; PA = progressive aphasia; SD = semantic dementia.

We next compared survival in FTLD with AD, the most common neurodegenerative dementia. Published estimates of survival in AD vary considerably owing to differences in patient selection and study design. To ensure the most accurate comparison, we used identical methods to examine survival in consecutive patients presenting with probable AD during the same period as the FTLD cohort. Median survival was longer in AD than in FTLD (11.7 ± 0.6 years from onset, p < 0.005 [figure 2A]; 5.7 ± 0.1 years from presentation, p < 0.0001 [figure 2B]). This difference was driven by the FTD subtype, which had increased mortality relative to AD (p < 0.0001), whereas the survival curves for SD and AD overlapped (see figure 2).

• Download figure
• Open in new tab
• Download powerpoint

Figure 2. Kaplan–Meier curves demonstrating reduced survival in frontotemporal dementia (FTD) compared with Alzheimer disease (AD), plotting survival from symptom onset (A) and survival from initial presentation (B). Survival in semantic dementia (SD) overlaps AD. The p value for log-rank comparisons between survival curves is shown.

We also compared survival in FTLD with CBD/PSP. Because of the significant clinical and pathologic overlap between these conditions, we predicted that they would reduce survival to a similar degree. Indeed, the survival curves for CBD/PSP overlapped FTLD, with median survival of 11.8 ± 3.9 years from onset and 3.6 ± 0.4 years from presentation in CBD/PSP (see figure E-1 on the Neurology Web site; go to www.neurology.org).

Neuropsychological tests.

Scores on the MMSE, digit span, and category, letter, and design fluency tests were obtained for most patients at their first visit (n = 101 to 157). Neither the total MMSE score nor the subscores for orientation, registration, attention, recall, or language affected survival (table 2). The visuospatial MMSE item correlated with survival from presentation but not with overall disease duration. Poor performance on reverse digit span and letter fluency, which both depend on the integrity of frontal–subcortical circuits,^21–24 was associated with significantly reduced survival from presentation (see table 2). Interestingly, impairment on either of these tests was also associated with reduced survival from the onset of symptoms, indicating that patients presenting with disease in frontal–subcortical circuits had a shorter course from first symptom to death than those in whom the temporal or parietal lobes were more selectively affected. These two measures were correlated; in a stepwise Cox regression model, only letter fluency was included and reverse digit span did not add to the association (χ² = 4.438, p < 0.05). Survival curves for subgroups able to generate more than five vs five or fewer “D” words are shown in figure 3. We considered the possibility that these tests may be associated with reduced survival in FTLD simply because they serve as a surrogate marker for the poorer-prognosis FTD subtype. However, the effect was also observed in the FTD group studied alone, suggesting that even within this group, greater frontal–subcortical dysfunction correlates with higher mortality.

View this table:

• View inline
• View popup
• Download powerpoint

Table 2 Neuropsychological hazards

• Download figure
• Open in new tab
• Download powerpoint

Figure 3. Kaplan–Meier curves demonstrating reduced survival in patients with frontotemporal lobar degeneration with lower scores on a letter fluency task, with plots of survival from onset (A) and survival from initial presentation (B). The p value for log-rank comparisons between survival curves is shown.

Neurologic findings.

We focused on two neurologic abnormalities commonly associated with FTLD that might reduce survival: parkinsonism and ALS. The analysis for both factors was based on the presence of clinically significant findings on neurologic examination at the initial evaluation.

Parkinsonism was present in 16 of 177 FTLD patients (9%). Although there was a trend toward reduced survival from presentation in patients with extrapyramidal signs, there was no significant effect of parkinsonism on mortality (see figure E-3 on the Neurology Web site).

Signs of ALS were present in 21 of 177 patients with FTLD (12%), of whom 20 had died. Comorbid ALS dramatically increased mortality, reducing median survival from onset to 4.9 ± 1.6 years (vs 11.5 ± 0.4 years in patients without ALS) (figure 4A) and survival from presentation to 1.4 ± 0.2 years (vs 4.1 ± 0.5 years) (figure 4B). The majority of patients with FTLD and ALS (18/21) had the FTD subtype, 2 had SD, and 1 had PA, although 2 other FTD-ALS cases began with a short PA-like language-predominant syndrome. To determine whether the difference in survival between FTD and AD could be accounted for solely by the effect of ALS, we compared the groups after excluding patients with FTD-ALS. FTD patients without ALS still showed significantly reduced survival compared with AD (see figure E-4 on the Neurology Web site).

• Download figure
• Open in new tab
• Download powerpoint

Figure 4. Kaplan–Meier curves demonstrating reduction of survival in patients with frontotemporal lobar degeneration (FTLD) with comorbid ALS, with plots of survival from onset (A) and presentation (B). The p value for log-rank comparisons between survival curves is shown.

To search for other associated conditions possibly contributing to the difference in mortality between FTD and AD, we examined causes of death as coded on death certificates. Apart from the frequency of ALS, there were no significant differences between the groups (data not shown).

Neuropathology.

There are at least three common neuropathologic substrates for FTLD, which are distinguished by characteristic intracellular inclusions: 1) Pick disease, with prominent tau-positive inclusions; 2) FTLD-U, with ubiquitin-positive but tau-negative inclusions; and 3) DLDH, with no tau- or ubiquitin-positive inclusions. All three syndromes show a combination of frontotemporal atrophy, neuronal cell loss, and spongiosis of the superficial cortical layers at autopsy.^25,26 We compared survival in patients who died with each of these three conditions. Survival was shortest in FTLD-U (3.9 ± 1.5 years), intermediate in DLDH (7.1 ± 1.4 years), and longest in Pick disease (12.4 ± 2.9 years; p < 0.05) (figure 5A).

• Download figure
• Open in new tab
• Download powerpoint

Figure 5. Kaplan–Meier curves showing the effect of underlying neuropathology on survival from onset. (A) Comparison of survival with the three common pathologic substrates of frontotemporal lobar degeneration (FTLD). Progression was fastest in FTLD with ubiquitin-positive inclusions (FTLD-U), slowest in Pick disease, and intermediate in dementia lacking distinctive histology (DLDH). (B) Tau-negative cases progressed significantly faster than tau-positive cases. The difference persisted if tau-negative patients with ALS were excluded. The p value for log-rank comparisons between survival curves is shown.

Interpretation of this difference is confounded somewhat by the higher frequency of ALS in the FTLD-U group (4/6 cases), so we also analyzed the data excluding those cases with clinical ALS. We compared the 14 tau-positive cases (mostly Pick disease; see Methods) with those that were tau-negative (21 cases, mostly FTLD-U and DLDH). Tau-negative cases progressed faster than tau-positive cases (median survival from onset: 5.6 ± 0.3 vs 10.2 ± 5.0 years; p < 0.05) (see figure 5B), even when those with clinical ALS, all of which were tau-negative, were excluded from the analysis.

Discussion.

In this longitudinal study of survival in a large cohort of FTLD patients, we report the discovery that FTD progresses faster than AD. There are several published studies of median survival in AD.^{1–5,27–29} Estimates of survival from presentation range from 3.5 to 6 years.^1–5 The figure of 5.7 years that we observed is consistent with these reports and suggests that our patient population and referral patterns are similar to those of prior studies. Published figures for survival from onset in AD have been more variable, ranging from 4.3 to 9.3 years,^1,2,27–29 shorter than the 11.7 years that we observed. This is partly because the mean age of our AD cohort, 70 years, is at the lower end of the spectrum, and survival in AD is age dependent (longer with lower age at onset).^4,30 But methodologic differences are probably the main reason for the difference. We defined onset as the date of the first symptom, as determined retrospectively. Because of our center’s long-standing interest in the first symptoms of different types of dementia,^13,14 our history taking encourages patients and families to recall the very first time anyone noticed something wrong, an approach that is likely to lead to earlier estimates of onset.

Early studies of Pick disease suggested a total duration of about 7 years.^31,32 Survival in familial forms of FTD and parkinsonism linked to chromosome 17 is highly variable depending on the mutation, family, and methods for ascertainment of diagnosis.³³ In the best prior study on survival in FTLD, median survival in FTLD was 3 years from presentation and 6 years from onset; FTD with ALS progressed significantly faster, but no difference was observed between the other FTLD subtypes.³⁴ Overall, our results for survival from presentation are similar, except for the significantly longer survival we observed in SD. However, our results for survival from onset are consistently longer than those observed in the prior study.³⁴ There are several methodologic differences that explain these discrepancies. First, the larger size of our study may enable detection of differences (e.g., the SD group) not possible in smaller studies. Second, as mentioned above, our method of determining symptom onset is likely to yield higher estimates of disease duration. Third, the prior study used only autopsy-confirmed cases, which may improve diagnostic certainty but could bias toward exclusion of longer-surviving patients, as death was required for entrance into the study. By including survivors in the analysis, our methodology may capture a broader spectrum of FTLD cases.

Although many studies have contrasted the clinicopathologic characteristics of FTD and AD, few have compared their natural history. One recent report found similar mortality rates in the two diseases;³⁵ however, the study included only 15 FTD deaths, and it is unclear whether patients with comorbid ALS or tau-negative neuropathology, two of the fastest progressing subgroups, were included. Two other studies have compared subsets of FTLD with AD using parameters other than survival. One examined the rate of change in neuropsychological tests and functional rating scales and found that patients with Pick disease declined at a faster pace than AD patients,⁸ whereas another found a trend toward higher cerebral atrophy rates in FTD than AD, although this difference was not statistically significant.⁹ Our study adds to a growing body of literature indicating that, by several measures, FTD progresses faster than AD.

On the other hand, we observed a striking similarity in the survival curves between FTD and CBD/PSP. The congruence of their natural history adds to the increasingly appreciated clinical, genetic, and pathologic overlap between these conditions.³⁶

Understanding why FTD progresses more rapidly than AD may contribute to a better understanding of the underlying disease mechanisms and could highlight aspects of the disease that should receive high priority as treatment targets.

Referral bias is unlikely to explain the difference, as there is no evidence that longer-living FTD patients or rapidly progressing AD patients were selectively excluded from referral to our center. The large size of the cohorts and the consistency of our results with those of prior studies in terms of both demographic profile and survival from presentation argue against such selective exclusion.

Some have suggested that poor awareness of symptoms in FTD patients and their families relative to those with AD might result in a delay in perceived onset or presentation and thus reduce calculated survival in FTD. However, this is unlikely for several reasons. First, determining the date of first symptom in retrospect enables reporting of events considered abnormal only in hindsight, after the full disease phenotype has emerged. Second, there is no reason to assume that the behavioral symptoms of FTD (or for that matter the motor symptoms in CBD/PSP, which had similarly reduced survival) are less striking to families than the memory symptoms of AD, which are considered by some as part of normal aging. Finally, the FTD and AD cohorts did not differ in the delay between first symptom and presentation to clinic, as might be expected if one group’s early symptoms were deemed less concerning.

The FTD cohort had a higher male predominance than the AD group, which might have contributed to the observed difference in survival, but sex did not have a significant effect on survival in the hazards analysis. The other demographic difference was lower age at onset in FTD, but there was no significant effect of age in hazards analysis; moreover, survival in AD increases with younger age at onset,^4,30 so an AD cohort age-matched to FTD would be expected to display an even greater difference in survival.

Our data identified three factors that did significantly contribute to the survival difference between FTLD and AD. First, the association of FTLD with ALS had a striking effect on survival from both onset and presentation. Comorbid ALS, however, did not explain all of the reduction in survival with FTD; only 16% of our FTD cohort presented with clinically significant ALS, and there was still a significant difference between FTD and AD when these cases were excluded. Although some of the remaining cases might have had subclinical signs of ALS,³⁷ few progressed to clinically significant ALS before death.

A second contributor to the difference in survival between FTLD and AD appears to be differences in the brain regions affected by the two diseases; specifically, degeneration of frontal–subcortical circuits in FTD might increase mortality more than the temporal lobe disease of AD. Our analysis of neuropsychological hazards (see table 2) showed that difficulties with two frontal–subcortical tests, letter fluency and reverse digit span, were associated with accelerated disease from onset to death, an effect not observed with tests of temporal lobe function. In addition, survival was similar in two diseases that predominantly affect the temporal lobe, AD and SD (see figure 2), and these progressed significantly slower than two conditions where the pathology is more concentrated in frontal and subcortical areas, FTD and CBD/PSP (see figure E-1 on the Neurology Web site). This idea is consistent with a prior study of cerebral atrophy rates in FTLD showing that the fastest-progressing cases had a frontal predominance, whereas atrophy rates in cases with a temporal predominance matched those in AD.⁹ Frontal lobe disease might hasten mortality more than temporal degeneration either because its attendant apathy contributes to the akinetic–mute state that is a common final pathway to death in neurodegenerative disease or because behavioral problems associated with frontal lobe disease lead to overmedication or other differences in the quality of medical care delivered.

Finally, we observed differences in survival based on the underlying molecular pathology. Progression was significantly faster in tau-negative than in tau-positive cases, as observed in a prior study,³⁴ an effect that was not explained by the higher frequency of ALS in the tau-negative cases. Shorter survival in FTLD-U and DLDH accounted for much of the difference between FTD and AD, whereas the survival curve for Pick disease was similar to that in AD. In light of growing evidence that inclusion body formation can be protective in neurodegenerative disease,³⁸ it is interesting that one type of inclusion (tau-positive) is associated with longer survival in FTLD, whereas another (tau-negative, ubiquitin-positive) is not.

These findings have several implications for understanding FTLD and related disorders. First, an accurate understanding of the natural history of FTLD is important for patient care and counseling about prognosis. Our hazards analysis points to factors such as signs of motor neuron disease and impaired letter fluency that might be clinically useful for identifying patients at higher risk of rapid progression. Second, this research has implications for future clinical studies; for example, longitudinal therapeutic trials must consider the more rapid course of FTD, possibly following patients at shorter intervals. Finally, these data suggest important directions for basic research in FTD, such as elucidating the pathophysiology of tau-negative cases and understanding the link between FTD and ALS.

Acknowledgment

The authors thank the patients and families who participated in the study, John Neuhaus for assistance with statistics, Gary Howard and Stephen Ordway for editorial review, Rosalie Gearhart for administrative assistance, and Elizabeth Head and William Seeley for help with neuropathology.