J Neurol Neurosurg Psychiatry 81:441-445 doi:10.1136/jnnp.2009.195719
  • Research paper

Survival in progressive supranuclear palsy and frontotemporal dementia

  1. John C van Swieten
  1. 1Department of Neurology, Erasmus Medical Centre, Rotterdam, The Netherlands
  2. 2Department of Neuropathology, Vrije Universiteit Medical Centre, Amsterdam, The Netherlands
  1. Correspondence to Dr John C. van Swieten, Department of Neurology, Erasmus University Medical Centre Rotterdam, Room Hs 611, ‘s-Gravendijkwal 230, 3015 CE Rotterdam, The Netherlands; j.c.vanswieten{at}
  • Received 21 September 2009
  • Revised 23 November 2009
  • Accepted 3 December 2009


Objective To compare survival and to identify prognostic predictors for progressive supranuclear palsy and frontotemporal dementia.

Background Progressive supranuclear palsy (PSP) and frontotemporal dementia (FTD) are related disorders. Homozygosity for H1 haplotype is associated with PSP, whereas several MAPT mutations have been identified in FTLD-τ. Survival duration probably reflects underlying pathophysiology or disease.

Methods Patients with PSP and FTD were recruited by nationwide referral. Survival of 354 FTD patients was compared with that of 197 PSP patients. Cox regression analysis was performed to identify prognostic predictors. FTLD-τ was defined as Pick disease and FTDP-17 with MAPT mutations. Semiquantitative evaluation of τ-positive pathology was performed on all pathologically proven cases.

Results The median survival of PSP patients (8.0 years; 95% CI 7.3 to 8.7) was significantly shorter than that of FTD patients (9.9 years; 95% CI 9.2 to 10.6). Corrected for demographic differences, PSP patients were still significantly more at risk of dying than FTD patients. In PSP, male gender, older onset-age and higher PSP Rating Scale score were identified as independent predictors for shorter survival, whereas in FTD a positive family history and an older onset-age were associated with a poor prognosis. The difference in hazard rate was even more pronounced when comparing pathologically proven cases of PSP with FTLD-τ.

Conclusion Survival of PSP patients is shorter than that of FTD patients, and probably reflects a more aggressive disease process in PSP. Independent predictors of shorter survival in PSP were male gender, older onset-age and higher PSP rating scale score, whereas in FTD a positive family history and higher onset-age were predictors for worse prognosis.


Progressive supranuclear palsy (PSP) is clinically characterised by parkinsonism, supranuclear gaze palsy and cognitive decline,1 2 and shows clinical, pathological and genetic overlap with frontotemporal dementia (FTD).3–5 A frontal presentation has been identified in 20% of PSP cases,6 whereas FTDP-17 associated with microtubule associated protein τ (MAPT) mutations may present with the clinical picture of PSP.7 8 Neuronal and glial τ-positive inclusions are found in PSP and consist mainly of hyperphosphorylated four-repeat τ isoforms. In contrast to this, a subset of FTLD with Pick bodies (so-called Pick disease) is characterised by the accumulation of three-repeat τ isoforms,9 whereas inclusions in FTDP-17 with MAPT mutations variably consist of three- and four-repeat τ isoforms, depending on the location of the mutation.10 The relevance of clinical and pathological overlap is further emphasised by the strong association between MAPT H1/H1 genotype and PSP.11 Determining survival within this FTLD-PSP spectrum is of important clinical relevance and may give insight into the underlying disease process. However, only a few small studies compared survival between PSP and FTD and did not find any differences.12 13 Small pathological series of PSP and τ-positive and τ-negative FTLD patients have shown conflicting results regarding the effect of τ pathology on survival.14–16 In a recent study, specific neuropsychological profiles in FTLD have been correlated to disease duration, whereas onset-age or positive family history were not.12 Early falls and gaze palsy have been found as prognostic features in retrospective studies on PSP,17–20 whereas the PSP Rating Scale (PSPRS) has also proven to be of predictive value in survival in a prospective longitudinal study.21 However, this still has to be replicated. Severity of τ pathology has shown an inverse correlation with prognosis in PSP,22 23 whereas conflicting results have been reported in FTLD.14–16 24

The aim of this study is to prospectively investigate the survival in two large cohorts of PSP and FTD patients in relationship to demographic and clinical features, and to the presence and severity of τ pathology in a subset of patients who underwent brain autopsy.


Patients with PSP and FTD were recruited by nationwide referral from neurologists and by visiting patients in nursing homes.6 25 Detailed clinical history, including the first presentation of symptoms, was obtained from patients and their family members, and by reviewing medical records. The onset-age was defined as the age at which the first symptom attributable to PSP and FTD appeared according to the patient's care giver and from medical records. In case of discrepancies, data from medical records were used. Data on family history were obtained using a structured questionnaire provided by spouse or first-degree relative. Family history was defined positive if at least one first-degree relative suffered from dementia, parkinsonism or motor neuron disease. All available hard copies of neuroimaging of both PSP and FTD patients were reviewed by the investigators in order to exclude other structural causes of both conditions and to semiquantitatively measure the severity of lobar atrophy.

PSP patients were neurologically examined and videotaped, and the severity of their cognitive and motor functioning was scored by means of Mini-Mental State Examination (MMSE), Frontal Assessment Battery (FAB), Unified Parkinson's Rating Scale-III (UPDRS-III) and PSPRS.

FTD patients underwent neurological examination, neuropsychological evaluation and neuroimaging (CT, MRI or SPECT with 99mTc-hexamethyl propyleneamine oxime (HMPAO)). The clinical diagnosis of all patients was established in a consensus meeting according to the National Institute for Neurological Diseases and Stroke-Society for PSP (NINDS-SPSP) criteria2 and the Lund and Manchester criteria for FTD.26 PSP patients were subdivided according to phenotype as described by Williams et al.27 One hundred and twenty-one patients were classified as Richardson syndrome (RS), and seven cases of PSP-parkinsonism (PSP-P) were identified in our cohort. Of 18 patients, there were insufficient data available on the first 2 years after onset. The remainder of the patients (n=51) could not be subdivided into a phenotype. Both studies on PSP and FTD patients were approved by the Medical Ethics Committee of the Erasmus Medical Centre of Rotterdam. Informed consent for participation (including blood collection) was obtained from the spouse or a first-degree relative of each patient. MAPT, CHMP2B and GRN genes were sequenced in all familial FTD patients, as has been previously described.28–30 In PSP patients with a positive family history, screening of MAPT, GRNand LRRK2 was performed according to previously described methods.29 31 32

The possibility of post-mortem examination was discussed with patients and their relatives during follow-up. Brain autopsy of patients who gave consent and who died during follow-up was conducted by The Netherlands Brainbank according to their Legal and Ethical Code of Conduct. All brains that became available for autopsy were processed for routine staining and immunohistochemistry with AT8 (1:40, Innogenetics, Ghent, Belgium), ubiquitin (1:500, Dako, Glostrup, Denmark), three-repeat tau isoform (RD3, Upstate, Charlottesville, VA; 1:3000) and four-repeat tau isoform (RD4, Upstate, Charlottesville, VA; 1:100), p62 (BD Biosciences Pharmingen, San Diego, California, USA; 1:200, following 80°C antigen retrieval), TDP-43 (Proteintech, Chicago, Illinois; 1:100, following pressure-cooking), β-amyloid (anti-β-amyloid, DAKO, Glostrup, Denmark, 1:100, following formic acid pretreatment) and α-synuclein (anti-α-synuclein, Zymed Laboratories, San Francisco, California, USA; undiluted, following formic acid pretreatment). These were incubated overnight at 4°C. Endogenous peroxidase activity was inhibited by 30 min incubation in PBS–hydrogen peroxide–sodium azide solution (100 ml 0.1 M PBS, 2 ml 30% H2O2, 1 ml natriumazide). The Histostain-Plus broad-spectrum kit DAB (Zymed, San Francisco, California, USA) was used as a detection system. Slides were counterstained with Mayer haematoxylin and mounted in Entellan.

The neuropathological diagnosis FTLD was classified into FTLD-τ and FTLD-U (with or without TDP-43-positive inclusions).33 FTLD-τ was defined as Pick disease and FTDP-17 with MAPT mutations. Cases with FTD-MND were excluded from this study. In FTLD-τ cases, neuronal loss and τ-staining reactive neurons and glial cells were visually quantified (none, mild, moderate and severe) in the following regions: frontal lobe, temporal lobe, hippocampus, parietal lobe, caudate nucleus and substantia nigra.

The neuropathological diagnosis PSP was established according to international criteria,34 and a semiquantitative assessment of neurofibrillar tangles (NFT), tufted astrocytes (TA), oligodendroglial coiled bodies (CB) and thread pathology (Th) in all regions was carried out by two raters (WK, JvS) using a five-point grading scale according to Williams et al.23 The PSP-τ score was calculated from the combined grade of coiled bodies and thread lesions in the substantia nigra, and caudate and dentate nucleus.

Follow-up of PSP and FTD patients was performed by visits to the outpatient department of the Erasmus Medical Centre or by telephone interview with relatives up to 1 August 2008.

Statistical analysis

SPSS 15.0 for Windows (SPSS, Chicago, Illinois, USA) was used for analysis. Onset-age, gender and family history were analysed by independent sample t test or χ2 test. The actuarially corrected median survival was calculated, as well as the mean survival in deceased cases. Survival analysis was performed using the Cox proportional hazard model, using a backward selection procedure model. Only results of multivariate analyses are shown, with variables that were significant in the univariate analysis. As the early occurrences of clinical symptoms are incorporated into the PSPRS, these symptoms were not analysed together with the PSPRS in one model. However, different sections of the PSPRS (history, mentation, bulbar, ocular, limb and gait sections) were analysed separately. Entry date was set as the time of the first symptoms. Censoring date was either date of death or end of follow-up (1 August 2008). The assumption of proportionality of hazards was examined by log–log plots. Hazard ratios (HRs) and 95% CIs were calculated. Onset-age and PSPRS score were categorised into quartiles. Correlation between τ pathology and disease duration and onset-age was examined using Spearman calculation. All statistical testing took place at a 0.05 level of significance (two-tailed).


The demographic data of patients with PSP and FTD are summarised in table 1. FTD-MND patients (n=30) were excluded, due to their known shorter disease duration. Two PSP patients died of non-natural cause and have not been included in the survival analyses. The mean onset-age and age at death of PSP patients were significantly higher than that of FTD patients. During follow-up, 133 of 197 patients with PSP died at a mean disease duration of 7.2±2.6 years, whereas 242 out of 354 FTD patients had died after a mean disease duration of 9.2±4.1 years.

Table 1

Demographic characteristics of patients with progressive supranuclear palsy and frontotemporal dementia

Survival and hazard analysis of PSP and FTD

The median disease duration in PSP patients (8.0 years; 95% CI 7.3 to 8.7) was significantly shorter than in FTD patients (9.9 years; 95% CI 9.2 to 10.6) (χ2 17.1, p<0.001) (figure 1). This worse prognosis for PSP patients than FTD patients in a univariate analysis (HR 0.634; 95% CI 0.509 to 0.788) remained significant after adjustment for gender, onset-age and family history (HR 0.766; 95% CI 0.603 to 0.975). Comparing PSP phenotypes, RS (6.8 years; 95% CI 6.3 to 7.4) was found to have a shorter median survival than PSP-P (10.9 years; 95% CI 7.5 to 14.2) and the non-conclusive group (8.8 years; 95% CI 8.2 to 9.3).

Figure 1

Kaplan–Meier survival curve for progressive supranuclear palsy (PSP) and frontotemporal dementia (FTD) patients.

A Cox proportional hazards regression model of PSP patients (table 2) revealed male gender, older onset-age (>72 years) and higher score on the PSPRS to be independent predictors for shorter disease duration, after adjustment for the interval between onset and ascertainment (mean of 5.3±2.6 years).

Table 2

Multivariate Cox models in progressive supranuclear palsy patients (HR with 95% CI)

When entering separate sections of the PSPRS in the model, only supranuclear ocular motor exam (HR 1.195; 95% CI 1.090 to 1.310) remained significant, whereas bulbar exam (HR 1.144; 95% CI 0.997 to 1.312) and gait exam (HR 1.063; 95% CI 0.999 to 1.131) were almost significant.

In FTD patients, positive family history (HR 1.438; 95% CI 1.114 to 1.858) and onset age >64 years (HR 1.656; 95% CI 1.160 to 2.363) were significantly associated with poor survival. Looking into family history in FTD in more detail, the mean disease duration of deceased FTD patients with a negative family history (9.9 years; 95% CI 9.1 to 10.6) was significantly longer than that of FTD patients with a positive family history (8.4 years; 95% CI 7.7 to 9.1; p=0.006). In this latter group, a trend towards a longer mean disease duration of patients with a MAPT mutation (n=36 from 10 families; 9.3 years; 95% CI 7.8 to 10.8) was found compared with patients without a MAPT mutation (n=83, including 17 patients with a GRN mutation from three families; 8.1 years; 95% CI 7.3 to 8.8; p=0.105). Of the MAPT mutations, L315R had the shortest mean disease duration (n=5; 5.7±1.9 years), followed by P301L (n=20; 8.2±3.0 years), whereas R406W had the longest mean disease duration (n=4; 17.5±3.2 years). The remaining MAPT mutations, S320F (n=1), G272V (n=5), and ΔK280 (n=1) all had a mean disease duration of just above 10 years. Patients with a GRN mutation had a survival of 7.7±2.8 years.


Pathological examination was available for 24 PSP patients (all RS) and 61 FTLD patients (FTLD-τ n=32 and FTLD-U n=29). Men were over-represented (70.8%) in the PSP series, and the FTLD series showed a higher percentage of a positive family history (57.6%) and younger onset-age (55.3 years) than the total group, due to a significantly lower onset-age for cases with MAPT mutations (50.9 years).

After adjustment for gender, onset-age and family history, FTLD-τ patients remained less at risk than PSP patients (HR 0.524; 95% CI 0.282 to 0.974), and a trend towards longer survival was found compared with FTLD-U patients (HR 0.608; 95% CI 0.361 to 1.024).

The FTLD-τ group consisted of 15 sporadic cases, all of which showed pure three-repeat τ pathology, and 17 cases with MAPT mutation, with pure three-repeat (G272V and ΔK280), pure four-repeat (P301L) or a mix of three-repeat and four-repeat (S320F, R406W and L315R) τ pathology depending on the location of the mutation. All PSP cases showed four-repeat τ pathology. The mean disease duration of sporadic Pick disease cases was 12.1 years and was similar to that in MAPT cases with three-repeat τ pathology (n=6) of 10.2 years, whereas a trend (p=0.098) could be observed towards shorter survival in MAPT cases with four-repeat τ pathology (n=7) of 8.6 years. Disease duration in MAPT mutations with a mix of three-repeat and four-repeat τ pathology varied considerably.

The FTLD-U cohort consisted of nine type 1, 16 type 2 and four type 3 (all four with GRN mutation) cases. Survival of pathological GRN cases did not differ significantly from the total group of deceased GRN cases or FTLD-U type 2 cases (8.4±3.1 years) but was significantly shorter than survival of FTLD-U type 1 cases (11.6±5.0 years).

τ Pathology quantification

Neuronal loss in PSP cases was most prominent in subthalamic nucleus, globus pallidus, dentate nucleus and substantia nigra. τ pathology consisting of globoid neurofibrillar tangles, tufted astrocytes and glial coiled bodies varied considerably between cases, with the subthalamic nucleus, thalamus, substantia nigra, basal pontine nuclei, locus coeruleus and dentate nucleus regions most severely involved. The severity of τ pathology, expressed in the PSP-τ score, showed a significant negative correlation with disease duration (figure 2) but was not correlated with onset-age.

Figure 2

Disease duration of pathologically proven progressive supranuclear palsy (PSP) cases according to PSP-τ score (Spearman rho −0.44, p=0.045).

For the FTLD-τ group, neuronal loss in frontal, temporal and hippocampus regions was severe in most cases, whereas parietal, caudate nucleus and substantia nigra regions showed a variable neuron loss. τ-positive inclusions showed a similar pattern of topographic distribution with severe τ pathology in frontal and temporal cortex and hippocampal regions, whereas the severity of τ pathology was more variable in parietal cortex, caudate nucleus and substantia nigra. Astrocytic τ pathology was severe in the L315R mutation but only mild in other MAPT mutations and sporadic Pick disease. No significant correlation could be found between disease duration and either neuronal loss or τ-reactivity in any region.


This study is the largest prospective population-based study comparing the survival between patients with PSP and FTD, and showed a significantly shorter disease duration in PSP. This difference was even more pronounced when comparing pathologically proven cases of PSP with FTLD-τ. This study replicates, for the first time, the prognostic value of the PSPRS with a sharp increase in probability of death above a score of 60. In PSP patients, male gender and older onset-age were also independent predictors for a shorter disease duration, whereas a positive family history and an older onset-age were associated with a poor prognosis for FTD.

Our observation of a shorter disease duration in PSP than in FTD contrasts with two other studies,12 13 in which the small number of PSP patients may explain the lack of correlation. Our findings are probably close to true survival rates, as the patients were population-based ascertained. Looking into the natural history of PSP separately, the mean disease duration of deceased cases of 7.2 years in the present study comes very close to 6.8 years found in the only other large prospective study by Golbe et al,21 whereas a large retrospective study20 showed a shorter survival of 5.7 years. This was also true for RS cases in the clinicopathological study by O'Sullivan et al (6.2 years),35 whereas a much longer survival was found for PSP-P patients (11.6 years). Although our PSP-P group consists of only seven cases, due to the strict use of NINDS-SPSP criteria, the difference in survival compared with our RS cases was striking as well. The effect of higher onset-age on survival in the present study was also found in retrospective studies,20 21 35 whereas our observed predictive value of gender contrasted to a weak or absent effect on survival in several other studies,18 20 21 but not all.35 The prognostic significance of older onset-age in PSP resembles observations made in Alzheimer disease36 and Parkinson disease (PD), whereas there is conflicting evidence regarding effect on prognosis of male gender in PD.37 A good explanation for lower survival in men with PSP in our study is lacking. The finding might perhaps be explained by differences for gender in comorbidity at higher age. However, such data are not available in our study.

The predictive value of the PSPRS score for survival in PSP patients confirms the first observations made in a tertiary referred cohort of Golbe et al21 and also proves its predictive value in a population-based cohort. In line with Golbe's observations, a sharp rise in mortality risk was seen in patients with a PSPRS score above 60. Only the subsections supranuclear ocular motor exam, bulbar exam and gait exam were of prognostic value in our study. The replication of Golbe's findings on the PSPRS has implications for its potential use in clinical trials.

Shorter survival in FTD patients with a positive family history in this study contrasts with other studies on the natural history of FTD,12 13 24 38 39 and may suggest a more malignant disease process for hereditary forms. This is especially true for patients with GRN mutations and hereditary FTLD with an unknown genetic defect, both groups exhibiting ubiquitin pathology,40 whereas MAPT mutations showed a trend towards longer disease duration. However, as several mutation carriers were related, we cannot exclude other familial genetic factors influencing the disease duration within the families. The absence of an association between positive family history and survival in other studies may be explained by a low number of patients or an unknown family history.

The longer survival of FTLD-τ group than with pathologically proven PSP cases supports the hypothesis of a different disease process. The mean disease duration in the present series of 11.1 years is similar to that in the study by Hodges et al (9.0 years).24 The shorter survival of τ-positive cases (6 years) in the study by Xie et al16 can be explained by the inclusion of PSP and CBD cases. Our findings are very similar to the observations made by Hu et al,14 which showed that three-repeat FTLD-τ have a longer survival than four-repeat FTLD-τ and four-repeat controls, comprising PSP and CBD patients, and supports the idea that FTLD-τ patients tend to have a more indolent disease course than PSP.

The observed negative correlation between the severity of glial τ pathology and disease duration in PSP patients is in line with the study by Josephs et al.22 The severity of oligodendroglial τ pathology in the substantia nigra and caudate and dentate nucleus represented the overall τ pathology reliably in the study by Williams et al, which again correlated negatively with disease duration,23 and was shown to be higher in RS than in PSP-P. Due to the absence of PSP-P in our pathological cohort, we could not replicate the latter finding. The correlation between the type and severity of τ pathology indicates that pathophysiological mechanisms determine the disease progression. Small sample size, a semiquantitative method of scoring and different MAPT mutations with different functional effects may have hampered our analysis in FTLD. The association in FTLD between shorter survival and abundant τ pathology in basal ganglia in the study by Xie et al16 could not be confirmed by our study and should probably be explained by the inclusion of PSP and CBD cases in their analysis. The best strategy would be to extend the survival analysis to a much larger series of pathologically proven FTD cases, which have been prospectively ascertained during life in order to collect reliable clinical information.

One of the drawbacks of the present study is a selection bias towards typical cases, and therefore missing cases with PSP-P, a subgroup that usually has a longer disease duration and an atypical presentation. Furthermore, as the population of the study consists of cases alive at the time of entry, there may be some degree of survival bias. A final drawback is that there was pathological confirmation in only 24 of 133 of the deceased patients. However, the NINDS-SPSP criteria show a good positive predictive value for probable PSP (100%) and possible PSP (83%) in patients presenting with parkinsonism,34 but also in patients presenting with dementia (96% for combined possible and probable PSP).41 Also, no large differences were found between our clinical and our pathological cohort.

In conclusion, this large prospective study showed that survival in PSP is shorter than in FTD. This difference in prognosis was even more pronounced when comparing pathological PSP cases with FTLD-τ. Within the PSP group, male gender, older onset-age and higher PSPRS score were independent predictors for shorter disease duration, whereas a positive family history and an older onset-age were associated with a poor prognosis in FTD. The significant effect of diagnosis on survival may suggest that the underlying pathophysiology in PSP is more aggressive than in FTD. This perspective should help clinicians anticipate disease progression of patients with PSP and FTD.


  • Funding This study was financially supported by: Prinses Beatrix Fonds (grant no 01-0128) Javastraat 86, 2585 AS Den Haag—the Hersenstichting (project no 13F05(2).14), Koediefstraat 5, 2511 CG Den Haag- stichting Dioraphte (grant no 07 01 05).

  • Competing interests None.

  • Ethics approval Ethics approval was provided by the Erasmus Medical Centre, Rotterdam.

  • Provenance and peer review Not commissioned; externally peer reviewed.


Visit the full archive of podcasts for JNNP here >>

Free sample
This recent issue is free to all users to allow everyone the opportunity to see the full scope and typical content of JNNP.
View free sample issue >>

Don't forget to sign up for content alerts so you keep up to date with all the articles as they are published.