Survival estimates and their predictors in genetic frontotemporal dementia: an international, retrospective, cohort study

BACKGROUND: What drives the heterogeneity of survival estimates in genetic frontotemporal dementia is unknown. We sought to understand the natural history and predictors of disease trajectory, which are crucial not only for effective care but also for the design of therapeutic clinical trials and efficacy evaluation. METHODS: In this international, cohort study, we used the Kaplan-Meier method to retrospectively assess survival estimates in patients enrolled in the GENFI cohort, which included 32 research sites located in Belgium, Canada, Finland, France, Germany, Italy, the Netherlands, Portugal, Spain, Sweden, and the UK, and comprised participants carrying a causal C9orf72 expansion or a causal mutation in GRN or MAPT genes. Survival was calculated as the time from symptom onset to time of death or censoring date; median survival estimate for all patients was the primary endpoint. Cox proportional hazards models were used to identify predictors of survival, which were subsequently externally validated in an independent cohort. We further designed a structural equation model to assess the relationships between predictors, applying a least absolute shrinkage and selection operator method. FINDINGS: Of 278 participants of the GENFI cohort included in this study, 160 (58%) were men and 118 (42%) were women. 162 died during follow-up (58%) and 116 were still alive (42%) on June 1, 2024, the chosen censoring date. 138 participants carried a C9orf72 expansion, 94 carried a GRN mutation, and 46 a MAPT mutation. 179 participants were diagnosed with behavioural variant frontotemporal dementia, 46 with primary progressive aphasia, and 31 with frontotemporal dementia-amyotrophic lateral sclerosis. 22 participants had other diagnoses. The median survival estimate for all patients with genetic frontotemporal dementia was 6·94 years (95% CI 6·59-7·80) from symptom onset. The median survival estimate for patients with GRN mutations was 6·63 years (6·08-7·98), for patients with a C9orf72 expansion was 7·04 years (6·45-8·77), and for patients with MAPT mutations was 8·56 years (7·06-13·50). Older age at onset, shorter disease duration from onset to enrolment in the GENFI study, clinical presentation (ie, frontotemporal dementia-amyotrophic lateral sclerosis), domain of first symptom (ie, motor or language onset), and geographical area of residency (ie, central and southern Europe) were associated with poorer prognosis. Genetic group did not directly affect survival estimates; rather its effect was mediated by age at onset and clinical phenotype. We computed a genetic frontotemporal dementia survival risk index, which can be used at an individual patient level. INTERPRETATION: Our results highlight that motor impairment in addition to cognitive and behavioural symptoms should be considered when estimating prognosis in genetic frontotemporal dementia. Individual risk scores might be of help for patient stratification in future therapeutic trials, although refinement and prospective validation are now needed. FUNDING: Italian Ministry of Health (Ricerca Corrente), Fondation Philippe Chatrier, and Fondation Vaincre Alzheimer.