Background Cognitive performance shows a marked deterioration in close proximity to death, as postulated by terminal decline hypothesis. However, the association between education and terminal decline remains highly controversial. This study investigated the role of education on terminal decline in healthy and incident dementia cases from two European longitudinal studies of oldest-old.

Methods Participants were from the Newcastle 85+, UK (N = 702) and from OCTO-Twin, Sweden (N = 845). They were assessed biannually over 3 and 5 consecutive waves respectively. In a coordinated analysis, multilevel models were employed to examine terminal decline in Mini-Mental State Examination (MMSE), controlling for education, age at baseline, dementia incidence, sex, and time to death from the study entry within each cohort. Cognitive change was modelled as a linear function of time to death in both cohorts and as a quadratic function in the OCTO-Twin study. Education was used as a continuous measure (ranging 6–20 yrs in Newcastle 85+ and 0–23 yrs in OCTO-Twin).

Results The results suggest that a typical British man, aged 85 at baseline, with 10 years education, entered the terminal phase at around 2.5 years before death, and the rate of decline was -1.04 (0.25), p < 0.001 with each year closer to the time of death. In contrast, a Swedish man, aged 83, with an average of 7 years education, entered the study at around 8 years from death, after which the rate of cognitive decline steepened by -1.70 (SE = 0.20), p < 0.001 and accelerated by -0.11 (0.01) p < 0.001 points per year closer to the time of death. Incident dementia cases experienced a steeper decline compared to healthy individuals in both studies with (-2.70 (SE = 0.21) and (-0.13 (SE = 0.02) acceleration), p < 0.001 units in OCTO-Twin and -1.73 (SE = 0.33), p < 0.001 in Newcastle 85+). Education was positively associated with the estimated MMSE scores prior to the time of death in OCTO (0.43 (SE = 0.15) p < 0.005), but not in the Newcastle 85+ and did not attenuate the rate of terminal decline in either of the two cohorts investigated.

Conclusion As postulated by terminal decline hypothesis, decline and acceleration of this decline were detectable in both of these studies prior to death, with steeper rates of decline observed in the Swedish cohort. However, this process was not lessened by education itself. It will be useful to extend these analyses in studies with longer follow-ups periods, in order to allow a better understanding of the transition from the subtle cognitive changes accompanying age decline to those of neurological substance.