The age at which cognitive decline begins has been the subject of much debate. The Seattle longitudinal study has provided most of the evidence that it doesn’t begin until age 60. A more recent, much larger study that allows both longitudinal and cross-sectional analysis suggests that, depressingly, mid-to-late forties might be closer to the mark.
A long-term British study known as Whitehall II began in 1985, when all civil servants aged 35-55 in 20 London-based departments were invited to participate. In 1997-9, 5198 male and 2192 female civil servants, aged 45-70 at this point, were given the first of three rounds of cognitive testing. The second round took place in 2002-4, and the third in 2007-9.
Over these ten years, all cognitive scores except vocabulary declined in all five age categories (45-49, 50-54, 55-59, 60-64, and 65-70 at baseline). Unsurprisingly, the decline was greater with increasing age, and greatest for reasoning. Men aged 45-9 at baseline showed a 3.6% decline in reasoning, compared to a 9.6% decline for those aged 65-70. Women were less affected by age: while showing the same degree of decline when younger, the oldest showed a 7.4% decline.
None of the other cognitive tasks showed the same age-related deterioration as reasoning, which displayed a consistently linear decline with advancing age. The amount of decline over ten years was roughly similar for each age group for short-term memory and phonemic and semantic fluency (although the women displayed more variability in memory, in a somewhat erratic pattern which may perhaps reflect hormonal changes — I’m speculating here). Moreover, the amount of decline in each decade for these functions was only about the same as reasoning’s decline in the younger decades — about -4% in each decade.
Men and women differed significantly in education (33% of men attended university compared to 21% of women; 57% of women never finished secondary school compared to 39% of men). It is therefore unsurprising that men performed significantly better on all cognitive tests except memory (noting that the actual differences in score were mostly quite small: 16.9/35 vs 16.5 for phonemic fluency; 16.7/35 vs 15.8 for semantic fluency; 25.7/33 vs 23.1 for vocabulary; 48.7/65 vs 41.6 for reasoning).
The cognitive tests included a series of 65 verbal and mathematical reasoning items of increasing difficulty (testing inductive reasoning), a 20-word free recall test (short-term verbal memory), recalling as many words as possible beginning with “S” (phonemic fluency) and recalling members of the animal category (semantic fluency), and a multi-choice vocabulary test.
The design of the study allowed both longitudinal and cross-sectional analyses to be carried out. Cross-sectional data, although more easily acquired, has been criticized as conflating age effects with cohort differences. Generations differ on several relevant factors, of which education is the most obvious. The present study semi-confirmed this, finding that cross-sectional data considerably over-estimated cognitive decline in women but not men — reflecting the fact that education changed far more for women than men in the relevant time periods. For example, in the youngest group of men, 30% had less than a secondary school education and 42% had a university degree, and the women showed a similar pattern, with 34% and 40%. However, for those aged 55-59 at baseline, the corresponding figures were 38% and 29% for men compared to 58% and 17% for women.
The principal finding is of course that measurable cognitive decline was evident in the youngest group, meaning that at some point during that 45-55 decade, cognitive faculties begin to decline. Of course, it should be emphasized that this is a group effect — individuals will vary in the extent and timing of any cognitive decline.
(A side-note: During the ten year period, 305 participants died. The probability of dying was higher in those with poorer cognitive scores at baseline.)
 . Timing of onset of cognitive decline: results from Whitehall II prospective cohort study. BMJ [Internet]. 2012 ;344(jan04 4):d7622-d7622 - d7622-d7622. Available from: http://www.bmj.com/content/344/bmj.d7622
Full text available at http://www.bmj.com/content/344/bmj.d7622