Study participants were from the Medical Research Council (MRC) National Survey of Health and Development (NSHD), a national sample initially consisting of 5362 British births occurring during 1 week in March 1946 that has to date been regularly followed-up to age 69 years [13]. Most participants (79%) included in the home visit phase of the NSHD 24th data collection in 2015–16 [13] were invited to participate in the Vertical Impacts on Bone in the Elderly (VIBE) study [8, 14], which was initially set up to investigate the health consequences of higher impact PA in older people. Relevant ethics approval has been granted for each data collection; ethical approval for the most recent assessment in 2014–2015 was obtained from the Queen Square Research Ethics Committee (14/LO/1073) and the Scotland A Research Ethics Committee (14/SS/1009). Study participants provided written informed consent.
During the home visit at age 69, participants were invited to participate in the VIBE study. If they agreed, the nurse provided them with a GCDC X15-1c triaxial accelerometer (Gulf Coast Data Concepts, Waveland, Mississippi), custom designed elasticated belt, a time log and a stamped addressed package along with instructions. Accelerometers were configured to a sampling frequency of 50 Hz, a deadband setting of 0.1 g and a timeout setting of 10 s. We instructed participants to wear the accelerometer securely positioned in the belt over their right hip pointing toward the centre of their body for seven continuous days, removing only for sleeping, washing and swimming. Participants were asked to record the times at which the monitor was put on in the morning and taken off at night for each monitoring day and to state reasons, if any, why that day had not been reflective of their normal activity.
Standardised cleaning and processing of raw accelerometer data was carried out by the study coordinating centre and is described in detail elsewhere [8]. Briefly, data were cleaned to remove movement artefacts and non-wear time, and activity data were normalised based on seven valid days of ≥10 h recording time. Vertical (i.e. Y-axis) accelerations peaks were then calculated based on accelerations higher than the preceding and subsequent reading. Participants were grouped into three bands reflecting low (0.5 < g < 1.0), medium (1.0 < g < 1.5 g) and higher (≥1.5 g) impact. The ≥1.5 g cut-point for higher impacts was selected as very few impacts were observed within higher g bands [8, 14]. Periods of inactivity were removed by excluding accelerations ≤0.5 g8. All g values represent above 1 g from earth’s gravitational force.
Cognitive function was assessed at age 69 by tests of processing speed and verbal memory, and by the Addenbrooke’s Cognitive examination-III (ACE-III) scale. Processing speed was assessed by a timed visual search task requiring cancellation of target letters P and W embedded among non-target letters; the speed score was derived from the position reached at the end of 1 minute. Verbal memory was assessed by a 15-word list learning task with three learning trials and free written recall at the end of each trial, therefore the maximum score achievable was 45. The ACE-III scale is the most comprehensive test of cognitive state, developed for use in clinical settings. It includes five subdomains that assess attention, memory, fluency, language and visuospatial ability, and has a maximum score of 100, with a quasi-normal distribution. Recent studies demonstrate the validity of ACE-III for diagnosing mild cognitive impairment, Alzheimer’s disease and dementia [15]. Each cognitive measure was standardised to mean = 0 and standard deviation (SD) = 1.
Childhood cognition, own socioeconomic position (SEP), and contemporaneous BMI and depression were identified as potential confounders. Childhood cognitive ability was tested at age 15 using the Heim AH4 test of verbal and non-verbal ability [16] Watts-Vernon reading comprehension test [17] and a test of mathematical ability [18]. Test scores were combined to derive an overall standardised score (mean = 0 and standard deviation (SD) = 1). Own SEP was based on highest Registrar General’s occupational class at age 53 years (and if missing, then taken from earlier ages), categorised as professional or intermediate; skilled non-manual; skilled manual; and semi-skilled or unskilled manual. BMI (kg/m2) was calculated from heights and weight measured by nurses at age 69; heights were measured to the nearest millimetre using a Leicester stadiometer (Marsden Group, UK) and weights to the nearest 100 g using Tanita weighing scales (Tanita UK Ltd., Uxbridge, UK).
Depression was assessed at age 69 using responses to questions in the depression subscale of the General Health Questionnaire-28, a screening tool used to detect risk of psychiatric disorders [19]. Responses to each question (Been thinking of yourself as a worthless person? Felt that life is entirely hopeless? Felt that life isn’t worth living? Though of the possibility that you might make away with yourself? Found at times you couldn’t do anything because your nerves were too bad? Found yourself wishing you were dead and away from it all? Found that the idea of taking your own life kept coming into your mind?) were assigned a score (0 = not at all, 1 = no more than usual, 2 = rather more than usual, 3 = much more than usual) and summed to derive a total score with potential range from 0 to 21.
We initially examined how vertical impacts related to cognitive function by plotting mean scores for each cognitive test across quartiles of low, medium and high impacts, and tested trends using an extension of the Wilcoxon rank-sum test (Cuzick’s test for trend). Separate linear regression models were then used to examine associations between each PA impact measure (low, medium and high impacts) and each cognitive score. Accelerometer data were log-transformed due to their skewed distributions, and model estimates presented as SD difference in each cognitive score per doubling in the number of impacts. Interaction terms were used to test sex differences, and subsequently men and women were combined, with adjustment made for sex after little evidence of interaction was found. Three models were fitted to test associations between PA within each impact band and each cognitive score; a sex-adjusted model which was subsequently further adjusted for SEP, BMI and depression, and then additionally for childhood cognition. Models were fitted after multiple imputation of missing confounders (n = 72 participants) using 20 multiply imputed datasets which were combined with Rubin’s combination rules [20]. Analyses were performed in STATA 15.
We investigated if any associations found for specific impact levels were independent of total PA by fitting additional models with mutual adjustment for PA within other impact bands. We also examined if musculoskeletal or functional problems influenced findings by repeating the main analyses after excluding in turn those with difficulties walking i.e. noticeable limp (n = 57), walking restricted due to pain (n = 114), regular mobility aid use (n = 29), falls in the past year (n = 136) and fractures since age 45 (n = 231). This information was captured by a self-reported questionnaire left with participants to complete and return with their accelerometer. Finally, we compared multiple imputation results to complete-case analyses.