%0 Journal Article %A LD Westbury %A HE Syddall %A EM Dennison %A C Cooper %T OP54 Describing change in musculoskeletal aging: a comparison of techniques using data from the health, aging and body composition study %D 2019 %R 10.1136/jech-2019-SSMabstracts.55 %J Journal of Epidemiology and Community Health %P A26-A26 %V 73 %N Suppl 1 %X Background Preventive strategies for musculoskeletal disorders require a better understanding of age-related changes in muscle strength, physical function and body composition (including bone). Many publications use simple change measures from observations at two time-points which do not account for measurement error and assumes change is linear. Sophisticated techniques for analysing change are available but are rarely implemented in this field.Methods Changes in grip strength, walking speed, lean mass and hip bone mineral density (BMD) were explored among 3075 men and women from the Health, Aging and Body Composition Study; each measure was assessed at least 5 times during a median 9 year follow-up period. The following techniques were implemented: linear mixed effects models (LMEM) (applied to raw data and age-specific z-scores from generalised additive models for location, scale and shape); growth mixture models (GMM); and latent class trajectory models (LCTM). LMEM use random effects to capture inter-individual variation in level and change around a population-average trajectory; GMM extend LMEM by identifying clusters of individuals with similar trajectories and deriving cluster-specific average trajectories; and LCTM are simplified GMM with no random effects, assuming all individuals in a cluster have the same trajectory.Results Mean (SD) age at baseline was 74.1 (2.9) years. Mean annual percentage declines for walking speed and grip strength were 2.1% and 1.5% respectively; declines were smaller for hip BMD (0.6%) and lean mass (0.5%). Trajectories from LMEM (applied to raw data) for grip strength, walking speed and hip BMD were quadratic in relation to age such that declines accelerated with advancing age; decline in lean mass was linear. Random slopes from LMEM applied to z-scores were weakly correlated with baseline levels for all characteristics (-0.36<r<0.17), resulting in person-specific measures of change that were broadly independent of level. All GMM contained a group comprising at least 80% of the sex-specific sample with sparse numbers of participants in other groups, suggesting that a LMEM with a single population average trajectory describes most of the change in the sample. LCTM derived subgroups with much larger differences in levels of the characteristics rather than in rates of loss.Conclusion LMEM enable a more comprehensive analysis of change compared to methods using data from only two time-points. However, inter-individual differences in rates of change regarding musculoskeletal parameters in this age group and duration of follow-up may be too small to be identified using more complex techniques such as GMM or LCTM. %U https://jech.bmj.com/content/jech/73/Suppl_1/A26.2.full.pdf