Longitudinal Data Analysis of Bone Health During Lactation and Weaning

This thesis looked at the effects of lactation on bone health. The analysis was used to determine if there was a significant decrease in bone measurements with lactation, a significant increase after lactation and whether or not there is full recovery from the pre-lactation to post-lactation measurements. Research has been done on BMD (bone mineral density) measurements, so statistical methods were applied to see if the same trends applied to pQCT (peripheral quantitative computed tomography) measurements. The data used are from the South Dakota Rural Bone Health Study, an ongoing study through the Ethel Austin Martin Program at South Dakota State University.;The methods applied in the paper include multiple linear regression and a piecewise linear mixed effects model. In multiple linear regression, the bone differences between time points (-18 to 0 months, 0 to 18 months, & -18 to 18 months, with time 0 being lactation) are regressed on the set of predictors, whereas in the linear mixed effects model the whole sequence of observations is used to model the change as a function of time. With multiple linear regression methods, the pQCT measurements that appeared significant included a decrease in total cross-sectional area at the 4% distal radius (tot_a4) during lactation and an increase in total cross-sectional area at the 20% distal radius (tot_a20) in overall recovery. Multiple regression showed that change in weight, parity, and age (time) are significant covariates for some bone measurements. The piecewise linear mixed effects model showed significant decrease during lactation for total cross-sectional area at the 4% distal radius, but none of the pQCT sites showed significant increase in the bone measurements during weaning.