To analyse the rate of joint destruction, a multivariate normal regression model for longitudinal data with log-transformed SHS as response variable was used as previously described [ 23].
In the cohorts with multiple sets of radiographs over time (EAC, Umeå, HCSC-RAC and Wichita) a multivariate normal regression model for longitudinal data was used with radiographic scores as response variable.
A multivariate normal regression model for longitudinal data [ 27] was used with radiographic scores as the outcome and age as the continuous independent variable.
For steps 2 and 3 the multivariate normal regression model was used as described.
41 Multivariate normal regression models fit using SAS Proc Mixed were used to compare older and middle-aged participants with regards to changes in metabolic risk factors over time.
Linear regression and multivariate normal regression models were used.
A multivariate Cox regression model with adjustment for treatment, tumour stage, nodal stage, grade, ER, PR, and age was performed for the comparison of overweight/obese vs normal weight patients with regard to DFS, distant recurrence-free survival, and OS.
We used a multivariate logistic regression model, with stepwise variable selection, to determine predictors for SVR.
To compensate for the missing values in the logistic regression models, multiple imputation was employed using multivariate normal regression, based on age, sex, and smoking data; multiple imputation results in less biased findings when dealing with missing covariate data [ 13].
