To eliminate the effect of confounders, logistic regression
analysis was conducted. All of the potential
confounders were tested for multicollinearity diagnostics.
A formal-detection tolerance of all of these
ranged from 0.6 to 0.9, so there was no multicollinearity
problem. After adjustment for many confounders,
odds ratio in those who had few pain sites
showed no statistical significance (1.89 (95 % CI 0.93
– 3.80)), whereas those who had many pain sites still
showed statistical significance although its odds ratio
showed a small decrease (2.44 (95 % CI 1.06 – 5.66)).