Table 6

Results of a multiple binary logistic regression model for use of medication for knee pain.

Variable

Ever used pain meds


Female

1.9 (1.2 - 3.2), P = 0.009

Non-white

1.2 (0.6 - 2.2), P = 0.671

Age

P = 0.136

65 to 74 versus <65

0.8 (0.5 - 1.3), P = 0.320

>74 versus <65

0.5 (0.2 - 1.0), P = 0.053

KL grade

P = 0.151

3 versus 2

1.3 (0.8 - 2.2), P = 0.262

4 versus 2

2.0 (1.0 - 4.3), P = 0.061

Pain NRS

P = 0.007

4 to 7 versus 0 to 3

1.6 (0.9 - 2.6), P = 0.090

8 to 10 versus 0 to 3

4.6 (1.8 - 12.0), P = 0.002

Co-morbidity

1.8 (1.0 - 3.2), P = 0.066

BMI

P = 0.021

26 to 30 versus <26

1.3 (0.7 - 2.5), P = 0.360

>30 versus <26

2.4 (1.2 - 4.8), P = 0.010

Pain NRS samea or worse

1.7 (1.0 - 2.8), P=0.043


Number of patients included

806

Overall model significance:

Likelihood ratio test

χ2=53.4, df = 12, P <0.001

Goodness-of-fit test:

Hosmer & Lemeshow

χ2=5.48, df = 8, P = 0.713


aor improved by less than 2 units; BMI, body mass index kg/m2; KL, Kellgren-Lawrence; NRS, numerical rating scale; OTC, over the counter. For categorical variables the joint significance of the combined categories is presented in addition to the results for each category relative to the reference category. Values presented are odds ratio (95% CI) of reporting use of any form of medication for pain, aching or stiffness in either knee at any time during the 36-month follow-up period unless otherwise stated; values in bold type indicate statistical significance.

Kingsbury et al. Arthritis Research & Therapy 2013 15:R106   doi:10.1186/ar4286

Open Data