Data from 2,627 individuals were available from the Johnston County Osteoarthritis Project, including 558 African-American women, 273 African-American men, 1,016 Caucasian women, and 780 Caucasian men. Overall, the mean age of the men was 60.6 ± 10.2 years, and the mean age of the women was 60.8 ± 10.6 years. Thirty-eight percent of men and 37% of women had low educational attainment (<12 years). Of the 1,330 postmenopausal women, 455 were African American, and 875 were Caucasian. Mean age for all women in this subset was 62.4 ± 10.1 years. Of these, 39% had low educational attainment, and 24% were current HRT users.
For the complete sample, 29% had radiographic OA, 17% had bilateral radiographic OA, 16% had symptomatic OA, and only 6.5% had bilateral symptomatic OA. Prevalences of all OA outcomes were somewhat higher for women than for men.
Those with low educational attainment were more likely to be older, African American, to work in more physically demanding occupations, to have more radiographic and symptomatic knee OA, and to be less likely to report current alcohol use (Table ). Men with low educational attainment were more likely to smoke and have lower BMIs than were men with higher educational attainment, whereas the opposite was true for women and postmenopausal women. Postmenopausal women with low educational attainment were less likely to report use of an HRT (Table ).
Characteristics of the study sample according to gender and educational attainment
For men, results of multivariable modeling indicate that the associations of low educational attainment with radiographic knee OA outcomes were explained by age and race, but the association with symptomatic knee OA persisted (Table ). Men with low educational attainment have 86% higher odds of having symptomatic knee OA than do men with higher educational attainment, after additional adjustment for BMI, knee injury, current smoking, current alcohol use, and occupational factors.
Odds ratios [95% CI] for educational attainment associated with four knee OA outcomes
For women, the results of multivariable modeling showed that the association of low educational attainment with all four knee OA outcomes persisted after adjustments for all covariates (Table ). Compared with women with higher educational attainment, women with low educational attainment had about 50% to 85% higher odds of having the knee OA outcomes (Table ).
To form the subset of postmenopausal females, about 240 women were removed (average age of 50 years and 23% low educational attainment). Overall, the results from the subset tend to show smaller odds ratios (ORs) for education than those seen from the set of all females (Table ), but associations with bilateral radiographic OA and unilateral and bilateral symptomatic OA remain significant. The OR for education with radiographic knee OA narrowly misses significance (P = 0.066). However, adjustment by the occupation variables actually results in nonsignificance (P = 0.054), and adjustment by HRT use effects only a minor additional change. In general, the findings for the subset of postmenopausal women were similar to those for all women, and adjustment for HRT did not have any substantial influence on the results.
The ORs for education presented in Table are calculated from a series of models ranging from unadjusted to fully adjusted by age, race, lifestyle, clinical, and occupation factors, and HRT for postmenopausal women. The ORs for all the independent variables from the fully adjusted models are displayed in Table and include the ORs for education appearing in Table . Among all the covariates, age, knee injury, and BMI were significant in all models. Smoking was protective for men in radiographic OA and bilateral radiographic OA (but not for women). Alcohol was protective for women in OA and symptomatic OA (but not for men). Race did not appear to be an important factor. Occupation variables were never significant, with P values > 0.2 in all models. Finally, HRT was protective for postmenopausal women in knee OA (OR = 0.58) and bilateral knee OA (OR = 0.44).
Odds ratios [95%CI] for all independent variables associated with four knee OA outcomes
In exploratory models with all potential covariates, run separately by gender, no interaction terms for educational level with age group or race were found to be statistically significant at the 0.10 level, by using the Bonferroni correction for multiple tests, and so no further stratification was necessary.
Finally, as a check on the effect of missing values, particularly the Census occupation code, characteristics prepared from the original sample of 3,047 participants showed no notable differences from those presented in Table . A modeling analysis was repeated but with removal of the grouping of occupational variables, leaving a sample of 2,914. Results were similar in OR magnitudes to those of Table , and the findings of significance were unchanged, except, in the case of postmenopausal women adjusted for HRT, formal significance of the OR for education associated with knee OA was restored (P = 0.034).