In a large population of participants with radiographic-confirmed OA of the knee, overall CAM use is prevalent, but less so among AA. After adjustment for sociodemographic and clinical characteristics, CAM use was less common among AA than CA, while no differences in use of conventional medications remained. In addition to differences in the overall prevalence of use, we observed differences in the specific types of CAM used by race. The cross-sectional data used in this study do not permit any evaluation of the effectiveness of CAM approaches on symptom management in knee OA, nor do we suggest these data are documenting health disparities. Rather, the unique data collected as part of the OAI (symptoms, severity, radiographic confirmation, detailed CAM assessments) are useful to further our understanding of racial differences in CAM use.
Our finding that AA are less likely to use CAM therapies than CA is consistent with some previous studies
[13, 29, 30]. Our work extends similar reports based on the general US population
[29, 30], as well as persons with self-reported arthritis
, to a population with radiographically confirmed knee OA. Our work is not consistent with other reports
[14, 17], however. Comparing results of CAM use across studies is challenging because researchers often use different definitions of CAM. The detailed questionnaire of the OAI related to CAM permitted us to place our study into context with the extant literature. We confirmed that indeed the discrepancies between our findings and results from several other studies may primarily be due to different definitions of CAM use. Katz and Lee reported that CAM was more common among AA than CA, which was driven by biologically based diets to treat OA (75.2% of AA used biologically based diet relative to 60.9% of CA)
. In our study, however, we found only 1% of AA and CA used this method. Another study reporting greater CAM use among AA relative to CA did not include massage and chiropractic service as CAM therapy
. Consistent with other studies
, we found that CA used more biologically based supplements, especially glucosamine and chondroitin, and chiropractic services relative to AA. Also consistent with the literature
[14, 16, 17], we found that AA used more spiritual and religious activities and topical agents relative to CA.
We found that, without adjusting sociodemographic and clinical characteristics, AA were nearly twice as likely as CA to use conventional medications. Greater scrutiny of the actual medications revealed that the largest differences were observed with over-the-counter medications such as acetaminophen and NSAIDs, whereas CA reported greater use of prescription COX-2 inhibitors relative to AA (9.3% versus 5.7%). This is consistent with previous work
[18, 19], as well as in equal access health systems requiring minimal co-payments for medications
. In our study and in others
[31, 32], AA experienced more severe pain and other symptoms than CA despite having less severe radiographic evidence of disease. Nevertheless, after controlling for clinical factors, we found similar use of conventional medications by race. Yet, despite these adjustments, the finding that AA were less likely than CA to use CAM either exclusively or with conventional medications persisted.
While the reasons for racial differences in CAM use are likely multifactorial, one possible reason for the differences we observed may be differential access to care by race. Compared to AA, CA have better access to health care services
. In equal access systems, no racial differences in overall frequency of OA related physician visits or visits to rheumatologists were observed among veterans with OA
. But in our study, CA were more likely to have reported X-ray examination of their knees prior to study entry relative to AA which may in turn have led to more treatment using CAM. Moreover, relative to AA, CA did appear to report more CAM approaches that required access to the medical system (practitioners), e.g., chiropractic service. Further, in our study, AA had less favorable socioeconomic positioning (as measured by marital status, education and income) relative to CA. CAM therapies are not covered by insurance. Indeed, among AA, those with income over $50,000 were more likely to use CAM therapies. Some of CAM approaches, such as glucosamine
 and acupuncture
, have been shown beneficial in relieving symptoms among OA patients. However, as found in our study, both glucosamine and acupuncture were less frequently used among AA than among CA. If further studies confirm the effectiveness in delaying disease progression and ameliorating symptoms, CAM therapies should be promoted and made accessible to minority populations.
We found an association between age and CAM use among AA, but not CA. Older AA were more likely than those younger than 65 years of age to report use of CAM, conventional medications or both, but this age effect was not found among CA. Our findings align with studies that report older AA hold more positive views about CAM and use CAM more frequently than younger AA
[15, 36]. The differential effects of age on using conventional medication may be explained by different perceptions between AA and CA on side effects of conventional medications
. AA are less likely than CA to recognize any risk associated with over-the-counter and prescription NSAIDs
. Indeed, AA are less likely than CA to report that their doctors have discussed NSAID-related gastrointestinal problems
. Given that NSAIDs constitute the majority of conventional medications used for treating OA, this could explain why there was no increase in conventional medication use among older CA compared to younger CA.
Our study has several limitations. First, we could only evaluate CAM practices in AA and CA. While others report that Hispanics and Asian Americans are more likely to use CAM therapies
[14, 38], we were unable to evaluate this in our data. Second, it would have been preferable to identify correlates for each type of CAM treatment, but we were unable to because of inadequate sample size. Such analyses would have been more useful for understanding potential areas for intervention
. Lastly, only self-reported information on CAM and conventional medication therapies was available. Use of treatments was based on a 30-day or 6-month recall. As such, it is possible that participants did not accurately report the use of treatments. Further, persons with OA of the knee are prone to recall bias of treatments
. Nevertheless, if present, the misclassification of participants to treatment approaches is likely to have been non-differential which would have diluted any observed associations.