Rheumatology patients' use of complementary therapies: Results from a one-year longitudinal study


  • Jaya K. Rao,

    Corresponding author
    1. Durham Veterans Affairs Medical Center and Duke University Medical Center, Durham, North Carolina
    Current affiliation:
    1. Health Care and Aging Studies Branch, Centers for Disease Control and Prevention, Atlanta, Georgia
    • CDC, 4770 Buford Highway, NE, MS K-51, Atlanta, GA 30341
    Search for more papers by this author
  • Kurt Kroenke,

    1. Indiana University School of Medicine and Regenstrief Institute for Health Care, Indianapolis, Indiana
    Search for more papers by this author
  • Kimberly A. Mihaliak,

    1. Roudebush Veterans Affairs Medical Center, Indianapolis, Indiana
    Search for more papers by this author
  • Steven C. Grambow,

    1. Durham Veterans Affairs Medical Center and Duke University Medical Center, Durham, North Carolina
    Search for more papers by this author
  • Morris Weinberger

    1. Roudebush Veterans Affairs Medical Center, Indiana University School of Medicine, and Regenstrief Institute for Health Care, Indianapolis, Indiana
    Current affiliation:
    1. University of North Carolina School of Public Health, Chapel Hill, North Carolina
    Search for more papers by this author

  • The views expressed herein are those of the authors and do not necessarily represent those of the Department of Veterans Affairs.



To examine the natural history of complementary and alternative medicine (CAM) use and its impact on outcomes within a cohort of rheumatology patients.


Consecutive patients were recruited from 3 university and 3 private rheumatology practices. Baseline chart reviews provided demographic information and rheumatic diagnoses. Patients answered questions on CAM use and health status during 1 year. We identified correlates of 4 CAM usage patterns (started, maintained, stopped, nonuse) and compared outcomes among these groups.


Of 232 baseline participants, 203 (87%) and 177 (76%) responded to the 6- and 12-month surveys. In each survey, approximately 34% reported currently using CAM. During the year, 44% of patients remained nonusers whereas 12% started, 22% maintained, and 22% stopped use. The most frequent reasons for stopping CAM were lack of effectiveness and expense. CAM users and nonusers had no difference in outcomes.


Arthritis patients' usage behavior varied substantially, but CAM use was not associated with a difference in outcomes.


Complementary and alternative medicine (CAM) has attracted widespread interest among the public, health care providers, and policy makers. In 1990, 1 in 3 Americans reportedly used CAM, and they made more visits to CAM practitioners than primary care physicians (1). In a 1997 update, 4 in 10 Americans reportedly used CAM and spent an estimated $27.0 billion on these treatments (2).

Despite numerous studies documenting the prevalence (3–13) and associated costs (1–3, 9, 12, 14) of CAM, gaps remain in our knowledge of this phenomenon. First, because previous studies obtained information from patients at a single time point, we have limited understanding of the natural history of CAM use and the secular trends in use of certain CAMs that may occur as newer therapies emerge. Second, although several studies assessed patients' reasons for using CAM (1–5, 13, 15–19), only 1 (14) described why they stop using CAM. It is unclear whether costs or adverse effects influence these decisions to discontinue use. Finally, except for a few clinical trials, the relationship between use of most CAMs and clinical outcomes is uncertain. Although randomized controlled trials are necessary to rigorously evaluate the efficacy of individual CAMs, epidemiologic surveys may provide information on whether CAM use in general influences clinical outcomes.

Rheumatic conditions, such as osteoarthritis, rheumatoid arthritis, and fibromyalgia, offer an optimal disease framework for providing insights into the natural history of CAM use and its impact. These conditions are prevalent (20), are characterized by a variable course, and often adversely impact functional status. Furthermore, arthritis patients often use CAM (7, 21–32). In 1999, we reported a cross-sectional description of CAM use among 232 patients with rheumatic conditions (33). We subsequently followed this cohort for 1 year, administering the same survey at 6 and 12 months. In this article, we describe 1) patterns of current use of individual CAMs at baseline, 6 months, and 12 months; 2) patients' reasons for discontinuing CAM; 3) who starts, stops, or continues use; and 4) the impact of CAM use on functional status, provider satisfaction, learned helplessness, and severe pain.



The institutional review board of Indiana University approved this study. Consecutive established outpatients who kept scheduled appointments with a rheumatologist during a 2-week period in 1997 were potentially eligible. The patients were recruited from 3 university and 3 private rheumatology outpatient practices. The university sites included a municipal hospital that provides care to medically and socioeconomically vulnerable patients, a Veterans Affairs Medical Center, and a fee-for-service specialty practice. Patients were excluded if they were new to the practice, <18 years old, demented, or a nursing home resident.

At study enrollment, 2 research assistants screened patients' medical charts for eligibility prior to each clinic session (33). Patients agreeing to participate underwent the baseline telephone survey, which occurred within 2 weeks of their visit. The 6-month survey was administered by telephone and the 12-month survey was a written questionnaire.

CAM definition.

CAM was operationally defined as any intervention not typically prescribed or recommended by physicians (i.e., herbs, magnets, dietary supplements, etc.). Relaxation techniques, exercise programs, over-the-counter salves (i.e., Aspercreme), daily multivitamins, folate, or calcium were not included because physicians might recommend such modalities for pain management (34–36) or health maintenance. Because we were interested in interventions with costs attached, we did not consider self prayer as CAM.

Study procedures.

At study enrollment, chart audits were performed to obtain demographic information and all rheumatic diagnoses for every eligible patient. In each survey, respondents answered questions on their functional status using the Modified Health Assessment Questionnaire (37), a 10-point numerical pain scale (38), medications prescribed for their rheumatic conditions, 4-item psychological scale (4 items on sleep, anxiety, stress, and depression) (39), learned helplessness scale (40), and provider satisfaction (41). Finally, patients were asked if they used any other treatments for their rheumatic condition that their doctor did not prescribe. The survey contained a structured list of 12 different CAMs with examples of each and open-ended questions to elicit information on treatments not listed. We avoided using “alternative,” “unconventional,” or “complementary” when describing CAM.

In each survey, patients who reported using CAM for their rheumatic condition were asked if they were using CAM currently or were past users. In the 12-month survey, patients reporting any past use of a CAM at any time were asked to describe all reasons for stopping use.

CAM usage patterns.

At baseline, 6 months, and 12 months, we defined current users as those who reported using CAM at the time of the survey. Patients reporting past CAM use were considered past users, even if they used CAM within the year. Those who reported never using CAM were considered never users for that time point. With this classification scheme, we found that 32 patients had inconsistent CAM usage patterns during the year. For example, 1 patient reported current use at baseline, past use at 6 months, and never using at 12 months; 2 patients reported current use at baseline and 6 months but never using at 12 months. Except for site of care (P = 0.02), inconsistent and consistent responders did not differ in terms of severe pain, disease type, functional status, education, race, sex, or age.

Because we had more confidence in reports of current use than past or never use, we classified patients as current CAM users or nonusers (i.e., past or never used) at each time point. Based on this classification, patients were then classified into 1 of 4 usage patterns (Table 1) during the year. Patients who were baseline nonusers but currently using by 12 months were categorized as starting use. Those who were baseline current users but nonusers by 12 months were considered as stopping use. Current users at baseline and 12 months were classified as maintaining use and nonusers at baseline and 12 months were categorized as nonusers.

Table 1. Classification of complementary therapy usage patterns during the followup period*
PatternStatus at baselineStatus at 12 months
  • *

    Patients' usage status was categorized as current use (reported CAM use at time of survey) or nonuse (reported either past CAM use or never used). CAM = complementary and alternative medicine.

Started CAM useNonuserCurrent user
Maintained CAM useCurrent userCurrent user
Stopped CAM useCurrent userNonuser

Disease-related variables.

We examined relationships between CAM usage and disease-related factors, such as rheumatic diagnoses, rheumatic therapy, and severe pain. Because patients could have multiple diagnoses, we classified patients according to a 4-level hierarchical disease variable (33): rheumatoid arthritis, fibromyalgia, osteoarthritis, or other rheumatic conditions based on the rheumatologist's diagnosis at study enrollment. Patients were classified as receiving disease-modifying/corticosteroid treatment if they reported taking 1 of the following medications: gold salts, corticosteroids, methotrexate, hydroxychloroquine, azathioprine, sulfasalazine, cyclosporine, cyclophosphamide, or penicillamine. Severe pain was defined, a priori, as ≥5 on the 10-point numerical pain scale.

Statistical analyses.

All analyses were performed using PC-SAS version 6.12 for Windows (SAS Institute, Cary, NC). We compared baseline characteristics of responders and those lost to followup using chi-square tests (categorical variables) or t-tests (continuous variables).

The analyses were conducted in 2 phases. First, we performed bivariate comparisons between CAM usage patterns (started, stopped, maintained, nonuse) and the following variables: demographics (age, sex, race, education), rheumatic diagnosis, disease duration, disease modifying/corticosteroid therapy, psychological status, and site of care. We used Pearson's chi-square tests (exact P values whenever low cell counts were present) for categorical variables and one-way analysis of variance tests for continuous variables. These analyses were limited to patients who responded to all 3 surveys.

Second, we fit models for 4 outcomes: functional status, learned helplessness, provider satisfaction, and severe pain. The first 3 outcomes were continuous variables and analyzed using linear mixed models (42), and the fourth, severe pain, was a categorical variable and analyzed using a marginal model (43). Each model examined the associations between the outcome variable and CAM usage pattern at baseline and during the year. Because both modeling strategies are appropriate for longitudinal data and can accommodate missing data, case-wise deletion of patients with missing responses was not necessary. Thus, all baseline participants were included in these analyses. As a secondary analysis, we repeated these models after excluding the 32 patients with inconsistent CAM usage patterns during the year.


Current CAM users.

Of the 232 baseline participants, 203 (88%) and 177 (76%) responded to the 6- and 12-month surveys, respectively. Of the 12-month survey participants, 174 responded to all 3 surveys, and 3 responded to only the baseline and 12-month surveys. The 55 nonrespondents were similar to the responders according to age, sex, race, education, rheumatic diagnosis, disease duration, site of care, functional status, and satisfaction. The respondents' average age was 55.5 years; 74% were female, and 89% were white. Nearly half (48%) were seen in a university clinic; 19% had fibromyalgia, 15% had osteoarthritis, and 39% had rheumatoid arthritis. The mean disease duration was 11.1 years; 51% reported severe pain; and 123 patients (74%) were receiving disease-modifying/corticosteroid therapies.

Overall, the number of patients currently using at least 1 CAM at baseline, 6 months, and 12 months was 82 (35%), 69 (34%), and 61 (34%), respectively. Because some patients started and others stopped CAM use during the 1-year period, those reporting current CAM use at any 1 time point were not exactly the same patients: 29 patients reported current CAM use in all 3 surveys, 30 reported current use in 2 surveys (either consecutive or nonconsecutive), and 39 reported current use in only 1 survey. Although current use of individual CAM treatments was consistent across time (Table 2), we observed a trend in 1 treatment. Among those reporting dietary supplement use, 10 (63%), 11 (79%), and 13 (81%) were using glucosamine and/or chondroitin sulfate at baseline, 6 months, and 12 months, respectively.

Table 2. Frequency of current use of individual complementary therapies reported in the 3 surveys*
 Baseline (N = 232) n* (%)6 months (N = 203) n* (%)12 months (N = 177) n* (%)
  • *

    Because some patients reported using more than 1 complementary and alternative medicine (CAM), and usage behavior changed during the year, those reporting current CAM use in the 3 replications were not exactly the same people. Overall, 29 patients reported current use of a CAM in all 3 surveys, 30 reported current use in 2 surveys (consecutive or nonconsecutive), and 39 reported current use in 1 survey only. N = total number of participants in the survey; n = current users only.

Acupuncture2 (1)1 (1)0 (0)
Chiropractors11 (5)14 (7)13 (7)
Copper bracelets/magnets9 (4)8 (4)6 (3)
Dietary supplements16 (7)14 (7)16 (9)
Electrical stimulators15 (7)6 (3)12 (7)
Herbal remedies17 (7)15 (7)17 (10)
Minerals/megavitamins13 (6)16 (8)8 (5)
Natural healer2 (1)6 (3)1 (1)
Salves11 (5)9 (4)5 (3)
Special diets19 (8)13 (6)13 (7)
Spiritual healer3 (1)5 (3)5 (3)
Vinegar mixtures7 (3)5 (3)6 (3)
Other20 (9)14 (7)6 (3)

CAM usage patterns.

Overall, 21 of 174 patients started using CAM during the year. Another 38 patients (22%) were current users at baseline, but discontinued use by 12 months. Finally, 39 patients (22%) maintained use, and 76 patients (44%) were nonusers at all 3 time points. With respect to the usage patterns for individual CAMs, equal numbers of patients started, stopped, or maintained visits to chiropractors during the study period (Table 3). Although patients most frequently started herbal remedies and dietary supplements, an equal number discontinued these therapies. Megadose vitamins or minerals were most often discontinued.

Table 3. Patterns of use of individual complementary therapies during the 1-year followup*
 Started nMaintained nStopped n
  • *

    N = 174 participants in all 3 surveys. Some patients reported using more than 1 complementary therapy. Overall, 21 patients started, 38 stopped, and 39 maintained complementory therapy use during the study period. Started = nonuse at baseline and current use at 12 months; maintained = current use at baseline and 12 months; stopped = current use at baseline and nonuse at 12 months.

Copper bracelets/magnets335
Dietary supplements12410
Electrical stimulators737
Herbal remedies11610
Natural healer 15
Special diets8511
Spiritual healer23
Vinegar mixtures333

Why patients discontinued use.

In the 12-month survey, patients reporting any past use of a CAM at any time were asked their reasons for discontinuing use (Table 4). The 2 most common reasons patients gave for stopping use were perceived ineffectiveness and expense. For certain CAMs, some patients reported concerns about the therapy, including potential interactions with prescribed therapies, side effects, or the practitioner's credentials. For vinegar mixtures and topical therapies, patients reported adverse effects, such as stomach discomfort or skin irritation. Finally, several reported using CAM on a short-term basis for severe pain, or stopping after CAM alleviated their symptoms.

Table 4. Reported reasons for discontinuing individual complementary therapies (CAM)*
Type (number of patients)Made worse nDidn't help nToo costly nUsed short term nConcern about CAM nAdverse effects nOther n
  • *

    In the 12 month survey, patients who reported any past use of a CAM at any time were asked to provide all reasons for stopping use. Some patients reported multiple reasons.

  • The following patients did not provide their reasons for discontinuing CAM use: chiropractor (3 patients), copper bracelets/magnets (1 patient), herbal remedies (1 patient), minerals/megavitamins (1 patient), and vinegar mixtures (1 patient).

Acupuncture (10) 65    
Chiropractor (51)1019654 8
Copper bracelets/magnets (28)124    2
Dietary supplements (24)11711   2
Electrical stimulators (17) 942  2
Herbal remedies (21)1146 2 1
Minerals/megavitamins (12) 65   1
Natural healer (11)1451  1
Salves (23)21631 41
Special diets (6) 41  1 
Spiritual healer (4) 21   1
Vinegar mixtures (23)413   32
Other (10) 244  2

Correlates of usage patterns.

The baseline characteristics of the 4 usage groups did not differ with respect to sex, race, disease duration, disease-modifying/corticosteroid treatment, psychological status, or site of care (Table 5). Younger patients (P = 0.03) and college graduates (P = 0.001) tended to be current CAM users by 12 months (started or maintained). Rheumatoid arthritis patients were significantly less likely to use CAM (P = 0.005).

Table 5. Baseline characteristics of study participants according to their subsequent usage of complementary therapy during the 1-year followup*
VariableNonuser (n = 76)Started (n = 21)Maintained (n = 39)Stopped (n = 38)P
  • *

    N = 174 participants in baseline, 6-, and 12-month surveys. P values ≤ 0.2. NS = not significant.

Age, mean, years585251570.03
Female sex, n (%)52 (68)16 (76)31 (79)29 (76)NS
White race, n (%)69 (91)18 (86)35 (90)34 (89)NS
College degree, n (%)16 (21)9 (43)18 (46)3 (8)0.001
Disease duration, mean, years11.210.011.810.8NS
Disease modifying/corticosteroid therapy, n (%)55 (72)14 (67)24 (62)29 (76)NS
Psychological status, mean score1.
Site of care, university (%)37 (49)11 (52)19 (49)21 (55)NS
Type of disease    0.005
 Rheumatoid arthritis, n (%)38 (50)8 (38)9 (23)14 (37)
 Osteoarthritis, n (%)6 (8)2 (10)8 (21)10 (26)
 Fibromyalgia, n (%)12 (16)1 (5)13 (33)5 (13)
 Other disease, n (%)20 (26)10 (48)9 (23)9 (24)

Association between CAM use and outcomes.

In the marginal models, we examined the proportions of patients with severe pain based on their CAM usage pattern. Although all 4 groups ultimately had similar proportions with severe pain at 12 months, there appeared to be 3 different patterns during the year. Patients who stopped CAM had the highest proportion with severe pain at baseline (62%) but that proportion decreased (to 46%) by 12 months. Those starting CAM had the lowest proportion (33%) with baseline severe pain, but the proportion increased to 55% by 12 months. Finally, maintainers and nonusers had intermediate proportions with baseline severe pain (maintainers = 57%, nonusers = 53%) that remained relatively constant during followup. In the unadjusted model for severe pain, we compared the 3 usage groups (started, stopped, and maintained) to nonusers and found no significant differences either at baseline or over time.

For functional status and learned helplessness, the mean scores for the 4 usage groups were similar at baseline, and all appeared to increase slightly during the year. In the linear mixed models for functional status and learned helplessness, we found no significant differences between the usage groups at baseline or in the rate of change over time. For satisfaction, the 4 usage groups showed no baseline differences. Although the rate of change in satisfaction among the 4 groups was similar and significant (P = 0.01), the degree of change was not clinically meaningful (i.e., <5% change). When we excluded the 32 patients with inconsistent usage patterns from the models, we obtained the same results for all 4 outcomes.


This study contributes to the existing CAM literature in several important ways. First, we studied a clinically representative sample of arthritis patients from community and university rheumatology practices. Second, in contrast to previous cross-sectional surveys (1, 2, 6, 7, 12), we studied these patients longitudinally to assess changes in CAM use and patient-centered outcomes. Third, we asked former CAM users to describe their reasons for discontinuing use. At baseline, 6 months, and 12 months, one-third of patients reported currently using CAM. There was variation, however, in precisely who was using CAM at any particular time: some patients started, others maintained, and some stopped use during the year. Although lack of effectiveness and expense were the most frequently reported reasons for discontinuing use, some patients reported that CAM was effective in alleviating their symptoms and was no longer needed. Finally, similar to the few studies that have assessed outcomes in relation to CAM use (8, 44, 45), we found no clinically significant differences between users and nonusers with respect to pain, learned helplessness, functional status, or provider satisfaction.

Our longitudinal study revealed a potential measurement issue that is not apparent in cross-sectional surveys. Many investigations have made prevalence estimates based on self-reported CAM use during the past year (1–3, 6, 9, 12, 13, 28, 45, 46). Although patients' inability to recall distant “past” use might be expected, we were surprised to find that some did not recall their “current” CAM use 6 months later. Because the literature indicates that patients' recall of salient life events (i.e., hospitalizations) (47, 48) may fade with time, a shorter time window (i.e., <4 weeks) may provide more accurate estimates of CAM use for a particular time.

Because the same patients were surveyed over time, we also observed a trend in the use of glucosamine and chondroitin sulfate. These therapies were described in The Arthritis Cure (49), a book that received lay media attention immediately before our baseline survey. After our data collection ended in 1998, studies describing the modest treatment effects of glucosamine and chondroitin sulfate for osteoarthritis (50, 51) were published. Perhaps with a longer followup, we might have observed a larger increase in use of these or other therapies, such as S-adenosylmethionine (52).

The lack of relationship between CAM use and outcome has several potential explanations. One possible interpretation is that CAMs are ineffective (44). The patients did report that they discontinued use if, from their perspective, the CAM was ineffective. Second, 1 year may not be long enough to see significant changes in outcome, but this period is usually adequate to detect outcome changes for most rheumatic disorders. Third, perhaps a complex relationship exists between the patients' symptoms and decisions to use CAM. For example, a rapid increase in pain may influence some to seek CAM, while others who use CAM without seeing an appreciable difference clinically may be concerned that discontinuing treatment would cause a flare. Fourth, perhaps some patients' CAM use was fleeting so that its long-term impact was minimal. In such instances, a diary study that details daily symptoms and CAM use would be more precise than periodic questionnaires. Finally, the positive outcomes for the few patients using effective CAMs may have been overshadowed by negative outcomes for the larger numbers using ineffective treatments.

We must acknowledge several limitations of our study. First, we included only adult patients who had established relationships with rheumatologists. Although our results may not apply to individuals who have eschewed traditional arthritis care, prior studies indicate that many use CAM concurrently with conventionally prescribed therapy (3, 9, 14, 15, 19, 27, 28, 30, 32, 45, 53). Second, we did not ask detailed questions on number of doses, lifetime use, or CAM-related expenditures. We believed that patients' recall might not be accurate for such data. Because 18% of patients gave inconsistent responses to our general CAM questions during the 1-year period, this concern may be reasonable. Third, some patients might have been reluctant to reveal their CAM use. We tried to limit this potential bias by blinding patients to the study's intent. We did find a consistent reporting of overall use and use of individual CAMs across time. Finally, our sample size limited our ability to predict patterns of overall use or use of individual CAMs.

Our findings of variation in CAM usage behavior among arthritis patients have implications for researchers and clinicians. Researchers who analyze longitudinal survey data should consider such behavioral changes, particularly when relating long-term outcomes to CAM use at a given point. Similarly, since arthritis patients often use CAM, clinicians may need to periodically review their patients' current CAM regimens. Our observational data did not demonstrate an association between CAM use and outcomes. Patient-centered research on the natural history of CAM use among arthritis patients should continue in other clinical venues, such as primary care practices or other specialty settings. Such investigations will complement rigorously designed clinical trials and provide a more complete picture of CAM use to health care providers, policy makers, and patients.


The authors thank the 6 rheumatologists in Indianapolis who kindly allowed us to recruit their patients for this study. We also thank Maureen Reindl, BS, for her assistance with patient recruitment and data collection.