Important advances in rheumatoid arthritis (RA) treatment began with the availability of effective disease-modifying antirheumatic drugs (DMARDs), such as methotrexate (1). After use of methotrexate became common, studies showed that its effectiveness was related to dose, and that in the absence of adverse effects, 20 mg weekly was the appropriate dosage level (1). In contrast, the effects of most earlier DMARD therapies were relatively dose independent. This advance in effective, dose-dependent therapy was followed by the general availability of biologic therapy and the later demonstration that combinations of methotrexate and biologic agents are more effective than either treatment alone in relieving symptoms, improving function, and retarding radiographic erosion (2–6). Studies have also shown that simultaneous use of multiple DMARDs leads to improved outcomes (7, 8), and, in some studies, treatment has resulted in profound improvement, even remission, in many users of biologic agents or combinations of biologic agents and DMARDs (2, 6, 9). Taken as a whole, these data imply that aggressive, high-dose, multiple-drug therapy, often including biologic agents, should be used to optimize treatment of RA.
A second important treatment-related advance is the use of validated measurement tools to assess response to therapy. When response is judged to be inadequate, more aggressive therapy is said to be indicated. Tools for measuring response include the Disease Activity Score (10), the Simplified Disease Activity Index (11, 12), the Clinical Disease Activity Index (11), and patient-based assessments such as the Health Assessment Questionnaire (HAQ) (13), the Rheumatoid Arthritis Disease Activity Index (14), and the Patient Activity Scale (PAS) (15). Such tools are suitable for use in randomized clinical trials and clinical practice. The PAS is a 0–10 measure of RA disease activity in which higher values indicate greater RA disease activity. It is computed by multiplying the HAQ by 3.33 and then dividing the sum of the scores on a visual analog scale (VAS) for pain, VAS for global assessment, and (multiplied) HAQ by 3.
Given the promise of effectiveness described above, one might expect that high-dose methotrexate, combination therapy, and biologic agents would be the treatment norm in clinical practice. However, that has not been the case. A number of factors might influence physicians' current treatment recommendations. Published literature on large clinical studies of biologic agents has been available for only ∼6 years, and this might simply not be a long enough time to communicate the putative benefits of these agents to physicians and patients, despite marketing efforts by manufacturers of the drugs. It might also be that physicians do not believe the results of clinical trials can be extrapolated to clinical practice, or they may be concerned about possible short- or long-term side effects, or about costs (16–18).
Another reason therapy is not as aggressive as often advocated may reflect issues or preferences that are important to patients but are not reflected in expert recommendations or in results of clinical trials. In this investigation we addressed patient preferences concerning RA therapy, starting with the often-heard patient comment, “As long as I don't get worse I'll stay with my current treatment.” We studied 6,135 RA patients with regard to their willingness or unwillingness to change therapies, their stated reasons for these decisions, including physician recommendations, and associated characteristics.
PATIENTS AND METHODS
- Top of page
- PATIENTS AND METHODS
- AUTHOR CONTRIBUTIONS
In January 2006 we surveyed 6,135 patients with RA who were enrolled in a National Data Bank for Rheumatic Diseases longitudinal observational study regarding satisfaction with RA therapy and factors that might lead patients to change therapy. Participants were recruited from the practices of US rheumatologists, and are followed up prospectively with semiannual detailed, 28-page questionnaires, as previously described (17, 19–22). RA was diagnosed by the patients' rheumatologists. We obtained sociodemographic data including age, sex, education level, income, and employment status. We also obtained scores on disease-related patient outcomes, including HAQ disability scores (23), pain, and global severity, and with these values, assessed RA disease severity according to the PAS. Information on current and past RA therapy was recorded.
Eleven questions specifically addressed issues regarding change of therapy and satisfaction with therapy. In question 1 (Table 1), patients reported whether or not they would change therapy, by endorsing the statement “As long as I don't get worse I wouldn't want to change my arthritis medications.” For simplicity in the descriptions below, we sometimes refer to the groups as those who would and those who would not change therapy, omitting the qualifier “as long as I don't get worse.” Questions 2–9 (Table 1), referred to below as “treatment predictor” questions, addressed reasons for patient unwillingness to change therapy. In question 10, patients rated their satisfaction with current treatment according to the following categories: very satisfied, somewhat satisfied, neither satisfied nor dissatisfied, somewhat dissatisfied, or very dissatisfied. Finally, in question 11, patients indicated, as a percent, how much more effective a medication would have to be before they would switch therapies.
Table 1. Treatment predictor variables and preference for not changing therapy
|Question||All patients (n = 6,135) (100.0%)*||Responses according to whether the statement “As long as I don't get worse I wouldn't want to change my arthritis medications” (question 1) was answered “true” or “false”|
|True (n = 3,914) (63.8%)*||False (n = 2,221) (36.2%)*||% difference (true − false) (95% CI)†|
|2. I don't need new medications because I am satisfied with the control I have over my arthritis.||53.3||73.6||17.4||56.5 (54.1–58.3)|
|3. I don't want the risk of side effects that might come from taking new medications.||72.5||87.1||46.2||40.1 (38.6–43.3)|
|4. I want to follow my doctor's suggestions, and my doctor thinks I don't need to change medications.||71.5||84.4||48.3||36.1 (33.3–38.8)|
|5. I am concerned that new treatments might not work as well and that I might lose control of my arthritis.||68.1||80.8||45.8||35.0 (32.5–37.3)|
|6. I don't think there are medications currently available that are better than the medications I am using now.||66.3||76.0||48.7||27.2 (24.9–26.7)|
|7. I don't want to take treatments that require injections or IVs.||35.7||41.5||25.5||16.0 (13.7–18.3)|
|8. I can't afford the cost of new medications.||42.7||45.7||35.7||8.1 (5.6–10.5)|
|9. Getting approval from my insurance company and the hassle of tests and medical visits for new drugs are important problems for me.||54.6||56.1||52.1||4.0 (1.0–6.8)|
Multivariable analyses of the role of treatment predictors (questions 2–9) and other variables in relation to willingness to change therapy (question 1) were performed by logistic regression (Table 2). We calculated Akaike's information criterion (AIC), a method for selecting among nested models by applying a penalty for each included variable. In general, the preferred model is the one with the lowest AIC value. The percent of missing values per variable in Table 2 ranged from 0 to 3.5% except for total household income, for which the percent of missing values was 8.7%. To perform the multivariable regression optimally, we used multiple imputation of missing values by chained equations (MICE), as described by Royston and Cox (24, 25). The 5 imputed data sets created by the MICE procedure were analyzed simultaneously.
Table 2. Multivariable predictors of unwillingness to change therapy*
| ||Odds ratio|
|Model 1||Model 2||Model 3||Model 4|
|RA control is satisfactory||7.32†||7.45†||7.42†||6.83†|
|Risk of side effects||4.49†||4.39†||4.39†||4.41†|
|Doctor thinks current medications are best||1.95†||1.90†||1.91†||1.86†|
|Fear loss of control||1.82†||1.82†||1.84†||1.84†|
|Can't afford new medication||1.51†||1.30‡||1.30‡||1.32†|
|There is no better medication||1.33†||1.34†||1.38†||1.38†|
|Don't want injections/IVs||1.16§||1.12||1.09||1.07|
|Too much hassle||0.85§||0.85§||0.85§||0.85§|
|Age (per 10 years)|| ||1.05||1.04||1.03|
|Sex (male)|| ||1.04||1.06||1.05|
|College graduate|| ||0.82§||0.81§||0.80‡|
|Total income (per $10,000 [US])|| ||0.95†||0.95†||0.94†|
|RA duration (years)|| ||1.00||1.00||1.00|
|Lifetime no. of DMARDs/biologic agents|| || ||1.03||1.03|
|Currently taking a biologic agent|| || ||0.84||0.82§|
|Currently taking prednisone|| || ||0.92||0.93|
|Currently taking methotrexate|| || ||1.20§||1.19§|
|VAS pain (0–10)|| || || ||0.95‡|
|HAQ disability (0–3)|| || || ||1.02|
Group comparisons for other variables were performed using linear regression or chi-square analysis when appropriate. Correlation coefficients were determined by the Spearman method.
The associations between treatment predictor variables and HAQ and PAS scores were measured using Kendall's tau-a and the area under the receiver operating characteristic (ROC) curve (26) (Table 3). ROC values can be interpreted as indicating no association (0.50–0.60), poor (>0.60–0.70), fair (>0.70–0.80), good (>0.80–0.90), or excellent (>0.90–1.00). Kendall's tau is related mathematically to the area under the ROC curve. The tau-a allows rough understanding of the degree to which variables such as satisfaction with arthritis control are simultaneously associated with variables such as HAQ and PAS scores. Kendall's tau has a simple interpretation, i.e., the percent concordance between variables. For example, the value of 0.155 shown in the first cell of Table 3 indicates that it is 15.5% more likely that a person with a low HAQ score will report being satisfied with control over arthritis activity than that a patient with a high HAQ score will report being satisfied with control over arthritis activity.
Table 3. Association of activity-associated treatment predictor variables with HAQ and PAS scores*
| ||Tau-a (AUC ROC)||95% CI|
|Question 2. I don't need new medications because I am satisfied with the control I have over my arthritis.|| || |
| HAQ||0.155 (0.656)||0.142–0.169|
| PAS||0.226 (0.726)||0.213–0.238|
|Question 3. I don't want the risk of side effects that might come from taking new medications.|| || |
| HAQ||0.026 (0.532)||0.013–0.039|
| PAS||0.054 (0.567)||0.040–0.067|
|Question 4. I want to follow my doctor's suggestions, and my doctor thinks I don't need to change medications.|| || |
| HAQ||0.073 (0.591)||0.060–0.086|
| PAS||0.119 (0.646)||0.106–0.132|
|Question 5. I am concerned that new treatments might not work as well and that I might lose control of my arthritis.|| || |
| HAQ||0.011 (0.514)||0.003–0.024|
| PAS||0.050 (0.557)||0.036–0.064|
Data were analyzed using Stata, version 9.2 (StataCorp, College Station, TX). P values less than 0.05 were considered significant. Confidence intervals (CIs) were established at 95%. All tests were 2-tailed.
- Top of page
- PATIENTS AND METHODS
- AUTHOR CONTRIBUTIONS
The median age of the study participants was 62.7 years, and the median duration of RA was 14.8 years. Of the 6,135 study participants (20.1% male, 79.9% female), 63.8% reported that they would not want to change therapy as long as their condition didn't get worse (Table 1), including 61.8% of the patients who were currently using biologic agents and 65.9% of those who were not. Because this difference was small, we combined current users and nonusers of biologic agents in the analyses described below, except as described. Of the 55.6% of patients who were taking methotrexate, the mean and median weekly dosage was 15 mg.
Table 1 presents the questionnaire statements about new (future) treatment and the relationship between such statements and unwillingness to change therapy. Generally, patients were concerned about the risk of side effects (72.5%) and losing control of their arthritis (68.1%). They indicated that they didn't want to change therapy because they were satisfied with their current arthritis control (53.3%), or felt that their doctor thought they did not need to change medications (71.5%), that there were no better medications than those they were currently taking (66.3%), and that the hassle of new tests and insurance approval was an important problem (54.6%). Fewer than half (42.7%) indicated that they couldn't afford new medications, and 35.7% did not want to take medications that required intravenous (IV) administration or injection.
For each of the above statements, the percent difference between patients who would change therapy and those who would not was further examined (Table 1). The greater the difference, the stronger the association with unwillingness to change therapy. A difference of 56.5% was noted for satisfaction with arthritis control, followed by 40.1% for risk of side effects. Cost of medications (8.1%) and hassle factor (4.0%) appeared to play almost no role in differentiating those who would change medications from those who would not. Following their physicians' instructions (36.1%), concern about loss of control (35%), no availability of better medication (27.2%), and not wishing to use IV medications or injections (16.0%) held the middle positions.
The associations noted in Table 1 were not independent; therefore, we examined these factors in multivariable logistic regression models (Table 2). In the simplest model (model 1), satisfaction with RA control (odds ratio [OR] 7.3) and risk of side effects (OR 4.5) were the dominant correlates of unwillingness to change therapy. Also important, though less so, were the physician's opinion (OR 2.0), fear of loss of control (OR 1.8), and inability to afford new treatment (OR 1.5). To this simple model, we added demographic and psychosocial factors (model 2). Greater income (OR 0.95 per $10,000 [US]) and college education (OR 0.8) were associated with willingness to change therapy, while being married was associated with not wanting to change therapy. There was only a slight change in the OR of treatment predictor variables as an effect of the additional variables in this model.
Addition of current use of prednisone, biologic agents, and methotrexate (model 3) resulted in essentially no change to the model. However, addition of HAQ and VAS pain scores to the model (model 4) did result in some changes. The use of biologic agents was associated with willingness to change therapy, as were higher pain scores. Methotrexate use was associated with unwillingness to change therapy. The final ORs for the treatment predictor variables were as follows: satisfaction with RA control 6.8 (95% CI 5.8–8.0), risk of side effects 4.4 (95% CI 3.8–5.2), physician's opinion 1.9 (95% CI 1.6–2.2), fear of loss of control 1.8 (95% CI 1.6–2.1), no better medication 1.4 (95% CI 1.2–1.6), and inability to afford new treatment 1.3 (95% CI 1.1–1.6).
To gain insight into the role of adverse effects in willingness or unwillingness to change therapy, we examined patient-reported side effects. Of interest, 22.4% of the patients in this study reported experiencing side effects attributed to specific treatments during the previous 6 months, and 14.1% reported side effects attributed to DMARDs, biologic agents, or prednisone. Of the DMARD, biologic agent, or prednisone side effects, 34.2% were reported to be severe and 42.7% to be moderate. Approximately two-thirds (65.5%) of the RA patients reported having had a side effect to an arthritis medication at some point during their lifetime. Patients reporting side effects were more likely to be unwilling to change therapy (OR 1.8 [95% CI 1.6–2.1]) and to be concerned about the risk of side effects (OR 1.2 [95% CI 1.1–1.3]).
We attempted to gauge quantitatively how much better a medication would have to be before patients would want to change therapy. Question 11 asked, “Imagine that a medication is being developed for your rheumatic disease condition. How much more effective would the medication have to be compared with your current medications to make you switch to it?” Choices of 0–100%, in 10 10-point ranges, were presented. Patients who reported not wanting to change therapy (question 1) indicated that on average the medication would have to be 75.8% better before they would change therapy, compared with an average of 52.3% among patients who would change therapy (P < 0.001). In addition, irrespective of patients' replies regarding willingness to change therapy, users of biologic agents responded that they would have to have a medication that was 67.4% better than their current therapy in order for them to be willing to switch therapies, compared with 65.7% for nonusers of biologic agents (P = 0.093).
Associations between treatment predictors and overall satisfaction with measures of RA activity and severity.
Table 3 presents selected treatment predictor variables and their relationship to RA activity as measured by the PAS, and to RA activity/severity as measured by the HAQ. All associations were weak. The strongest association was between the PAS score and satisfaction with RA control (tau-a = 0.226 [area under the ROC curve 0.726]). Weaker still were the associations of the PAS score with physician instructions (tau-a = 0.119 [ROC 0.646]), fear of side effects (tau-a = 0.054 [ROC 0.567]), and loss of arthritis control (tau-a = 0.050 [ROC 0.557]). For all questions examined, HAQ scores showed even weaker associations. These data indicate that patient measures of disease activity/severity are only weakly linked to decisions about therapy.
We also assessed the relationship between overall satisfaction with current therapy as measured by the 5-category satisfaction question (question 10) and the PAS measure of RA activity and similar measures. The correlations between satisfaction with current therapy and scores on such measures were as follows: PAS 0.465, patient global assessment 0.442, pain 0.439, Short Form 36 physical component scale (27) 0.386, HAQ 0.344. The weakness of the association between satisfaction with RA control (question 2) and RA activity/severity as assessed with the HAQ and the PAS is also illustrated in Figures 1 and 2, respectively. Considerable overlap is seen in these figures, i.e., many patients with HAQ or PAS scores that indicate unsatisfactory function or disease activity levels were satisfied with their RA control, while others with “good” scores for function or disease activity were dissatisfied with their RA control.
Figure 1. Distribution density of Health Assessment Questionnaire scores in relation to patients' reports of their degree of satisfaction with control of their rheumatoid arthritis (RA). Vertical lines at 0.5, 1.0, and 1.625 divide the histograms into compartments of mild, moderate, severe, and very severe functional states.
Download figure to PowerPoint
Figure 2. Distribution density of Patient Activity Scale scores in relation to patients' reports of their degree of satisfaction with control of their rheumatoid arthritis (RA). Vertical lines at 1.9, 3.8, and 5.6 divide the histograms into compartments of mild, moderate, severe, and very severe disease activity.
Download figure to PowerPoint
Use of biologic agents, preference for therapy change, and satisfaction with therapy.
After adjustment for age, sex, and duration of RA, current biologic agent users had a lifetime use of 4.5 DMARDs or biologic agents (95% CI 4.5–4.6), compared with 2.5 (95% CI 2.5–2.6) among patients not currently taking biologic agents. Current users of biologic agents were more likely to want to change therapy than those not currently taking these agents (OR 1.2 [95% CI 1.1–1.3]); however, this difference disappeared after adjustment for age, sex, and duration of RA (OR 1.1 [95% CI 1.0–1.2], P = 0.130).
Overall, 77.3% of all patients were very satisfied or somewhat satisfied with their medications (question 10) while 9.4% were somewhat or very dissatisfied. Satisfaction was greater among patients not taking biologic agents. On a 1–5 scale of satisfaction (with higher scores representing greater dissatisfaction), the average score was 2.01 among current biologic users and 1.88 among non–biologic agent users. This small difference was statistically significant by ordered logistic regression (P < 0.001) and remained significant after adjustment for age, sex, and RA duration or for age, sex, RA duration, HAQ score, and pain score.
- Top of page
- PATIENTS AND METHODS
- AUTHOR CONTRIBUTIONS
More than three-fourths of the patients in this study (77.3%) were satisfied with their RA medications. A lesser proportion (53.3%) indicated that they did not need new medications because they were satisfied with the control they had over their arthritis. However, 71.3% of the patients who were satisfied with their therapy had moderate or greater arthritis activity as assessed by the PAS, 67.3% had HAQ scores >0.5, and 46.9% had HAQ scores >1.0 (Figures 1 and 2). The correlations of the PAS and the HAQ scores, respectively, with satisfaction with therapy were 0.465 and 0.344, and correlations of PAS and HAQ scores with satisfaction with RA control were even weaker (Table 3). These data indicate that there is an important discrepancy between declared satisfaction with therapy and measured activity and functional status, and that clinical activity is not an adequate explanation for satisfaction with therapy. Such a discrepancy between overall satisfaction and RA disease assessments has been noted previously in an insightful study by Leeb and colleagues (28).
Being satisfied with therapy for RA appears to represent much more than current disease status. In particular, satisfaction with therapy may incorporate the idea of not getting worse. Stated otherwise, one might say, “Even if therapy did not improve me as much as I might wish, I am satisfied with my therapy because it appears to have stabilized my RA.” Given relatively stable RA and prior experience with RA treatments, maintenance of current status (not getting worse) appears to be given high priority by patients. In addition, changing treatment implies that there will be improvement, i.e., that there is a better treatment than the one now received. Of interest, 66.3% of patients did not think there was a better treatment than what they were now receiving, including 57.7% of those not receiving biologic agents.
The major results of our study appear in Table 2. The belief that RA control is adequate (OR 6.8) and the fear of side effects (OR 4.4) dominate the predictors of treatment change after adjustment for all sociodemographic and RA predictors. Prior studies have indicated that patients with RA have great concern about treatment side effects, often with the implication that these fears may be irrational. We noted that 22.4% of study patients reported experiencing, within the last 6 months, side effects attributed to specific treatments and 14.1% reported side effects attributed to DMARDs, biologic agents, or prednisone. The RA-related side effects were rated severe in 34.2% of the cases and moderate in 42.7%. Sixty-six percent of the RA patients reported having had a side effect to an arthritis medication at some point in their lifetime. These data suggest that concerns about side effects may be appropriate even if beliefs about side effects sometimes represent an inappropriate risk–benefit analysis by patients (29). The literature suggests that RA patients are risk aversive (30). Our data are consistent with this assessment but also suggest that within the categorization of risk aversion is the fear of tipping the balance from stable RA to worsening RA.
Aversion to treatment changes might also be related to nonmedical factors, such as cost, inconvenience, and administrative hassles. We found that although such issues were a general concern of RA patients, they played a small role in the thinking about changing therapy (Table 2).
A number of studies have addressed medication use from the patient's perspective (29–34), including a detailed questionnaire study by Neame and Hammond (35). The current study differs from earlier investigations in several important ways. First, our study focused on maintaining therapy and the reasons for not changing. In addition, we used patient-derived activity/severity scores to further understand patient preferences, and we addressed patients' concerns about therapy rather than approaching RA treatment from the perspective of physicians' concerns about adherence and compliance. Adherence and compliance may be much more important concepts in nonsymptomatic illnesses such as hypertension and osteoporosis. However, in symptomatic illnesses such as RA, ascertaining patients' impressions of the worth of treatment may be a more appropriate and realistic approach to understanding treatment value.
This study has a number of limitations. First, the questionnaire inquired about attitudes rather than behaviors, and it is possible that respondents might say one thing and do another. Second, patients' beliefs about effectiveness and side effects might be wrong. Even so, patients make decisions about their health and in the long run decide the worth of therapies. Patients' reluctance to change therapy may stem in part from wisdom about homeostasis that goes beyond the experimentation of clinical trials, and that is based on personal experience and preferences for the future. The idea of “not getting worse” may be an important outcome that has not been considered sufficiently in treatment-related decision making.
In summary, there is substantial discrepancy between RA patients' declared satisfaction with their treatment and measured disease activity and functional status. Most patients are satisfied with their therapy, even many with abnormal scores. Fear of loss of RA control and fear of side effects are major patient concerns. Maintenance of current status, rather than future improvement, appears to be a high priority for patients.
- Top of page
- PATIENTS AND METHODS
- AUTHOR CONTRIBUTIONS
Dr. Wolfe had full access to all of the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.
Study design. Wolfe, Michaud.
Acquisition of data. Wolfe, Michaud.
Analysis and interpretation of data. Wolfe, Michaud.
Manuscript preparation. Wolfe.
Statistical analysis. Wolfe, Michaud.