Maintenance of exercise in women with fibromyalgia

Authors


Abstract

Objective

To identify predictors of maintenance of exercise for women with fibromyalgia (FM).

Methods

Women with FM who had been randomized to the exercise arm of a clinical trial were studied prospectively during and 3 months following treatment. Subjects completed exercise logs weekly and returned the data via postal mail. Outcome variables were duration of aerobic and stretching exercises. Two separate multivariate models for longitudinal data were built with adjustment for in-treatment adherence and time. Pretreatment characteristics (self efficacy, pain, disability, stress, exercise barriers and benefits, and age) and changes during treatment (pain, disability, stress, and exercise barriers and benefits) were considered potential predictors of exercise maintenance.

Results

Stretching significantly decreased in the 3 months following treatment. High stress at baseline and increases in stress during treatment were associated with poor maintenance of stretching. Disability at baseline (measured with the Fibromyalgia Impact Questionnaire), an increase in barriers to exercise during treatment, and increases in upper-body pain during treatment were associated with worse maintenance of aerobic exercise in the 3 months following treatment.

Conclusion

The maintenance of an exercise program in women with FM appears to be contingent on being able to deal with stress, pain, barriers to exercise, and disability.

INTRODUCTION

Fibromyalgia (FM) is a rheumatic condition that is commonly diagnosed in middle-aged women. Cardinal symptoms of FM are widespread body pain and a minimum of 11 out of 18 tender points evident for at least 3 months (1). One stream of research has focused on single modality treatments for FM such as medications (2, 3) or exercise (4–6). Although pessimism prevails with regard to prognosis (7, 8), a consensus has emerged that exercise is one critical component of treatment for patients with FM (9). Typically physicians recommend that the patient become more active. Unfortunately, during a clinic visit, how to begin a physical activity program may not be adequately discussed; moreover, how to sustain exercise once it is initiated is rarely covered.

The literature pertaining to behavior change reveals that these tasks are complex and factors that influence them are likely to vary depending on various stages of change (10, 11). King and Martin (12) note that in the general population, a mix of personal (e.g., past experience with exercise, health status), behavioral (e.g., skills), and environmental (e.g., access to facilities, type of program) factors influence both uptake and maintenance of exercise. For example, Litt et al (13) studied 189 older women for whom predictors of uptake of exercise (first 3 months) were readiness for exercise and social support for exercise; the only predictor of maintenance of exercise (over 12 months) was self efficacy for exercise adherence. In a large-scale study of community-dwelling older persons, Burton et al (14) reported that the initiation of physical activity was significantly related to age, current health status, and the patient's belief in the importance of physical activity. Predictors of maintenance (over 4 years) of physical activity were education, moderately good or excellent health, and the belief that exercise contributed to good health. Persons with emotional distress were less likely to continue to exercise. However, it is unknown if predictors of uptake and maintenance of exercise differ for nonhealthy populations, such as individuals with FM.

The time when one measures exercise is important because adopting new behaviors is a complex process and maintenance of change is dynamic (15). King and Martin (12) noted that among sedentary adults who began an exercise program, half dropped out between 3 and 6 months. Apparently this is the critical period when attrition is most likely to occur. McAuley's study (16) of 82 sedentary older adults who began an exercise program and were followed for 4 months posttreatment underscores the importance of this period.

Indeed, the learning theory perspective on lapse, relapse, and maintenance of behavior change (17) suggests that the determinants of uptake and maintenance of exercise are likely to differ. However, the dominant theoretical approaches to understanding behavior change (e.g., health belief model, social-cognitive theory, theory of reasoned action) offer little guidance as to how the stages of change may differ (18). Logically, it seems that although favorable expectations regarding future outcomes are crucial to initiation of action (e.g., start to exercise), the decision to maintain the behavior depends on the individual's perceived satisfaction with the outcome (e.g., less fatigue, increased fitness) as well as environmental contingencies that serve to reinforce or extinguish behaviors (e.g., stressful life events) (15).

Anecdotal reports suggest that patients with FM limit their exercise due to pain and fatigue (19, 20). Very little empirical work has been published that sheds light on uptake and maintenance of exercise in this patient population. It may be that predictors of these behavioral changes in patients with FM differ from those of healthy individuals because patients with FM often have comorbid conditions, perceive themselves as disabled, and often have psychological conditions that may further impede exercise. In an earlier study with the cohort presented herein, we noted that women with FM who started a home-based individualized exercise program for 12 weeks were more likely to engage in the program during treatment if they were less physically fit at baseline (21). Predictors of adherence differed between the 2 types of exercise that were studied: stretching and aerobic. Higher baseline lower-body pain predicted less stretching, whereas higher upper-body pain predicted more aerobic exercise over time. During the 12-week program, women with higher baseline physical fitness and/or older age reduced aerobic exercises more, as did women with higher baseline stress.

The purpose of this investigation was to extend this initial work by examining the variables that contributed to maintenance of these 2 types of exercise for the 3-month critical period following termination of the program, while statistically controlling for exercise during the program. We hypothesized that stress and pain following exercise would negatively influence maintenance of exercise, and that greater self efficacy for exercise and beliefs that exercise would be beneficial would positively influence maintenance of exercise.

SUBJECTS AND METHODS

Subjects.

Women who met the American College of Rheumatology criteria for primary FM (1) were included in the study. Subjects were recruited by rheumatologists through letters sent to patients followed at a hospital rheumatology clinic and community rheumatology practice, or through newspaper advertisements. Subjects were invited to participate in a randomized clinical trail (the results are reported elsewhere [22rsqb;). Exclusion criteria were concomitant diseases that precluded participation in an exercise program, contraindication to exercise identified by the examining physician, recent change in medication (prior 2 weeks), and regular participation in moderate intensity exercise (3.0–6.0 metabolic equivalents [METs], where one MET represents the metabolic activity of an individual at rest and is set at 3.5 ml of oxygen consumed per kilogram body mass per minute or ∼1 kcal/kg/hour) for at least 30 minutes, 3 or more times a week at the time of study entry (23).

Procedures.

The study was approved by the McGill University Health Centre Institutional Review Board prior to commencement. Subjects were informed about the study procedures and gave written consent. They were examined by a physician and subsequently underwent a cardiovascular fitness test. A certified exercise physiologist developed an individualized exercise program for each subject. Next, the self-report questionnaires were explained by the project coordinator with instructions on how and when to complete them. Exercise logs were returned via postal mail, along with selected measures pertaining to FM symptoms and variables hypothesized to be related to adherence to exercise.

Exercise program.

During the 12-week training phase (supervised period), patients met 4 times with the same exercise physiologist. The first visit lasted ∼90 minutes with 30-minute followups scheduled for weeks 1, 3, and 9 following the initial session. The initial visit included a review of the cardiovascular fitness test results, a brief overview on the benefits of exercise, an individualized exercise prescription, and a supervised exercise training session. The principles of warm-up and cool-down along with basic stretching exercises and general exercise precautions were reviewed to minimize the risk of injury

The exercise prescription was individualized and followed guidelines from the American College of Sports Medicine (ACSM) for developing and maintaining cardiorespiratory fitness (24). These guidelines suggest that individuals perform 60–120 minutes/week of aerobic exercise within their target heart rate zone (60–85% of maximal heart rate). Duration is dependent on the intensity of the activity, so that individuals exercising at a lower intensity exercise for a longer duration. The individualized approach allowed for flexibility not only in the intensity and duration, but also in the frequency of sessions and the mode of aerobic exercise. Programs were tailored to the individual depending on severity of FM, accessibility to equipment, time constraints, and enjoyment of various activities. The intensity of the exercise began at 60–70% of maximal heart rate for all individuals and was gradually increased to as high as 75–85% of maximal heart rate depending on the subject's adaptation to the exercise. The followup sessions with the exercise physiologist during the 12-week training phase consisted of providing guidance and support to the subjects, solving any difficulties, and gradually increasing the intensity of the exercise. An unsupervised period followed in which data were collected pertaining to exercise participation using the same methodology as described during the supervised period (21).

Exercise logs.

Subjects were asked to complete an exercise log following each exercise session, each week during the supervised phase of treatment and during the unsupervised phase. Each log included information pertaining to the type of exercise performed (stretching, aerobic), frequency, duration, and heart rate. This methodology has successfully been used for a period of 24 months (25) and has been adequately validated through examination of changes in treadmill exercise performance, use of a solid-state portable microprocessor (Vitalog) to record heart rate and body movement, and comparisons of recorded exercise heart rates and ratings of perceived exertion throughout the study period (26). A stamped envelope was provided for weekly return of the logs during treatment and monthly thereafter. The project coordinator was vigilant with regard to compliance with the research protocol; if logs were not promptly returned, she telephoned subjects and encouraged them to complete and return the exercise logs, irrespective of whether the subject had exercised for that week or month.

Outcomes.

Data from the exercise logs were summed to yield separate monthly measures of the total number of minutes spent exercising (stretching or aerobic) in the first, second, and third month following the end of the active treatment.

Putative determinants of maintenance of exercise.

Baseline characteristics.

Age, educational level, work status, income, and marital status were collected at baseline. Body weight was assessed to the nearest one-tenth kilogram with a balance-beam scale. Body mass index (BMI) was calculated by dividing weight (in kilograms) by squared height (in meters).

Cardiovascular fitness.

At baseline, all subjects performed a physician-supervised maximal graded exercise stress test on a treadmill to determine current level of fitness. Procedures used in the fitness screening were based on the ACSM guidelines (27). Using a protocol developed by Bruce et al (28), the test consisted of increasing workloads of ∼3 METs every 3 minutes until the participant reached volitional exhaustion or any of the ACSM indications for stopping an exercise test (29). A 12-lead electrocardiogram measured heart rate and rhythm at rest, continuously during the exercise test, and for 5 minutes during recovery. Fitness was evaluated by time on test and maximal MET capacity.

Physician assessment.

A rheumatologist assessed tender points, duration of symptoms, and time since diagnosis of FM. A physician global assessment (PGA) of disease activity using a 100-mm visual analog scale (VAS) (30) was obtained at baseline. A review of 24 randomized controlled trials in FM identified the PGA as the outcome most likely to respond to treatment (31), i.e., the measure that was most sensitive to change.

Self-reported FM symptoms.

Pain intensity was recorded at baseline and weekly during the 12-week training phase and followup period. Subjects were asked to indicate their pain intensity over the past week on a 100-mm VAS (30) in 6 areas of the body: neck and shoulders, chest, upper/lower back, arms, buttocks, and legs (where 0 = no pain and 100 = severe pain). The scores were averaged across body sites to reflect the mean upper and mean lower pain intensity score.

Disability.

The Fibromyalgia Impact Questionnaire (FIQ) (32) is a reliable, validated, self-administered measure of functioning in the past week. The first 10 items address ability to carry out tasks that require physical strength; these items are summed and divided by the number of valid scores to yield a physical functioning score. Two items ask respondents to circle the number of days they felt good, as well as the number of days of missed work. Seven items pertaining to common FM symptoms (e.g., morning stiffness, fatigue) are measured on 100-mm VAS scales. A total score is created with higher scores indicating greater disability. Test-retest reliability coefficients for each item ranged from 0.56 to 0.95 (32). The FIQ was completed at baseline, posttreatment, and 3 months posttreatment.

Weekly Stress Inventory (33).

The Weekly Stress Inventory (WSI) is a self-report instrument assessing the frequency and stressfulness of minor stressors that respondents have experienced over the past week. For all 87 items on this questionnaire, respondents are asked to indicate whether the event occurred in the past week and to rate the perceived stressfulness of the experienced event on a 7-point scale (where 1 = occurred but was not stressful and 7 = extremely stressful). The WSI yields a total WSI impact derived by summing the perceived stress ratings. The questionnaire was designed to avoid problems such as items being confounded with psychological symptoms of distress, insensitivity to subtle fluctuations in stress levels, and contamination from retrospective reports. The WSI has good psychometric properties, as seen in a cardiac rehabilitation patient population (34). This measure was returned along with the exercise logs.

Self efficacy.

Self efficacy was measured using the Arthritis Self-Efficacy Scale (35). Two of the 3 subscales, self efficacy for pain management and self efficacy for other (FM) symptoms, were used in this study. The construct and concurrent validity of this scale has been demonstrated (36). Self efficacy assessed with this scale has been linked to better outcomes among patients with FM (37). This measure was administered to all subjects at baseline, posttreatment, and 3 months posttreatment, with the term “arthritis” being replaced with “fibromyalgia.”

Exercise Beliefs Questionnaire.

The Exercise Beliefs Questionnaire, an instrument developed by Gecht et al (38) for patients with arthritis (replaced herein with FM), includes 20 items that address self efficacy for exercise, barriers to exercise, benefits of exercise, and impact of exercise on FM. This measure was administered to all subjects at baseline, posttreatment, and 3 months posttreatment.

Statistical analyses.

Descriptive statistics used to summarize the distribution of baseline values of all relevant characteristics included mean, median, and SD for quantitative variables and proportions for categorical variables. Primary analyses focused on patient characteristics associated with better maintenance of the exercise during 3 months after the end of the supervised training phase. Separate analyses were carried out for 2 measures of exercise intensity: duration of aerobic and stretching. The values of each outcome, observed during the first, second, and third months after the end of the supervised phase, were considered repeated measures of a quantitative dependent variable. Each subject contributed to the analysis only for those times when the subject was actually assessed for the outcome of interest. To account for the likely correlation between outcomes observed at subsequent assessments of the same subject, we relied on generalized estimating equations (GEE) extension of conventional multiple linear regression for longitudinal data (39). In GEE analyses, we assumed the autoregressive order 1 structure of the covariance matrix of the residuals, which implies that correlations between observations that are closer in time are stronger than for observations several months apart, a standard assumption in longitudinal studies (40).

With a relatively small sample size, we could not include all available subject characteristics as independent variables in the same model. Therefore, the decisions regarding which candidate independent variables were included in a given multivariable model were based on a combination of a priori considerations and statistical criteria. First, it was decided a priori that the posttreatment adherence to a given type of exercise would depend on the intensity of the corresponding exercise during the supervised treatment phase. Secondly, it was expected that the exercise intensity may systematically change with increasing posttreatment time (i.e., unsupervised phase). Based on these considerations, the average value of the respective outcome across the 12 weeks of the training phase, referred to as “in-treatment adherence,” and an ordinal variable indicating the number of months elapsed since the end of treatment, referred to as “time,” were included in all multivariable models, regardless of their statistical significance. Adjustment for the in-treatment adherence implied that the estimated effects of other variables represented their effects among subjects who had the same level of exercise intensity during the supervised training, so that the resulting model really focused on the predictors of exercise maintenance. In contrast, inclusion of the time effect allowed us to assess if there was a systematic trend to increase or decrease exercise with increasing time since the end of the supervised phase, and to ensure that outcomes of different subjects were compared at the same time.

The decisions regarding inclusion of other characteristics (age, working status, depression, FIQ score, upper- or lower-body pain, fatigue, stress, barriers, beliefs, PGA, and METs at baseline fitness test) in the final multivariable model were based on a multistep procedure, resembling the stepwise selection process. To better understand how characteristics such as disability (FIQ), changes in upper- or lower-body pain, stress, barriers, and beliefs may affect the intensity of the unsupervised exercise phase, we represented their effects by 2 interrelated variables, one representing the unsupervised phase baseline (pretreatment) intensity value and another representing the corresponding change during the 12-week training phase, calculated as the difference between the values observed at 12 weeks and at baseline. The effects of age, depression, PGA, and METs at baseline stress test were limited to their baseline values because they were not measured thereafter. First, we estimated a series of relatively simple models, each including the measures of in-treatment adherence and time, as well as 1 (e.g., age) or 2 variables (e.g., baseline stress and change in stress during training) related to a given putative predictor of posttreatment exercise intensity. Then, a multivariable model was built by selecting those “candidate predictors” that were statistically significant or marginally significant (P < 0.15) in the simpler models, with a 2-variables group being entered whenever the baseline and/or the change variable met the significance criterion. Therefore, the final multivariable model included only those putative predictors that were at least marginally significant, as well as the in-treatment adherence and time that were forced into the model regardless of their statistical significance. A significance level of 0.05 was used for testing hypotheses based on the final multivariate model. All GEE analyses were performed using the GENMOD procedure available in the SAS version 8 statistical package (SAS Institute, Cary, NC).

RESULTS

Subjects.

Thirty-nine women were randomized to the home-based exercise program. At 3 months posttreatment, 33 (85%) women were still participating in the study protocol. The average age of subjects was 49.2 years. At study entry, the mean symptom duration was 10.5 years and the mean time since diagnosis of FM was 3.8 years. The average BMI was 28.0, indicating that, on average, subjects were overweight. The mean ± SD METs achieved on the cardiovascular stress test were 8.9 ± 2. The distribution of sociodemographic, disease-related, and fitness variables at study entry are summarized in Table 1. There were no significant differences between the 33 women who remained in the study compared with the 6 who dropped out (data not shown).

Table 1. Participant characteristics at baseline (n = 39)*
CharacteristicsBaseline value
  • *

    Values are the mean ± SD unless otherwise indicated.

  • Visual analog scale 0–100. Higher scores indicate greater disease activity.

  • Higher metabolic equivalents (METs) indicate better cardiovascular fitness.

Age, years49.2 ± 8.7
Education, years14.0 ± 2.8
Marital status, no. (%) 
 Single9 (23.1)
 Married/cohabiting25 (64.1)
 Divorced/separated4 (10.3)
 Widowed1 (2.6)
Work status, no. (%) 
 Not working16 (41.0)
 Working full time14 (35.9)
 Working part time9 (23.1)
Fibromyalgia duration, years10.5 ± 8.4
Time since diagnosis, years3.8 ± 4.5
Physician global assessment49.3 ± 18.3
Tender points12.8 ± 4.6
Body mass index28.0 ± 6.0
METs at baseline8.9 ± 2.0

As is evident in Table 2, the weekly average of minutes spent stretching decreased from almost an hour to slightly more than one-half hour for the 3 months following treatment. The average weekly number of minutes of aerobic exercise was stable for the first 2 months posttreatment (∼2 hours), but decreased to ∼1.5 hours by the third month.

Table 2. Duration (mean ± SD) of aerobic and stretching exercise at 1, 2, and 3 months post active training phase
 1st month2nd month3rd month
Stretching   
 Total minutes228.5 ± 219.8157.2 ± 179.9146.40 ± 179.9
 Average weekly57.1 ± 55.039.3 ± 45.036.6 ± 45.0
Aerobic   
 Total minutes459.7 ± 328.3463.0 ± 457.4380.5 ± 299.1
 Average weekly114.9 ± 82.1115.8 ± 114.395.1 ± 74.8

The results of 2 GEE analyses are shown in Table 3, each focusing on the predictors of a different measure of posttreatment exercise maintenance. Analyses were limited to the 33 subjects who provided data. Each model included only the variables with at least marginally significant effects, selected through the procedure described in the Subjects and Methods section, as well as the effects of in-treatment participation and time. For each of the selected variables, Table 3 shows the estimated regression coefficient, i.e., the expected change in a respective outcome associated with a 1-unit increase in a given prediction, together with the corresponding 95% confidence interval (95% CI). The model for stretching (Table 3) showed a very significant finding towards decreasing the duration of stretching with increasing time; with every additional month the mean duration decreased by ∼42 minutes (95% CI 13–71 minutes). As expected, higher in-treatment activity was associated with significantly higher posttreatment duration of stretching exercise. Among other putative predictors considered in our analyses, stress was the strongest predictor of maintenance for stretching. Both higher baseline stress level and greater increases in stress during the treatment predict, independently of each other, greater reduction in posttreatment stretching duration. The fact that these effects were adjusted for the mean in-treatment duration of stretching implies that higher stress before and during the treatment predicts worse maintenance, even among subjects who had the same level of in-treatment activity. Table 3 shows a marginally nonsignificant (P = 0.08) effect of increasing beliefs in the benefits of exercise, i.e., subjects who increased their beliefs during the treatment showed better maintenance even if the baseline beliefs had no significant effect (P = 0.62). Finally, even if upper-body pain was selected into a model at an earlier stage, in the final model neither baseline pain (P = 0.16) nor change in pain (P = 0.84) showed a significant association with stretching maintenance, after having adjusted for all other variables in the model.

Table 3. Results of generalized estimating equations models*
PredictorSDUnitsOutcome (Exercise type at 4, 5, and 6 months), minutes
StretchingAerobic
Parameter estimate95% CIParameter estimate95% CI
  • *

    95% CI = 95% confidence interval; FM = fibromyalgia; FIQ = Fibromyalgia Impact Questionnaire; MET = metabolic equivalent.

  • P < 0.01

  • 0.01 ≤ P < 0.05

  • §

    0.05 ≤ P < 0.15

Time  −42.03−71.43, −12.62−36.07−96.80, 24.65
In-treatment adherence      
 StretchingMinutes3.452.56, 4.34  
 Aerobic exerciseMinutes  2.301.16, 3.43
FM upper-body pain      
 3rd month–baseline23.161 SD3.67−32.69, 40.04−108.00−188.66, −27.33
 Baseline15.361 SD27.97−11.02, 66.9730.04−74.59, 134.67
Stress      
 3rd month–baseline62.311 SD−38.58−63.42, −13.75  
 Baseline61.161 SD−28.29−54.26, −2.32  
Exercise benefits      
 3rd month–baseline6.151 SD63.31§−7.06, 133.68  
 Baseline3.791 SD17.62−52.04, 87.28  
Exercise barriers      
 3rd month–baseline3.041 SD  −120.97−236.79, −5.16
 Baseline1.911 SD  −98.64§−216.22, 18.94
FIQ      
 3rd month–baseline20.641 SD  35.84−52.27, 123.94
 Baseline14.841 SD  −89.06§−180.78, 2.67
Baseline      
 Age8.631 SD  −38.54§−89.57, 12.49
 METs stress test1.961 SD    

Results for the prediction of aerobic exercise maintenance are shown in Table 3. Although there was a trend toward a decrease in the duration of aerobic exercise over time, it did not reach statistical significance (P = 0.24). As expected, in-treatment activity was a strong predictor of posttreatment, unsupervised aerobic duration. The results also confirmed an important negative impact of barriers on exercise; greater increases in barriers during treatment predicted a significant decrease in posttreatment aerobic participation (P = 0.04), while the negative effect of higher pretreatment barriers was only marginally nonsignificant (P = 0.10). In addition, an increase in upper-body pain during the treatment was associated with a significant decrease in posttreatment participation (P = 0.009) even if the baseline pain had no effect (P = 0.57). By contrast, for the FIQ measure of disability, the higher pretreatment FIQ predicted worse maintenance (P = 0.057) whereas the in-treatment change in FIQ had no impact (P = 0.43). Finally, after adjusting for all the above effects, older age seemed to be associated with lower posttreatment aerobic participation, but this effect failed to reach statistical significance (P = 0.13). Results for these outcomes are summarized in Figure 1.

Figure 1.

Relationships between baseline and in-treatment variables and outcomes.

DISCUSSION

The findings from the randomized clinical trial for this cohort indicated that the 12-week, individualized, home-based exercise program significantly improved functional capacity at 3 and 9 months following treatment for women in the exercise group who were more functionally disabled at baseline (22). When we examined adherence during treatment, we found that for both types of exercise, women who were less physically fit at baseline engaged in more exercise during the program (21). However, these results could not be directly extrapolated to identify predictors of exercise maintenance. Because maintenance is key to continued benefits, understanding what hinders or helps a person continue to exercise is important.

This report, the third and final article in the series from the trial, addresses maintenance of exercise in this cohort. The study was exploratory in that we did not know if findings stemming from nonpatient populations could be generalized to FM (15). Nonetheless, given our results on adherence during the treatment phase, we expected that the predictors would vary across exercise modalities. We took advantage of the fact that we had measures of both baseline characteristics and changes during treatment to explore the dynamic process of initiating and maintaining these behavioral changes. Although the sample size was relatively small, we compensated for this limitation with data from 3 posttreatment consecutive months.

There was a significant decrease in stretching over the 3 months posttreatment. Not surprisingly, high in-treatment adherence predicted better posttreatment maintenance of stretching exercises (41). Consistent with the literature from nonpatient populations (25, 42), high stress was the best predictor of poor maintenance of stretching, both for stress at baseline and for increases in stress during treatment. The relationship observed between stress and poor adherence to exercise may be relevant to understanding potential mechanisms by which stress may exacerbate FM-related symptoms. In contrast to our results for in-treatment adherence to stretching, neither pain at baseline nor changes in pain predicted maintenance of stretching over the 3-month, posttreatment, unsupervised period.

High in-treatment adherence predicted better posttreatment maintenance of aerobic exercise. Moreover, higher baseline disability predicted worse maintenance of aerobic exercise over the 3-month followup period. This latter finding is consistent with that of Burton et al (14) who found that maintenance of physical activity was higher for community-dwelling older persons who reported having moderately good or excellent health (i.e., were not disabled). Increases in barriers to exercise during treatment were associated with less aerobic exercise, as were increases in upper-body pain. No other candidate variables (unlike McAuley's finding for self efficacy [16]) were found to predict sustaining aerobic exercise.

These results highlight the importance of being specific when studying adherence, because determinants vary even within a behavioral category such as exercise. We have described elsewhere (43) predictors of adherence to medications in a different cohort of women with FM. Patient-physician discordance on communication and satisfaction with the office visit predicted overall medication nonadherence. Dobkin et al (44) found that more general adherence (i.e., to medical directives) was predicted by lower patient-physician discordance on patient wellbeing and lower patient psychological distress. These results highlight the fact that predictors differed across the stage of behavior change (i.e., uptake versus maintenance of exercise). Therefore, one needs to ask, “Adherence to what, and when?” Moreover, as suggested by the transtheoretical model (45), our results underscore the importance of simultaneously evaluating a number of factors from various domains (e.g., patient characteristics, psychological variables, clinical characteristics) for understanding exercise adherence in patients with FM.

Our findings have clinical implications for health care providers working with patients with FM. As a consensus emerges with regard to exercise being an important component of treatment for FM (9, 46, 47), educating patients as to how to initiate and continue exercise is crucial to its success. Specifically, one could discuss barriers that may impede exercise and develop an action plan to prevent lapses in practice. Patients may benefit from cognitive-behavior therapy (CBT), which aims to reduce catastrophizing among patients with FM (48) such that they do not overreact to bodily sensations when exercising. CBT could also teach these patients stress-management techniques so that stress does not pose a barrier to exercise participation.

Although this study provides novel data, it is limited by a potential selection bias. Only women who were willing to start an exercise program entered the trial; they are probably not representative of all patients with FM. Also, no men participated in the study, further restricting generalization. Because our measures were based on self report of outcomes, the associations we found may reflect subjects' perceptions (49). Self report can be erroneous if the subject wants to be viewed positively by others, or if there is recall bias. We addressed these 2 potential problems by having the project coordinator (not the exercise physiologist) keep track of the daily logs; recall bias was minimized by having subjects complete the logs after each exercise session and return the form weekly. It is possible that self monitoring influenced the amount of exercise performed. Finally, we assessed pain and fatigue on a weekly basis rather than immediately following exercise; results may have been different if we did the latter.

In conclusion, one may conceptualize some patients with FM as being both physically and mentally deconditioned. A combined approach to the mind-body problem of FM (i.e., exercise plus CBT) may have a synergistic effect (50) such that CBT addresses the mental barriers to improvement whereas exercise changes physiologic factors that contribute to FM symptoms. Although the evidence is not conclusive with regard to the effectiveness of multidisciplinary rehabilitation for FM (46), high-quality studies that are adequately powered to provide solid findings may change widespread pessimism concerning prognosis for patients with FM.

Ancillary