Occupational Therapist–Delivered Cognitive–Behavioral Therapy for Knee Osteoarthritis: A Randomized Pilot Study

Study Design

We conducted a randomized pilot trial to determine intervention feasibility and estimate the effect size for physical function based on intervention completers. The study was approved by the University of Michigan institutional review board.

Participants

Community-living adults ages ≥50 yr were recruited between May 2013 and May 2014 by means of advertisements. Participants were included if they met American College of Rheumatology (ACR) clinical criteria for KOA (Altman et al., 1986) and reported knee pain of ≥3 mo duration with pain severity of 3 of 10 on the numerical rating scale at least 50% of days. Other inclusion criteria included having computer access and being ambulatory with or without a cane. People were excluded if they had a severe physical impairment, report of a psychiatric disorder, or other issues that could interfere with completion of study procedures; had cancer (other than skin) in the past year; had undergone invasive procedures for KOA in the past 3 mo; had participated in CBT or KOA rehabilitation within the past 12 mo; or were taking long-acting opioids.

Measures

Data Collection

The flow of participants through the study is depicted in Figure 1. Participants meeting preliminary eligibility criteria during a phone screening were scheduled for a baseline visit, at which time verbal and written consent was obtained. Participants were screened for KOA using ACR clinical criteria by trained clinical research staff, and other inclusion criteria were verified. If eligible, participants were administered questionnaires and physical performance testing. For the collection of physical activity data, participants were instructed to wear the Actiwatch-S on their nondominant wrist for 7 days and to take it off when there was a possibility of the device becoming wet. There were also end-of-day activity checklists organized in 2- to 4-hr time blocks for participants to mark the types of activities in which they were engaged. After 7 days, participants returned the Actiwatch-S and checklists in a prepaid envelope.

OPEN IN VIEWER

Figure 1.

CONSORT (Consolidated Standards of Reporting Trials) flow diagram for study participants.

Upon satisfactory completion of the actigraphy, participants were randomized into Engage or usual care in a 2:1 ratio (this ratio was chosen to better assess program feasibility and acceptability). All outcome assessors were blinded to group assignment at baseline, because participants were not yet randomized into a treatment arm, and often, but not always, blinded to outcome assessment at the 6-mo follow-up visit. We estimate that blinding was broken in three to five cases at the 6-mo time point as a result of participants revealing the group assignment to the assessor or in cases where the coordinator needed to conduct assessment visits.

Procedure

Three licensed occupational therapists underwent training as a group by a trained psychologist (David A. Williams) before the study. They met initially for two weekly 1-hr sessions before seeing their first participant and then met every 2–3 wk after conducting sessions with a few participants in the initial pilot sample. The meetings consisted of education about cognitive–behavioral principles, discussion of individual treatment sessions, and review of responses to completed homework by participants. Any issues that arose for the therapists were discussed as a group, and appropriate problem-solving strategies were offered by the psychologist.

Treatment fidelity was further enhanced by having the occupational therapists follow a standardized therapy manual (see Supplemental Appendix A, available online at http://otjournal.net; navigate to this article, and click on “Supplemental”). Subsequent supervision was provided by the first author (Murphy) throughout the study to ensure treatment fidelity, discuss problematic issues, and check progress. Therapists’ use of the evidence-based online modules to teach various CBT skills also aided in maintaining integrity of the cognitive and behavioral tenets of CBT for participants.

The Engage intervention consisted of eight weekly in-person, 1-hr sessions with an occupational therapist. In the first session, the therapist explained the program’s focus on lifestyle changes to help manage osteoarthritis symptoms through self-monitoring and goal setting. The therapist asked the participant how KOA had affected his or her daily life and about personal goals for the program. The therapist reviewed general information about osteoarthritis using large colored figures and diagrams and followed the standardized therapy manual. Participants were introduced to the online program.

Program modules were taken from a successful CBT program for people with fibromyalgia (Williams et al., 2010) and included exercise, sleep hygiene, pleasant activity scheduling, relaxation, activity pacing, problem solving, and a wrap-up session that focused on further goal attainment. Each module included a video presentation by a health professional (e.g., rheumatologist, psychologist), written materials, and printable worksheets. In the original program (Williams et al., 2010), the modules were not accompanied by sessions with a health professional. In the Engage program, occupational therapists worked with participants to practice CBT skills and assisted them in identifying how to integrate new behaviors in their everyday life. These therapists were ideal facilitators for this program because they were trained in activity analysis and intervention for issues that arise among a person, the activity, and the environment while involved in daily routines. Each participant was asked to choose one module to work on at home and complete the associated worksheets that involved self-monitoring and practice of a particular skill, such as relaxation or activity pacing.

At each subsequent session, the therapist oriented the participant to the topic from the last session, discussed the homework, and assisted with troubleshooting barriers and goal setting. Participants could choose new modules each week or continue with the same module. The therapist sometimes suggested particular modules for participants to try. In the final session, participants established an action plan for maintaining new skills. Participants in the control condition continued with their usual osteoarthritis care and were asked not to initiate any new behavioral programs over the 6-mo study period.

Data Analysis

We first calculated descriptive statistics on demographic and health characteristics of the sample at baseline, comparing Engage and usual-care groups using independent-sample t tests (for continuous variables) and χ² tests for categorical variables. We then used analysis of covariance to compute an effect size for the primary outcome (physical function measured with WOMAC–PF) and to explore program effects on secondary outcomes. Predictors were treatment group (Engage vs. usual care), the baseline value of the model outcome as a covariate, and a Baseline Score × Group interaction term to test whether program-associated improvement on a given outcome depended on the value of that outcome at baseline. When this interaction term was nonsignificant, it was omitted and the model was rerun with main effects only. Residual plots were examined to verify model fit, and Cook’s D values were inspected for influential outliers.

We then used χ² tests to assess differences in the proportion of participants in Engage versus usual-care groups who achieved a clinically significant improvement in physical function (17%; Angst et al., 2002) on the WOMAC–PF and who reported feeling much or very much improved. Finally, we compiled data from a post-program survey that included closed- and open-ended items about program satisfaction. IBM SPSS Statistics (Version 21; IBM Corp., Armonk, NY) was used for all analyses.

Baseline Characteristics

Participants who dropped out (n = 11), compared with study completers (n = 46), were more likely to have the least educational attainment (high school or some college; 73% vs. 24%; χ² = 9.52, p = .009) and higher CES–D depression scores (13.2 vs. 7.0), t(55) = –2.2, p = .03. Attrition was similar in the Engage and usual-care groups (18% and 21%, respectively; p = .54). Most program completers were female (76%), White non-Hispanic (85%), college-educated (76%), and retired (57%; Table 1). The average BMI was 31.9, on the low end of the obese category. Participants walked an average of 1,167 feet (356 m) on the Six-Minute Walk Test (below the range cited for healthy women older than age 60, i.e., 448–503 m; Bohannon, 2007), had an average of 8.9 symptoms lasting ≥3 mo in the past year on the CMSI, and had a mean CES–D score of 7.0, substantially below the cutoff of 16, indicating likely depression (Radloff, 1977).

Primary and Secondary Outcomes

Table 2 shows the means for Engage and usual-care groups at baseline and 6-mo follow-up for primary and secondary outcomes. It also shows the effect sizes ( η p ²) for these outcomes associated with program participation. In the Engage group, the mean WOMAC–PF score, representing difficulty in daily activities, decreased from 21.0 at baseline to 15.3 at follow-up, with a smaller decrease in the usual-care group (from 22.9 to 18.5). We observed an effect size for Engage of 0.01 on the WOMAC–PF. The observed power to detect this effect was low at 0.11. For both the BPI and the BFI, indicating degree of pain and fatigue, respectively, mean scores dropped slightly in the Engage group (2.7 to 2.2 for BPI and 2.5 to 2.3 for BFI) and increased slightly in the usual-care group. Effect sizes for the BPI and BFI were 0.04 and 0.01, respectively. We did not calculate an effect size for accelerometer-measured physical activity, because a decline in activity was observed in both groups, with the usual-care group declining less.

Figure 2 shows the proportion of participants in each group who achieved a clinically significant or other threshold level of change on primary and secondary outcomes. For the WOMAC–PF, 58% of Engage participants vs. 47% of controls met the criteria for the minimally important clinical difference of 17% (χ² = 0.53, p = .34). This threshold was set at 30% improvement for pain (according to BPI score; achieved by 39% of Engage participants vs. 33% of controls; χ² = 0.13, p = .49) and for fatigue (according to BFI score; 29% of Engage participants vs. 13% of controls; χ² = 1.37, p = .22). On the Patient Global Impression of Change measure, 45% of Engage participants compared with 13% of controls indicated that they were very much or much improved (χ² = 4.5, p = .03).

OPEN IN VIEWER

Figure 2.

Proportion of Engage and usual-care groups meeting criteria for meaningful change.

Feasibility and Participant Satisfaction

Regarding program engagement, 30 of 31 participants (97%) attended six or more sessions. The two most frequently used modules, as recorded by the occupational therapists, were exercise/physical activity (chosen 44 times across 31 participants) and pleasant activities (chosen 37 times). The least used was communication (chosen 9 times). Participants (n = 25) were asked in postprogram surveys to select up to three modules that they found the most useful; the three most commonly chosen modules were pacing (80% rated as most useful), relaxation (36%), and goal setting (36%). According to feedback in optional open-ended questions, the element of the program that participants perceived as most useful was the sessions with the occupational therapist (n = 14), which participants described as helpful for explanations, clarifications, and support, and tracking and planning (n = 12). Suggested changes to the program included simplifying and reducing materials, spreading out sessions with the occupational therapist, and providing a follow-up meeting after the last session. No adverse reactions or events were reported in either treatment condition.

Effects on Primary and Secondary Outcomes

Our hypothesis that Engage would have a positive impact on self-reported physical function over a 6-mo period was supported, although the effect size (η_p ²; Cohen, 1988) was small at 0.01 (Table 2). Only a slightly higher proportion of Engage participants (58% vs. 47%) met criteria for clinically significant improvement on this measure. This trial was not powered to detect significant results. Improving physical function by promoting patient engagement in valued and necessary activities was the central goal of the Engage program. To this end, occupational therapists worked with patients to examine the role of beliefs, behaviors, and symptoms in activity engagement; set long-term and short-term therapy goals; and practice new skills (e.g., activity pacing). Although the observed impact of Engage on function was modest, it is possible that the skill set learned by participants will result in more durable health behavior change than traditional rehabilitation approaches and a greater effect on daily functioning over time. It is also possible that a ceiling effect may have been present in this particular sample of patients, who tended to be highly educated and who may have already been practicing some of the recommended techniques.

In contrast with the evidence of a small improvement in daily physical functioning, we did not observe any intervention-related impact on physical activity as measured at baseline and follow-up during a 7-day home-monitoring period by a wrist-worn accelerometer. This result suggests that, consistent with Engage’s emphasis on the influence of beliefs on activity engagement, participants were able to improve daily function in spite of no change in objective physical activity, possibly as a result of enhanced confidence in their ability to improve. Future studies should include participation measures to determine whether participants are able to engage in their life roles despite symptoms or are more satisfied with their ability to engage.

An important element of Engage was its attention to myriad symptoms that can contribute to disability, such as fatigue and sleep disturbance, as opposed to the more narrow focus of traditional rehabilitation on biomechanical factors (Sharma et al., 2003). Our results provide preliminary evidence that Engage may help reduce fatigue, because more than twice as many Engage participants as controls (29% vs. 13%) reached the threshold for clinically meaningful improvement on the BFI. Notably, the pacing module, which may have helped in reducing fatigue, was rated as the most useful by patients.

Feasibility and Participant Satisfaction

Several findings were positive indicators of Engage’s feasibility and acceptability. Among participants who completed the program, virtually all attended at least six of eight sessions. The majority of participants volunteered in an open-ended response that the occupational therapy sessions were the most helpful part of the program. The following quotes provide examples of what participants felt that they gained from these sessions: “The sessions helped me see how I was improving what I was supposed to do or not do to decrease my pain.” “The sessions helped clarify points and provided structure for the week’s activities.”

Process data also revealed some challenges that could be addressed in a larger efficacy trial of Engage. First, study attrition was high at 19%. Reasons for dropping out were varied (Figure 1) but included “changed mind” and scheduling and transportation difficulties. Sample size calculations for a larger study will need to incorporate the likely attrition rate, and strategies should be developed to address scheduling- and transportation-related barriers.

Second, the online modules were most often named—albeit by a minority of participants—as the least useful component. These modules were incorporated into the program design for two main purposes: (1) to standardize CBT content, some of which was new to the occupational therapists delivering the program, and (2) to enable patients to continue to learn and practice new skills between sessions with supportive content and worksheets. In general, evidence is accumulating for the efficacy of Internet-based chronic pain management programs in improving pain and activity limitation (Bender et al., 2011). It is likely that participant satisfaction with the Engage online component could be improved with some changes to the modules as suggested in participant comments, for example, reducing the amount of content in each module.

Finally, process data revealed that most participants did not use all of the modules equally, but rather selected specific modules that were most appealing or relevant to them. Our past studies have demonstrated a substantial amount of variability within KOA patients regarding their most problematic symptoms (pain, fatigue, or stiffness) and in the associations between these symptoms and activity. Taken together, these results support the need for programs to be individually tailored to optimize functioning among patients.

Study Limitations

The primary goal of our trial was to test feasibility, and it was underpowered to detect clinically significant effects. More participants would be needed to examine the effectiveness of this intervention in a larger trial. In addition, the convenience sample of participants recruited from the community may have contributed to the lack of significant findings because they did not have severe pain or other symptoms. Further studies of Engage should include participants with more severe symptoms, perhaps in conjunction with conventional treatments to determine the added value of CBT on outcomes. Despite limitations, the intervention was well received by participants and trended positively toward supporting efficacy, and delivery by occupational therapists could provide an important point of access to this intervention for people with KOA.