To examine the effectiveness of cognitive-behavioral and pharmacologic treatment of depression in rheumatoid arthritis (RA).
To examine the effectiveness of cognitive-behavioral and pharmacologic treatment of depression in rheumatoid arthritis (RA).
Subjects (n = 54) with confirmed diagnoses of both major depression and RA were randomly assigned to 1 of 3 groups: 1) cognitive-behavioral/pharmacologic group (CB-PHARM), 2) attention-control/pharmacologic group, or 3) pharmacologic control group. Measures of depression, psychosocial status, health status, pain, and disease activity were collected at baseline, posttreatment (10 weeks), 6-month followup, and 15-month followup. Data were analyzed to compare the treatment effectiveness of the groups; data also were aggregated to examine the effects of antidepressive medication over time. Lastly, a no-treatment control group was defined from a cohort of persons who declined participation.
Baseline comparisons on demographic and dependent measures revealed a need to assess covariates on age and education; baseline scores on dependent measures also were entered as covariates. Analyses of covariance revealed no statistically significant group differences at postintervention, 6-month followup, or 15-month followup, except higher state and trait anxiety scores for the CB-PHARM group at the 15-month followup. In the longitudinal analyses of the effects of antidepressive medication, significant improvement in psychological status and health status were found at posttreatment, 6-month followup, and 15-month followup, but no significant improvements were shown for pain or disease activity. In addition, the comparison of the aggregated pharmacologic group with a no-treatment group revealed a statistically significant benefit for the 3 groups that received the antidepressive medication.
In persons with RA, cognitive-behavioral approaches to the management of depression were not found to be additive to antidepressant medication alone, but antidepressant intervention was superior to no treatment.
Depression has been identified as a problem for a sizeable number of persons with rheumatoid arthritis (RA) (1, 2); most investigators have concluded that depressive symptoms are identifiable in one-quarter or more of persons with RA (1, 3–5). However, estimates of the prevalence of the specific diagnosis of major depression vary substantially, depending upon the manner in which potential overlap symptoms are counted; some symptoms (e.g., fatigue, loss of energy) may present as secondary to RA itself, may reflect major depression, or may be related to both RA and major depression. Regardless of debates over specific diagnostic criteria, there is general agreement that depressive symptoms are both common in RA and associated with reduced health status (6, 7). Specifically, higher depression has been shown to be related to higher pain, higher fatigue, and reduced quality of life for persons with RA (7).
Fortunately, depression is a highly treatable condition. Approximately 70% of persons with major depression have a positive response to first-line antidepressant medications (8). For those persons who do not respond to the initial antidepressant, successive trials on alternative antidepressant medications yield a positive treatment response for the vast majority of patients; the incidence of treatment-resistant major depression is exceedingly low (8). Yet, little data exist to clarify whether antidepressive therapies confer their expected benefits in an RA population. In addition, the impact of antidepressant treatment on broad healthcare outcomes, such as disease activity, disability, and health status for persons with RA, is unknown.
The purpose of this study was to conduct a prospective, randomized trial of the effectiveness of an antidepressive treatment program in a cohort of persons with both RA and a concomitant major depression. The primary hypothesis was that a combined cognitive-behavioral/pharmacologic treatment program for major depression would result in greater improvement on psychosocial, health status, pain, and disease activity measures than would an attention-control intervention or antidepressive medication alone.
The study included 54 subjects (15 males and 39 females) with a diagnosis of classic or definite RA. Subjects were recruited from a midwestern Department of Veterans Affairs (VA) hospital (n = 8, 15%), a university medical center (n = 24, 44%), and a private rheumatology practice (n = 22, 41%) consistent with an institutional review board-approved protocol. At the end of the study (15-month followup), 41 subjects remained for statistical analyses. The diagnosis of RA was made by collaborating rheumatologists (WSI, GRK, SEW) using the 1987 diagnostic criteria of the American College of Rheumatology (ACR; formerly, the American Rheumatism Association) (9). Subjects also met the diagnostic criteria for major depression as diagnosed by a collaborating psychiatrist (JRS) using a structured diagnostic interview (Structured Clinical Interview for Diagnostic and Statistical Manual of Mental Disorders, Third Edition, Revised [SCID]) (10).
The exclusionary criteria were as follows: history of organic brain syndrome, presence of a psychotic disorder, presence of other uncontrolled medical disorders, presence of a major communication disorder, illiteracy, presence of less than an 8th grade education (due to inability to complete questionnaires), presence of a therapeutic dosage of an antidepressant medication, or presence of another autoimmune disease/disabling condition. Patients who were classified as functional class IV according to the criteria of Steinbrocker et al (11) were excluded due to the difficulty they would encounter meeting the demands of the protocol.
The mean age of the sample was 54.6 (SD = 11.4) years; the mean educational level was 12.6 (SD = 2.3) years. The median annual income was in the $15,000–$19,999 range. Sixty-nine percent (n = 37) were unemployed, disabled, or retired; 67% (n = 36) were married; 98% were white. The mean Hollingshead index of socioeconomic status was 40.7 (SD = 12.7), reflecting a predominantly middle class sample (12). The distribution of patients by functional class (11) was as follows: class I, 4 % (n = 2); class II, 48% (n = 26); and class III, 48% (n = 26). The distribution of patients by ACR anatomic stage (13), based on radiographs of hands or feet, was as follows: stage I, 15% (n = 6); stage II, 33% (n = 13); stage III, 43% (n = 17), and stage IV, 10% (n = 4). The mean disease duration for the sample was 12.6 (SD = 10.1) years.
Subjects were randomly assigned to 1 of the following 3 groups: 1) cognitive-behavioral/pharmacologic group (CB-PHARM; n = 17), 2) attention-control/pharmacologic group (AC-PHARM; n = 20), or 3) pharmacologic control group (CN-PHARM; n = 17). After dropouts, the group sizes were attenuated as follows: CB-PHARM = 14; AC-PHARM = 13; CN-PHARM = 14. For ethical reasons, the protocol was designed for all subjects to receive antidepressant medication; the primary question was whether cognitive-behavioral treatment would confer a benefit above and beyond the antidepressant medication alone. During randomization, subjects were stratified on the basis of clinic site and functional class.
Subjects in the CB-PHARM group received their ongoing rheumatologic care. They also received a therapeutic trial of an antidepressant medication and participated in a comprehensive, outpatient cognitive behavioral (CB) program that emphasized management of depression. The CB intervention was modeled after the procedures of Parker et al (14), which previously have been shown to be effective in a sample of persons with RA. The counselor for the CB program was a doctoral-level psychologist (MLP) with extensive experience in the treatment of depression using CB techniques. The intervention was provided individually via 10 weekly outpatient visits. Each session lasted approximately 1.5 hours and included a relaxation training component. Following the 10-week intervention, a 15-month maintenance program was provided to help sustain treatment gains. During the maintenance phase, antidepressive medication was continued, and subjects were seen individually at least once every 3 months.
Subjects in the AC-PHARM group received their ongoing rheumatologic care. In addition, they received a therapeutic trial of an antidepressant medication and participated in a general patient education program that served as an attention-control (AC) condition; Sackett and Haynes (15) concluded that information alone does not produce a significant effect on health-related behavior. The same counselor (MLP) who delivered the CB-PHARM program also delivered the AC-PHARM program to control for potential provider effects. The number of sessions and followup visits for subjects in the AC-PHARM and the CB-PHARM groups were identical, but no active problem-solving or depression-management components were included in the AC-PHARM program.
Subjects in the CN-PHARM group received their ongoing rheumatologic care; they also received a therapeutic trial of an antidepressant medication, but they were not exposed to either the CB or the AC programs.
The CES-D is a 20-item self-report scale that measures depressive symptoms (during the past week) (16). The CES-D has high internal consistency, adequate test-retest reliability, and good factorial and discriminant validity (17). There is evidence that the CES-D is a valid measure of depression for individuals with arthritis (4). In this study, Cronbach's coefficient alpha for the CES-D was 0.89.
The HAM-D is a 17-item, interview-based inventory that yields a measure of depression severity (18, 19). The HAM-D provides continuous (rather than categorical) data; the HAM-D has been shown to be a reliable and valid measure of depression (20–22). In this study, Cronbach's coefficient alpha for the HAM-D was 0.66.
The GDS is a 30-item questionnaire designed to measure depression in an elderly population. The GDS has been shown to have adequate internal consistency and reliability (23). In this study, the Kuder-Richardson 20 for the GDS was 0.80.
The SCL-90-R composite measures (i.e., global severity index [GSI], positive symptom distress index [PSDI], and positive symptom total [PST]) and the SCL-90-R depression subscale were used in this study. The SCL-90-R has been shown to have excellent reliability and validity (24). In this study, Cronbach's coefficient alpha for the GSI and depression scores were 0.93 and 0.84, respectively; the Kuder-Richardson 20 for the PST score was 0.91.
The CSQ is a 50-item questionnaire with 9 subscales (25). Two CSQ factor scores have been identified: coping attempts (CA), which reflects the tendency to utilize pain coping strategies; and pain control and rational thinking (PCRT), which reflects an individual's confidence in his or her ability to manage pain (26). CSQ scores have been found to correlate with health status and depression in persons with arthritis (26–28). In this study, Cronbach's coefficient alphas for the CA and the PCRT scores were 0.89 and 0.76, respectively.
The DSI is a 15-item measure of stressful events in daily life (during the past 24 hours) that yields 3 scores: frequency of stressors, severity of stressors, and average impact of stressors. The DSI has good internal consistency and validity (29). In this study, the severity of stressors score was used; Cronbach's coefficient alpha was 0.94.
The HS is a 117-item measure of general life stress, which quantifies pressures, problems, and difficulties of everyday life (during the past month). Adequate reliability and validity for the HS have been reported (30). In this study, the severity of stressors subscore was used; Cronbach's coefficient alpha was 0.92.
The STAI is a 40-item measure used to assess state and trait anxiety (31). Both the state and trait anxiety subscores of STAI have been shown to have adequate reliability and validity (31). The STAI has been successfully used in previous research with RA patients (14, 32). In this study, Cronbach's coefficient alphas for the state and trait anxiety scores were 0.89 and 0.90, respectively.
The AHI is a 15-item scale designed to measure patients' perceptions of control regarding arthritis. Nicassio et al (33) have demonstrated adequate reliability and validity for the AHI. In this study, Cronbach's coeffecient alpha for the AHI was 0.58.
The ASES is a 20-item, self efficacy measure developed by Lorig et al (34). The ASES was developed specifically for persons with arthritis (35). The ASES yields 3 scores: self efficacy for physical function, self efficacy for pain management, and self efficacy for controlling other arthritis symptoms. Adequate reliability and validity for the ASES have been reported (35). In this study, an ASES total score was derived by summing the components; Cronbach's coefficient alpha for the ASES total score was 0.90.
The AIMS-2 is a 78-item questionnaire designed to measure health status for persons with arthritis. Five factor scores have been identified for the AIMS-2 as follows: 1) physical score, 2) affect score, 3) symptom score, 4) social interaction score, and 5) role score (36). The role score (which contains numerous work-related questions) was not used in this study due to the fact that a sizeable proportion of participants (69%) were unemployed, disabled, or retired. Data on the reliability and validity of the AIMS-2 have been provided (36). In this study, Cronbach's coefficient alphas for the physical, affect, symptom, and social interaction scores were 0.91, 0.84, 0.62, and 0.81, respectively.
The MAF is a 16-item questionnaire designed to measure fatigue. The MAF has been shown to have good reliability and concurrent validity (37). In this study, Cronbach's coefficient alpha for the MAF was 0.93.
The VAS for pain is a 10-cm line anchored with “no pain” at 1 extreme and “pain as bad as it can be” at the other extreme. Subjects were asked to indicate the point on the line that best represented the degree of pain experienced over the previous week. Adequate reliability and validity for the VAS have been reported (38, 39).
The MPQ is a self report pain questionnaire with 7 subscales that measure various aspects of the pain experience (40). In this study, the number of words chosen (NWC), percentage of body pain (Body %), present pain intensity (PPI), pain rating index (PRI), and average pain rating (these days) were used as pain measures.
The RADAR is a brief, self-administered questionnaire that measures perception of joint pain/tenderness and clinical status (41). The subscores used in this study were as follows: 1) arthritis activity today (AAT), 2) arthritis pain today (APT), and 3) painful and/or tender joint count (PTJC). The RADAR has been shown to have high subject-clinician agreement and to be both valid and sensitive to change (41). In this study, Cronbach's coefficient alpha for the PTJC was 0.88.
The systemic indicators consisted of erythrocyte sedimentation rate (ESR), grip strength on dominant and nondominant hands (average of 2 trials), 50-foot walking speed (average of 2 trials), and patients' estimates of daily morning stiffness (42). In addition, the average circumference of the proximal interphalangeal (PIP) joints on the dominant and nondominant hand were measured as an index of joint swelling.
Subjects with RA were consecutively recruited into the study following a 2-stage evaluation of depression. The first stage involved a screening for depressive symtomatology at regular intervals to identify persons with RA who were potentially depressed. Every 6 months, consenting individuals with RA completed the CES-D and provided information regarding their current medication regimen. Subjects for the study were recruited over a 4-year timeframe, so they completed the CES-D on a maximum of 7 occasions. A total of 2,673 depression screenings were administered to 638 persons with RA. Those with a CES-D ≥ 12 (n = 254) were invited to participate in a comprehensive evaluation for major depression, which included the SCID, administered by a board-certified psychiatrist (JRS); 84 subjects consented to proceed with the structured diagnostic interview. Persons meeting the criteria for MDD (n = 54) were enrolled in the study and randomized to 1 of 3 groups. Preintervention measures (i.e., comprehensive medical history, social history, and profile of current medications) were collected. Followup measures were collected at postintervention, 6-month followup, and 15-month followup; subsequent measures were collected at approximately the same time each day (± 1 hour) for each participant. Subjects were compensated for their travel ($0.11 per mile) and provided with antidepressant medication at no cost.
The psychiatrist (JRS), blinded to group assignments, initiated a trial of sertraline and provided pharmacologic management of depression for each participant. Prior to receiving sertraline, 12 participants were instructed to taper off low-dose antidepressants (i.e., amitriptyline) that were being used primarily for sleep or pain. Sertraline was titrated to 100 mg/day over a 3-week period. Dosage adjustments were made with consideration of the individual's side effect profile and clinical efficacy as judged by JRS. Medication dosage, tolerance, and the profile of side effects were monitored at weeks 1, 2, 3, 4, 8, 16, and at every 3-month interval thereafter. The average sertraline dosage was 99.5 mg/day (range = 25–150). For individuals who continued to experience sleep difficulties (n = 6), a trial of zolpidem was prescribed as needed. For 3 subjects, sertraline was deemed clinically ineffective; therefore, a trial of nortriptyline was prescribed as an alternative antidepressant (consistent with practice guidelines for management of depression); the average nortriptyline dosage was 75.0 mg/day. The regimen of antidepressant medication extended into the maintenance phase of the protocol, so all participants received a 15-month trial of either sertraline or nortriptyline.
No significant differences in program credibility were found between the CB-PHARM and the AC-PHARM groups at either preintervention or postintervention (see Table 1). In addition, no participants in any of the 3 groups reported seeking psychological assistance outside the study protocol.
|Do you think that this treatment program seems logical in its approach to managing the depression associated with rheumatoid arthritis?||5.3||6.1||5.7||6.4|
|How confident are you that this treatment program will be successful in helping you manage the depression associated with your rheumatoid arthritis?||5.5||5.9||5.5||6.1|
|How confident would you be in recommending this treatment program to a family member or friend who also had problems with rheumatoid arthritis?||5.2||6.2||5.5||6.5|
|Do you feel this treatment program will have a positive lasting effect for more than 6 months?||5.3||6.3||5.3||6.1|
The data were analyzed in the following 7 phases. First, normality tests were performed on all study variables. Second, descriptive statistics were generated, and the CB-PHARM, AC-PHARM, and CN-PHARM groups were compared on demographic characteristics and dependent measures at baseline. Third, baseline score, age, and education were used as covariates in a one-way (group) analysis of covariance (ANCOVA) to compare groups on the main outcome variable (CES-D score) at postintervention, 6-month followup, and 15-month followup; the alpha level was set at 0.05 for this main outcome analysis. Fourth, baseline scores, age, and education were used as covariates in either one-way (group) ANCOVAs or one-way multivariate analyses of covariance (MANCOVAs) to compare groups on the secondary outcome variables at postintervention, 6-month followup, and 15-month followup. Due to the relatively small sample size, the MANCOVAs were conducted using the subscores from specific measures or small clusters of conceptually related measures, as opposed to larger clusters of outcome variables. To compensate for these multiple comparisons, the alpha level for the secondary outcome analyses was set at 0.01. The Bonferroni correction for multiple comparisons was judged to be too conservative because the technique does not take into account the correlational structure of the variables (43). Fifth, additional analyses of group effects were performed to examine dropouts, medication changes, and program adherence. Sixth, the study groups were combined to analyze intervention effects in longitudinal fashion (using either analyses of variance [ANOVAs] or MANOVAs). Seventh, comparisons between the combined pharmacologic group (PHARM) and a no-treatment group (NO-TREAT) (i.e., persons who declined participation in the treatment arm of the study) were undertaken.
All demographic and dependent measures were examined for normality. Distributions were found to be acceptably normal or symmetric, so parametric analyses were used.
Baseline comparisons on demographic characteristics and study variables are shown in Tables 2, 3, and 4. Chi-square tests revealed no significant differences among the groups at baseline with regard to sex, marital status, clinic site (i.e., VA versus university versus private practice), anatomic stage, functional class, or medication classification. One-way ANOVAs (groups) revealed no significant differences among the groups at baseline with regard to education (P = 0.06), socioeconomic status (Hollingshead index), or disease duration; however, group differences were found for age (P = 0.03). The AC-PHARM group was significantly younger than both the CB-PHARM and the CN-PHARM groups. Given the ANOVA P value of 0.06 for education, post-hoc comparisons were conducted; the AC-PHARM group was found to have a significantly higher educational level than the CB-PHARM group, and this observed difference remained at followup intervals. A suggestion of group differences also was found at baseline for presence of comorbidities (P = 0.06), but this difference disappeared prior to followup analyses (due to dropouts). Hence, age and education were used as covariates in the subsequent group comparisons.
|Age||58.2 (9.6)||49.5 (10.7)||57.2 (12.1)||0.03†|
|Education, years||11.5 (2.2)||13.3 (1.7)||12.8 (2.7)||0.06†|
|Hollingshead index||44.8 (14.4)||36.8 (10.9)||41.1 (12.3)||0.16|
|Disease duration, months||12.5 (10.7)||11.6 (9.8)||14.0 (10.4)||0.78|
|Variable||CB-PHARM n (%)||AC-PHARM n (%)||CN-PHARM n (%)||P|
|Sex, female||12 (71)||14 (70)||13 (76)||0.89|
|Married||12 (71)||15 (75)||9 (53)||0.34|
|VAH||2 (12)||4 (20)||2 (12)|
|University||7 (41)||9 (45)||8 (47)|
|Private practice||8 (47)||7 (35)||7 (41)|
|I||2 (15)||3 (19)||1 (9)|
|II||3 (23)||6 (37)||4 (36)|
|III||5 (39)||7 (44)||5 (46)|
|IV||3 (23)||0 (0)||1 (9)|
|I||1 (6)||1 (5)||0 (0)|
|II||7 (41)||10 (50)||9 (53)|
|III||9 (53)||9 (45)||8 (47)|
|Variable||CB-PHARM n (%)||AC-PHARM n (%)||CN-PHARM n (%)||P|
|Presence of comorbidities||14 (82)||10 (50)||8 (47)||0.06|
|NSAIDs||13 (76)||19 (95)||14 (82)||0.27|
|Noncytotoxic DMARDs†||7 (41)||11 (55)||6 (35)||0.46|
|Cytotoxic DMARDs‡||11 (65)||10 (50)||9 (53)||0.65|
|Corticosteroids||10 (59)||9 (45)||11 (65)||0.46|
|Antibiotics||1 (6)||0 (0)||0 (0)||0.33|
To examine group effects on the main outcome variable at postintervention, 6-month followup, and 15-month followup, one-way ANCOVAs (group; covarying on baseline score, age, and education) were performed on the CES-D score. No significant group effects were found.
To examine group effects at postintervention, one-way ANCOVAs (group; covarying on baseline score, age, and education) were performed on the MAF score, the AHI score, and the ASES total score. No significant group effects were found at postintervention. One-way (group) MANCOVAs (covarying on baseline score, age, and education) were performed on the SCL-90-R composite scores, the stress severity scores (i.e., DSI and HS), the STAI subscores, the CSQ factor scores, the systemic indicators (i.e., estimates of morning stiffness, grip strength, 50-foot walking time, ESR, and PIP circumference), the AIMS-2 factor scores, the secondary depression scores (i.e., GDS, HAM-D, SCL-90-R depression score), the pain scores (i.e., VAS, NWC, Body %, PPI, PRI), and the RADAR scores (i.e., AAT, APT, PTJC). No significant group effects were found at postintervention (see Table 5).
|Variables||Preintervention||Postintervention||6-month F/U||15-month F/U|
|HS total severity||65.8||77.6||65.1||44.2||56.2||30.8||51.4||58.2||32.9||55.1||50.6||24.9|
|SCL-90-R GSI T-score||65.4||66.4||65.6||55.3||56.8||55.1||57.4||55.5||54.6||57.4||52.9||50.2|
|SCL-90R PSDI T-score||61.2||63.2||63.1||53.2||53.2||56.7||54.9||52.1||58.4||54.4||51.2||55.0|
|SCL-90R PST T-score||62.9||63.8||62.7||54.1||57.0||53.0||56.4||55.5||52.2||56.9||52.8||47.9|
|Present pain intensity||1.8||2.1||2.1||1.9||1.3||1.9||1.9||1.4||1.9||2.1||1.6||1.4|
|Pain rating index||20.0||17.2||18.9||20.1||13.7||16.1||20.6||13.1||15.5||20.1||12.8||11.7|
|Disease activity: RADAR|
|Disease activity: systemic|
|Grip strength (average)||12.9||21.7||13.7||12.3||21.2||13.5||11.4||20.7||11.8||10.0||18.4||13.0|
|Walk 50 ft||14.4||13.6||13.7||15.3||13.8||14.1||14.1||13.2||15.1||15.0||12.2||13.5|
To examine group effects at the 6-month followup, one-way (group) ANCOVAs were performed on the MAF score, the AHI score, and the ASES total score. No significant group effects were found at the 6-month followup. One-way (group) MANCOVAs were performed on the SCL-90-R composite scores, the stress severity scores, the STAI subscores, the CSQ factor scores, the systemic indicators, the AIMS-2 factor scores, the secondary depression scores, the pain scores, and the RADAR scores. No significant group effects were found at the 6-month followup (see Table 5).
To examine group effects at the 15-month followup, one-way (group) ANCOVAs were performed on the MAF score, the AHI score, and the ASES total score. No significant group effects were found at the 15-month followup. One-way (group) MANCOVAs were performed on the SCL-90-R composite scores, the stress severity scores, the STAI scores, the CSQ scores, the systemic indicators, the AIMS-2 composite scores, the secondary depression scores, the pain scores, and the RADAR scores. No significant group effects were found at the 15-month followup, except on the STAI scores (P = 0.008). ANCOVAs revealed a statistically significant group effect for both the state anxiety score (P = 0.02) and the trait anxiety score (P = 0.01) at the 15-month followup. For the state anxiety score, pairwise comparisons revealed that the CB-PHARM group reported significantly higher state anxiety scores than the AC-PHARM or the CN-PHARM groups, (P = 0.01 and P = 0.02, respectively). For the trait anxiety score, pairwise comparisons revealed that the CB-PHARM group reported significantly higher trait anxiety scores than the CN-PHARM group (P = 0.003) (see Table 5).
Of the 54 participants at postintervention, 10 were dropped from the study prior to the 6-month followup; specifically, 4 were dropped due to medical complications; 5 were dropped for noncompliance with program demands (e.g., not keeping appointments or discontinuing medication); and 1 was dropped due to side effects from sertraline and the concomitant inability to take the alternative antidepressant (nortriptyline). Of the remaining participants at the 6-month follow up (n = 44), 1 additional participant was dropped due to noncompliance. Of the remaining participants at the 15-month followup (n = 43), 2 additional participants were dropped due to noncompliance. The overall dropout rates for the CB-PHARM, AC-PHARM, and CN-PHARM groups were 18%, 35%, and 18%, respectively. Chi-square analysis revealed no significant differences in the dropout rate.
The primary medication regimen for each subject was recorded at preintervention (baseline) and at each subsequent followup. Medications were categorized as follows: nonsteroidal anti-inflammatory drugs, antibiotic treatment, noncytotoxic disease-modifying antirheumatic drugs (DMARDs), cytotoxic DMARDs, and corticosteroids. Over the 15 months of the study, the percentage of subjects remaining on their pretreatment medication regimen was 88% of the CB-PHARM group, 75% of the AC-PHARM group, and 88% of the CN-PHARM group. Chi-square analysis revealed no significant group differences with regard to medication changes.
The CB-PHARM counselor rated program adherence for each participant based on attendance records and a homework diary. Eight of 14 CB-PHARM participants were rated as highly adherent over the course of the 15-month intervention program. Accordingly, the group comparisons on the outcome variables were repeated using only the high-adherent participants in the CB-PHARM group, but ANCOVAs and MANCOVAs again failed to reveal significant group differences.
Because significant group differences were not found, the 3 groups (which all received antidepressant medication) were combined for longitudinal analyses of the overall pharmacologic effect.
A one-way ANCOVA (time) with repeated measures was performed on the CES-D scores at preintervention, postintervention, 6-month followup, and 15-month followup. A significant time effect was found for the CES-D score (P = 0.0001); paired t-tests revealed that the CES-D scores were significantly lower at postintervention (P ≤ 0.0001), at 6-month followup (P ≤ 0.0001), and at 15-month followup (P ≤ 0.0001) compared with the preintervention scores.
One-way ANOVAs (time) with repeated measures were performed on the AHI and the ASES total scores at preintervention, postintervention, 6-month followup, and 15-month followup. A significant time effect was found for the AHI score (P = 0.0001); paired t-tests revealed that the AHI scores were significantly lower at postintervention (P = 0.0001), at 6-month followup (P = 0.0001), and at 15-month followup (P = 0.0001) compared with the preintervention scores. A significant time effect also was found for the ASES total score (P = 0.0001); paired t-tests revealed that the ASES total scores were significantly higher at postintervention (P = 0.0002), at 6-month followup (P = 0.0009), and at 15-month followup (P = 0.0009) compared with the preintervention scores.
One-way MANOVAs (time) with repeated measures revealed significant longitudinal changes for the SCL-90-R composite scores (P = 0.0001), the stress severity scores (P = 0.0001), the STAI subscores (P = 0.0001), the CSQ factor scores (P = 0.0001), and the secondary depression scores (P = 0.0001). Paired t-tests revealed that the composite scores from the SCL-90-R (i.e., GSI, PSDI, PST) were all significantly lower at postintervention (all P values ≤ 0.0001), at 6-month followup (all P values ≤ 0.0001), and at 15-month followup (all P values ≤ 0.0001) compared with the preintervention scores. Paired t-tests revealed that the stress severity scores (i.e., DSI severity, Hassles severity) were both significantly lower at postintervention (all P values ≤ 0.0001), at 6-month followup (all P values ≤ 0.0002), and at 15-month followup (all P values ≤ 0.0002) compared with the preintervention scores. Paired t-tests revealed that the STAI scores (i.e., state and trait anxiety) were both significantly lower at postintervention (all P values ≤ 0.0001), at 6-month followup (all P values ≤ 0.0001), and at 15-month followup (all P values ≤ 0.0001) compared with the preintervention scores. Paired t-tests revealed that the CSQ scores (i.e., CA, PCRT) were both significantly higher at postintervention (all P values ≤ 0.001), at 6-month followup (all P values ≤ 0.001), and at 15-month followup (all P values ≤ 0.04) compared with the preintervention scores. Paired t-tests revealed that the secondary depression scores (i.e., GDS, HAM-D, SCL-90-R depression score) were all significantly lower at postintervention (all P values ≤ 0.0001), at 6-month followup (all P values ≤ 0.0001), and at 15-month followup (all P values ≤ 0.0001) compared with the preintervention scores (see Table 6).
|Variables||Pre-I||Post-I||6-m F/U||15-m F/U||F||df||P|
|Main depression score||29.0||(3,38)||0.0001|
|HS total severity||69.3||43.4||47.2||43.3|
|SCL-90-R GSI T-score||65.8||55.7||55.8||53.5|
|SCL-90-R PSDI T-score||62.5||54.4||55.2||53.6|
|SCL-90-R PST T-score||63.1||54.6||54.7||52.5|
|Present pain intensity||2.0||1.7||1.7||1.7|
|Pain rating index||18.7||16.7||16.5||14.9|
|Disease activity: RADAR||0.9||(9,32)||0.52|
|Disease activity: systemic||2.2||(15,25)||0.04|
|Grip strength (average)||15.9||15.5||14.5||13.7|
|Walk 50 ft||13.9||14.4||14.1||13.6|
One-way ANOVAs (time) with repeated measures were performed on the MAF scores at preintervention, postintervention, 6-month followup, and 15-month followup. A significant time effect was found for the MAF score (P = 0.006). Paired t-tests revealed that the MAF scores were significantly lower at postintervention (P = 0.007), at 6-month followup (P = 0.002), and at 15-month followup (P = 0.0008) compared with the preintervention scores.
A one-way MANOVA (time) with repeated measures revealed significant longitudinal changes for the AIMS-2 factor scores (P = 0.0001). Paired t-tests revealed that selected AIMS-2 factor scores (i.e., affect, symptom, social interaction) were significantly lower at postintervention (all P values ≤ 0.002) compared with the preintervention scores, but no significant difference was found for the AIMS-2 physical scores. Paired t-tests revealed that all AIMS-2 scores were significantly lower at 6-month followup (all P values ≤ 0.01) and at 15-month followup (all P values ≤ 0.05) compared with the preintervention scores (see Table 6).
A one-way MANOVA (time) with repeated measures was performed on the pain scores at preintervention, postintervention, 6-month followup, and 15-month followup, but no significant longitudinal changes were revealed.
A one-way MANOVA (time) with repeated measures was performed on the disease activity at preintervention, postintervention, 6-month followup, and 15-month followup, but no significant longitudinal changes were revealed.
The absence of a no-treatment control group constituted a major limitation for the interpretation of the longitudinal data. From an ethical standpoint, a study withholding antidepressant medication from persons with major depression is not viable. Therefore, an alternative strategy for defining a no-treatment control group was utilized.
Specifically, the list of persons with RA who were found to have a CES-D ≥ 12 (n = 254) was reexamined; a subgroup of 70 persons (excluding participants in the treatment arm of the study) who were followed with consecutive CES-D data for a minimum of 18 months was defined. To maximize the probability of severe depression (because SCID data were not available for persons who declined the intervention arm of the study), the 41 subjects (from the subgroup of 70) with the highest CES-D scores were selected; this strategy resulted in a cohort of subjects with CES-D ≥ 19. Baseline comparisons revealed that the PHARM group (n = 41) and the NO-TREAT group (n = 41) were not significantly different on the following potentially confounding variables: 1) initial CES-D score, 2) sex, 3) age, 4) disease duration, 5) MPQ average pain score, 6) education level, 7) functional class, and 8) clinic site (see Tables 7 and 8).
|Disease duration, months||145.2||148.4||0.89|
|MPQ average pain, these days||4.0||3.8||0.44|
|Variable||PHARM %||NO-TREAT %||P|
|Marital status, married||66||68||0.81|
The CES-D scores at long-term followup were examined to see if the PHARM group had lower depression scores than the NO-TREAT group. Long-term followup data were available at 15 months for the PHARM group and at 18 months for the NO-TREAT group; these intervals were judged to be sufficiently similar for a meaningful comparison. A one-way ANCOVA (group) was performed on the long-term followup CES-D scores, using the baseline scores as a covariate. A significant group effect was found (P = 0.03), revealing that the PHARM group had significantly lower CES-D scores at long-term followup than did the NO-TREAT group (see Figure 1).
Given the small sample size, an analysis of statistical power is critical for data interpretation. The main hypothesis was that the CB-PHARM intervention would yield an improvement in CES-D scores that was 13 points greater than the changes yielded by the AC-PHARM or the CN-PHARM interventions. The 13-point difference was selected because the target of the intervention was to reduce preintervention CES-D scores from approximately 28 to below the nondepressed cut-off score of 16 (for a decrease of approximately 13 CES-D points). Thus, at alpha 0.025 (derived by dividing 0.05 by 2 due to the comparison of the CB-PHARM group with both the AC-PHARM and the CN-PHARM groups) and with power set at 0.80, 13 participants per group is sufficient to detect a difference between the CB-PHARM change score and the CN-PHARM change score (based on a common standard deviation of 10.6) of 13.5 CES-D points (or 1.27 SD). The largest observed difference between the CB-PHARM change score and the CN-PHARM change score (at the 15-month followup) was only 6.8 CES-D points, which is far less than the 13-point target that was set for clinical meaningfulness (especially because the observed difference was in the unexpected direction). Therefore, a Type II statistical error does not appear probable in this case.
The study yielded a number of interesting findings. First, there was no difference in the overall effectiveness of the CB-PHARM, AC-PHARM, and CN-PHARM interventions. Specifically, subjects who received antidepressive medication alone improved just as much as subjects who received the combination cognitive-behavioral or attention-control components. Although the absence of group differences was not predicted in the original hypotheses, the finding is not entirely unexpected. The literature on combined pharmacologic and psychologic interventions for depression reveals mixed results; some studies have found benefits for combined treatments, whereas others have not (8). In the context of RA, the use of antidepressant medication alone appears to be equally as effective in the acute phase treatment of major depression as is the use of a pharmacologic/psychologic combination.
A second major finding pertains to the substantial long-term benefit of antidepressive treatment for RA patients with major depression. When the subjects from the 3 study groups (who all received an antidepressive medication) were aggregated, a significant clinical improvement was noted over a 15-month followup. Specifically, antidepressive medication appeared to induce improvements on a number of psychosocial and health status variables as follows: 1) depression, 2) helplessness, 3) self efficacy, 4) psychological distress, 5) perceived life stress, 6) anxiety, 7) coping, 8) fatigue, and 9) the AIMS affect, symptom, and social interaction scores. From a quality of life standpoint, the notable improvements on psychosocial and health status measures is exceedingly important for the subgroup of RA patients who have concomitant major depression.
A third key finding was that there was no impact on pain and/or disease activity measures following the antidepressive treatment. This finding is contrary to the original hypotheses that were based on an impressive research literature indicating associations between measures of pain and depression (44). Likewise, the hypothesized impact of antidepressive intervention on disease activity was not confirmed, even though the literature describes potential neurohormonal pathways in the pathophysiology of both RA and major depression (45). Clearly, further research will be required to more fully examine psychoneurologic aspects of RA and major depression.
A potential limitation of any study that does not include a randomized no-treatment group is that spontaneous recovery from depression cannot be equivocally ruled out. Therefore, steps were taken to examine 18-month outcomes for a group of patients, albeit nonrandomized, who exhibited symptoms of severe depression to a comparable degree as did the antidepressive intervention group. Hence, a fourth major finding was that persons with RA and concomitant depression who did not receive an antidepressive intervention were significantly more depressed at followup than were the persons who did receive antidepressant treatment. The data provide support for the effectiveness and utility of antidepressive treatment for persons with concomitant RA and major depression.
Several limitations of the study must be acknowledged. The sample size was relatively small, although a Type II statistical error does not appear likely. Similarly, the number of statistical comparisons was high, but an alpha adjustment to the 0.01 level was utilized. The generalizability of the midwestern sample is unknown. Lastly, for ethical reasons, a randomized no-treatment group was not possible, although a nonrandomized comparison group was defined to rule out spontaneous recovery from depression.
Future research is needed to explore the underlying mechanisms of antidepressive intervention and to identify alternative strategies for reducing pain and disease activity in the subgroup of persons with both RA and major depression.
The authors gratefully acknowledge the contributions of Barbara Cullen, Kimberli Holtmeyer, Karri Amelunke, and Patricia Cornwell.