Abstract
Objective:
To identify depression subtypes in participants with or at-risk for symptomatic knee osteoarthritis (OA) and to evaluate differences in pain and disability trajectories between groups.
Methods:
Participants (n=4486) were enrolled in the Osteoarthritis Initiative. Latent class analysis (LCA) was applied to the 20-Item Center for Epidemiological Studies Depression Scale measured at baseline to identify groups with similar patterns of depressive symptoms, and subtypes were assigned using poster probability estimates. The relationships between depression subtypes and Western Ontario and McMaster Universities Osteoarthritis Index (WOMAC) pain and disability subscales were modeled over four years and were stratified by baseline knee OA status [symptomatic (n=1626) or at-risk (n=2860)].
Results:
Four subtypes were identified: Asymptomatic (80.6%), Catatonic (5.3%), Anhedonic (10.6%), and Melancholic (3.5%). Catatonic and Anhedonic subtypes were differentiated by symptoms corresponding to psychomotor agitation and the inability to experience pleasure, respectively. The Melancholic subtype expressed symptoms related to reduced energy and movement, anhedonia, and other somatic complaints. Detectable mean differences in pain and disability compared to the Asymptomatic group were observed for the Anhedonic (1.52.3 WOMAC units) and Melancholic (4.8–6.6 WOMAC units) subtypes, and associations were generally larger in persons with symptomatic knee OA relative to those at-risk.
Conclusion:
Among persons with or at-risk for symptomatic knee OA, there is evidence of depression subtypes characterized by distinct clusters of depressive symptoms that have differential effects on reports of pain and disability over time. Thus, findings imply depression interventions could be optimized by targeting the specific symptomology these subtypes exhibit.
INTRODUCTION
Osteoarthritis (OA) is the most common joint disorder in the United States (US); symptomatic knee OA affects approximately 9.3 million American adults and 10% of men and 13% of women aged 60 years or older (1, 2). Knee OA represents a failure of normal joint repair and is often accompanied by symptoms of pain and disability (3). Worsening OA disease severity may lead to the development of psychiatric comorbidity, particularly depressive symptoms, which can exacerbate the course of pain, disability, and disease progression (4–8). The bidirectional relationship between OA disease severity and depressive symptoms may influence the severity of both, thus complicating medical management and contributing to higher health care costs, decreased quality of life, and greater mortality (9–11).
Depression presenting in chronic diseases is difficult to recognize; indeed, depressive symptoms in arthritis patients are under-diagnosed (12, 13). A contributing factor to the under-recognition of depression in OA patients is the overlapping somatic symptomology between the conditions; resultantly, many individuals are only treated for their chronic disease and not depressive symptoms (12). A meta-analysis estimated that depressive symptoms are present in 18.5% of adults with knee OA, a prevalence that is more than double that in the US general population and has remained unchanged over time, despite increases in depression treatment rates (14–16). Depression treatment in American adults predominantly consists of pharmacotherapy using antidepressant medications (15). However, many patients do not achieve symptomatic remission, and non-response is even more pervasive in those with chronic diseases (10, 17).
Increasingly, major depressive disorder is recognized to be heterogenous with respect to clinical presentation (12). Official classifications and corresponding treatments for major depressive disorder have traditionally used a “one size fits all” approach, yet it is becoming more accepted that such definitions and management strategies do not accurately reflect the immense heterogeneity of depressive symptomology (12, 18). In the research and clinical setting, depressive symptoms are generally evaluated in terms of symptom count or a dichotomous indicator, which does not differentiate patients with disparate symptomology (18). By contrast, nascent research has begun utilizing the “Depression Symptomics” framework, a methodology that evaluates how different constellations of depressive symptoms cluster and differentially impact health, well-being, and medical management (19).
Currently, variability in depressive symptoms among OA patients and the implications of distinct depression profiles on OA disease severity is poorly understood. Understanding heterogeneity in depressive symptoms among persons with or at-risk for symptomatic knee OA could lead to more personalized interventions that target individuals’ symptomatic profiles in order to improve clinical outcomes for both conditions. Study objectives were to 1) identify depression subtypes in persons with or at-risk for symptomatic knee OA based on patterns of depressive symptoms and 2) examine the impact of depression subtypes on pain and disability stratified by individuals with or at-risk symptomatic knee OA.
MATERIALS & METHODS
Study Data & Sample
The sample included participants from the Osteoarthritis Initiative (OAI), an observational cohort study designed to identify risk factors and biomarkers for the onset and progression of knee OA, and methodologic details have been published previously (20). Institutional Review Boards at each site approved the OAI study, and all participants provided informed consent. The OAI cohort (n=4,796) was restricted to participants (n=4,486) with baseline symptomatic knee OA data and baseline radiographs that were read centrally at Boston University by trained, certified, radiologic technicians (21). This sample (Figure 1) was used to examine heterogeneity in baseline depressive symptoms, and there were 1,626 and 2,860 persons with or at-risk (respectively) for symptomatic knee OA at baseline, defined as “pain most days of a month in past 12 months” (20). Only one knee per participant selected by random sampling was included in the analysis. Participants with complete baseline data and at least one follow-up observation were used for longitudinal analyses. Complete data on covariates and outcomes at study enrollment were available for 4,300 participants, and information on pain and disability during follow-up were available for 4,204, 4,043, 3,990, and 3,970 participants at the first, second, third, and fourth annual follow-up visits, respectively, in persons with fully observed baseline data. Depressive episodes last between 6 to 12 months, and a four year follow-up period provided sufficient time to examine associations between baseline depression subtypes and pain and disability (22).
Figure 1.
Study sample flow diagram
Depressive Symptoms
The 20-Item Center for Epidemiologic Studies Depression (CES-D) scale was used to assess an array of depressive symptoms: psychomotor agitation/retardation, poor appetite, restless sleep, sadness, feelings of loneliness, social interactions, impaired concentration, and anhedonia (23). CES-D items have a reference time frame corresponding to the occurrence of depressive symptoms in the prior week, and response options ranging from 0 to 3, where increasing score is representative of greater symptomatic frequency (23). Binary indicators were created for each symptom, classifying individuals responding “2” (i.e., occasionally or a moderate amount of time) or “3” (i.e., most or all of the time) as having a given symptom, an approach which has be used in prior research (24). These 20 binary symptom indicators were used to identify depression subtypes.
Pain and Disability
The Western Ontario and McMaster Universities Arthritis Index (WOMAC; Likert version 3.1) was used to assess knee OA disease severity. The WOMAC assessed three distinct domains: stiffness (2 items), pain (5 items), and disability (17 items) (25). WOMAC item-responses are on a Likert scale, ranging from zero (“none”) to four (“extreme”), with higher scores indicating greater severity. Item scores are summed across subscales to calculate summary scores for each domain. In the current study, pain and disability subscales were used as outcomes and assessed at baseline and four annual follow-up visits. Pain and disability scores were rescaled to range from 0–100 so that estimates can be compared to minimal perceptible clinical differences. Clinically significant differences on rescaled WOMAC pain and disability scores are approximately 9.7 and 9.3 units, respectively (26).
Confounders
Potential confounders measured at study baseline were selected a priori based on review of the research literature. Sociodemographic and behavioral measures were age (years), sex, race (white or non-white), marital status (married, widowed, divorced, separated, or never married), educational attainment (high school, college graduate, or graduate degree), health insurance (insured or uninsured), employment (employed or unemployed), alcohol consumption (none, minimal, or moderate), and smoking (never, former, or former). Clinical characteristics included body mass index (BMI; kilograms per meters squared), comorbidity, history of knee injuries, use of analgesic medications, OA disease severity (Kellgren-Lawrence [K-L] grade) and total WOMAC score. Comorbidity was measured using the Charlson Comorbidity Index, a composite scale comprising 22 different comorbid conditions that does not incorporate major depression (27). History of knee injuries was assessed as “ever injured badly enough to limit ability to walk for at least two days.” Analgesic medications were operationalized as the use of acetaminophen, nonsteroidal anti-inflammatory drugs, or opioids within the previous 30 days. K-L grade is an ordinal scale ranging from zero to four, and higher values represent worse structural disease (28).
Latent Class Analysis
Latent class analysis (LCA) is an approach that identifies subgroups (i.e., classes) based on multiple indicators of a given construct (e.g., depression) and was used to identify classes of individuals with similar patterns of depressive symptoms (29). LCA assumes mutually exclusive and exhaustive classes of individuals that are differentiated within a population by values of observed indicators (29). The LCA model estimates the prevalence of each class in the overall sample and item-response probabilities within each class; namely, the probabilities of endorsing each indicator given membership in a specific class (29). For each individual, posterior class probabilities are estimated to provide a person’s likelihood of membership in each subtype, given their indicator response pattern (29).
LCA models were implemented sequentially with one to six classes and were evaluated regarding fit, parsimony, and clinical interpretability. Akaike Information Criterion (AIC) and Bayesian Information Criterion (BIC) fit statistics were used to compare LCA models with different numbers of classes (30, 31). Model uncertainty was assessed using relative entropy, with values ranging from zero to one, and higher estimates indicate greater classification certainty. Class prevalence estimates and item-response probabilities were used to qualitatively examine and describe the different classes and select the optimal number of depression subtypes. LCA model fitting was conducted in the overall study sample (n=4486) and then stratified by participants with (n=1626) or at-risk (n=2860) for symptomatic knee OA to assess measurement invariance. LCA estimation allows missing values in response variables and were assumed to be missing at random. After identifying the optimal LCA model, posterior probability estimates were used to assign individuals to the subtype for which they had the highest probability of membership. Chi-square tests and analysis of variance were used to evaluate between-group differences in baseline covariates by depression subtype. Analyses were conducted using R statistical software (version 3.4.1).
Propensity Scores
Multiple-group propensity score weights were used to balance between-group differences in potential baseline confounders by depression subtype in order to promote causal interpretations regarding their effect on pain and disability (32). Propensity score weights were estimated using boosted regression, which has been shown to outperform other methods of estimation concerning bias reduction (32). The generalized boosted model is a flexible machine learning estimation routine that fits multiple regression trees to account for potentially complex and non-linear relationships between exposure and covariates without overfitting data (32). The boosting algorithm optimizes balance on covariates across groups; indicators for missing covariates were automatically included in the propensity model, such that depression subtypes were balanced regarding observed covariates as well as missing data patterns (32). Given that longitudinal analyses were stratified by symptomatic knee OA status at baseline, weights were estimated separately in at-risk and symptomatic OIA participants. Propensity score weights in the subsamples were stabilized with the marginal probability for each depression subtype. Standardized covariate differences were used to assess balance in the weighted and unweighted samples, and differences of ≥ 0.2 standard deviations were considered evidence of imbalance.
Weighted Estimating Equations
Weighted estimating equations (WEE) were used to assess the relationship between depression subtype and pain and disability over four years stratified by participants with and at-risk for symptomatic knee OA. WEE account for missing data using weighting, where weights are the inverse probability of observation conditional on predictors of missing data and were stabilized using the time-specific marginal probabilities for response. Final weights were the product of the time-invariant propensity score weights and time-specific non-response weights. WEE were implemented using survey analysis methods that are appropriate for clustered data and can be used to estimate population-average exposure effects while accounting for between-person heterogeneity in standard error estimates. Models included categorical indicators for depression subtype, follow-up time, and their interaction to determine whether there were differences in pain and disability across subtypes over time. Differences in pain and disability by depression subtype at each time point were estimated with 95% confidence intervals (CI), and an alpha level of 0.05 was used to define statistical significance.
RESULTS
Depression Subtypes
A four-class LCA model was chosen based on clinical interpretability of the subtypes, prevalence estimates, and fit statistics. Indicators of model fit implied that a five-class model provided a more optimal solution; however, the fifth class was not uniquely distinct when compared to one of the other identified subtypes. Moreover, class prevalence structure and item-response probabilities were consistent by symptomatic knee OA status at baseline, and the relative model entropy of 0.86 indicated a high degree of classification certainty. Four depression subtypes (Table 1) were identified: Asymptomatic (80.6%), Catatonic (5.3%), Anhedonic (10.6%), and Melancholic (3.5%). CES-D score was lowest (Table 2) in Asymptomatic participants, comparable between the Catatonic and Anhedonic subtypes, and highest in the Melancholic group. Similarly, the proportions of individuals meeting CES-D screening criteria for probable depression were 0.4%, 36.7%, 36.2%, and 100% for the Asymptomatic, Catatonic, Anhedonic, and Melancholic subtypes, respectively.
Table 1.
Baseline CES-D item response probabilities by depressive symptom subtype among participants who had or were at risk for knee osteoarthritis.
 |
Indicates CES-D item was reversed scored.
Table 2.
Baseline sample characteristics by depressive symptom subtype before propensity score weighting.
| Variable (mean (sd) or n (%)) | Asymptomatic (n=3,615) |
Catatonic (n=239) |
Anhedonic (n=474) |
Melancholic (n=158) |
P Value |
|---|---|---|---|---|---|
| Age | 61.24 (9.09) | 61.06 (9.37) | 61.87 (9.81) | 57.44 (8.27) | <0.001 |
| Female | 2054 (56.8) | 157 (65.7) | 283 (59.7) | 113 (71.5) | <0.001 |
| White | 630 (17.4) | 82 (34.6) | 122 (25.7) | 53 (33.5) | <0.001 |
| Marital status | <0.001 | ||||
| Married | 2537 (70.8) | 124 (52.1) | 267 (56.3) | 75 (47.8) | |
| Widowed | 250 (7.0) | 30 (12.6) | 59 (12.4) | 13 (8.3) | |
| Divorced | 467 (13.0) | 41 (17.2) | 80 (16.9) | 34 (21.7) | |
| Separated | 45 (1.3) | 11 (4.6) | 11 (2.3) | 8 (5.1) | |
| Never Married | 285 (8.0) | 32 (13.4) | 57 (12.0) | 27 (17.2) | |
| Education | <0.001 | ||||
| No Degree | 1265 (35.3) | 137 (57.6) | 252 (53.2) | 97 (61.4) | |
| College Degree | 1111 (31.0) | 53 (22.3) | 123 (25.9) | 37 (23.4) | |
| Graduate Degree | 1208 (33.7) | 48 (20.2) | 99 (20.9) | 24 (15.2) | |
| Employment | 2272 (62.9) | 133 (55.6) | 279 (59.0) | 88 (55.7) | 0.022 |
| Health insurance | 3507 (97.9) | 219 (92.0) | 442 (94.0) | 138 (87.3) | <0.001 |
| Smoking | <0.001 | ||||
| Never | 1944 (54.6) | 104 (43.9) | 232 (49.5) | 78 (50.0) | |
| Current | 177 (5.0) | 29 (12.2) | 41 (8.7) | 30 (19.2) | |
| Former | 1438 (40.4) | 104 (43.9) | 196 (41.8) | 48 (30.8) | |
| Alcohol consumption | 0.015 | ||||
| None | 641 (17.9) | 55 (23.1) | 104 (22.0) | 42 (26.6) | |
| Minimal | 2634 (73.5) | 163 (68.5) | 338 (71.5) | 105 (66.5) | |
| Moderate | 307 (8.6) | 20 (8.4) | 31 (6.6) | 11 (7.0) | |
| Charlson comorbidity | 0.34 (0.78) | 0.54 (1.04) | 0.54 (1.04) | 0.68 (1.03) | <0.001 |
| BMI | 28.40 (4.68) | 29.33 (5.27) | 29.17 (5.14) | 29.94 (5.79) | <0.001 |
| Symptomatic knee OA | 1224 (33.9) | 116 (48.5) | 202 (42.6) | 84 (53.2) | <0.001 |
| K-L grade | <0.001 | ||||
| 0 | 1455 (40.2) | 74 (31.0) | 163 (34.4) | 52 (32.9) | |
| 1 | 650 (18.0) | 42 (17.6) | 81 (17.1) | 32 (20.3) | |
| 2 | 916 (25.3) | 61 (25.5) | 127 (26.8) | 52 (32.9) | |
| 3 | 484 (13.4) | 54 (22.6) | 79 (16.7) | 19 (12.0) | |
| 4 | 110 (3.0) | 8 (3.3) | 24 (5.1) | 3 (1.9) | |
| History of knee injury | 938 (26.2) | 69 (29.0) | 133 (28.4) | 48 (30.4) | 0.419 |
| Analgesic use | 1238 (34.3) | 108 (45.6) | 196 (41.4) | 95 (60.5) 21.97 | <0.001 |
| WOMAC | 10.18 (13.17) | 17.40 (18.81) | 15.51 (17.64) | (20.28) | <0.001 |
| CES-D score | 4.03 (3.56) | 13.92 (4.59) | 13.85 (4.47) | 28.97 (7.2) | <0.001 |
| CES-D score ≥ 16 | 15 (0.4) | 87 (36.7) | 170 (36.2) | 158 (100.0) | <0.001 |
BMI: Body mass index; CES-D: Center for Epidemiologic Studies Depression scale; K-L grade: Kellgren-Lawrence grade; OA: Osteoarthritis; WOMAC: Western Ontario and McMaster Universities Osteoarthritis Index.
The Asymptomatic group was the most prevalent subtype and had low item-response probabilities for every depressive symptom. The second most common subtype was the Anhedonic group, which was characterized by high item-response probabilities for symptoms corresponding to the inability to experience pleasure and happiness. The Catatonic subtype was the third highest in prevalence and was more likely to endorse symptoms related to psychomotor agitation and somatic complaints; in particular, decreased energy and movement, difficultly concentrating, and restless sleep. The least common subtype was the Melancholic group that had high item-response probabilities for the widest spectrum of symptoms, including sadness, loneliness, anhedonia, psychomotor agitation, and other somatic complaints.
Subtype Characteristics
When compared to those classified as Asymptomatic, other subtypes, particularly Catatonic and Melancholic, were more likely to be female, non-white, not married, and of lower socioeconomic status as measured by educational attainment and employment and health insurance status. Age was comparable across the Asymptomatic, Catatonic, and Anhedonic subtypes but lower in the Melancholic group (57 years versus 61 years). The Catatonic, Anhedonic, and Melancholic subtypes had a higher likelihood of being current smokers but were less likely to consume alcohol than the Asymptomatic group. In addition, Asymptomatic participants had fewer comorbid conditions, lower BMI, lower probability of symptomatic knee OA, and less analgesic medication use compared to every other subtype. Similarly, K-L grade and total WOMAC score were worse in the Catatonic, Anhedonic, and Melancholic subtypes compared to the Asymptomatic group. Subtype characteristics and covariate trends of the overall sample were similar to the stratified samples (Supplementary Table 1 and Supplementary Table 2, respectively). In general, propensity score weights reduced the magnitude of between-subtype standardized covariate differences at baseline below 0.2 standard deviations (Supplementary Table 3 and Supplementary Table 4, respectively).
OA Pain and Disability
In persons at-risk for symptomatic knee OA, the Asymptomatic group had almost no change in pain and disability, while other subtypes generally sustained small increases (Figure 2B & 2D). Time-specific differences in outcomes were smallest in magnitude between the Catatonic and Asymptomatic groups, ranging over four years from −0.20 (95% CI: −2.23, 1.84) to 1.66 (95% CI: −0.59, 3.91) for pain and −1.06 (95% CI: −2.69, 0.57) to 0.70 (95% CI: −1.07, 2.48) for disability (Table 3). However, the between-subtype differences in pain and disability for the Anhedonic and Asymptomatic groups increased in magnitude from baseline (0.45 [95% CI: −0.73, 1.64] and −0.17 [95% CI: −1.17, 0.82], respectively) and were as high as 2.28 (95% CI: 0.33, 4.22) and 2.63 (95%: 0.92, 4.35) during follow-up, respectively. Similarly, the Melancholic subtype had worse outcomes across all four annual follow-up visits, but the time-specific differences in pain and disability were not statistically significant.
Figure 2.
Pain (A and B) and disability (C and D) trajectories by baseline depression subtypes among participants with (A and C) or at-risk (B and D) for symptomatic knee OA.
Table 3.
Time-specific differences (Reference=Asymptomatic) in pain and disability by baseline depression subtype among participants at-risk for symptomatic knee OA.
| Time Point |
Catatonic | Anhedonic | Melancholic | ||||||
|---|---|---|---|---|---|---|---|---|---|
| β | 95% CI | P | β | 95% CI | P | β | 95% CI | P | |
| Pain | |||||||||
| Baseline | 0.01 | −1.74, 1.76 | 0.994 | 0.45 | −0.73, 1.64 | 0.451 | −1.02 | −3.17, 1.12 | 0.349 |
| Year One | 1.54 | −0.91, 4.00 | 0.219 | 1.59 | 0.07, 3.12 | 0.041 | 2.77 | −1.99, 7.53 | 0.254 |
| Year Two | 0.32 | −2.02, 2.66 | 0.790 | 0.97 | −0.62, 2.57 | 0.231 | 0.40 | −2.73, 3.53 | 0.802 |
| Year Three | 1.66 | −0.59, 3.91 | 0.148 | 2.28 | 0.33, 4.22 | 0.022 | 2.59 | −2.23, 7.42 | 0.292 |
| Year Four | −0.20 | −2.23, 1.84 | 0.850 | 1.56 | −0.12, 3.23 | 0.069 | 1.95 | −1.87, 5.77 | 0.316 |
| Disability | |||||||||
| Baseline | −0.26 | −1.79, 1.26 | 0.734 | −0.17 | −1.17, 0.82 | 0.737 | 0.50 | −1.78, 2.78 | 0.667 |
| Year One | 0.70 | −1.07, 2.48 | 0.437 | 1.12 | −0.16, 2.40 | 0.087 | 4.17 | −0.52, 8.86 | 0.082 |
| Year Two | −0.20 | −1.75, 1.34 | 0.798 | 1.70 | 0.17, 3.23 | 0.030 | 0.80 | −2.19, 3.80 | 0.599 |
| Year Three | 0.32 | −1.57, 2.21 | 0.742 | 2.63 | 0.92, 4.35 | 0.003 | 1.55 | −1.55, 4.64 | 0.327 |
| Year Four | −1.06 | −2.69, 0.57 | 0.204 | 1.57 | 0.00, 3.14 | 0.050 | 2.12 | −1.73, 5.96 | 0.281 |
Among participants with symptomatic knee OA, every group experienced improvement from baseline except the Melancholic subtype, which had persistently greater pain and disability during the follow-up period (Figure 2A & 2C, respectively). When compared to the Asymptomatic group, time-specific differences in pain and disability among the Catatonic and Anhedonic subtypes were small in magnitude and generally ≤ 1 rescaled WOMAC units (Table 4). By contrast, differences in pain between the Melancholic and Asymptomatic groups increased from 0.47 (95% CI: −3.68, 4.62) at baseline to 4.79 (95% CI: −1.77, 11.35) at the fourth annual follow-up visit. Similarly, differences in disability increased from 2.80 (95% CI: −1.84, 7.44) at baseline to time-specific differences as large as 6.56 (95% CI: 1.72, 11.40) rescaled WOMAC units in the Melancholic subtype during follow-up.
Table 4.
Time-specific differences (Reference=Asymptomatic) in pain and disability by baseline depression subtype among participants with symptomatic knee OA.
| Time Point |
Catatonic | Anhedonic | Melancholic | ||||||
|---|---|---|---|---|---|---|---|---|---|
| β | 95% CI | P | β | 95% CI | P | β | 95% CI | P | |
| Pain | |||||||||
| Baseline | −1.51 | −5.06, 2.04 | 0.404 | 0.39 | 2.31, 3.09 | 0.779 | 0.47 | −3.68, 4.62 | 0.824 |
| Year One | −0.14 | −3.96, 3.68 | 0.942 | 0.57 | −2.32, 3.46 | 0.701 | 1.33 | −3.40, 6.05 | 0.582 |
| Year Two | −0.33 | −4.34, 3.69 | 0.873 | 0.95 | −2.18, 4.08 | 0.551 | 4.25 | −0.45, 8.95 | 0.077 |
| Year Three | −0.17 | −4.07, 3.72 | 0.932 | −0.64 | −3.58, 2.30 | 0.668 | 2.98 | −2.49, 8.44 | 0.286 |
| Year Four | 0.82 | −3.34, 4.98 | 0.699 | −0.31 | −3.18, 2.55 | 0.830 | 4.79 | −1.77, 11.35 | 0.152 |
| Disability | |||||||||
| Baseline | 1.27 | −2.07, 4.60 | 0.457 | 0.48 | −2.17, 3.12 | 0.723 | 2.80 | −1.84, 7.44 | 0.237 |
| Year One | 0.58 | −3.12, 4.27 | 0.760 | 1.24 | −1.60, 4.08 | 0.393 | 2.76 | −1.74, 7.27 | 0.230 |
| Year Two | −0.54 | −4.26, 3.19 | 0.778 | −0.08 | −2.80, 2.64 | 0.954 | 4.82 | 0.31, 9.33 | 0.037 |
| Year Three | −0.01 | −3.57, 3.54 | 0.994 | 0.14 | −2.68, 2.96 | 0.921 | 6.56 | 1.72, 11.40 | 0.008 |
| Year Four | 1.21 | −2.92, 5.34 | 0.567 | 0.31 | −2.32, 2.94 | 0.817 | 5.35 | −0.90, 11.60 | 0.094 |
DISCUSSION
The current study identified four distinct depression subtypes based on patterns of depressive symptoms in persons with or at-risk for symptomatic knee OA. Consistent with a previous meta-analysis, findings indicate that approximately 80% of OAI participants expressed few symptoms of depression (16). However, findings demonstrated moderate heterogeneity in the 20% of participants reporting more depressive symptoms at baseline, and these subtypes were qualified as Catatonic, Anhedonic, and Melancholic. Moreover, detectable effects on pain and disability across four years of follow-up were limited to the Anhedonic and Melancholic subtypes and were largest in persons with symptomatic knee OA who exhibited both somatic and cognitive symptomology. These results imply there is variability in both the expression and severity of depressive symptoms among participants with or at-risk for symptomatic knee OA that may lead to differences in knee OA outcomes.
Few studies have examined depressive symptom heterogeneity in patients with chronic physical diseases, and the current study highlights three symptomatic depression subtypes identified in the OAI cohort (33). The Catatonic and Anhedonic subtypes were differentiated by somatic and cognitive symptoms, while the Melancholic group had a broader constellation of symptomology. Prior work in cancer patients identified a mild depression subtype presenting with concentration and sleep problems and psychomotor agitation that is like the Catatonic group reported in the current study (34). Research also provides support for a non-dysphoric (i.e., without sadness) depression subtype in older adults typified by slowness of movement, but unlike the Catatonic group, exhibited other cognitive symptoms (35–37). Anhedonic subtypes epitomized solely by the absence of happiness have not been widely reported (38). Research conducted in the general population derived an Anhedonic subtype with similar characteristics to current results: older age and more proportionate sex distribution (39). However, that study suggested the Anhedonic class had fewer future depressive episodes and stressful events than other subtypes; surprising, given Anhedonia is associated with chronic stress and is a risk factor for major depression (39, 40). Subtypes comprised of cognitive and somatic symptoms are consistently reported, and melancholic depression characterized by sadness, anhedonia, decreased energy and movement, difficulty concentrating, restless sleep, and other physical and emotional problems has been identified in general population and elderly samples (33, 41, 42). Melancholic depression is a more severe phenotype, evidenced by higher CES-D scores satisfying screening criteria for every participant in this group, and is associated with an almost two-fold response time to pharmacological treatment compared to other subtypes (43). Nonetheless, one third of individuals in the Catatonic and Anhedonic groups screened positive for probable depression, and these subtypes not only represent differences in severity, but illustrate different patterns of symptomology and may explain (in part) why almost 60% of depressed OA patients do not receive mental health care (44).
Differences in pain and disability by depression subtype further highlight the difficulty in managing psychosomatic factors in knee OA patients. The Catatonic subtype was the only group that did not experience significantly greater pain and disability than Asymptomatic participants. This finding is contrary to research showing that older adults who report somatic symptoms of depression but deny feelings of sadness are at an increased risk for functional impairment (35). Perhaps the Catatonic subtype represents a mild phenotype, but it may also indicate somatic symptom overlap, where CES-D items detect symptomology of knee OA or another unrelated medical condition (e.g., fatigue) that is not predictive of pain and disability (34, 45, 46). By contrast, participants at-risk for symptomatic knee OA in the Anhedonic subtype had statistically greater pain and disability (1.5–2.3 WOMAC units) compared to Asymptomatic individuals that was small in magnitude and not clinically significant. However, prior research indicates that OAI participants experience minimal changes in their knee OA symptoms; therefore, psychosomatic factors that contribute to persistently higher disease severity may be relevant at the population level (47). The cognitive symptoms of the Anhedonic subtype may represent perseverative thought, a process of repetitive pessimistic thinking associated with negative affect and other traits closely-related to depression, which could act as a response-shift and lead to reports of worse pain and disability (48). Alternatively, the Anhedonic subtype may constitute a mild mood disorder which, following the onset of a chronic physical disease, decompensates in to melancholic depression in some individuals with symptomatic knee OA. This premise is supported by persistently higher disease severity (4.8–6.6 WOMAC units) among symptomatic knee OA participants in the Melancholic subtype, perhaps indicative of substantive increases in pain severity and decreased physical performance associated with a greater number and spectrum of depressive symptoms (49). Unfortunately, it is not possible to determine causality regarding knee OA and depression in the current study, and the melancholic subtype may represent more severe depressive symptoms that are a consequence of pain, disability, or other related factors. Notwithstanding, Anhedonia, and to a larger extent, Melancholic depression, may be a modifiable risk factor for worsening knee OA disease severity and potential target for intervention.
There are limitations that should be considered when interpreting study results. First, the 20-Item CES-D lacks assessments of symptoms (e.g., hallucinations, risk taking behaviors, etc.) commonly found in Psychotic and Atypical depression subtypes (23). Nevertheless, the CES-D is a valid and reliable measure of depressive symptoms in persons with knee OA that has been used in prior OAI studies on depression (6, 7, 49). Second, a three-step design was used, where the measurement model was estimated, participants were assigned to a subtype, and a structural model of outcomes was fit; an approach that may lead to misclassification of class assignment. However, relative model entropy suggested a minimal level of class uncertainty and any resulting bias would attenuate associations to the null. Lastly, the potential for confounding by unmeasured factors, such as depression treatments, cannot be eliminated here and is the case for all observational studies. These limitations are mitigated by the study’s strengths. First, this is one of the largest LCA studies on depression heterogeneity in persons with or at-risk for chronic musculoskeletal disorders. Second, modern statistical techniques that leveraged machine learning and inverse probability weighting methods were used to overcome model misspecification when adjusting for potential confounders and missing data. Finally, the OAI is a well-documented, prospective cohort, and measured comprehensive sociodemographic and clinical characteristics relevant to assessments of depression subtypes and their influence OA pain and disability.
To conclude, study findings indicate that depression subtypes among individuals with or at-risk for symptomatic knee OA are differentiated primarily by psychomotor agitation, anhedonia, and other somatic complaints. Moreover, Anhedonic and Melancholic depression subtypes may be risk factors for increased pain and disability in persons who develop and experience symptomatic knee OA. These results demonstrate the advantages of using a datadriven approach to identify distinct depression subtypes that represent unique clinical phenotypes presenting with different spectrums of symptoms. Consistent with clinical recommendations, our findings support the need for depression screening in knee OA patients during primary care and rheumatology encounters with simple, reliable, and valid instruments (e.g., 2-item Patient Health Questionnaire) (50). If depression screening becomes routine in patients with musculoskeletal disorders, more comprehensive symptomatic assessments could be conducted after referral to mental health professionals, where patients then receive treatments that are tailored to their specific depressive symptomology. However, the feasibility and cost-effectiveness of more complex assessment methods in the secondary care setting that are required for depression subtype identification are unknown. Ultimately, findings highlight the difficulty of using standard depression treatments in individuals with musculoskeletal disorders, and protocols that address the specific symptomatology of different subtypes of depression that may present in knee OA patients are needed.
SIGNIFICANCE & INNOVATION.
Depressive symptoms in persons with or at-risk for symptomatic knee OA present as one of four unique subtypes
Depression subtypes are differentiated primarily by psychomotor agitation, anhedonia, and other somatic complaints
Anhedonic and Melancholic depression subtypes may be risk factors for increased pain and disability
Results highlight the need for protocols designed to address the specific symptomatology of different depression subtypes
ACKNOWLEDGEMENTS
The OAI is a public-private partnership comprised of five contracts (N01-AR-2–2258; N01-AR-2–2259; N01-AR-2–2260; N01-AR-2–2261; N01-AR-2–2262) funded by the National Institutes of Health, a branch of the Department of Health and Human Services, and conducted by the OAI Study Investigators. Private funding partners include Merck Research Laboratories; Novartis Pharmaceuticals Corporation, GlaxoSmithKline; and Pfizer, Inc. Private sector funding for the OAI is managed by the Foundation for the National Institutes of Health. This manuscript was prepared using an OAI public use data set and does not necessarily reflect the opinions or views of the OAI investigators, the NIH, or the private funding partners.
Funding: This study was supported by the Rheumatology Research Foundation’s Scientist Development Award. The sponsor had no role in the study design, collection, analysis and interpretation of data; in the writing of the manuscript; and in the decision to submit the manuscript for publication.
Disclosures: Dr. Alan M. Rathbun is supported by a grant from the Rheumatology Research Foundation. Drs. Megan S. Schuler, Elizabeth A. Stuart, Michelle D. Shardell, Michelle S. Yau, Joseph J. Gallo, Alice S. Ryan, and Marc C. Hochberg have no conflicts of interest to declare.
Footnotes
Publisher's Disclaimer: This article has been accepted for publication and undergone full peer review but has not been through the copyediting, typesetting, pagination and proofreading process, which may lead to differences between this version and the Version of Record. Please cite this article as doi: 10.1002/acr.23898
REFERENCES
- 1.Lawrence RC, Felson DT, Helmick CG, Arnold LM, Choi H, Deyo RA, et al. Estimates of the prevalence of arthritis and other rheumatic conditions in the United States: Part II. Arthritis & Rheumatism. 2008;58(1):26–35. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 2.Zhang Y, Jordan JM. Epidemiology of osteoarthritis. Clinics in geriatric medicine. 2010;26(3):355–69. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 3.Lane NE, Brandt K, Hawker G, Peeva E, Schreyer E, Tsuji W, et al. OARSI-FDA initiative: defining the disease state of osteoarthritis. Osteoarthritis and Cartilage. 2011;19(5):478–82. [DOI] [PubMed] [Google Scholar]
- 4.Rathbun AM, Harrold LR, Reed GW. Temporal associations between the different domains of rheumatoid arthritis disease activity and the onset of patient-reported depressive symptoms. Clinical rheumatology. 2014:1–11. [DOI] [PubMed] [Google Scholar]
- 5.Rathbun AM, Harrold LR, Reed GW. Temporal Effect of Depressive Symptoms on the Longitudinal Evolution of Rheumatoid Arthritis Disease Activity. Arthritis care & research. 2015;67(6):765–75. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 6.Rathbun AM, Stuart EA, Shardell M, Yau MS, Baumgarten M, Hochberg MC. Dynamic Effects of Depressive Symptoms on Osteoarthritis Knee Pain. Arthritis Care & Research. 2017. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 7.Rathbun AM, Yau MS, Shardell M, Stuart EA, Hochberg MC. Depressive symptoms and structural disease progression in knee osteoarthritis: data from the Osteoarthritis Initiative. Clinical rheumatology. 2017;36(1):155–63. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 8.Sugai K, Takeda Imai F, Michikawa T, Nakamura T, Takebayashi T, Nishiwaki Y. Association Between Knee Pain, Impaired Function, and Development of Depressive Symptoms. Journal of the American Geriatrics Society. 2018;66(3):570–6. [DOI] [PubMed] [Google Scholar]
- 9.Rosemann T, Gensichen J, Sauer N, Laux G, Szecsenyi J. The impact of concomitant depression on quality of life and health service utilisation in patients with osteoarthritis. Rheumatology international. 2007;27(9):859–63. [DOI] [PubMed] [Google Scholar]
- 10.Detweiler-Bedell JB, Friedman MA, Leventhal H, Miller IW, Leventhal EA. Integrating co-morbid depression and chronic physical disease management: identifying and resolving failures in self-regulation. Clinical psychology review. 2008;28(8):1426–46. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 11.Carstensen J, Andersson D, Andre M, Engstrom S, Magnusson H, Borgquist LA. How does comorbidity influence healthcare costs? A population-based cross-sectional study of depression, back pain and osteoarthritis. BMJ open. 2012;2(2):e000809–2011-. Print 2012. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 12.Goldberg D. The heterogeneity of “major depression02”. World Psychiatry. 2011;10(3):226–8. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 13.Rathbun AM, Harrold LR, Reed GW. A description of patient- and rheumatologist-reported depression symptoms in an American rheumatoid arthritis registry population. Clinical and experimental rheumatology. 2014;32(4):523–32. [PubMed] [Google Scholar]
- 14.Kessler RC, Berglund P, Demler O, Jin R, Koretz D, Merikangas KR, et al. The epidemiology of major depressive disorder: results from the National Comorbidity Survey Replication (NCS-R). JAMA : the journal of the American Medical Association. 2003;289(23):3095–105. [DOI] [PubMed] [Google Scholar]
- 15.Marcus SC, Olfson M. National trends in the treatment for depression from 1998 to 2007. Archives of general psychiatry. 2010;67(12):1265–73. [DOI] [PubMed] [Google Scholar]
- 16.Stubbs B, Aluko Y, Myint PK, Smith TO. Prevalence of depressive symptoms and anxiety in osteoarthritis: a systematic review and meta-analysis. Age and Ageing. 2016:afw001. [DOI] [PubMed] [Google Scholar]
- 17.Wang PS, Insel TR. NIMH-funded pragmatic trials: moving on. Neuropsychopharmacology. 2010;35(13):2489. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 18.Fried E. Moving forward: how depression heterogeneity hinders progress in treatment and research. Taylor & Francis; 2017. [DOI] [PubMed] [Google Scholar]
- 19.Fried EI, van Borkulo CD, Cramer AO, Boschloo L, Schoevers RA, Borsboom D. Mental disorders as networks of problems: a review of recent insights. Social Psychiatry and Psychiatric Epidemiology. 2017;52(1):1–10. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 20.Nevitt MC, Felson DT, Lester G. The Osteoarthritis Initiative. [Google Scholar]
- 21.Altman RD, Gold GE. Atlas of individual radiographic features in osteoarthritis, revised. Osteoarthritis and cartilage. 2007;15:A1–A56. [DOI] [PubMed] [Google Scholar]
- 22.Furukawa TA, Kitamura T, Takahashi K. Time to recovery of an inception cohort with hitherto untreated unipolar major depressive episodes. The British journal of psychiatry : the journal of mental science. 2000;177:331–5. [DOI] [PubMed] [Google Scholar]
- 23.Radloff LS. The CES-D scale a self-report depression scale for research in the general population. Applied psychological measurement. 1977;1(3):385–401. [Google Scholar]
- 24.Ulbricht CM, Rothschild AJ, Lapane KL. The association between latent depression subtypes and remission after treatment with citalopram: A latent class analysis with distal outcome. Journal of affective disorders. 2015;188:270–7. [DOI] [PubMed] [Google Scholar]
- 25.Bellamy N. Validation study of WOMAC: a health status instrument for measuring clinically-important patient-relevant outcomes following total hip or knee arthroplasty in osteoarthritis. J Orthop Rheumatol. 1988;1:95–108. [PubMed] [Google Scholar]
- 26.Ehrich EW, Davies GM, Watson DJ, Bolognese JA, Seidenberg BC, Bellamy N. Minimal perceptible clinical improvement with the Western Ontario and McMaster Universities osteoarthritis index questionnaire and global assessments in patients with osteoarthritis. The Journal of rheumatology. 2000;27(11):2635–41. [PubMed] [Google Scholar]
- 27.Charlson M, Szatrowski TP, Peterson J, Gold J. Validation of a combined comorbidity index. Journal of clinical epidemiology. 1994;47(11):1245–51. [DOI] [PubMed] [Google Scholar]
- 28.Kellgren JH, Lawrence JS. Radiological assessment of osteo-arthrosis. Annals of the Rheumatic Diseases. 1957;16(4):494–502. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 29.Lanza ST, Rhoades BL. Latent class analysis: an alternative perspective on subgroup analysis in prevention and treatment. Prevention Science. 2013;14(2):157–68. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 30.Schwarz G. Estimating the dimension of a model. The annals of statistics. 1978;6(2):461–4. [Google Scholar]
- 31.Akaike H. Factor analysis and AIC. Psychometrika. 1987;52(3):317–32. [Google Scholar]
- 32.McCaffrey DF, Griffin BA, Almirall D, Slaughter ME, Ramchand R, Burgette LF. A tutorial on propensity score estimation for multiple treatments using generalized boosted models. Statistics in medicine. 2013;32(19):3388–414. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 33.Van Loo HM, De Jonge P, Romeijn J-W, Kessler RC, Schoevers RA. Data-driven subtypes of major depressive disorder: a systematic review. BMC medicine. 2012;10(1):156. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 34.Zhu L, Ranchor AV, van der Lee M, Garssen B, Sanderman R, Schroevers MJ. Subtypes of depression in cancer patients: an empirically driven approach. Supportive care in cancer. 2016;24(3):1387–96. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 35.Gallo JJ, Rabins PV, Lyketsos CG, Tien AY, Anthony JC. Depression without sadness: functional outcomes of nondysphoric depression in later life. Journal of the American Geriatrics Society. 1997;45(5):570–8. [DOI] [PubMed] [Google Scholar]
- 36.Gallo JJ, Rabins P, Anthony J. Sadness in older persons: 13-year follow-up of a community sample in Baltimore, Maryland. Psychological Medicine. 1999;29(2):341–50. [DOI] [PubMed] [Google Scholar]
- 37.Gallo JJ, Rabins PV. Depression without sadness: alternative presentations of depression in late life. American family physician. 1999;60(3):820–6. [PubMed] [Google Scholar]
- 38.Penninx BW, Milaneschi Y, Lamers F, Vogelzangs N. Understanding the somatic consequences of depression: biological mechanisms and the role of depression symptom profile. BMC medicine. 2013;11:129-7015-11-129. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 39.Chen L-S, Eaton WW, Gallo JJ, Nestadt G. Understanding the heterogeneity of depression through the triad of symptoms, course and risk factors: a longitudinal, population-based study. Journal of affective disorders. 2000;59(1):1–11. [DOI] [PubMed] [Google Scholar]
- 40.Loas G. Vulnerability to depression: a model centered on anhedonia. Journal of Affective Disorders. 1996;41(1):39–53. [DOI] [PubMed] [Google Scholar]
- 41.Lamers F, Beekman A, Van Hemert A, Schoevers R, Penninx B. Six-year longitudinal course and outcomes of subtypes of depression. The British Journal of Psychiatry. 2016;208(1):62–8. [DOI] [PubMed] [Google Scholar]
- 42.Veltman E, Lamers F, Comijs H, De Waal M, Stek M, Van der Mast R, et al. Depressive subtypes in an elderly cohort identified using latent class analysis. Journal of affective disorders. 2017;218:123–30. [DOI] [PubMed] [Google Scholar]
- 43.Bühler J, Seemüller F, Läge D. The predictive power of subgroups: an empirical approach to identify depressive symptom patterns that predict response to treatment. Journal of affective disorders. 2014;163:81–7. [DOI] [PubMed] [Google Scholar]
- 44.Gleicher Y, Croxford R, Hochman J, Hawker G. A prospective study of mental health care for comorbid depressed mood in older adults with painful osteoarthritis. BMC psychiatry. 2011;11:147-244X-11-147. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 45.Addington A, Gallo J, Ford D, Eaton W. Epidemiology of unexplained fatigue and major depression in the community: the Baltimore ECA follow-up, 1981–1994. Psychological medicine. 2001;31(6):1037–44. [DOI] [PubMed] [Google Scholar]
- 46.Pincus T, Hassett AL, Callahan LF. Criterion contamination of depression scales in patients with rheumatoid arthritis: the need for interpretation of patient questionnaires (as all clinical measures) in the context of all information about the patient. Rheumatic Disease Clinics of North America. 2009;35(4):861–4. [DOI] [PubMed] [Google Scholar]
- 47.Collins JE, Katz JN, Dervan EE, Losina E. Trajectories and risk profiles of pain in persons with radiographic, symptomatic knee osteoarthritis: data from the osteoarthritis initiative. Osteoarthritis and Cartilage. 2014;22(5):622–30. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 48.Rathbun AM, Reed GW, Harrold LR. The temporal relationship between depression and rheumatoid arthritis disease activity, treatment persistence and response: a systematic review. Rheumatology (Oxford, England). 2013;52(10):1785–94. [DOI] [PubMed] [Google Scholar]
- 49.Rathbun AM, Shardell MD, Stuart EA, Yau MS, Gallo JJ, Schuler MS, et al. Pain Severity as a Mediator of the Association between Depressive Symptoms and Physical Performance in Knee Osteoarthritis. Osteoarthritis and Cartilage. 2018. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 50.Cg N. Osteoarthritis Care and Management in Adults. London: National Institute for Health and Care Excellence; 2014. [PubMed] [Google Scholar]