Background
Chronic obstructive pulmonary disease (COPD) is one of the leading causes of death worldwide, and affects 9-10% of adults aged 40 years or older [
1,
2]. It is a slowly progressive lung disease, characterized by chronic airway inflammation and not fully reversible airflow obstruction [
3]. Cigarette smoking is the main cause of COPD in developed countries, and patients typically present with shortness of breath, chronic cough and/or excessive sputum production. In addition to a higher risk of mortality and morbidity, COPD is associated with a substantial economic burden of illness to the health care system [
4‐
6].
The management of COPD is largely symptomatic, so patient-reported outcomes (PROs) that evaluate health-related quality of life (HRQL) are important to evaluating the treatment and management of COPD. As COPD progresses, poor symptom control and exacerbations can lead to limitations in functioning and impaired HRQL [
7]. Disease-specific measures can provide insight into specific aspects of HRQL, while generic HRQL measures have the advantage of being able to compare across different patient populations but they may be less sensitive to changes in HRQL compared to disease-specific measures [
8,
9].
Two generic HRQL measures, the Patient Reported Outcomes Measurement Information System (PROMIS) and EQ-5D-5L, were recently developed that have potential for broad use in evaluating COPD outcomes. PROMIS is a health measurement system designed for a wide variety of patient populations that utilizes banks of items belonging to specific domains of health [
10]. The PROMIS item banks were derived using item response theory (IRT) and developed through a rigorous process of literature review, focus groups across multiple diseases and sites, cognitive assessments, and expert consultation. In addition, fixed-length short forms, including the PROMIS-43, were developed as an alternative computer-adaptive testing (CAT). These short forms, while not providing as precise measurement estimates as using CAT, cover core dimensions of health from the PROMIS. Another measure, the EQ-5D-5L, expanded the EQ-5D from 3 Levels to 5 Levels in order to potentially improve upon the properties of the standard 3-level EQ-5D by enhancing sensitivity and reducing ceiling effects [
11].
As few studies have examined these recently developed measures in COPD, the aims of this study were: (1) to examine their psychometric properties in patients with COPD, and (2) to identify dimensions of HRQL that differ and do not differ by lung function.
Results
A total of 670 patients were included in our analysis. The mean age was 68.5 (SD 10.4) years, with 58% men and 78% self-identified as caucasian (Table
1). Most patients were current (n = 193, 28.8%) or past (n = 412, 61.5%) smokers, with an average of 44.2 (SD 31.2) pack-years. The study subjects were divided into 4 groups based on the severity of airflow limitation: mild (GOLD 1, n = 102), moderate (GOLD 2, n = 353), severe (GOLD 3, n = 165) and very severe (GOLD 4, n = 50). Patients included as GOLD 3 and 4 (with more severe disease) were significantly younger, more likely to be African-American, have lower education level, less household income, heavier smoking history, and less likely to have coronary heart disease as compared to those with mild to moderate disease (p < 0.05) (Table
1).
Table 1
Patient demographic and clinical characteristics (Total N = 670)
Age, mean (SD) | 0 | 68.5 (10.4) | 72.1 (11.3) | 68.3 (10.5) | 67.7 (9.7) | 65.1 (7.4) | 0.02§
|
Male | 0 | 387 (57.8) | 69 (67.7) | 203 (57.5) | 88 (53.3) | 27 (54.0) | 0.12†
|
Race | 0 | | | | | | <.0001‡
|
Caucasian | | 524 (78.2) | 91 (89.2) | 290 (82.2) | 114 (69.3) | 29 (58.0) | |
African American | | 121 (18.1) | 11 (10.8) | 52 (14.7) | 40 (24.1) | 18 (36.0) | |
Native American* | | 20 (3.0) | 0 (0) | 8 (2.3) | 9 (5.4) | 3 (6.0) | |
Asian/multiracial/other | | 5 (0.8) | 0 (0) | 3 (0.9) | 2 (1.2) | 0 (0) | |
BMI (Kg/m2), mean (SD) | 4 | 29.1 (7.6) | 29.1 (6.2) | 29.4 (7.0) | 29.4 (8.9) | 26.0 (9.3) | 0.26§
|
Main activity | 0 | | | | | | 0.30†
|
Employed (incl.self-employment) | | 106 (15.8) | 18 (17.7) | 61 (17.3) | 20 (12.1) | 7 (14.0) | |
Retired | | 451 (67.3) | 69 (67.7) | 234 (66.3) | 114 (69.1) | 34 (68.0) | |
Keeping house/student/seeking work | | 61 (9.1) | 10 (9.9) | 31 (8.8) | 17 (10.3) | 3 (6.0) | |
Disabled – on disability | | 41 (6.1) | 5 (4.9) | 22 (6.2) | 8 (4.9) | 6 (12.0) | |
Other | | 11 (1.6) | 0 (0) | 5 (1.4) | 6 (3.6) | 0 (0) | |
Education | 0 | | | | | | 0.04†
|
Some high school or less | | 92 (13.7) | 12 (11.8) | 48 (13.6) | 27 (16.4) | 5 (10.0) | |
High school graduate or GED | | 178 (26.6) | 23 (22.6) | 97 (27.5) | 44 (26.7) | 14 (28.0) | |
Vocational college or some college | | 256 (38.2) | 36 (35.3) | 126 (35.7) | 70 (42.4) | 24 (48.0) | |
College degree | | 86 (12.8) | 14 (13.7) | 53 (15.0) | 16 (9.7) | 3 (6.0) | |
Professional or graduate degree | | 58 (8.7) | 17 (16.7) | 29 (8.2) | 8 (4.9) | 4 (8.0) | |
Household total gross yearly income | 77 | | | | | | 0.002†
|
Less than $30,000 | | 275 (46.4) | 30 (36.1) | 137 (43.5) | 86 (57.3) | 22 (48.9) | |
$30,001 to $50,000 | | 164 (27.7) | 25 (30.1) | 94 (29.8) | 33 (22.0) | 12 (26.7) | |
$50,001 to $75,000 | | 82 (13.8) | 14 (16.9) | 39 (12.4) | 23 (15.3) | 6 (13.3) | |
Over $75,000 | | 72 (12.1) | 14 (16.9) | 45 (14.3) | 8 (5.3) | 5 (11.1) | |
Smoking status | 0 | | | | | | 0.09†
|
Current smoker | | 193 (28.8) | 21 (20.6) | 109 (30.9) | 49 (29.7) | 14 (28.0) | |
Past smoker | | 412 (61.5) | 64 (62.8) | 209 (59.2) | 104 (63.0) | 35 (70.0) | |
Never smoker | | 65 (9.7) | 17 (16.7) | 35 (9.9) | 12 (7.3) | 1 (2.0) | |
Smoke pack-years, mean (SD) | 2 | 44.2 (31.2) | 36.3 (30.3) | 44.2 (30.9) | 48.0 (31.5) | 48.2 (31.8) | 0.03§
|
Comorbid medical conditions | | | | | | | |
Coronary heart disease** | 0 | 220 (32.8) | 42 (41.2) | 119 (33.7) | 51 (30.9) | 8 (1.2) | 0.04†
|
Hypertension | 0 | 416 (62.1) | 57 (55.9) | 221 (62.6) | 110 (66.7) | 28 (56.0) | 0.44†
|
Diabetes | 0 | 154 (23.0) | 20 (19.6) | 85 (24.1) | 39 (23.6) | 10 (20.0) | 0.93†
|
Stroke or cerebrovascular disease | 0 | 96 (14.3) | 13 (12.8) | 53 (15.0) | 25 (15.2) | 5 (10.0) | 0.85†
|
Cancer history | 0 | 174 (26.0) | 22 (21.6) | 96 (27.2) | 44 (26.7) | 12 (24.0) | 0.98†
|
Dementia | 0 | 10 (1.5) | 2 (2.0) | 7 (2.0) | 1 (0.6) | 0 (0) | 0.18‡
|
Depression | 0 | 229 (34.2) | 34 (33.3) | 131 (37.1) | 52 (31.5) | 12 (24.0) | 0.10†
|
All levels of the EQ-5D-5L were utilized by the overall cohort, but only a relatively small proportion of patients reported severe or extreme problems on each dimension (range 1.4-8.7%) (Table
2). More than 50% of patients reported no problems with self-care and anxiety/depression (in all COPD grades) and usual activities (GOLD 1 only). More severe COPD was associated with significantly more problems with mobility, self-care and usual activities (p < 0.01). Approximately 30% of all COPD patients reported at least moderate pain/discomfort and 15% reported at least moderate anxiety/depression, but differences across COPD grades were not significant (p = 0.26 and p = 0.15, respectively). The response of ‘11111’ (no problems in any dimension) was reported in a dimishing proportion of respondents as grades got more severe: 19.6% (GOLD 1), 18.4% (GOLD 2), 11.5% (GOLD 3), and 4% (GOLD 4). Overall, the mean (SD) score was 0.79 (0.15) for the EQ-5D index, and 70.6 (19.6) for the EQ-VAS (Table
3). When stratified by GOLD grade, EQ-5D index-based mean scores ranged from 0.81 (GOLD 1) to 0.74 (GOLD 4) (p-value = 0.004), and EQ-VAS mean scores ranged from 76.6 (GOLD 1) to 61.1 (GOLD 4) (p-value < 0.001). Patients with more severe COPD had lower mean EQ-5D index scores and EQ-VAS scores, although the index-based score did not discriminate between the milder grades of COPD.
Table 2
EQ-5D-5L profile of patients by GOLD grade
EQ-5D dimension |
Mobility | | | | | | 0.003†
|
No problems | 244 (36.4) | 44 (43.1) | 141 (39.9) | 41 (24.9) | 18 (36.0) | |
Slight problems | 179 (26.7) | 28 (27.5) | 93 (26.4) | 46 (27.9) | 12 (24.0) | |
Moderate problems | 189 (28.2) | 25 (24.5) | 92 (26.1) | 58 (35.2) | 14 (28.0) | |
Severe problems | 52 (7.8) | 5 (4.9) | 25 (7.1) | 16 (9.7) | 6 (12.0) | |
Extreme problems | 6 (0.9) | 0 (0) | 2 (0.6) | 4 (2.4) | 0 (0) | |
Self-care | | | | | | <.0001‡
|
No problems | 539 (80.5) | 87 (85.3) | 297 (84.1) | 127 (77.0) | 28 (56.0) | |
Slight problems | 88 (13.1) | 13 (12.8) | 43 (12.2) | 23 (13.9) | 9 (18.0) | |
Moderate problems | 34 (5.1) | 2 (2.0) | 9 (2.6) | 11 (6.7) | 12 (24.0) | |
Severe problems | 5 (0.8) | 0 (0) | 4 (1.1) | 0 (0) | 1 (2.0) | |
Extreme problems | 4 (0.6) | 0 (0) | 0 (0) | 4 (2.4) | 0 (0) | |
Usual activities | | | | | | <.0001†
|
No problems | 303 (45.2) | 61 (59.8) | 175 (49.6) | 56 (33.9) | 11 (22.0) | |
Slight problems | 183 (27.3) | 22 (21.6) | 96 (27.2) | 49 (29.7) | 16 (32.0) | |
Moderate problems | 135 (20.2) | 13 (12.8) | 64 (18.1) | 44 (26.7) | 14 (28.0) | |
Severe problems | 32 (4.8) | 4 (3.9) | 12 (3.4) | 10 (6.1) | 6 (12.0) | |
Extreme problems | 17 (2.5) | 2 (2.0) | 6 (1.7) | 6 (3.6) | 3 (6.0) | |
Pain/discomfort | | | | | | 0.26†
|
No problems | 255 (38.1) | 35 (34.3) | 131 (37.1) | 62 (37.6) | 27 (54.0) | |
Slight problems | 207 (30.9) | 33 (32.4) | 118 (33.4) | 48 (29.1) | 8 (16.0) | |
Moderate problems | 160 (23.9) | 27 (26.5) | 81 (23.0) | 42 (25.5) | 10 (20.0) | |
Severe problems | 43 (6.4) | 7 (6.9) | 21 (6.0) | 11 (6.7) | 4 (8.0) | |
Extreme problems | 5 (0.8) | 0 (0) | 2 (0.6) | 2 (1.2) | 1 (2.0) | |
Anxiety/depression | | | | | | 0.15†
|
No problems | 427 (63.7) | 63 (61.8) | 237 (67.1) | 101 (61.2) | 26 (52.0) | |
Slight problems | 139 (20.8) | 20 (19.6) | 64 (18.1) | 41 (24.9) | 14 (28.0) | |
Moderate problems | 79 (11.8) | 16 (15.7) | 40 (11.3) | 15 (9.1) | 8 (16.0) | |
Severe problems | 20 (3.0) | 2 (2.0) | 11 (3.1) | 5 (3.0) | 2 (4.0) | |
Extreme problems | 5 (0.8) | 1 (1.0) | 1 (0.3) | 3 (1.8) | 0 (0) | |
Table 3
Patient clinical and HRQL measurements
FACIT-Dyspnea | 6 | 44.6 (8.4) | 41.0 (7.9) | 43.3 (8.0) | 47.6 (8.0) | 51.6 (6.9) | 31.19 | Ref | <.0001 | <.0001 |
mMRC dyspnea | 60 | 1.50 (0.99) | 1.20 (0.95) | 1.38 (0.95) | 1.79 (0.95) | 1.88 (1.06) | 11.71 | 0.38 | <.0001 | <.0001 |
Borg dyspnea (at rest) | 55 | 0.67 (1.08) | 0.55 (0.91) | 0.60 (1.07) | 0.92 (1.19) | 0.64 (0.97) | 3.52 | 0.11 | 0.01 | 0.02 |
Borg dyspnea (during 6MWT) | 60 | 2.52 (1.94) | 2.07 (1.73) | 2.27 (1.90) | 3.02 (1.78) | 3.78 (2.37) | 13.06 | 0.42 | <.0001 | <.0001 |
6MWD | 60 | 335.6 (110.4) | 371.9 (111.2) | 345.5 (109.4) | 301.2 (103.1) | 297.8 (101.3) | 11.24 | 0.36 | <.0001 | <.0001 |
EQ-5D index | 0 | 0.79 (0.15) | 0.81 (0.14) | 0.81 (0.14) | 0.76 (0.17) | 0.74 (0.15) | 6.13 | 0.20 | 0.0004 | 0.002 |
EQ-VAS | 0 | 70.6 (19.6) | 76.6 (17.5) | 72.6 (19.1) | 65.7 (20.2) | 61.1 (19.7) | 12.24 | 0.39 | <.0001 | <.0001 |
PROMIS | | | | | | | | | | |
Physical function (P-PF) | 0 | 40.6 (7.6) | 43.2 (8.2) | 41.7 (7.7) | 38.3 (6.0) | 35.3 (5.5) | 21.41 | 0.69 | <.0001 | <.0001 |
Anxiety (P-A) | 0 | 49.7 (9.2) | 49.5 (9.6) | 49.3 (9.0) | 50.0 (9.4) | 52.3 (8.7) | 1.69 | 0.05 | 0.17 | 0.15 |
Depression (P-D) | 0 | 48.3 (9.3) | 49.3 (10.1) | 47.3 (8.8) | 48.5 (9.6) | 51.8 (8.9) | 4.08 | 0.13 | 0.007 | 0.01 |
Fatigue (P-F) | 0 | 50.8 (9.2) | 51.2 (9.9) | 50.0 (9.2) | 51.3 (8.8) | 53.9 (8.4) | 3.01 | 0.10 | 0.03 | 0.04 |
Sleep disturbance (P-SD) | 0 | 49.8 (9.4) | 51.5 (10.2) | 49.3 (9.0) | 50.0 (9.6) | 48.9 (9.4) | 1.64 | 0.05 | 0.18 | 0.23 |
Social roles (P-SR) | 2 | 48.1 (9.3) | 49.3 (9.0) | 49.0 (9.4) | 46.3 (9.3) | 44.7 (7.9) | 6.07 | 0.19 | 0.0004 | 0.001 |
Pain interference (P-PI) | 3 | 53.9 (9.6) | 53.8 (9.5) | 54.3 (9.5) | 53.3 (9.7) | 53.1 (9.8) | 0.56 | 0.02 | 0.64 | 0.74 |
Pain intensity scale (P-P) | 0 | 3.49 (2.67) | 3.49 (2.69) | 3.54 (2.62) | 3.43 (2.70) | 3.28 (2.90) | 0.18 | 0.01 | 0.91 | 0.85 |
Regarding PROMIS-43, physical function had a overall mean score of 40.6 (SD 7.6), and for the rest of domains, the mean scores ranged from 48.1 (social roles) to 53.9 (pain interference) (Table
3). The mean pain intensity score was 3.49 (SD 2.67) on a scale of 0 to 10. The PROMIS physical function, depression, fatigue, and social roles had p-values <0.05, but only physical function and social roles demonstrated a statistically significant decline that was monotonically associated with decreasing lung function (Table
3). No differences in mean scores by GOLD grade were observed for the PROMIS domains of anxiety, sleep disturbance, pain interference and pain intensity.
About sixty patients refused or did not complete the 6MWT and/or dyspnea severity assessment due to health issues (e.g., wheelchair or walker dependent, body pain, discomfort while doing the test). All the dyspnea measures and 6MWD demonstrated discriminative ability when mean scores were compared among subgroups with different levels of COPD severity (Table
3). The FACIT-Dyspnea provided the highest relative efficiency (RE) to discriminate among subgroups of COPD severity, followed by PROMIS physical function, Borg dyspnea during 6MWT, EQ-VAS, mMRC dyspnea, 6MWD, and EQ-5D index .
Cronbach’s alpha for PROMIS domains ranged from 0.89 to 0.95, demonstrating acceptable internal consistency reliability [
30]. Strong correlations between related domains on the EQ-5D-5L and PROMIS-43 were observed as expected (Table
4). EQ-5D usual activities (EQ-UA) showed strong correlations with the PROMIS physical function (P-PF) (
r
s
= -0.65), fatigue (P-F) (
r
s
= 0.54), and social roles (P-SR) (
r
s
= -0.55), and moderate correlations with the rest of PROMIS domains. EQ-5D pain/discomfort (EQ-PD) was strongly correlated with PROMIS pain interference (P-PI) and pain intensity (P-P) (
r
s
= 0.67 and 0.63, respectively), and the EQ-5D anxiety/depression (EQ-AD) with PROMIS anxiety (P-A) and depression (P-D) (
r
s
= 0.60 and 0.59, respectively). EQ-5D mobility (EQ-MO) was moderately to strongly related to four domains of the PROMIS (physical function [P-PF], fatigue [P-F], social roles [P-SR], and pain interference [P-PI]) and the pain intensity item (P-P). The EQ-5D dimension of self-care (EQ-SC) was moderately correlated with P-PF and P-SR.
Table 4
Correlations between domains of EQ-5D-5L and PROMIS-43 (all with
p
<0.001)
EQ-MO | 1 | | | | | | | | | | | |
EQ-SC | 0.39 | 1 | | | | | | | | | | |
EQ-UA | 0.52 | 0.45 | 1 | | | | | | | | | |
EQ-PD | 0.47 | 0.23 | 0.38 | 1 | | | | | | | | |
EQ-AD | 0.26 | 0.22 | 0.38 | 0.34 | 1 | | | | | | | |
P-PF | -0.65 | -0.46 | -0.65 | -0.40 | -0.34 | 1 | | | | | | |
P-A | 0.23 | 0.21 | 0.34 | 0.29 | 0.60 | -0.40 | 1 | | | | | |
P-D | 0.28 | 0.26 | 0.37 | 0.30 | 0.59 | -0.44 | 0.78 | 1 | | | | |
P-F | 0.41 | 0.27 | 0.54 | 0.40 | 0.41 | -0.62 | 0.53 | 0.59 | 1 | | | |
P-SD | 0.20 | 0.15 | 0.34 | 0.28 | 0.32 | -0.34 | 0.41 | 0.45 | 0.55 | 1 | | |
P-SR | -0.44 | -0.34 | -0.55 | -0.32 | -0.37 | 0.65 | -0.45 | -0.49 | 0.61 | -0.42 | 1 | |
P-PI | 0.46 | 0.23 | 0.44 | 0.67 | 0.35 | -0.52 | 0.42 | 0.43 | 0.57 | 0.41 | -0.47 | 1 |
P-P | 0.40 | 0.19 | 0.34 | 0.63 | 0.32 | -0.43 | 0.39 | 0.38 | 0.45 | 0.41 | -0.40 | 0.79 |
In examining the relationship between clinical measures and the EQ-5D index, EQ-VAS, PROMIS subscales, only the FACIT-Dyspnea and the PROMIS physical function (P-PF) showed at least moderate correlations with % of predicted FEV
1 (
r
s
= -0.36 and 0.32, respectively) (Table
5). EQ-5D index scores and VAS scores were both moderately to strongly correlated with PROMIS domains, dyspnea scales, and the 6MWD, but the magnitude of these correlations was smaller with EQ-VAS scores than EQ-5D index scores. All subscales of PROMIS-43 had moderate to strong correlations with at least one of the dyspnea scores. Among all the HRQL measures, the PROMIS physical function (P-PF), fatigue (P-F), social roles (P-SR), EQ-5D index and EQ-VAS, in general, had stronger correlations with the symptom severity. All measures were at least (or nearly) moderately correlated with 6MWD, except for PROMIS anxiety (P-A), depression (P-D) and sleep disturbance (P-SD) (absolute
r <0.3).
Table 5
Correlations between clinical and HRQL measures
% pred. FEV1
| 1 | | | | | | | |
FACIT-dyspnea | -0.36 | 1 | | | | | | |
mMRC dyspnea | -0.29 | 0.59 | 1 | | | | | |
Borg dyspnea (at rest) | -0.13 | 0.46 | 0.33 | 1 | | | | |
Borg dyspnea (during 6MWT) | -0.26 | 0.56 | 0.48 | 0.36 | 1 | | | |
6MWD | 0.28 | -0.44 | -0.45 | -0.25 | -0.34 | 1 | | |
EQ-5D index | 0.19 | -0.58 | -0.48 | -0.38 | -0.37 | 0.46 | 1 | |
EQ-VAS | 0.26 | -0.43 | -0.41 | -0.34 | -0.35 | 0.29 | 0.55 | 1 |
P-PF | 0.32 | -0.71 | -0.62 | -0.39 | -0.58 | 0.56 | 0.68 | 0.51 |
P-A | -0.09 | 0.42 | 0.23 | 0.27 | 0.23 | -0.19 | -0.46 | -0.33 |
P-D | -0.08 | 0.45 | 0.25 | 0.26 | 0.29 | -0.19 | -0.49 | -0.34 |
P-F | -0.09 | 0.58 | 0.43 | 0.38 | 0.42 | -0.29 | -0.55 | -0.48 |
P-SD | 0.02†
| 0.35 | 0.28 | 0.26 | 0.23 | -0.10 | -0.37 | -0.31 |
P-SR | 0.18 | -0.57 | -0.41 | -0.37 | -0.42 | 0.37 | 0.54 | 0.47 |
P-PI | -0.01†
| 0.41 | 0.33 | 0.31 | 0.22 | -0.29 | -0.65 | -0.37 |
P-P | 0.01†
| 0.35 | 0.26 | 0.27 | 0.21 | -0.29 | -0.58 | -0.33 |
Discussion
Our study results provide evidence to support the validity of two recently developed generic measures of HRQL, EQ-5D-5L and PROMIS-43, in COPD. The convergent construct validity of the two measures was supported by the moderate to strong correlations between related domains, and between the domain and summary scores of the generic measures and the dyspnea measures. EQ-5D-5L index, EQ-VAS, and two domains of PROMIS (physical function and social roles) had higher RE ratios among the HRQL measures, suggesting that these scores provide greater statistical power (discriminative ability) to capture differences in HRQL in relation to disease severity as measured by lung function.
Level of dyspnea is a strong predictor for health status [
31‐
33]. Both EQ-5D and PROMIS had moderate associations with at least one measure of dyspnea, with the correlations varying across the PROMIS-43 subscales. Our results concur with previous reports that spirometric parameters (% of predicted FEV
1), unlike severity of breathlessness, does not correlate well with HRQL [
31,
32,
34,
35]. While lung function test with spirometry serves as an important clinical tool to measure the degree of airflow limitation, a number of studies have demonstrated that it provides an incomplete assessment of health burden to the patient, which can include physical and psychosocial functioning. This discernment coincides with the new COPD assessment tool recently proposed by the GOLD, which recommends evaluation of COPD based on not only lung function, but also the assessment of symptoms and exacerbation risk [
3]. This also reinforces the importance of evaluating patient-reported outcomes along with clinical measures (e.g., lung function test) when gauging the effect of health interventions.
COPD severity has been shown to influence the degree of physical disability, impairing the ability to perform activities of daily living, and contributing to poor HRQL [
36]. Patients with COPD had relatively worse self-rated HRQL in multiple PROMIS domains as compared with individuals without COPD or any condition [
37]. The negative impact of COPD is more pronounced on the physical aspect of health than on the mental component [
31]. Consistent with the study by Gonzalez-Moro and colleagues [
36], our findings suggest that, in general, physical functioning tends to be affected in all grades of COPD patients while mental health is impaired only in patients at more severe stages. The mean scores of PROMIS domains indicated that physical function, among all the domains measured in the PROMIS, was the aspect of health status most affected by COPD; physical function was considerably impaired even in patients with mild COPD, as the mean domain score of PROMIS physical function in patients with GOLD grade 1 was less than 50 (the mean score of the general population). The mean domain scores of PROMIS anxiety and depression were higher than 50 only in patients with very severe COPD (i.e. GOLD 4).
Evidence on the properties of the EQ-5D-5L is only beginning to emerge. The first paper was a multi-country study by Janssen et al. in 2012 that compared the measurement properties of the 5-level and 3-level EQ-5D, including 342 patients with respiratory disease (COPD or asthma) as one of the eight patient groups with chronic conditions [
38]. The 5-level EQ-5D descriptive system (EQ-5D-5L) reduced ceiling effects of the 3-level EQ-5D (EQ-5D-3L) and improved the discriminatory power and convergent validity. In our study, broad use of 4 of the 5 Levels of the EQ-5D-5L suggests that it could provide higher discriminative power than the standard EQ-5D-3L in COPD, although the most severe category appears to be rarely utilized. A previous study showed that EQ-5D-3L index score (both UK and US) failed to differ across COPD severity stages [
35]. The mean EQ-5D-5L index score significantly decreases as COPD severity deteriorates, particularly in the advanced stages of disease (Table
3), which may suggest better discriminatory power of EQ-5D-5L than EQ-5D-3L to distinguish COPD patients of different severity. Similar to studies of the EQ-5D-3L in COPD, self-care is the dimension least affected by COPD [
39,
40]. In accordance with a study by Punekar et al. [
40], about 80% of COPD patients reported no problems in self-care. As the severity of COPD increased, COPD patients reported more problems with mobility, self-care, and usual activities. However, pain/discomfort and anxiety/depression tended not to differ by disease severity using the EQ-5D-5L or the PROMIS. Our study also suggested that EQ-5D-5L index scores were less able to discriminate among patients with milder disease, i.e. GOLD grades 1 and 2. This is consistent with a previous study by Antonelli-Incalzi et al. who observed that health status dramatically declined when predicted FEV
1 was 49% or less (upper limit of GOLD grade 3) [
41]. Alternatively, the lack of discrimination between grade 1 and 2 may also suggest that the EQ-5D-5L descriptive system does not entirely address some of the limitations of the three-level EQ-5D [
39], assuming there is a meaningful difference in self-reported health based on GOLD grades 1 and 2. Unlike the EQ-5D index score which is derived based on the five dimensions using population preference weights, the EQ-VAS provides a direct rating of health from the patient’s point of view. Consistent with previous reports [
35], EQ-VAS had a more monotonic relationship with disease severity and better ability to discriminate according to disease severity compared to EQ-5D index.
Among the PROMIS subscales, physical functioning was most strongly associated with disease grades and measures of breathing difficulty and functioning. Only physical function (P-PF), depression (P-D), fatigue (P-F), and social roles (P-SR) varied significantly across COPD grades and the magnitude of differences in the PROMIS scores of depression and fatigue across different GOLD grades were smaller than half of their SD, a commonly used cutoff for interpreting important differences in HRQL scores [
42]. Anxiety, sleep, and pain domains of PROMIS, although moderately related to other HRQL measurements and dyspnea scores (mainly FACIT-Dyspnea), did not vary by COPD GOLD grade. The lack of correlation between pain, anxiety, and sleep disturbance and the degree of COPD severity does not preclude the importance of these HRQL parameters in COPD patients. In fact, it has been reported that 35%, 37% and 51% of advanced COPD patients suffer from sleep disturbance, pain and anxiety, respectively, arguably among the most prevalent symptoms associated with advanced COPD [
43]. Despite the inability of these domains to discriminate patients with different level of airflow limitation, the domains present convergent validity and it suggests that they may capture patient-reported outcomes other than those associated with spirometry. In addition, the observation that the parameters of physical or physiological measures (dyspnea scores; mobility, self-care and usual activities in EQ-5D-5L; physical function in PROMIS) deteriorate more with the increase in COPD severity, as compared to psychosocial measures (anxiety/depression in EQ-5D-5L; anxiety, depression and social roles in PROMIS), suggests the possibility of adapation and coping mechanisms developed in COPD patients as the disease severity progresses, which is often observed with chronic illnesses and disabling conditions [
44].
EQ-5D and PROMIS, both generic measures of HRQL, are distinctive in several ways. While EQ-5D index and VAS scores both provide summary scores for evaluating general health status as a whole, PROMIS describes different aspect of health status individually using domain scores. The domains of anxiety, depression, and pain are apparently covered by both of the measures, but it is arguable if fatigue (PROMIS) overlaps with pain/discomfort (EQ-5D), as well as the extent of overlap between physical functioning, fatigue, sleep disturbance, or social roles (PROMIS) and mobility, self-care, or usual activities (EQ-5D). EQ-5D index-based scores are generated from societal preferences for health that can be applied to economic evaluations. Although PROMIS-43 does not include global items and was not designed as a preference-based measure as EQ-5D, at least one scoring function is available to convert PROMIS selected domain scores into a single index value by mapping onto the EQ-5D [
45]. Comparing to PROMIS, EQ-5D is presumably briefer to administer as it contains 6 items (including VAS) rather than 43, but the PROMIS-43 contains more items in each domain, thereby providing the potential of a higher level of precision and sensitity than EQ-5D. Alternatively, even briefer short-form versions of the PROMIS are available.
This study has several limitations. Since patients did not complete EQ-5D-3L, we could not directly determine whether the EQ-5D-5L improves upon the properties of the EQ-5D-3L in COPD. In addition, longitudinal data are needed to examine and compare the responsiveness of the measures to detect meaningful change following interventions. Lastly, in our study, patients with more severe COPD (GOLD 3 and 4) were younger than those with milder disease, which was contrary to our expectation but may be due to the eligibility of study participation or possibly a survivor effect. The representativeness of patients included in this analysis could also be restricted by the relatively low response rate (36%) for participating in the in-person evaluations. Age is a known factor that could confound the association between HRQL and disease severity [
7]. In order to rule out the confounding effect, we also conducted an analysis of covariance (ANCOVA) to control for age when comparing the responses in EQ-5D, PROMIS domain scores, dyspnea measures, and 6MWD among patients with different GOLD grades (data not shown). Similar results (F-statistic and significance level) were found as in Table
3 after controlling for age effect, except that the discriminative ability of 6MWD and PROMIS sleep disturbance (P-SD) to distinguish COPD patients of different severity was improved.
Acknowledgements
The COPD Outcomes-based Network for Clinical Effectiveness and Research Translation (CONCERT) was funded by the National Heart, Lung, and Blood Institute (NHLBI RC2 HL101618). Fang-Ju Lin was supported by the Graduate College of the University of Illinois at Chicago (2012/2013 Dean’s Scholar Award). The sponsors had no role in the design of the study, collection and analysis of the data, interpretation of results or the writing of this manuscript. Dr. David Au received funding from the Veteran Affairs Health Services Research & Development (VA HSR&D). The opinions expressed represent those of the authors and do not necessarily reflect those of the Department of Veterans Affairs.
Notation of prior abstract presentation
This study was presented at the American Thoracic Society International Conference, May 17-22, 2013, Philadelphia, Pennsylvania, USA. An earlier draft of this manuscript was presented as a discussion paper at the 30th EuroQol Plenary Meeting in Montreal, Canada, in September 2013.
Consent
Written informed consent was obtained from the patient for the publication of data in this report and any accompanying images.
Competing interests
Dr. Simon Pickard is a member of the EuroQol group. No authors had financial conflicts of interest to disclose regarding the contents of this manuscript.
Authors’ contributions
FJL participated in the study design, data analysis, data interpretation, and manuscript writing. ASP contributed to the supervision and design of the study, data interpretation, and revising the manuscript. TAL contributed in the study design, data collection, data interpretation, and manuscript revision. JAK, MJJ, DHA, SSC, SG, AGH, PKL, MAM, RAM, ETN, and WMV participated in the data collection and manuscript revision. All authors have read and approved the final manuscript.