Systemic lupus erythematosus in a multiethnic lupus cohort (LUMINA). XVII. Predictors of self-reported health-related quality of life early in the disease course

Authors


Abstract

Objective

To determine the baseline factors predictive of self-reported health-related quality of life (HRQOL) early in the course of systemic lupus erythematosus patients (SLE) from a multiethnic LUMINA (Lupus in Minorities: Nature versus nurture) cohort.

Methods

LUMINA patients with ≥2 visits were studied. Self-reported HRQOL was examined with the 8 subscales and 2 summary measures (the Physical Component Summary [PCS], and the Mental Component Summary [MCS]) of the Short Form 36 (SF-36). Bivariable and multivariable analyses were done with the PCS, MCS and 8 subscales as the dependent variables. The analyses were performed including and excluding the corresponding SF-36 measure from the independent variables. Age, sex, and ethnicity were included in all models. Time was modeled in all regressions.

Results

A total of 1,351 visits (346 patients [80 Hispanics-Texas, 34 Hispanics-Puerto Rico, 126 African Americans, and 106 Caucasians]) were included in these analyses. Mean ± SD PCS and MCS scores were 36.7 ± 12.0 and 46.6 ± 11.5, respectively. The scores for the eight subscales of the SF-36 were also lower than those for the general population. Baseline SF-36 measures were highly predictive of subsequent HRQOL. In the same set of regressions, older age was found to consistently predict poor self-reported HRQOL whereas fibromyalgia, helplessness, fatigue, and abnormal illness-related behaviors were also predictive, but less consistently. Estimated adjusted variances in these regressions ranged from 23% (Role-Emotional [RE]) to 43% (Physical Functioning [PF]).

Conclusion

In patients with SLE, poor baseline HRQOL was highly predictive of subsequent poor HRQOL. Other predictive variables of poor functioning were primarily psychological/behavioral and socioeconomic-demographic.

INTRODUCTION

Although patients with systemic lupus erythematosus (SLE) have a decreased life expectancy as compared with the general population, most live with their disease for many years (1–3). Studies of SLE outcome cannot be limited to those of survival/mortality, disease activity, and damage. International convened groups dealing primarily with lupus (Systemic Lupus International Collaborating Clinics [SLICC]) or with outcome (Outcome Measures for Arthritis Clinical Trials), have recommended that a measure of quality of life/self-reported functioning be included in SLE outcome studies (4—7). We included, therefore, a self-reported measure of function, the Short Form 36 (SF-36) from the Medical Outcomes Study (8, 9) in the assessment of patients from the multiethnic US Lupus in Minorities: Nature versus nurture (LUMINA) (10–12) cohort. The SF-36 self-reported functioning, a health-related quality of life (HRQOL) instrument, allows for comparison between patients with different chronic disorders and the general population (13–20). Correlates of better self-reported physical health functioning in cross-sectional SLE studies include younger age, better self-efficacy for disease management, adequate social support, and lower levels of disease activity and damage (21–26). Correlates of better self-reported mental health functioning in these SLE studies include adequate social support, higher family income, more knowledge about lupus, and paradoxically, higher levels of disease activity and damage. Ethnicity has not been found to consistently and directly impact on functioning; however, it has been thought to perhaps exert a modulatory effect via disease activity and damage and other socioeconomic variables associated with ethnicity (22).

We have previously reported the factors associated with poor functioning as patients entered the LUMINA cohort (12). Self-reported functioning was primarily associated with psychological and behavioral variables (e.g., coping with illness, helplessness) and to a lesser extent with disease-related variables (12). We are now presenting data for the baseline features predictive of poor HRQOL over the course of followup in our patients. We were particularly interested in learning to what extent poor self-reported functioning early in the course of the disease could impact on subsequent functioning and in determining what other early patient characteristics may be important in predicting HRQOL in lupus.

PATIENTS AND METHODS

The LUMINA cohort has been previously described (10, 11, 27). LUMINA is a longitudinal study of outcome of patients with SLE from 3 different ethnic groups (Hispanic, African American, and Caucasian). This ongoing study is being conducted in 3 US areas (Alabama, Texas, and Puerto Rico) at 4 institutions (The University of Alabama at Birmingham, The University of Texas Health Science Center-Houston, The University of Texas Medical Branch at Galveston, and The University of Puerto Rico Medical Sciences Campus). Patients meeting the American College of Rheumatology (ACR) revised criteria for SLE (28), with disease duration ≤5 years, who live in the catchment areas of the 4 institutions are eligible for inclusion in LUMINA. Patients are seen at recruitment or baseline (T0), at 6 and 12 months (T0.5 and T1, respectively), and yearly thereafter (T2, T3, etc. to TL [last visit available]). All visits include medical records review, interviews and questionnaires, physical examination, and phlebotomy. This study is approved by the Institutional Review Boards for the Protection of Human Subjects of the 4 institutions.

Variables.

Our database includes variables from the following domains: socioeconomic-demographic, clinical and immunologic, immunogenetic, and psychological, behavioral, and cultural. Only the following variables listed in this article were examined. Specific variables that were examined are listed as follows: from the socioeconomic-demographic domain, age, sex, ethnicity, education, occupation, marital status, health insurance, and income; from the clinical and immunological domain, disease onset type (considered acute, if ACR criteria (28) are accrued within 4 weeks, otherwise, insidious), disease duration (the date that a patient met ACR criteria for SLE or TD to T0), fibromyalgia (defined by ACR criteria) (29), disease activity (per the Systemic Lupus Activity Measure [SLAM]) (30, 31), and disease damage (per the SLICC damage index [SDI]) (32); from the psychological, behavioral, and cultural domain, social support (per the Interpersonal Support Evaluation List [ISEL]) (33), helplessness (per the Rheumatology Attitude Illness) (34), and coping with illness or illness-related abnormal behaviors (per the Illness Behavior Questionnaire [IBQ])(35).

Generic HRQOL was measured with the SF-36, version 1 (8, 36). The SF-36, a valid and reliable instrument, consists of 8 subscales (Physical Functioning, Role-Physical, Bodily Pain, General Health, Vitality, Social Functioning, Role-Emotional, and Mental Health) and 2 summary measures (the Physical Component Summary [PCS] and the Mental Component Summary [MCS]) (8, 37). The PCS and MCS are calculated from the 8 subscales. The PCS includes Physical Functioning, Role-Physical, Bodily Pain, and General Health subscales; the MCS includes the Vitality, Social Functioning, Role-Emotional, and Mental Health subscales. Higher scores indicate better HRQOL. In the general US population, the mean ± SD score for both the PCS and MCS is 50 ± 10, whereas the mean ± SD scores of the 8 subscales are Physical Functioning 83.3 ± 23.8, Role-Physical 82.5 ± 25.5, Bodily Pain 71.3 ± 23.7, General Health 70.9 ± 21.0, Vitality 58.3 ± 20.0, Social Functioning 84.3 ± 22.9, Role-Emotional 87.4 ± 21.4, and Mental Health 75.0 ± 17.8 (38).

Statistical analyses.

Patients with only one visit (n = 106) or with 2 or more visits, but who did not have SF-36 data for at least 2 of these visits, were excluded from the analyses (n = 361 visits). Excluded patients had overall similar baseline characteristics as patients included in these analyses. Bivariable analyses for the 8 subscales scores, PCS, and MCS at TL were examined in relation to T0 socioeconomic-demographic, clinical, and psychological, behavioral and cultural variables as well as T0 scores for the 8 subscales, the PCS, and the MCS. Means and SD were obtained for discrete variables, and correlation coefficients for continuous variables.

For the multivariable analyses, generalized estimating equations (GEE) were used to account for the dependence of the SF-36 scores for each of eight subscales, the PCS, and MCS at each study visit for each patient. Independent variables included in the GEE analyses were those found to be significant at P ≤ 0.05 in the bivariable analyses. Ethnicity, age, and sex were included in all models. Time was modeled in all regressions. Although some of the variables in the models are correlated, the correlations are not strong enough to suggest collinearity. Nevertheless, when constructing the models, attention was paid to instances of failed convergence and variance inflation that would tend to indicate collinearity. In all cases, the GEE models were examined including and excluding T0 self-reported functioning for the corresponding subscale or summary measure being examined. All analyses were performed using SAS software, Version 10 (SAS Institute, Cary, NC).

RESULTS

Three hundred forty-six patients (1,351 visits, or a mean of 4.0 visits per patient) were studied; there were 80 Hispanics from Texas (Hispanics-TX), 34 Hispanics from Puerto Rico (Hispanics-PR), 126 African Americans, and 106 Caucasians. The main socioeconomic-demographic and clinical features of the patients studied correspond to the overall features of LUMINA patients as described in detail previously (11, 12). However, in contrast to previous publications, we are including data for 2 distinct Hispanic sub-groups, one from Texas (primarily of Mexican and Central American descent) and the other from Puerto Rico (39), and thus, they will be examined independently. As in previous publications, the majority of the patients are women (∼91%), middle-aged (mean ± SD age 37.7 ± 13.0 years), with 5.5 ± 1.3 meeting ACR criteria for SLE (28) and mean ± disease duration of 18 ± 16.5 months at T0. The mean and median followup times (TL-T0) for patients included in these analyses were 50.0 and 44.2 months, respectively. Hispanic-TX and African American patients had lower socioeconomic status than the Caucasian and Hispanic-PR patients (data not shown).

Figure 1 depicts the scores for the PCS, the MCS, and the eight subscales. For comparison, population means are also depicted. Mean ± SD scores were 36.7 ± 12.0 for the PCS and 46.6 ± 11.5 for MCS. For the eight subscales the scores ranged from a low of mean ± SD 44.0 ± 24.7 for vitality to a high of 67.5 ± 21.5 for Mental Health.

Figure 1.

Means and standard deviations for the summary measures and subscales of the Short Form 36. The triangles above each bar correspond to the population means (9, 42). PCS = Physical Component Summary; PF = Physical Functioning; RP = Role-Physical; BP = Bodily Pain; GH = General Health; MCS = Mental Component Summary; VT = Vitality; SF = Social Functioning; RE = Role Emotional; MH = Mental Health.

Bivariable analyses.

Poverty, greater disease activity (SLAM), fibromyalgia, fatigue, helplessness, abnormal illness-related behaviors (IBQ total score), and less social support (ISEL total score), were highly associated with lower scores for the 8 subscales and the PCS and MCS (P < 0.0001, for all). In contrast, there were no differences in all the scores as a function of sex, and acute versus insidious onset type. Other variables such as ethnicity, age, education, home ownership, health insurance, marital status, and damage (SDI), were not as consistently associated with the 8 subscales and the PCS and MCS as the aforementioned variables. These data are presented in Tables 1 and 2 for the socioeconomic-demographic variables and Tables 3 and 4 for the clinical and psychological, behavioral, and cultural variables.

Table 1. Mean values for the Physical Component Summary (PCS) and related Short Form 36 subscales and selected baseline socioeconomic-demographic and clinical variables*
VariablePCSPhysical functioningRole-physicalBodily painGeneral health
  • *

    NS = not significant.

  • P ≤ 0.05 as compared with the reference group (Hispanic-Puerto Rico).

  • Comparisons not significant, P values not given.

  • §

    As defined by the American College of Rheumatology criteria (29).

Ethnicity     
 Hispanic-Texas39.459.556.258.945.2
 Hispanic-Puerto Rico39.861.657.454.954.9
 African American34.851.038.048.641.1
 Caucasian33.952.437.549.342.5
Sex     
 Men34.152.938.650.039.1
 Women36.354.544.451.944.1
Poverty, below line     
 No37.359.148.654.546.7
 Yes33.544.634.846.538.0
 P0.0012<0.00010.00100.00350.0001
Home ownership     
 No36.053.345.052.045.2
 Yes36.254.241.951.640.9
Insurance     
 Yes36.255.143.752.744.3
 No35.851.844.649.541.2
Marital status, married/together     
 No37.457.447.255.244.7
 Yes34.851.840.948.942.8
 P0.0201NSNS0.0152NS
Acute disease onset     
 No35.854.443.051.643.5
 Yes36.454.444.952.243.7
Fibromyalgia§     
 No36.455.144.852.644.0
 Yes24.530.58.320.130.4
 P<0.0001<0.0001<0.0001<0.00010.0223
Table 2. Mean Values for the Mental Component Summary (MCS) and related Short Form 36 subscales and selected baseline socioeconomic-demographic and clinical variables*
VariableMCSVitalitySocial functioningRole-emotionalMental health
  • *

    NS = not significant.

  • P ≤ 0.05 as compared with the reference group (Hispanic-Puerto Rico).

  • Comparisons not significant, P values not given.

  • §

    As defined by the American College of Rheumatology criteria (29).

Ethnicity     
 Hispanic-Texas45.850.867.960.265.3
 Hispanic-Puerto Rico49.555.073.271.770.9
 African American43.842.358.347.965.0
 Caucasian46.429.863.063.268.3
Sex     
 Men46.642.164.355.370.5
 Women45.643.263.157.766.1
Poverty, below line     
 No47.844.267.566.170.5
 Yes42.240.755.542.559.1
 P<0.0001NS<0.0001<0.0001<0.0001
Home ownership     
 No46.943.565.460.768.3
 Yes43.442.259.652.063.2
 P0.0015NS0.03240.03570.0165
Insurance     
 Yes46.242.463.559.268.0
 No44.744.762.453.862.0
 PNSNSNSNS0.0099
Marital status, married/together     
 No45.745.864.757.367.1
 Yes45.440.662.057.665.9
 PNS0.0227NSNSNS
Acute disease onset     
 No45.842.163.660.366.8
 Yes45.444.462.758.666.0
Fibromyalgia§     
 No45.943.863.958.266.9
 Yes36.715.636.731.149.7
 P0.0028<0.00010.00010.03580.0024
Table 3. Association between the Physical Component Summary (PCS) and related Short Form 36 subscales and selected baseline socioeconomic-demographic, clinical, and psychological/behavioral variables*
VariablePCSPhysical functioningRole-physicalBodily painGeneral health
  • *

    SLAM = Systemic Lupus Activity Measure; SDI = Systemic Lupus International Collaborating Clinics (SLICC) Damage Index; IBQ = Illness Behavior Questionnaire; ISEL = Interpersonal Support Evaluation List.

  • Same Short Form 36 variable at a different time point.

Age     
 r−0.26−0.29−0.18−0.18−0.10
 P<0.0001<0.0001<0.00010.00010.0003
Education     
 r0.020.100.020.060.15
 P0.54020.00030.56100.0380<0.0001
Disease duration     
 r−0.07−0.03−0.03−0.040.06
 P0.020.20710.26740.11810.0321
SLAM     
 r−0.16−0.14−0.12−0.11−0.26
 P<0.0001<0.00010.0002<0.0001<0.0001
SDI     
 r−0.12−0.13−0.15−0.05−0.07
 P<0.0001<0.0001<0.00010.06500.0104
Fatigue     
 r−0.39−0.39−0.34−0.29−0.37
 P<0.0001<0.0001<0.0001<0.0001<0.0001
Helplessness     
 r−0.34−0.33−0.29−0.29−0.39
 P<0.0001<0.0001<0.0001<0.0001<0.0001
IBQ     
 r−0.11−0.18−0.19−0.15−0.31
 P<0.0001<0.0001<0.0001<0.0001<0.0001
ISEL total     
 r0.120.180.150.130.22
 P<0.0001<0.0001<0.0001<0.0001<0.0001
Short Form 36     
 r0.580.570.460.440.51
 P<0.0001<0.0001<0.0001<0.0001<0.0001
Table 4. Association between the Mental Component Summary (MCS) and related Short Form 36 subscales and selected baseline socioeconomicdemographic, clinical, and psychological/behavioral variables*
VariableMCSVitalitySocial functioningRole-emotionalMental health
  • *

    SLAM = Systemic Lupus Activity Measure; SDI = Systemic Lupus International Collaborating Clinics (SLICC) Damage Index; IBQ = Illness Behavior Questionnaire; ISEL = Interpersonal Support Evaluation List.

  • Same Short Form 36 variable at a different time point.

Age   −0.090.02
 r0.04−0.11−0.110.00150.4960
 P0.12630.0430<0.001  
Education     
 r0.20−0.020.100.180.23
 P<0.00010.45770.0002<0.0001<0.0001
Disease duration     
 r0.110.020.010.070.08
 P0.00010.55060.64160.00750.0046
SLAM*     
 r−0.15−0.11−0.16−0.12−0.16
 P<0.0001<0.0001<0.0001<0.0001<0.0001
SDI     
 r−0.04−0.07−0.08−0.08−0.08
 P0.20620.00640.00420.00270.0074
Fatigue     
 r−0.21−0.34−0.33−0.22−0.21
 P<0.0001<0.0001<0.0001<0.0001<0.0001
Helplessness     
 r−0.29−0.33−0.31−0.24−0.31
 P<0.0001<0.0001<0.0001<0.0001<0.0001
IBQ     
 r−0.43−0.21−0.30−0.29−0.46
 P<0.0001<0.0001<0.0001<0.0001<0.0001
ISEL total     
 r0.290.120.230.220.12
 P<0.0001<0.0001<0.0001<0.0001<0.0001
Short Form 36     
 r0.500.480.400.390.55
 P<0.0001<0.0001<0.0001<0.0001<0.0001

Multivariable analyses by GEE with self-reported HRQOL included.

Poorer self-reported HRQOL at T0 constituted the strongest predictor of later poor self-reported functioning for the LUMINA patients; as noted in Tables 5 and 6, that was the case for the scores of the 8 subscales and the PCS and MCS. With regards to the PCS and its related subscales, poor self-reported HRQOL was also associated with older age (PCS and all subscales). In contrast, African American ethnicity, disease activity, fibromyalgia, increasing fatigue, and helplessness were predictive of poor HQROL in some but not all subscales. Time was significant only in the General Health subscale regression. As noted in Table 5 the percentage of the estimated adjusted variance explained in these models ranged from 27% for Bodily Pain to 43% for Physical Functioning. With regards to the MCS and its related subscales, African American ethnicity, older age, increasing fatigue, helplessness, and abnormal illness-related behaviors were associated with poorer HRQOL in some, but not all subscales. Time was a significant variable only in the Mental Health subscale regression. As noted in Table 6, the percentage of the estimated adjusted variance explained in these models ranged from 23% for Role-Emotional to 36% for Mental Health.

Table 5. Generalized estimating equation regression models for the Physical Component Summary (PCS) and related Short Form 36 (SF-36) subscales. Baseline predictors with corresponding SF-36 variables included*
VariablePCSPhysical functioningRole physicalBodily painGeneral health
Direction of AssociationPDirection of AssociationPDirection of AssociationPDirection of AssociationPDirection of AssociationP
  • *

    Only variables with P ≤ 0.05 are noted. SLAM = Systemic Lupus Activity Measure; Estimated adjusted variance (R2%) for PCS = 42, Physical functioning = 43, Role physical = 29, Bodily pain = 27, General health = 36.

  • As defined by the American College of Rheumatology criteria (29).

Ethnicity          
 African American−8.680.0175
 Caucasian−7.030.0430
Age−0.13<0.0001−0.40<0.0001−0.52<0.0001−0.250.0056−0.210.0020
Poverty−8.900.0009
SLAM−0.370.0240
Fibromyalgia−4.260.0320−18.17<0.0001
Helplessness−0.660.0300−0.430.0401
Fatigue−3.080.0347−2.190.0232
SF-360.52<0.00010.46<0.00010.31<0.00010.35<0.00010.44<0.0001
Time        0.550.0305
Table 6. Generalized estimating equation regression models for the Mental Component Summary (MCS) and related Short Form 36 (SF-36) subscales. Baseline predictors with corresponding SF-36 variables included*
VariableMCSVitalitySocial functioningRole emotionalMental health
Direction of AssociationPDirection of AssociationPDirection of AssociationPDirection of AssociationPDirection of AssociationP
  • *

    Only variables with P ≤ 0.05 are noted. IBQ = Illness Behavior Questionnaire. Estimated adjusted variance (R2%) for MCS = 30, Vitality = 31, Social functioning = 27, Role emotional = 23, Mental health = 36. Fatigue not included in the vitality regression.

  • As defined by the American College of Rheumatology criteria (29).

Ethnicity          
 African American−3.710.046−10.390.0144−18.130.011
 Caucasian−10.900.0041
Age−0.150.0432−0.300.0011−0.460.0004−0.130.0283
Education0.300.02921.120.0402
Fibromyalgia−4.260.0338−12.96<0.0001−16.100.0018
Fatigue−1.900.0340
Helplessness−0.620.0003
IBQ−0.310.0004−0.660.00270.910.0055−0.530.0010
SF-360.30<0.00010.37<0.00010.210.00010.23<0.00010.38<0.0001
Time        0.900.0007

Self-reported HRQOL excluded from the models.

When T0 self-reported HRQOL was not included, most variables significant in the previous analyses remained as important T0 predictors of later self-reported HRQOL. For example, older age, fatigue, and helplessness were consistent predictors of poor self-reported functioning; that was the case for the PCS and all its related subscales. Fibromyalgia was a predictor of poor HRQOL; this applied to the PCS and 2 of its subscales. Poverty was also a predictor of lower levels of HRQOL, but this applied only to the PCS and one of its subscales. Disease activity, disease damage, and abnormal illness-related behaviors only predicted one of the subscales contributing to the PCS, but not the PCS per se. Time was significant only in the General Health subscale. The percentage of the estimated adjusted variance explained in these models ranged from 20% for Bodily Pain to 33% for Physical Functioning (data not shown).

In terms of the MCS and its 4 subscales, the data were less consistent across all variables with the exception of abnormal illness-related behaviors that were significant predictors of the MCS and its 4 subscales. African American ethnicity, age, fibromyalgia, and helplessness were predictive of poor self-reported mental health functioning (MCS and some of its 4 subscales) indicating that these constructs associate with lower levels of HRQOL, but not consistently. Education was predictive of better levels of functioning, whereas time was associated with the MCS and 2 of its subscales (Role-Emotional and Mental Health). The percentage of the estimated adjusted variance explained in these models ranged from 20% for Role-Emotional to 29% for Mental Health (data not shown).

DISCUSSION

Patients in the LUMINA cohort continue to exhibit relatively low levels of self-reported HRQOL over the course of their disease as assessed by the SF-36 (36). This applies to its 8 subscales, to the PCS, and the MCS. As noted in Figure 1, all 8 subscales, the PCS, and to a lesser extent the MCS, were several points below normative US data. This overall level of poor HRQOL corresponds to those levels described in other chronic disorders and in lupus patients from other cohorts (13–20, 40–45). Our data, taken in conjunction with those from the literature, emphasize the functional limitations imposed by SLE, despite the fact that patients may not have overt disease activity and may have not yet accrued significant organ system damage.

In our previous analyses we examined the variables associated with the 2 summary components of the SF-36 at study entry, and were impressed with the relatively minor impact of disease-related variables, particularly, disease activity and damage accrued on these outcomes. Our analyses are now more exhaustive due to several reasons including: factors predictive of the 8 subscales have also been examined; the effect of time has now been ascertained by modeling for it in all regressions; the contribution of early HRQOL in subsequent HRQOL has been considered; and a second Hispanic subgroup has been included, making the overall data more applicable to the US Hispanic population.

As with our initial cross-sectional analyses, disease activity and damage contributed only to one or more dependent variables each. Disease activity has been variably associated with poor self-reported levels of functioning. This may be in part due to the instrument used to measure disease activity, whether it is the SLAM (22, 25, 26, 46–48), the Systemic Lupus Disease Activity Index (21, 25, 49, 50), or the British Isles Lupus Activity Group (51). Dobkin et al, (24) have suggested that disease activity per se does not exert a direct effect on functioning, but rather it is modulated by sociodemographic (age, education) and behavioral (coping with illness) patient characteristics. In fact, our analyses are consistent with such modulatory effect, as these 3 variables were relevant in the bivariable, but also in some of the multivariable analyses. The only other disease-related variable that was found to be a predictor of more than one subscale and the PCS and MCS was fibromyalgia (as defined by the ACR criteria [(29]). In fact, patients with fibromyalgia alone exhibit levels of self-reported functioning as low or lower than patients with other chronic rheumatic disorders, SLE included (18, 52–54). It is not surprising, therefore, that even though relatively few (5%) LUMINA patients had fibromyalgia as defined by the ACR criteria (55), fibromyalgia emerged as an important predictor of poor self-reported HRQOL.

During the early years of the LUMINA cohort, we were not ascertaining fibromyalgia; for these analyses, patients in whom fibromyalgia had not been ascertained at T0 were considered not to have fibromyalgia. This approach would have minimized the association between fibromyalgia and poor functioning. Fatigue was a very consistent and significant predictor for some of the subscales that contribute to the PCS and to a lesser extent to those contributing to the MCS, but not for the PCS and MCS per se as previously shown by Bruce et al (56) and ourselves (12). We have not found fatigue to be significantly associated with fibromyalgia; thus, its effect on HRQOL appears to be independent of fibromyalgia (55). Moreover, because of the relationship between fatigue and vitality, fatigue was not included in the vitality regression.

Other important predictors of overall poor levels of self-reported HRQOL include some socioeconomic-demographic variables (such as older age, poverty, lower levels of education), and psychological, behavioral, and cultural variables (such as higher degrees of helplessness, abnormal illness-related behaviors, and inadequate social support). Ethnicity was not a consistent predictor of self-reported functioning. These data concur with our previously reported analyses, and emphasize the importance that overall poor socioeconomic status and personal characteristics have in the outcome of chronic disorders, SLE included. Karlson et al (47) found that inadequate social support was associated with poor functioning in a multicenter, cross-sectional ethnically and socioeconomically balanced multicenter lupus study. Investigators from the same group (22) have further examined the influence of social support on functioning and have suggested that some social support strategies may exert a positive effect, whereas others may exert a negative effect and that these relationships may vary by ethnic group. Coping with illness was an important and consistent predictor of functioning in our patients, as measured by the IBQ total score, a score reflective of inadequate behaviors in dealing with illness and health. These data are difficult to compare with data from other studies because different instruments were used to ascertain this construct; nevertheless, poor mental and physical health functioning have been associated with inadequate coping styles (18, 24, 44).

Self-reported HRQOL has been rarely examined across time in lupus patients; in fact there is only one such study (51). In this European study, functioning was found to be relatively stable across time; however, in contrast with our study, the possible predictive value of functioning on later functioning was not examined; furthermore, this study lacked detail about the socioeconomic and demographic features of the patients studied. In short, ours is the first study in which self-reported HRQOL has been examined using data from a longitudinal, multiethnic US lupus cohort, and taking into consideration the different times during the course of the disease that the SF-36 was measured in all patients. We found that the patients' level of functioning early in the disease course exerts a very important influence on the levels of functioning later in the course of the disease; time per se appears to be a less important and consistent predictor of functioning overall in our patients. Given the stability of self-reported functioning over time and the comparable features between those patients whose visits were entered into the analyses, and those in whom they were not, we do not think our data are biased by the exclusion of some patients. Our data and those from the literature underscore the importance of comprehensive treatment strategies that encompass modifications of inadequate illness-related behaviors (e.g., behavioral cognitive therapy) along with conventional pharmacologic or biologic therapy aimed at modifying clinical manifestations associated with disease activity and damage accrual. Such interventions need to occur early if patient outcomes are to be optimized.

Acknowledgements

Thanks to Stacey Kovac, PhD and Catarina Kiefe, PhD, MD for their valuable advice, to our patients without whom this study would not have been possible, and to Ella Henderson, A.A. for her most expert technical assistance in the preparation of this manuscript.

Current LUMINA investigators and staff:

At the University of Alabama at Birmingham: Graciela S. Alarcón, MD, MPH, Holly M. Bastian, MD, MSPH, Barri J. Fessler, MD, Gerald McGwin, Jr., MS, PhD, Jeffrey Roseman, MD, PhD, MPH, América G. Uribe, MD, Sergio Toloza, MD, Bonnie S. Agee, PhD, Martha L. Sanchez, MD, MPH, Ellen Sowell, AA, and Bernadette Johnson, BS.

At the University of Texas-Houston Health Science Center: John D. Reveille, MD, Alan W. Friedman, MD, Khanh Ho, MD, Chul Ahn, PhD, Robert Sandoval, BA, Julio Charles, BS, and Li-Lung Wang, BS.

At the University of Texas Medical Branch at Galveston: Bruce A. Baethge, MD and Sonia Hunnicutt, BS.

At the University of Puerto Rico Medical Sciences Campus: Luis M. Vilá, MD, William Borges, AA, and Carmine Pinilla, BS.

Ancillary