Improvement in Health-Related Quality of Life for Children With Juvenile Idiopathic Arthritis After Start of Treatment With Etanercept

Authors


Abstract

Objective

Improvement in health-related quality of life (HRQOL) is an important therapy goal in the treatment of patients with juvenile idiopathic arthritis (JIA). We investigated the 12-month course of HRQOL in patients with JIA after the start of therapy with etanercept and identified its determining factors.

Methods

Children with JIA were enrolled in the BiKer (Biologics in Pediatric Rheumatology) registry at the start of etanercept treatment. Children were prospectively followed in the first year of treatment and completed the Pediatric Quality of Life Inventory (PedsQL) at each occasion. The change in HRQOL was investigated by random-effect regression models. The time-varying variables pain and inactive disease were used for predicting the change in HRQOL. Inactive disease was defined by the Wallace et al criteria and pain was assessed on a visual analog scale (range 0–100).

Results

The children (n = 61) had a mean age of 10.6 years and a mean disease duration of 3.4 years at the start of etanercept. The mean PedsQL total score was 75. The PedsQL total score increased at a rate of 2.8 units per month (P < 0.001) in the first 6 months of treatment, up to a level of 89.7. A low HRQOL score was significantly highly associated with the number of tender joints, functional restrictions, pain, disease activity, and the existence of a comorbid condition at baseline. Inactive disease and reduced pain predicted better HRQOL under etanercept treatment.

Conclusion

HRQOL was dramatically improved in children who started etanercept treatment. Inactive disease and lower pain were important predictors for improvement of HRQOL over time.

INTRODUCTION

Juvenile idiopathic arthritis (JIA) is the most common chronic inflammatory rheumatic disease and a major cause of chronic disability in children. It comprises a clinically heterogeneous group of disorders characterized by persistent joint inflammation and onset before age 16 years. JIA is associated with functional disability ([1-3]) due to joint manifestations, morning stiffness, and fatigue ([4]). The concept of health-related quality of life (HRQOL) has been widely accepted as a patient-relevant burden of disease measure in understanding the impact of chronic illness in recent years ([5]). Given the current inability to cure JIA, the primary aim of rheumatology care is to ensure the best possible lifelong HRQOL of patients with JIA ([2, 6-9]). HRQOL is a multidimensional concept that incorporates measures of physical symptoms, functional status, and disease impact on psychological and social functioning ([10, 11]), reflecting the definition of health proposed by the World Health Organization ([12]). Many disease-specific and generic HRQOL instruments have been developed and used to measure the subjective burden of disease, as well as the improvement caused by therapeutic interventions. The Pediatric Quality of Life Inventory (PedsQL) ([13]), with its generic and disease-specific module, has been one of the most commonly used measures in children with rheumatic disorders.

The introduction of etanercept (a tumor necrosis factor α antagonist) has appeared to have a major impact on outcome ([14, 15]) and has become an important treatment option for children with JIA who did not respond or were intolerant to the conventional disease-modifying antirheumatic drug (DMARD) methotrexate. Medical interventions affect not only clinically measurable signs or symptoms assessed by the physician, such as current disease activity and organ system damage, but also the patient's experience of disease, assessable via HRQOL measures ([9, 16]). Little is known about the relationship of medical interventions, clinical parameters, and HRQOL ([16]). Furthermore, changes in all aspects of HRQOL are not well studied in children with JIA after the start of etanercept treatment.

The objective of this study was to describe HRQOL in all its domains and determining factors in children with JIA at the start of etanercept. In addition, change of HRQOL in patients taking etanercept was investigated over the first 12 months of treatment. The role of clinical parameters as well as patient-reported outcome measures was studied in the 12-month course of HRQOL. The already existing network of the German BiKer (Biologics in Pediatric Rheumatology) registry was used to identify patients with polyarticular JIA starting etanercept treatment.

Box 1. Significance & Innovations

  • Health-related quality of life (HRQOL) steadily improved in the first 6 months after the start of etanercept treatment in children with juvenile idiopathic arthritis.
  • Significant predictors of impaired HRQOL at baseline were pain, high disease activity, and functional limitations.
  • Strong predictors for the improvement in HRQOL were a lower pain level and the attainment of an active disease over the 12-month study period.

PATIENTS AND METHODS

Patients

This study was conducted as part of the national prospective observational cohort BiKer study ([17]). The BiKer registry started recruitment of patients in 2001 after licensing of etanercept as the first biologic drug in the treatment of JIA. The study was approved by the ethics committee of the Aerztekammer Nordrhein-Westfalen, Dusseldorf, Germany. A total of 61 randomly selected children with JIA were enrolled at the start of etanercept treatment between January 2008 and May 2010 for a more extensive investigation of HRQOL. Included patients underwent the usual assessment of BiKer by a physician and a patient report of outcomes ([17]). They were additionally asked to complete the PedsQL 4.0 and PedsQL rheumatology module ([13]). Data were collected at the beginning of the etanercept treatment (baseline), monthly over the first 6 months of treatment (T1–T6), and after 9 (T9) and 12 months (T12). All parents gave their informed consent.

Measures

The age at the first symptom onset of JIA was recorded by the pediatric rheumatologist. Disease duration was calculated as the duration between symptom onset and the date of the baseline visit. Clinical characteristics of the patients were documented by the responsible rheumatologist at each visit. All patients were classified according to the criteria of the International League of Associations for Rheumatology (ILAR) ([18]). At each visit, the following variables were documented: disease activity on a 100-mm visual analog scale (VAS; where 0 = no disease activity and 100 = very severe disease activity), number of swollen and tender joints, joints with limited range of motion, duration of morning stiffness, erythrocyte sedimentation rate (ESR), and C-reactive protein (CRP) level. Moreover, information was collected with regard to the parameters required to define inactive disease or remission according to Wallace et al (no joint with active arthritis; morning stiffness of ≤15 minutes; no active uveitis, fever, rash, hepatosplenomegaly, serositis, or generalized lymphadenopathy attributable to JIA; ESR or CRP level within the normal limits; and physician's global assessment of disease activity score of the best possible value on the scale used) ([19]). The presence of antinuclear antibodies, HLA–B27 antigen, the patient's previous medication, and comorbidities were documented at baseline. The responsible rheumatologist recorded change in treatment and concomitant therapy with glucocorticoids or nonsteroidal antiinflammatory drugs (NSAIDs). Adverse events or serious adverse events were documented throughout the followup period. American College of Rheumatology (ACR) Pediatric 50 (Pedi 50) and 70 (Pedi 70) criteria for improvement were calculated ([20]).

Functional status

The German version of the Childhood Health Assessment Questionnaire (C-HAQ) ([21]) was used to measure functional ability. The resulting disability index may range from 0–3. A value of 0 indicates no functional disability and higher scores indicate a higher level of disability. The C-HAQ includes an evaluation of pain and overall well-being. Both ratings were assessed on a 100-mm VAS (where 0 = no pain/very good overall well-being and 100 = very strong pain/very poor overall well-being). The C-HAQ was administered to the parents. The C-HAQ score was categorized as suggested by Dempster and colleagues (where 0 = none, 0.1–0.63 = mild, 0.75–1.75 = moderate, and >1.75 = severe) ([22]). The German version of the C-HAQ has shown excellent reliability, validity, and responsiveness ([21]).

HRQOL

HRQOL was assessed by the German version of the PedsQL 4.0 generic core scales ([13]). The PedsQL covers a broad age range by a patient self-report (5–18 years) as well as by a parent proxy report (0–18 years). A 5-point Likert scale (range 0–4, where 0 = never a problem, 1 = almost never a problem, 2 = sometimes a problem, 3 = often a problem, and 4 = almost always a problem) is utilized for the items of the PedsQL generic core scales in the patient self-report and parent proxy report. The PedsQL generic core scales cover the 4 physical, emotional, social, and school functioning domains. The psychosocial health summary score is the composite score of the items answered in the emotional, social, and school functioning scales. The total summary score is based on all 4 domains of the PedsQL. The PedsQL 4.0 generic core scales have shown very good psychometric properties ([13]). The patient's self-report was administered to patients ages >5 years and the proxy report was administered to the parents of all children. The analyses of HRQOL were based on the parent proxy report for children ages ≤5 years and the patient self-report for children ages >5 years.

Statistical analyses

All summary scores and subscores were calculated according to the corresponding scoring algorithms. Categorical variables were reported as absolute and relative frequencies, and continuously distributed variables were reported as means ± SDs. The PedsQL total score and subscores were approximately normally distributed. Linear regression analyses were conducted for studying the association of HRQOL and clinical characteristics at baseline. The predictor variables were standardized (mean ± SD 0 ± 1). The resulting standardized regression coefficients provide a measure for the strength of the association of HRQOL and predictor variables. The strength of association can be categorized into small (β <0.1), medium (β ≥0.1 and ≤0.3), and large (β >0.3) for continuously distributed variables ([23]). Accordingly, categories for categorical variables are given by small (β <0.3), medium (β ≥0.3 and ≤0.8), and large (β >0.8). Significant predictor variables from univariate analyses were entered into a multivariable linear regression model. Random-effect regression models ([24]) were used for investigating the course of HRQOL over time. The development of a longitudinal model for the change in HRQOL was done by a stepwise procedure. This process included the introduction of a random slope parameter for change, inclusion of baseline covariates, and inclusion of the time-dependent predictor variables pain and inactive disease.

Additionally, different settings for the covariance matrix were tested. Models were compared by the Bayesian information criterion (BIC). The BIC balances the decreasing of the log-likelihood value by adding new parameters and a penalty for increasing the number of model parameters. The lower the BIC value is, the better the model fits. Preliminary data analyses suggested that the change in HRQOL showed a nonlinear pattern. A quadratic change parameter was added to the random-effect regression model to adjust for the nonlinear change in HRQOL. Random-effect regression models allow for investigating 2 sources of variability in HRQOL: the within-subject variability over time and the between-subject variability in mean responses. Random-effect regression models result in reliable effect estimates in the presence of missing data that are missing at random ([25]). An unstructured covariance matrix for the random-effect parameters resulted in best model fit based on the BIC. The time-dependent variables of inactive disease and pain were included in the analyses. The effect of the 2 variables on HRQOL was estimated by linear combinations of parameter estimates at each occasion. The intercept and growth parameters were adjusted for the time-invariant covariates disease duration and ILAR category of JIA. Finally, latent growth curve mixture modeling (LGMM) ([26]) was used to investigate whether a subgroup of patients who had a rapid improvement in HRQOL existed. LGMM aims to uncover heterogeneity in a population and to identify clinically meaningful groups characterized by a similar pattern of change in the response variable. The slope of the growth curves was modeled by a piecewise linear function to capture different growth rates over time (intervals include baseline to T3, T3 to T6, and T6 to T12). The intercept and growth parameters were adjusted for the time-invariant covariates disease duration and ILAR category of JIA. Models with 2, 3, and 4 groups were tested and the best-fitting model was determined by considering a combination of fit indices. The following indices ([27]) were used: the BIC, the Vuong–Lo–Mendell–Rubin likelihood ratio test (LRT), and the bootstrap likelihood ratio test (BLRT). The maximum likelihood principle was used for parameter estimation. Multiple sets of starting values were used to avoid local solutions in the estimation procedure. All statistical analyses were conducted using SAS software, version 9.3.

RESULTS

Sociodemographic information and baseline characteristics

The baseline characteristics of the 61 children with JIA are reported in Table 1. More than half of the patients (n = 32 [52.5%]) had rheumatoid factor–negative polyarthritis and 18.0% (n = 11) had extended oligoarthritis. A total of 40 children (65.6%) were female and the mean ± SD age of the sample was 10.0 ± 3.9 years. The disease was diagnosed at a mean ± SD age of 7.3 ± 4.6 years and the mean ± SD disease duration was 3.4 ± 3.0 years. Approximately one-third of the patients (n = 19 [31.1%]) had a comorbid condition at baseline, including allergic diseases (n = 6), uveitis (n = 6), and attention deficit hyperactivity disorder (ADHD; n = 2). Almost all of the patients (n = 58 [95.1%]) were treated at least once with a DMARD before baseline. Among these were methotrexate (n = 54), azathioprine (n = 3), chloroquine/hydroxychloroquine (n = 7), and cyclosporin A (n = 1). Only 13 children (21.3%) reported no functional limitation (C-HAQ score of 0) at baseline.

Table 1. Sociodemographic and clinical characteristics of the study population at baseline*
 Total (n = 61)Patients with complete followup data (n = 53)Patients with incomplete followup data (n = 8)P
  1. BMI = body mass index; JIA = juvenile idiopathic arthritis; RF = rheumatoid factor; VAS = visual analog scale; CRP = C-reactive protein; ESR = erythrocyte sedimentation rate; ANA = antinuclear antibody; NSAIDs = nonsteroidal antiinflammatory drugs; DMARD = disease-modifying antirheumatic drug; C-HAQ = Childhood Health Assessment Questionnaire.
  2. aAmong others: allergic diseases (n = 6), uveitis (n = 6), and attention deficit hyperactivity disorder (n = 2).
  3. bAzathioprine, chloroquine/hydroxychloroquine, cyclosporin A, and methotrexate.
  4. cAzathioprine, chloroquine/hydroxychloroquine, and cyclosporin A.
Age, mean ± SD (range) years10.6 ± 3.9 (3–17)10.8 ± 4.0 (3–17)9.3 ± 3.6 (4–14)0.293
Female sex, no. (%)40 (65.6)35 (66)5 (62.5)0.788
BMI, mean ± SD kg/m220.5 ± 10.420.9 ± 10.917.2 ± 3.30.379
JIA subtype, no. (%)   0.061
Systemic JIA1 (1.6)0 (0.0)1 (12.5) 
RF-negative polyarthritis32 (52.5)26 (49.1)6 (75.0) 
RF-positive polyarthritis9 (14.8)9 (17.0)0 (0.0) 
Enthesitis-related arthritis3 (4.9)3 (5.7)0 (0.0) 
Psoriatic arthritis2 (3.3)1 (1.9)1 (12.5) 
Persistent oligoarthritis2 (3.3)2 (3.8)0 (0.0) 
Extended oligoarthritis11 (18.0)11 (20.8)0 (0.0) 
Other arthritis1 (1.6)1 (1.9)0 (0.0) 
Disease duration, mean ± SD years3.4 ± 3.23.6 ± 3.31.8 ± 1.70.135
Age at disease onset, mean ± SD years7.3 ± 4.67.3 ± 4.77.4 ± 4.20.973
Physician-assessed disease activity, mean ± SD VAS score56.9 ± 19.556.1 ± 20.162.0 ± 15.70.433
No. of active joints, mean ± SD7.2 ± 6.27.3 ± 6.67.1 ± 3.70.958
Current morning stiffness ≥15 minutes, no. (%)30 (49.1)26 (49.1)4 (50.0)0.960
CRP level, mean ± SD mg/liter13.5 ± 22.610.5 ± 11.618.4 ± 17.20.153
ESR, mean ± SD mm/hour25.3 ± 25.925.8 ± 27.121.4 ± 14.70.681
Positive ANA test at baseline, no. (%)29 (47.5)28 (52.8)1 (12.5)0.044
Positive HLA–B27 test, no. (%)12 (19.7)12 (22.6)0 (0.0)0.216
Any comorbidity, no. (%)a19 (31.1)17 (32.1)2 (25.0)0.687
History of uveitis, no. (%)4 (6.7)4 (7.6)0 (0.0)0.421
Current use of glucocorticoids, no. (%)35 (57.4)27 (50.9)8 (100.0)0.009
Current use of NSAIDs, no. (%)56 (91.8)50 (94.3)6 (75)0.090
Previous DMARD treatment    
Any DMARD, no. (%)b58 (95.1)50 (94.3)8 (100.0)0.490
Any other DMARD, no. (%)c13 (21.3)13 (24.5)0 (0.0)0.114
Methotrexate, no. (%)54 (88.5)47 (88.7)7 (87.5)0.977
No. of previous DMARDs used, mean ± SD1.2 ± 0.61.2 ± 0.61.0 ± 0.00.398
C-HAQ total score, mean ± SD0.7 ± 0.60.7 ± 0.60.7 ± 0.60.972
C-HAQ no disability, no. (%)13 (21.3)11 (20.8)2 (25.0)0.785
Patient-reported pain, mean ± SD VAS score36.4 ± 27.135.9 ± 27.240.0 ± 28.00.692
Patient-reported well-being, mean ± SD VAS score38.3 ± 25.537.7 ± 25.141.8 ± 29.20.682

Predictors of HRQOL at baseline assessment

Univariate and multivariate linear regression analyses were used for investigating the association of HRQOL with clinical variables and patient-reported outcome variables at the start of etanercept treatment (Table 2). Lower HRQOL was strongly associated with functional disability, patient-reported pain, and overall well-being. Higher disease activity, a higher number of tender joints, and having a comorbid condition were negatively associated with HRQOL. Children with a comorbid condition reported lower HRQOL (mean ± SD PedsQL sum score 63.6 ± 17.8 for allergic disease and 65.5 ± 9.8 for uveitis). The 2 patients with ADHD had a remarkably low PedsQL sum score (mean ± SD 50.2 ± 2.8). All significant predictor variables from the univariate analyses were entered into a multivariable regression model predicting HRQOL at baseline. The C-HAQ showed a consistently large effect on HRQOL in the multivariable model. Comparable results were found for the PedsQL subscores physical health, emotional functioning, social functioning, school functioning, and psychosocial functioning.

Table 2. Univariate and multivariable regression analyses for clinical predictors of health-related quality of life at baseline assessment using the Pediatric Quality of Life Inventory sum score*
 Univariate analyses, β (95% CI)Multivariable analyses, β (95% CI)a
  1. β = standardized regression coefficient; 95% CI = 95% confidence interval; RF = rheumatoid factor; JIA = juvenile idiopathic arthritis; VAS = visual analog scale; C-HAQ = Childhood Health Assessment Questionnaire.
  2. aSignificant predictor variables from univariate analyses were entered into the multivariate model.
  3. bSystemic JIA, enthesitis-related arthritis, psoriatic arthritis, other arthritis, and persistent oligoarthritis.
  4. cP < 0.05.
  5. dPatient-reported well-being was omitted in multivariate analyses due to collinearity.
Age0.11 (−0.17, 0.39) 
Female sex0.23 (−0.03, 0.50) 
Body mass index0.04 (−0.23, 0.32) 
RF-negative polyarthritis−0.09 (−0.37, 0.18) 
RF-positive polyarthritis0.22 (−0.04, 0.47) 
Extended oligoarthritis0.04 (−0.24, 0.32) 
Other JIA subtypeb−0.17 (−0.46, 0.11) 
Disease duration−0.12 (−0.41, 0.16) 
Age at disease onset0.16 (−0.14, 0.45) 
Physician-assessed disease activity (VAS score)−0.36 (−0.62, −0.11)c−0.19 (−0.41, 0.03)
No. of swollen joints0.05 (−0.34, 0.43) 
No. of tender joints−0.50 (−0.83, −0.16)c0.00 (−0.33, 0.33)
No. of joints with limited range of motion−0.21 (−0.54, 0.12) 
Use of corticosteroids−0.09 (−0.36, 0.18) 
Current morning stiffness ≥15 minutes−0.29 (−0.56, −0.02)c0.12 (−0.11, 0.35)
Any comorbid condition−0.32 (−0.58, −0.06)c−0.20 (−0.40, 0.00)c
C-HAQ total score−0.66 (−0.87, −0.46)c−0.49 (−0.72, −0.25)c
Patient-reported pain (VAS score)−0.57 (−0.79, −0.35)c−0.27 (−0.52, −0.02)c
Patient-reported well-being (VAS score)d−0.61 (−0.83, −0.40)c 

Characteristics of patients over the 12-month study period

A total of 8 children (5 females, 3 males) were lost to followup. Reasons for study dropout were ineffectiveness of therapy in 6 patients, remission on medications in 1 patient, and occurrence of uveitis (adverse event) in 1 patient (Figure 1). Patients who dropped out due to ineffective therapy had moderate disease activity (mean ± SD 29.6 ± 16.1), a high CRP level (mean ± SD 23.6 ± 37.9), and an increased ESR (mean ± SD 23.3 ± 18.7) at the last visit. Sociodemographic and other clinical characteristics of children who were lost to followup did not significantly differ from participants with complete data over the study period (Table 1), except for the current use of glucocorticoids (50.9% for children with complete data versus 100.0% for children with incomplete data; P = 0.009).

Figure 1.

Study overview: patient flow chart showing the patient flow throughout the different study visits, including baseline, month 1 (T1), month 2 (T2), month 3 (T3), month 4 (T4), month 5 (T5), month 6 (T6), month 9 (T9), and month 12 (T12). A total of 8 patients left the study during the 12-month study interval. Reasons for dropout and the individual health-related quality of life (HrQoL) values at the last available visit are shown in the lower half of the figure.

At baseline and 12-month followup, almost all patients were treated with NSAIDs (n = 56 [91.8%] and n = 47 [88.7%], respectively) and NSAIDs or glucocorticoids (n = 60 [98.4%] and n = 49 [92.4%], respectively). All patients received etanercept and comedication (NSAIDs, glucocorticoids, or DMARDs) at baseline. One patient (1.9%) was treated only with etanercept at 12-month followup. A total of 47 patients (77%) were treated with a combination of etanercept and a nonbiologic DMARD (n = 45 [95.8%] taking methotrexate and n = 2 [4.2%] taking leflunomide) at baseline. Compared to patients treated with a combination of etanercept and methotrexate, patients treated solely with etanercept had a lower rate of uveitis (4.4% versus 13.3%), a lower frequency of rheumatoid factor–positive polyarthritis (6.7% versus 15.6%), a higher rate of extended oligoarthritis (20.0% versus 13.3%), a shorter disease duration (mean ± SD 2.9 ± 3.0 versus 4.8 ± 3.5 years), and a lower number of active joints (mean ± SD 4.9 ± 4.4 versus 8.0 ± 6.6) at baseline. The frequency of the combined prescription of etanercept and DMARDs steadily decreased to 39% at 12 months.

ACR Pedi 50 and 70 improvements were seen in 79.0% and 73.7% of children at 12-month followup, respectively. In contrast, ACR Pedi 50 and 70 improvements were lower at the last visit in patients who left the study (n = 3 [37.5%] and n = 2 [25.0%], respectively). The functional status improved under etanercept. Two-thirds of the patients (n = 32 [64.0%]) reported no functional disability (C-HAQ score of 0) after 6 months of treatment (P < 0.001 for change from T0 to T6). The mean ± SD pain level significantly decreased over time (36.4 ± 27.1 at baseline to 9.5 ± 13.3 at T12; β = −1.55 per month [95% confidence interval (95% CI) −1.97, −1.13]). The mean ± SD ESR decreased during the study period (25.3 ± 25.9 mm/hour at baseline to 11.8 ± 12.4 mm/hour at T12; β = −0.67 per month [95% CI −1.03, −0.30]). A normal ESR (≤20 mm/hour) and CRP level (≤10 mg/liter) were seen in 55.7% and 60.7% of patients at baseline and in 63.4% and 78.1% of patients at 12-month followup, respectively. The mean ± SD number of active joints (7.2 ± 6.2 at baseline and 1.0 ± 2.0 at T12; β = −0.40 per month [95% CI −0.50, −0.31]) and mean ± SD disease activity (56.9 ± 19.5 at baseline and 11.6 ± 20.6 at T12; β = −2.70 per month [95% CI −3.22, −2.17]) were significantly lower at 12-month followup compared to baseline. One-third of the children (n = 17 [32.1%]) had inactive disease after 12 months of treatment with etanercept. Five patients attained remission on medications according to the Wallace et al criteria during the observation period ([19]).

Improvement of HRQOL in followup

A total of 8 patients left the study. Individual HRQOL values at the last visit are reported in Figure 1. Only 2 patients reported very low HRQOL before they left the study (HRQOL 47.3 for a patient with an adverse event at T3 and HRQOL 69.8 for ineffectiveness of therapy at T9) in comparison to patients with complete data for that assessment. The change in HRQOL over time was modeled by a random-effect model including a quadratic change parameter. This model resulted in best fit based on the BIC compared to a fixed-effect model with a linear change. HRQOL significantly improved in the first 6 months of etanercept treatment (rate of improvement per month: β = 3.13, P < 0.001 for total score; β = 2.91, P < 0.001 for psychosocial functioning; β = 3.79, P < 0.001 for physical health; β = 3.50, P < 0.001 for emotional functioning; β = 2.38, P < 0.001 for social functioning; and β = 2.80, P < 0.001 for school functioning). However, during the interval T6 to T12, there was no further increase (rate per month: β = 0.59, P = 0.081 for total score; β = 0.52, P = 0.132 for psychosocial functioning; β = 0.77, P = 0.074 for physical health; β = 0.63, P = 0.190 for emotional functioning; β = 0.49, P = 0.238 for social functioning; and β = 0.42, P = 0.298 for school functioning). We investigated whether inactive disease and pain were associated with improvement of HRQOL across the study period. A random-effect regression model for change in HRQOL over time was estimated with the time-dependent predictor variables of inactive disease and pain (Table 3). The change in the PedsQL total score was positively associated with inactive disease at T3 to T6. A lower level of pain predicted better PedsQL total scores at each occasion. A status of inactive disease under etanercept resulted in a significant increase in emotional functioning and physical health (Table 3). We could not find a statistically significant association between inactive disease and the improvement in social or school functioning. The decrease in perceived pain was consistently associated with an increase in HRQOL across all PedsQL domains.

Table 3. Inactive disease and pain as independent predictors for improvement in HRQOL under etanercept over 12 months in 61 children with JIA*
 ValueTotal scorePhysical healthEmotional functioningSocial functioningSchool functioning
Mean ± SDβ (95% CI)Mean ± SDβ (95% CI)Mean ± SDβ (95% CI)Mean ± SDβ (95% CI)Mean ± SDβ (95% CI)
  1. The level of health-related quality of life (HRQOL) at baseline (intercept) and the change in HRQOL over time (linear and quadratic growth parameters) were adjusted for disease duration and International League of Associations for Rheumatology category of juvenile idiopathic arthritis (JIA). The influence of the time-varying predictor variables inactive disease and pain on HRQOL was investigated at each occasion. β = regression coefficient; 95% CI = 95% confidence interval; PedsQL = Pediatric Quality of Life Inventory.
  2. aUnadjusted growth curve model for HRQOL (omitting time-invariant and time-dependent covariates).
  3. bGrowth parameters are the estimated change per month in HRQOL.
  4. cSignificant.
  5. dGrowth curve model for HRQOL with time-invariant (disease duration and JIA subtype) and time-dependent covariates inactive disease and pain.
  6. eAccording to the Wallace et al criteria for inactive disease ([19]).
PedsQL           
T0 72.2 ± 16.8 66.6 ± 22.6 71.5 ± 22.4 79.6 ± 17.5 72.0 ± 16.5 
T1 79.2 ± 15.4 77.0 ± 18.4 79.2 ± 19.2 83.8 ± 16.8 77.2 ± 19.0 
T2 82.7 ± 14.1 82.7 ± 15.3 83.1 ± 17.9 84.9 ± 17.1 80.1 ± 16.5 
T3 81.4 ± 16.5 80.6 ± 19.5 82.3 ± 18.1 84.2 ± 19.2 79.1 ± 18.7 
T4 86.3 ± 11.9 84.9 ± 14.4 88.3 ± 13.6 89.4 ± 14.7 83.2 ± 16.4 
T5 89.7 ± 10.7 88.8 ± 15.2 89.3 ± 14.4 92.4 ± 11.3 88.8 ± 11.6 
T6 88.8 ± 11.6 87.5 ± 14.7 88.0 ± 15.9 91.2 ± 17.9 89.2 ± 12.5 
T9 92.8 ± 7.6 91.4 ± 11.4 93.1 ± 12.3 94.4 ± 10.6 91.0 ± 11.1 
T12 92.0 ± 9.2 90.4 ± 11.1 91.5 ± 17.9 93.4 ± 11.1 93.0 ± 8.8 
Model 1a           
Linear growth parameterb  3.83 (3.03, 4.63)c 4.66 (3.64, 5.69)c 4.23 (3.09, 5.37)c 2.86 (1.87, 3.86)c 3.55 (2.59, 4.50)c
Quadratic growth parameterb  −0.21 (−0.27, −0.14)c −0.26 (−0.34, −0.17)c −0.24 (−0.34, −0.15)c −0.14 (−0.23, −0.06)c −0.18 (−0.26, −0.10)c
Model 2d           
Linear growth parameterb  0.72 (0.34, 1.10)c 0.66 (0.21, 1.12)c 0.56 (−0.07, 1.19) 0.91 (0.41, 1.42)c 0.75 (0.26, 1.24)c
Quadratic growth parameterb  2.99 (−0.95, 6.93) 2.61 (−2.10, 7.32) 5.97 (−0.11, 12.06) 2.87 (−2.29, 8.03) 0.15 (−5.03, 5.33)
Inactive disease, no. (%)e          
T00 (0.0)          
T16 (10.0) 2.89 (−0.57, 6.35) 2.73 (−1.41, 6.87) 5.35 (0.02, 10.67)c 2.65 (−1.88, 7.19) 0.43 (−4.13, 4.99)
T28 (13.8) 2.78 (−0.25, 5.81) 2.85 (−0.77, 6.47) 4.72 (0.08, 9.37)c 2.43 (−1.54, 6.40) 0.70 (−3.30, 4.70)
T315 (26.8) 2.68 (0.01, 5.35)c 2.97 (−0.22, 6.16) 4.10 (0.02, 8.17)c 2.21 (−1.28, 5.71) 0.98 (−2.54, 4.49)
T411 (19.6) 2.58 (0.18, 4.98)c 3.09 (0.22, 5.96)c 3.47 (−0.21, 7.15) 1.99 (−1.15, 5.14) 1.25 (−1.91, 4.42)
T513 (23.2) 2.48 (0.20, 4.75)c 3.21 (0.50, 5.93)c 2.84 (−0.66, 6.35) 1.78 (−1.21, 4.76) 1.53 (−1.46, 4.51)
T620 (36.4) 2.37 (0.07, 4.67)c 3.33 (0.58, 6.08)c 2.22 (−1.38, 5.81) 1.56 (−1.47, 4.58) 1.80 (−1.20, 4.80)
T916 (30.2) 2.06 (−1.10, 5.23) 3.70 (−0.10, 7.49) 0.34 (−4.75, 5.43) 0.90 (−3.28, 5.08) 2.63 (−1.47, 6.72)
T1217 (32.1) 1.76 (−2.85, 6.37) 4.06 (−1.48, 9.59) −1.54 (−8.99, 5.91) 0.24 (−5.86, 6.35) 3.45 (−2.52, 9.42)
Pain, mean ± SD          
T036.4 ± 27.1          
T116.0 ± 18.5 −0.35 (−0.40, −0.30)c −0.53 (−0.59, −0.47)c −0.37 (−0.45, −0.29)c −0.22 (−0.29, −0.14)c −0.30 (−0.38, −0.23)c
T212.5 ± 17.1 −0.34 (−0.39, −0.29)c −0.52 (−0.58, −0.46)c −0.34 (−0.42, −0.26)c −0.22 (−0.29, −0.15)c −0.29 (−0.36, −0.22)c
T314.6 ± 20.9 −0.33 (−0.38, −0.27)c −0.51 (−0.58, −0.45)c −0.32 (−0.40, −0.23)c −0.21 (−0.29, −0.14)c −0.28 (−0.35, −0.21)c
T412.2 ± 16.3 −0.31 (−0.38, −0.25)c −0.51 (−0.58, −0.43)c −0.29 (−0.39, −0.19)c −0.21 (−0.30, −0.13)c −0.27 (−0.35, −0.19)c
T512.8 ± 17.3 −0.30 (−0.37, −0.23)c −0.50 (−0.58, −0.41)c −0.26 (−0.37, −0.15)c −0.21 (−0.31, −0.12)c −0.26 (−0.35, −0.16)c
T610.4 ± 14.0 −0.29 (−0.37, −0.21)c −0.49 (−0.59, −0.39)c −0.23 (−0.37, −0.10)c −0.21 (−0.32, −0.10)c −0.24 (−0.35, −0.14)c
T99.1 ± 11.9 −0.25 (−0.37, −0.13)c −0.46 (−0.61, −0.32)c −0.15 (−0.35, 0.05) −0.21 (−0.37, −0.05)c −0.21 (−0.36, −0.05)c
T129.5 ± 13.3 −0.21 (−0.37, −0.05)c −0.44 (−0.63, −0.24)c −0.07 (−0.33, 0.20) −0.20 (−0.42, 0.01) −0.17 (−0.38, 0.04)

The improvement in HRQOL for patients treated with a combination of etanercept and methotrexate was smaller (mean ± SD total score at baseline 71.8 ± 16.7, mean ± SD total score at 12 months 86.9 ± 11.9, rate of improvement per month 1.24; P < 0.001) than for patients treated with etanercept without concomitant methotrexate (mean ± SD total score at baseline 73.5 ± 17.5, mean ± SD total score at 12 months 94.4 ± 6.5, rate of improvement per month 1.44; P < 0.001). The difference in the PedsQL total score between the 2 groups was statistically significant at the 12-month assessment after adjusting for baseline differences in disease activity (difference in HRQOL 7.43; P = 0.009). Patients treated with a combination of etanercept and methotrexate reported lower emotional functioning (difference in HRQOL 12.3; P = 0.044) and social functioning (difference in HRQOL 8.4; P = 0.023) after 12 months.

Significantly lower HRQOL at baseline was reported by patients with a comorbid disease (mean ± SD 63.9 ± 16.4 versus 75.9 ± 15.8) and patients reporting any functional disability (mean ± SD 67.7 ± 16.1 versus 86.6 ± 9.2) compared to those without the condition (Table 2). This difference was not present at the 12-month followup (mean ± SD 90.8 ± 10.0 versus 92.7 ± 8.8 with and without a comorbid disease; P = 0.569 and mean ± SD 95.3 ± 5.1 versus 91.4 ± 9.6 with and without functional limitation; P = 0.379).

LGMM analysis was used to identify possible trajectories of HRQOL over the 12-month study period. We chose a 2-class model based on the BIC, LRT, and BLRT. The LRT and BLRT tests were not significant for the 3-class model, indicating no improvement in model fit by adding a third group. Most patients (n = 44 [72.1%]) were classified into a trajectory characterized by a rapid increase (rate of improvement per month: β = 4.2, P < 0.001) in HRQOL in the first 3 months of treatment (Figure 2). In contrast, the second group included children (n = 17 [27.9%]) who did not report a rapid improvement in HRQOL. Children in the second group reported a significant increase in HRQOL after 3 months of treatment (rate of improvement per month: β = 7.7, P < 0.001). A rapid increase in HRQOL in the first 3 months of treatment was associated with a lower number of tender joints (mean ± SD 5.9 ± 5.7 for group 1 and 9.3 ± 8.5 for group 2; P = 0.045), a lower pain level (mean ± SD 32.1 ± 27.4 for group 1 and 50.4 ± 22.9 for group 2; P = 0.022), and less functional disability (mean ± SD 0.5 ± 0.6 for group 1 and 1.1 ± 0.5 for group 2; P = 0.003) at baseline. Both groups did not significantly differ in other clinical characteristics.

Figure 2.

Trajectories of health-related quality of life (HrQoL) measured by the Pediatric Quality of Life Inventory total score over 12 months. Parameter estimates for the 2 trajectories and significant predictors at baseline for each trajectory of HrQoL are shown in the table. n.s. = not significant; 95% CI = 95% confidence interval; Δ = absolute change per month in HrQoL for the study intervals baseline (T0) to month 3 (T3), month 3 (T3) to month 6 (T6), and month 6 (T6) to month 12 (T12); C-HAQ = Childhood Health Assessment Questionnaire; VAS = visual analog scale.

DISCUSSION

We studied the diversification of HRQOL and its determining factors at the start of etanercept treatment. Our study showed that functional disability and pain were strongly associated with reduced HRQOL in children with severe and active JIA. Other correlates for lower HRQOL at baseline were high disease activity, number of painful joints, morning stiffness of at least 15 minutes, and a comorbid condition. HRQOL significantly increased in the first 6 months of treatment with etanercept. It remained at a high level for the period from 6 months to 12 months, with some minor fluctuations. LGMM identified a group of children with a rapid increase in HRQOL in the first 3 months of treatment. Children with a moderate number of tender joints, low functional restrictions, and moderate pain at baseline showed a rapid response in HRQOL.

All patients included in our study were prospectively followed for a 12-month period after the start of etanercept treatment. The study cohort included cases with a disease severe enough to be treated with a biologic drug. The sample was mainly composed of patients with polyarticular or systemic onset of JIA (68%), with almost all cases having a polyarticular course of JIA. Only 21% of the children had no functional limitations at baseline. Thirty-two percent of the patients had inactive disease at 12-month followup. A similar rate was reported from the Dutch Arthritis and Biologicals in Children registry ([28]) and patients with polyarticular JIA ([29]). The number of patients with inactive disease steadily increased during the first 6 months of treatment with etanercept (from 10% at T1 to 36.4% at T6). Emotional functioning was significantly improved if a patient attained inactive disease within the first 3 months of treatment. Inactive disease yielded better physical functioning for the period from 4 to 6 months. The children's pain level was significantly reduced after 1 year of etanercept treatment. A lower pain level was associated with higher levels of all HRQOL domains across the 12-month course.

Prospective long-term studies of HRQOL changes are scarce in patients with JIA ([30-32]). To our knowledge, this is the first study using the generic PedsQL questionnaire to examine determining factors for the change in HRQOL in patients treated with etanercept. Most of the previous studies used proxy reporting for HRQOL ([30, 33, 34]), such as the VAS for well-being from the C-HAQ. The self-report of HRQOL has appeared to be more reliable than proxy reporting ([35, 36]).

HRQOL in all domains increased in the first 6 months after the start of treatment with etanercept. Similar results were reported by Prince et al ([30]). This is the first study that investigated predictor variables for the change in HRQOL over time. A lower pain level predicted higher HRQOL over the 12-month study period. Furthermore, moderate pain at baseline resulted in a rapid increase in HRQOL in the first 3 months of treatment, as shown by LGMM, which supports the findings of other studies ([33, 34, 37]) that pain is an important determinant of HRQOL. Children with inactive disease reported significantly higher levels of HRQOL in the first 6 months of etanercept treatment. Inactive disease is characterized among others by no joint with active arthritis. It suggests that effective control of the arthritis by etanercept yields improved HRQOL.

VAS well-being, VAS pain, C-HAQ total score, and number of tender joints were highly correlated with HRQOL in univariate cross-sectional analyses. Disease activity, a comorbid condition, and morning stiffness of at least 15 minutes were predictor variables with a moderate effect size. In multivariable cross-sectional analysis, only pain, a comorbid disease, and the C-HAQ total score were significant predictor variables. Considering this, our study identified similar determining factors of HRQOL to the study by Haverman and colleagues ([23]). This is remarkable because their study included children with the full spectrum of JIA, whereas our study included the severe end of the JIA spectrum. Therefore, pain and functional status may be global predictor variables for HRQOL in JIA.

We showed that patients treated by a combination of etanercept and methotrexate had a smaller increase in HRQOL compared to patients under monotherapy with etanercept. Methotrexate is the first choice of DMARD in the treatment of polyarticular JIA. However, it has been shown that the intake of methotrexate is associated with gastrointestinal adverse effects such as nausea, abdominal pain, and vomiting ([38, 39]). Children with combination therapy reported significantly lower emotional and social functioning. This is in line with a previous study by van der Meer et al ([39]), who reported more frequent psychological problems under methotrexate use. Our results about differences in HRQOL between monotherapy and combination therapy with etanercept with methotrexate should be considered with caution. Patients with combination therapy were more likely severe cases with higher disease activity at baseline. It cannot be excluded that this result is biased by treatment indication, although we controlled for baseline differences between the 2 groups.

HRQOL was measured by patient self-report in children ages >5 years. The parent proxy report was used for children ages <5 years. We used the age-specific PedsQL self-report instrument for the age groups 5–7 years, 8–12 years, and 13–18 years. Varni et al ([40]) showed that pediatric patients ages >5 years can provide reliable and valid HRQOL estimates in case of an age-appropriate instrument. Parent proxy report was used when the children were unable to respond to the HRQOL instrument ([41]). One should consider that proxy HRQOL estimates may not be sufficiently accurate as self-reports. Ratings may be biased by the parent's perception and attitudes ([41]).

The data have to be seen in light of some limitations. This is an observational study about the improvement of HRQOL in children with JIA under etanercept treatment over 12 months. The lack of a control group that had never taken biologic agents limits the results of the study. Other important factors such as additional physical therapy or psychotherapy may be associated with the improvement in HRQOL in addition to treatment with etanercept. The separate analysis of HRQOL for the JIA subtypes is limited by the small sample. It would be interesting to see the change of HRQOL and its determining factors for the JIA subtypes because of their heterogeneity. We adjusted the statistical models for baseline differences in HRQOL between the JIA subtypes. The small sample size also limits the statistical modeling. We could only analyze 2 time-varying predictor variables in the growth curve model. Both variables did not explain all of the variance of change in HRQOL over time. Other important predictor variables for the change in HRQOL may exist.

In conclusion, improvement of HRQOL is an important goal in the treatment of JIA. HRQOL was dramatically improved in children who started etanercept treatment. Inactive disease and lower pain were important predictors for improvement of HRQOL over time.

AUTHOR CONTRIBUTIONS

All authors were involved in drafting the article or revising it critically for important intellectual content, and all authors approved the final version to be published. Dr. Klotsche had full access to all of the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.

Study conception and design. Klotsche, Minden, Urban, Horneff.

Acquisition of data. Klotsche, Minden, Thon, Ganser, Urban, Horneff.

Analysis and interpretation of data. Klotsche, Minden, Thon, Ganser, Horneff.

ROLE OF THE STUDY SPONSOR

Pfizer had no role in the study design or in the collection, analysis, or interpretation of the data, the writing of the manuscript, or the decision to submit the manuscript for publication. Publication of this article was not contingent upon approval by Pfizer.

Acknowledgments

The authors would like to thank those physicians who enrolled the patients for this study: Frank Dressler, Hannover; Regina Hühn, Halle; Jasmin Kuemmerle Deschner, Tübingen; Juergen Quietzsch, Plauen ([2]); Rolf Kuester, Bad Bramstedt; Maria Haller, Gundelfingen; Johannes Peter Haas, Garmisch Partenkirchen; Christina Mokross, Oldenburg; and Ivan Foeldvari, Hamburg.

Ancillary