Patient-reported outcomes following biologic therapy in a sample of adults with rheumatoid arthritis recruited from community-based rheumatologists

Authors


  • Because Drs. Katz and Yelin are Editors of Arthritis Care & Research, review of this article was handled by the Editor of Arthritis & Rheumatism.

Abstract

Objective

To examine self-reported symptoms and functioning in a community-based sample of persons with rheumatoid arthritis who did and did not initiate treatment with biologic agents.

Methods

Data were from annual telephone interviews (1998–2003) with an observational cohort identified through community rheumatologists. Self-reported function and symptoms of subjects who initiated biologic therapy (etanercept or infliximab) and reported consistent use at 2 annual interviews (continuous use; n = 64) were compared at 1 year prior to initiation of therapy (baseline), and years 1 and 2 of therapy to those with no biologic therapy (n = 183) and those who initiated biologic therapy but discontinued use (n = 42).

Results

At baseline, subjects taking biologic agents reported significantly worse function and symptoms on all measures except fatigue and pain severity. After 2 years, significant differences in the Health Assessment Questionnaire scores remained, but there were no other significant differences between the nonuser group and the continuous use group. The discontinued use group exhibited significantly greater pain severity and more painful joints than nonusers. Improvements in the number of painful (33.4% versus 16.2%; P = 0.004), and swollen (38.4% versus 18.7%; P = 0.003) joints, and morning stiffness (27.3% versus 10.4%; P = 0.001) were more frequent in the continuous use group than in the nonuser group.

Conclusion

Results suggest that biologic treatment was initiated based on severe disease. Over ∼17 months of treatment, differences in some but not all symptoms between the continuous use group and the nonuser group narrowed to statistical nonsignificance.

INTRODUCTION

Clinical trials of biologic agents have established the efficacy of these agents in improving outcomes among individuals with moderate to severe rheumatoid arthritis (RA) (1–3). Results of these trials are usually based on American College of Rheumatology 20%, 50%, and 70% criteria for improvement (4) or European League Against Rheumatism improvement criteria (5), which include some patient-reported outcomes; however, patient-reported outcomes have not traditionally been the primary focus of clinical trials. More recently, some studies have begun to focus on patient-reported outcomes such as joint pain, pain severity, fatigue, and function (6–8).

A limited number of these trials have evaluated results beyond 1 year. For example, of the trials evaluated by Chen and colleagues for a systematic review of the effectiveness of various biologic agents, only 2 of 11 trials of etanercept extended beyond 12 months, and the longest noted trials of infliximab were 54 weeks (4 of 9 trials) (1). Similarly, Gartlehner et al noted in another systematic review of 26 trials of biologic agents that the longest trial was 52 weeks (9).

In addition, patients enrolled in clinical trials are selected based on very strict inclusion and exclusion criteria, and treatment is defined by the clinical trial protocol. Patients are randomly assigned to treatments to accurately assess the efficacy of the treatment being studied and to minimize differences between treatment groups. However, in community practice, treatment assignment is not random and care is not as strictly controlled, so effectiveness may differ from the efficacy seen in clinical trials.

In 1998, 2 biologic agents, etanercept and infliximab, were approved for the first time for the treatment of individuals with moderate to severe RA. Given the high cost of these agents, it is important to assess how treatment is being allocated and whether individuals receiving long-term treatment are experiencing meaningful benefits from treatment. Some community-based studies suggest that patients who receive biologic agents have worse function and pain than those who do not (8).

The purpose of this analysis was to examine self-reported symptoms and functioning, and changes in symptoms and functioning after up to 2 years of treatment, of persons with RA treated by community rheumatologists who initiated biologic agents compared with a group of individuals from the same cohort who did not use biologic agents.

SUBJECTS AND METHODS

Overview.

Data were drawn from a cohort of individuals with RA who were interviewed annually by telephone. Disease-related characteristics of the individuals who reported initiation of biologic therapy (etanercept or infliximab) at one interview and reported consistent use at the following interview (n = 64) were compared with those individuals who did not use biologic agents (n = 183) and those who initiated biologic therapy but discontinued use (n = 42). For the treatment groups, the year prior to initiation of therapy was considered the baseline year, the first year of reported biologic treatment was year 1, and the subsequent followup year was year 2. For the comparison group, 1998 was used as the baseline year, 1999 as year 1, and 2000 as year 2. Comparisons of symptoms and functioning were made between the groups at each time point. Improvement scores, defined as a change from baseline by at least 0.5 SDs, were also calculated, and analyses compared the frequency of improvement among the 3 groups.

Subjects.

The sample for the present study was drawn from 5 waves of the University of California, San Francisco (UCSF) RA Panel Study, covering the years 1998 through 2003. The UCSF RA Panel was constructed in 1982 from a random sample of rheumatologists practicing in Northern California. Participants were recruited from lists maintained by participating rheumatologists of all persons with RA presenting to their offices over a 1-month period and expressing an interest in participating in the study. The original RA Panel consisted of 822 patients who were enrolled between June 1982 and July 1983. There were subsequently 4 additional enrollment periods in 1989 to 1990, 1995, 1999, and 2003, during which 203, 131, 122, and 169 individuals were enrolled, respectively. Retention from year to year has averaged 93%; the 7% attrition includes deaths. The principal data source for the RA Panel is an annual telephone interview that includes questions on demographics, medications, RA symptoms, comorbidities, and functioning. This study was approved by the UCSF Committee on Human Research.

Variables.

Primary independent variable: use of biologic agents.

Medication use is regularly assessed as part of the RA Panel telephone interview. Participants are asked if they have taken any of a list of medications during the preceding year. For each of the study years, they were asked if they had taken etanercept or infliximab for at least 1 month in the past year. Reported use of either etanercept or infliximab constituted use of biologic agents, which was determined for each year.

Dependent variables: self-reported symptoms and functioning.

Panel members were regularly queried about functioning and symptoms as a part of the annual telephone interview. Functioning was assessed using the Health Assessment Questionnaire (HAQ) (10), which was developed specifically to assess functioning among individuals with arthritis. HAQ scores range from 0 to 3, with higher scores representing more severe functional impairment. Panel members were also asked to rate the severity of their pain on a scale of 0–100, where 0 = no pain and 100 = very severe pain (10). The following symptom measures were also assessed: number of painful joints/joint groups from a list of 17 (11); number of swollen joints/joint groups from a list of 14 (11); duration of morning stiffness, dichotomized to <1 hour versus ≥1 hours; and severity of fatigue, rated as no fatigue, very mild, mild, moderate, severe, or very severe. Based on the distribution of responses, fatigue ratings were dichotomized to severe or very severe versus all other responses.

Covariates.

All multivariate analyses controlled for age, sex, years of education, baseline or 1998 number of comorbid conditions from a list of 7 conditions (hypertension, heart disease, stroke or neurologic condition, diabetes mellitus, lung disease, and kidney disease), and duration of RA.

Statistical analysis.

Subjects who initiated biologic therapy (etanercept or infliximab) were compared with those who did not take biologic agents. Initiation of therapy was defined as a report of taking one of the biologic agents with no previous use of biologic therapy. Subjects could report initiation of therapy in any of 4 years (1999, 2000, 2001, or 2002). The year prior to initiation of therapy was defined as the baseline year. Data from 3 years were examined: baseline and years 1 and 2 after initiation of therapy. Individuals who reported initiation of biologic therapy in year 1 and reported continuous use at the year 2 interview were included in the continuous use group (n = 64). Individuals who reported initiation of biologic therapy in year 1 but did not report continuous use at the year 2 interview were included in the discontinued use group (n = 42). The comparison group consisted of individuals who did not use biologic agents during any of the analysis years (1998 through 2003; n = 183). For the comparison group, 1998 was used as the baseline year, 1999 as year 1, and 2000 as year 2.

For these analyses, only individuals who were interviewed in 1998 (baseline year for analysis) and remained in the RA Panel for at least 2 additional years were included. Individuals who initiated biologic agents but were lost to followup before their year 2 interview were excluded (n = 4).

At each time point (baseline, year 1, and year 2), self-reported symptoms and functioning were compared among the 3 treatment groups (no use, discontinued use, and continuous use). Bivariate analyses (i.e., analyses of variance and chi-square analyses) were first conducted, followed by multiple linear and logistic regression analyses that controlled for age, sex, education, number of comorbid conditions, and duration of RA.

For a secondary set of analyses, improvement scores were computed. Improvement was defined for the HAQ score, pain rating, and numbers of painful and swollen joints as a change from baseline by 0.5 SDs or more, a proxy for clinically meaningful improvement (12). For the 2 binary symptom measures, morning stiffness and fatigue, improvement was defined as moving from the more severe group (i.e., severe or very severe fatigue, morning stiffness of ≥1 hours) to the less severe group (i.e., no fatigue or mild or moderate fatigue, morning stiffness of <1 hour). Chi-square analyses were conducted to determine whether there were differences in the frequency of improvement among the treatment groups. Multiple logistic regression analyses were then performed to determine if differences among the groups existed after controlling for age, sex, education, number of comorbid conditions, and RA duration. Some improvement might be expected even in individuals with no changes in treatment over the study period as a result of normal fluctuations in disease or gradual responses to therapy. Therefore, the frequency of improvement in the no use group is viewed in these analyses as a background rate of improvement.

RESULTS

Overall, 84% of the subjects were women, the mean age was 61 years, the mean duration of RA was 21 years, and 42.9% of the subjects had at least 1 comorbid condition (Table 1).

Table 1. Characteristics of the study sample*
 Total (n = 289)Use of biologic therapyP
None (n = 183)Discontinued (n = 42)Continuous (n = 64)
  • *

    Values are the mean ± SD unless otherwise indicated.

  • From chi-square analysis or t-tests comparing the no biologic therapy group with the 2 biologic therapy groups.

  • Significantly different from the no biologic therapy group (from the post hoc means test).

Women, no. (%)241 (84.0)153 (83.6)33 (78.6)55 (85.9)0.88
Age, years61.2 ± 12.162.5 ± 11.158.7 ± 13.659.1 ± 13.20.052
Education, years13.7 ± 2.613.7 ± 2.613.8 ± 3.013.8 ± 2.20.95
Comorbidities, no. (%)    0.97
 0165 (57.1)103 (56.3)25 (59.5)37 (57.8) 
 188 (30.5)56 (30.6)11 (26.2)21 (32.8) 
 ≥236 (12.5)24 (13.1)6 (14.3)6 (9.4) 
Disease duration, years20.7 ± 9.420.9 ± 9.420.7 ± 8.920.3 ± 10.00.93

Individuals who received biologic therapies were younger than those who did not (58.7 years [discontinued use] and 59.1 years [continued use] versus 62.5 years; P = 0.05), but there were no significant differences among the groups in sex, education, number of comorbid conditions, or disease duration.

Use of etanercept was more common than use of infliximab. Overall, 56 individuals reported taking etanercept only, 34 reported taking infliximab only, and 16 reported taking both biologic agents over the study period. Among the continuous use group, 61% took etanercept only, 31% took infliximab only, and 8% took both biologic agents. In contrast, among the discontinued use group, 40% took etanercept only, 33% took infliximab only, and 26% took both.

Baseline.

In bivariate analyses, at baseline, function was significantly worse for both biologic therapy groups compared with the comparison group (HAQ score 1.21 versus 0.90; P = 0.001) (Table 2). Other symptom measures (number of painful joints, number of swollen joints, and duration of morning stiffness) were also significantly worse in the biologic therapy groups with the exception of fatigue, for which no differences were noted, and severity of pain rating, for which the difference was marginal. Adjustment for age, sex, education, number of comorbidities, and duration of RA did not substantively change the results from the baseline comparisons.

Table 2. Year-to-year differences between the 2 biologic therapy groups and the no biologic therapy group*
Use of biologic therapyBivariate analysisAdjusted analysis
BaselineYear 1Year 2BaselineYear 1Year 2
  • *

    Values are the mean ± SD unless otherwise indicated. HAQ = Health Assessment Questionnaire.

  • Adjusted means/percentages (P [from multiple regression, with the no biologic therapy group as the reference]), adjusting for age, sex, education, number of comorbidities at baseline, and duration of rheumatoid arthritis. For duration of morning stiffness and severe/very severe fatigue, P values are from multiple logistic regression analyses, controlling for the same covariates.

  • Significantly different from the no biologic therapy group (from post hoc means test).

  • §

    From analysis of variance comparing treatment groups.

  • From chi-square analysis comparing treatment groups.

HAQ score      
 None0.90 ± 0.680.92 ± 0.680.92 ± 0.670.890.900.89
 Continuous1.21 ± 0.721.16 ± 0.721.20 ± 0.731.22 (0.0004)1.18 (0.002)1.22 (0.0004)
 Discontinued1.21 ± 0.601.15 ± 0.621.18 ± 0.601.22 (0.0024)1.15 (0.02)1.20 (0.005)
 P§0.0010.020.005   
Severity of pain rating      
 None25.1 ± 30.422.3 ± 27.025.2 ± 27.124.621.925.1
 Continuous33.1 ± 26.328.0 ± 24.330.5 ± 26.732.9 (0.05)28.3 (0.10)30.8 (0.14)
 Discontinued34.4 ± 25.831.5 ± 28.736.5 ± 28.133.0 (0.09)31.2 (0.05)37.0 (0.01)
 P§0.060.070.04   
Number of painful joints/  joint groups      
 None3.4 ± 4.43.2 ± 4.13.9 ± 4.33.43.23.9
 Continuous5.2 ± 4.25.2 ± 4.34.8 ± 4.35.2 (0.006)5.1 (0.002)4.8 (0.18)
 Discontinued6.3 ± 5.15.5 ± 5.35.9 ± 4.66.1 (0.0005)5.4 (0.003)5.9 (0.008)
 P§0.00010.00050.03   
Number of swollen joints/  joint groups      
 None2.2 ± 3.11.8 ± 2.82.3 ± 3.12.21.82.3
 Continuous3.3 ± 3.22.5 ± 2.62.8 ± 3.23.2 (0.02)2.4 (0.11)2.8 (0.23)
 Discontinued4.5 ± 3.73.3 ± 3.43.5 ± 3.44.4 (<0.0001)3.2 (0.004)3.5 (0.02)
 P§0.00010.0060.06   
Duration of morning stiffness  >1 hour, no. (%)      
 None42 (23.0)36 (19.7)33 (18.0)22.919.317.9
 Continuous22 (34.4)10 (15.6)10 (15.6)34.2 (0.07)16.1 (0.53)15.9 (0.70)
 Discontinued18 (42.9)10 (23.8)12 (28.6)40.6 (0.02)20.8 (0.81)30.4 (0.08)
 P0.020.570.16   
Severe/very severe fatigue,  no. (%)      
 None39 (21.3)32 (17.5)35 (19.1)21.117.418.7
 Continuous11 (17.2)17 (26.6)13 (20.3)17.4 (0.52)26.4 (0.10)20.6 (0.72)
 Discontinued11 (26.2)13 (31.0)10 (23.8)25.7 (0.49)28.0 (0.11)26.0 (0.27)
 P0.540.080.70   

Followup.

At the year 1 assessment, there was only a slight difference in the mean ± SD length of time receiving biologics for the discontinued use and continuous use groups (4.6 ± 3.3 months versus 6.1 ± 3.5 months). The biologic therapy groups still reported worse symptoms than the nonusers, with significant differences between the no use group and the 2 biologic therapy groups in HAQ score, number of painful joints, and number of swollen joints, and a marginal but not statistically significant difference in pain severity. However, there was no longer a statistically significant difference between the treatment groups in duration of morning stiffness.

At the year 2 assessment, the mean ± SD total length of time receiving biologics was 8.3 ± 5.6 months for the discontinued use group and 17.0 ± 4.7 months for the continuous use group. Bivariate analyses revealed that significant differences remained between both the biologic therapy groups and the no biologic therapy group in HAQ score. There were no significant differences between the continuous use group and the no use group in any of the symptom measures. The discontinued use group, however, exhibited significantly greater pain severity and more painful joints than the no use group. Adjustment for age, sex, education, comorbidities, and duration of RA yielded slightly different results in that the number of swollen joints was also significantly greater in the discontinued use group than in the no use group.

Improvement scores.

At year 1, after adjusting for covariates, individuals in the continuous use group were significantly more likely than those in the no use group to exhibit improvement from baseline in pain severity rating (35.3% versus 21.5%; P = 0.03), number of swollen joints (43.5 versus 21.1; P = 0.001), and duration of morning stiffness (25.7% versus 9.3%; P = 0.002) (Table 3). The discontinued use group, which had a similar amount of treatment time at year 1, also exhibited a significantly greater proportion of individuals who improved in HAQ score (29.1% versus 15.4%; P = 0.04), number of swollen joints (47.3% versus 21.1%; P = 0.001), and duration of morning stiffness (22.0% versus 9.3%; P = 0.04) compared with the no use group. The proportions of individuals who had improvements in number of painful joints and fatigue rating did not differ significantly between the groups.

Table 3. Improvement from baseline
Use of biologic therapyBivariate analysis*Adjusted analysis
Year 1Year 2Year 1Year 2
  • *

    Values are the percent who met the improvement criterion for each outcome.

  • Adjusted percentages (P value) calculated from multiple linear regression analyses, adjusting for age, sex, education, number of comorbidities at baseline, and duration of RA. P values are from multiple logistic regression analyses, controlling for the same covariates.

  • From chi-square analysis comparing the 2 biologic treatment groups and the no biologic treatment group.

Health Assessment Questionnaire    
 None15.912.615.412.3
 Continuous23.421.922.2 (0.23)21.4 (0.08)
 Discontinued28.622.529.1 (0.04)22.2 (0.12)
 P0.110.11  
Severity of pain rating    
 None21.921.621.521.3
 Continuous35.932.835.3 (0.03)31.8 (0.10)
 Discontinued31.017.527.0 (0.47)16.7 (0.52)
 P0.070.12  
Number of painful joints/joint groups    
 None22.416.421.716.2
 Continuous31.332.829.9 (0.17)33.4 (0.004)
 Discontinued35.720.032.1 (0.18)20.2 (0.55)
 P0.120.02  
Number of swollen joints/joint groups    
 None21.919.721.118.7
 Continuous43.839.143.5 (0.001)38.4 (0.003)
 Discontinued47.637.547.3 (0.001)36.4 (0.02)
 P0.00020.002  
Duration of morning stiffness >1 hour    
 None9.810.99.310.4
 Continuous26.628.125.7 (0.002)27.3 (0.001)
 Discontinued21.419.122.0 (0.04)14.7 (0.52)
 P0.0030.004  
Severe/very severe fatigue    
 None9.810.99.410.5
 Continuous4.79.44.2 (0.19)9.0 (0.68)
 Discontinued11.914.311.9 (0.70)9.5 (0.80)
 P0.360.73  

At year 2, after adjustment for covariates, the continuous use group was more likely than the no use group to exhibit improvements from baseline in number of painful joints (33.4% versus 16.2%; P = 0.004), number of swollen joints (38.4% versus 18.7%; P = 0.003), and duration of morning stiffness (27.3% versus 10.4%; P = 0.001). In comparison, a significant difference was noted between the discontinued use group and the no use group only for number of swollen joints (36.4% versus 18.7%; P = 0.02).

DISCUSSION

Results suggest that among patients of community-based rheumatologists, treatment with biologic agents was initiated based on severe disease. Individuals with RA who were treated with biologic agents had significantly worse functioning and more severe symptoms prior to initiation of treatment, which has been noted by other community-based studies. For example, Wolfe and Michaud (8) noted that patients who received treatment with biologic agents were younger and had worse baseline HAQ scores and pain.

We found that by year 2, differences between the continuous use group and the nonusers group in pain severity, number of painful joints, and number of swollen joints had narrowed to statistical nonsignificance. For this group of individuals who began the study period with significantly greater symptom levels than the nonusers group, reaching a nonsignificant difference in symptoms may represent an important change, in spite of the fact that function remained significantly worse in the biologic therapy groups.

With regard to HAQ score over time, we found that HAQ score for the no biologic therapy group was fairly stable, whereas for the biologic therapy groups the score decreased slightly (year 1) and then returned to baseline levels (year 2). The different patterns are worth noting for 2 reasons: first, for the no biologic therapy group, an increase of the HAQ score on average was smaller than the previously reported 0.02–0.03 per year by Welsing et al (13). Second, for the biologic therapy groups, the change in HAQ score (decreased at year 1 and then regressed to baseline level at year 2) demonstrated the effect of biologic treatment on function despite the fact that in patients with late RA (average disease duration of 20 years in this study), functional capacity is most associated with joint damage (13). The results are comparable with early findings that patients with established RA exhibit less improvement in HAQ score after initiation of biologic therapy (14).

The secondary analyses focusing on improvement showed that after initial treatment, individuals in the biologic treatment groups were more likely to demonstrate meaningful improvement, defined as a decrease in symptom severity rating of 0.5 SDs or more. It may not be surprising that individuals who received new treatments (biologic therapy groups) experienced more improvements than individuals who did not (nonusers group). However, we assumed that even in the nonusers group, some percentage of individuals would experience improvement. We compared the frequency of improvement of the biologic therapy groups with that of the nonusers group, considering the frequency of the nonusers group as a background rate of improvement. At the year 1 assessment, when the time receiving biologic therapy was similar for the discontinued use and continuous use groups, greater proportions of both biologic therapy groups exhibited improvement from baseline in number of swollen joints and duration of morning stiffness. By the year 2 assessment, however, when those in the discontinued use group had stopped taking the biologic agents, a significantly greater proportion of the continuous use group exhibited meaningful improvements in number of painful joints, number of swollen joints, and duration of morning stiffness compared with the no use group. In contrast, the discontinued use group exhibited a greater improvement only in the number of swollen joints.

We found no differences in the prevalence of severe fatigue between the groups at any point, which is consistent with the findings reported by Wolfe and Michaud (15). In clinical trial patients, Moreland et al noted improvements in fatigue among those treated with biologic agents (6). Farahani et al also noted differences in fatigue between patients treated and not treated with biologic agents after 6 months of treatment; at 12 months, however, those differences had disappeared (7). Differences in findings may be attributed to differences in the fatigue measure (e.g., single-item versus 4-item vitality battery from the Short Form 36) (16) or in the patient populations (e.g., community-based sample versus clinical trial cohorts, differences in disease duration). For example, Moreland et al found that fewer individuals with established RA achieved clinically meaningful improvement than individuals with early RA (6).

This study has important strengths and limitations to consider. The 2-year followup period of the current study is longer than most clinical trials examining the outcomes of biologic therapies. Data were obtained from a cohort of individuals with RA recruited from community-based rheumatologists rather than a clinical trial population, which should enhance the heterogeneity of the study sample and broaden the generalizability of the results. However, the mean duration of RA was 20 years, which may limit the ability to generalize results to individuals with early-onset RA. All subjects were recruited from rheumatology practices and therefore may be different from individuals who do not obtain care from rheumatologists. This limitation may be outweighed by the diagnostic certainty resulting from the recruitment source.

There is a potential for bias in both reports of treatments and symptoms; however, the symptom report measures used have been well validated, and in the past, reports of utilization have closely corresponded with utilization noted in medical records. Although we used a fairly standard method to estimate clinically meaningful improvement (12), our definition may have lacked precision. We do not have information regarding the reasons that patients did not start a biologic agent. It is possible that physicians may have wanted to prescribe such agents for some individuals, but access or payment issues precluded use. Finally, we do not have information as to why individuals in the discontinued use group stopped treatment, although the reason for discontinuation (e.g., access versus side effects) could have resulted in variations in outcomes.

In a cohort of individuals recruited through community-based rheumatologists, persons with RA who were selected for treatment with biologic agents had significantly worse functioning and more severe symptoms, suggesting that biologics were being reserved for individuals with more severe disease. Over 2 observation periods (an average length of treatment of 17 months) following initiation of biologic therapy, some but not all of the differences in symptoms between individuals treated continuously with biologic agents and nonusers narrowed and no longer reached statistical significance, representing what is probably an important change in symptoms. In contrast, individuals who initiated treatment with biologic agents but did not continue treatment maintained high levels of symptoms at the second followup period. In addition, individuals who continued use of biologic agents over the 2-year study period were significantly more likely to demonstrate meaningful improvement in most symptom measures than those who did not take biologic agents, whereas individuals who initiated biologic treatment but did not continue treatment did not achieve the same rates of improvement.

This study shows that although individuals selected for treatment with biologic agents exhibited more severe disease characteristics prior to initiation of therapy, those who received extended treatment with such therapy experienced a reduction in symptoms, leading to a narrowing of differences compared with individuals who did not receive biologic treatment. Individuals who received biologic agents for a shorter period of time did not achieve the same results and their symptoms remained significantly more severe; however, the reasons for treatment discontinuation are unknown.

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 submitted for publication. Dr. Katz 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. Katz, Yelin, Chiou.

Acquisition of data. Katz, Yelin.

Analysis and interpretation of data. Katz, Yelin, Patel, Huang, Chiou.

ROLE OF THE STUDY SPONSOR

Amgen had no role in the data collection. Co-authors from Amgen collaborated in the design of the study analyses and in the writing of the manuscript. All authors approved the final manuscript. The manuscript was reviewed by Amgen, but publication was not contingent upon the approval of Amgen.

Ancillary