Dr. Lewis has received consultancy fees, speaking fees, and/or honoraria (less than $10,000 each) from Abbott, Amgen, Eli Lilly, Janssen, Millennium Pharmaceuticals, Nestle, Pfizer, Prometheus, and Shire; has served on a Data and Safety Monitoring Board for clinical trials sponsored by Pfizer; has received research support from Shire, Centocor, and Takeda; and has provided expert opinion related to legal matters on behalf of Roche.
Jeffrey R. Curtis
University of Alabama, Birmingham
510 20th Street South, Faculty Office Towers 802 D, Birmingham, AL 35294. E-mail: email@example.com
Dr. Curtis has received consultancy fees, speaking fees, and/or honoraria (less than $10,000 each) and research grants from AbbVie, Bristol-Myers Squibb, Crescendo, and Pfizer, and fees (more than $10,000 each) and research grants from Amgen, the Consortium of Rheumatology Researchers of North America, Janssen, Roche/Genentech, and UCB.
Self-injectable biologic agents have been covered by Medicare Part D since 2006. We hypothesized that this coverage benefit and related financial considerations would lead to increased use of self-injectable biologic agents over time and would be influential in determining which rheumatoid arthritis (RA) patients received infusion versus self-injected biologic agents.
We used 100% of US Medicare data (2006–2009) to calculate the prevalence of use of different RA biologic agents and evaluated factors associated with receipt of infliximab versus etanercept or adalimumab among patients starting their first anti–tumor necrosis factor (anti-TNF) agent.
Through 2009, the prevalence of biologic agent use overall (∼27%) and via infusion (16–17%) or self-injection (10–11%) remained unchanged among Medicare beneficiaries with RA. After adjusting for patient characteristics, stronger physician preference for infused biologic agents was related to physician reimbursement and associated with an increased likelihood of using infliximab as the first anti-TNF agent (odds ratio [OR] comparing the highest to lowest quartile of physician preference 7.3, 95% confidence interval [95% CI] 6.4–8.3). Lower-income patients who received state assistance for Medicare coverage had lower out-of-pocket payments for injectable biologic agents ($4.10/prescription) and were less likely to use infliximab (OR 0.41, 95% CI 0.37–0.45) compared with etanercept or adalimumab.
The prevalence of injection and infusion biologic agents in RA remained stable in the Medicare program through 2009. The choice between an intravenous infusion versus an injectable anti-TNF agent as first-line treatment appeared to be strongly influenced by financial considerations affecting both patients and physicians.
Early and aggressive use of biologic and nonbiologic disease-modifying antirheumatic drugs (DMARDs) is recommended for the treatment of active rheumatoid arthritis (RA). Underutilization of these medications has been reported repeatedly and is associated with a number of non–disease-related factors ([1-5]). For example, in an investigation of RA patients enrolled in Medicare managed care plans, older age (≥85 years), male sex, African American race, low income, and enrollment in for-profit health plans were all found to be associated with a lower rate of DMARD utilization (). Not visiting a rheumatologist is another factor consistently associated with a reduced likelihood of receiving DMARDs ([2, 3, 5]). At least in some RA patient populations, we and others have suggested that there may be an increasing use of DMARDs and some biologic agents over calendar time, a trend that may reflect the increased availability of new agents and greater physician and patient familiarity with novel mechanisms of action ([6, 7]). Greater impetus to achieve low disease activity goals (i.e., treat-to-target approaches) may also motivate increased interest in using biologic agents ([2, 6, 7]), although whether these forces have had any meaningful impact on the overall prevalence of biologic agent use in the US is not clear. Additionally, although biologic agents are typically given in conjunction with nonbiologic DMARDs, monotherapy with biologic agents is of increasing interest and the size of the patient population that receives monotherapy with biologic agents has not been well characterized in a population-based US cohort of RA patients.
Because biologic agents are costly, the use of these medications is likely to be strongly influenced by health plan coverage (). Prior to 2006, RA patients with traditional Medicare fee-for-service coverage had very limited access to self-administered (i.e., subcutaneously injected) biologic agents unless they had a secondary insurance plan that covered prescription medications. As a result, infusion biologic agents were preferentially used because they were covered under Medicare Part B ([8, 9]). As expected and demonstrated in a US study of 1,663 RA patients, those with public insurance (Medicaid or Medicare) were more likely to receive infliximab than etanercept compared to those with commercial insurance, whereas RA-related characteristics such as measures of disease activity were not associated with infliximab use (). Beginning in 2006, the availability of Medicare Part D coverage for prescription medications was expected to increase access to subcutaneously injected, self-administered biologic DMARDs among RA patients, although the true impact of this coverage is not known.
The present study examined the patterns of biologic DMARD use using data from 100% of the US Medicare population with fee-for-service coverage from 2006–2009. Our specific aims were to 1) examine the prevalence of biologic agent use among Medicare beneficiaries with RA; 2) evaluate factors associated with receipt of an infused or self-injectable biologic agent among patients initiating anti–tumor necrosis factor (anti-TNF) therapy, with a focus on physician preference for intravenously administered infused biologic agents and patient-related costs; and 3) assess the prevalence of use of monotherapy with biologic agents, comparing anti-TNF switchers to those who were naive to therapy with biologic agents.
Box 1. Significance & Innovations
Using national US data, we found that infusion biologic agents (16% prevalence) were used more than self-injectable biologic agents (11% prevalence) among rheumatoid arthritis patients enrolled in Medicare with drug benefits.
Higher out-of-pocket payment for self-injectable biologic agents was associated with stronger preference for infusion biologic agents.
Physician preference was significantly associated with the choice between infusion and self-injectable biologic agents after controlling for patient characteristics.
Physicians who billed for infusion biologic agents had an increased preference for use of infusion biologic agents.
PATIENTS AND METHODS
We selected 2 cohorts of RA patients for the current study from 100% of US Medicare beneficiaries diagnosed with RA from 2006 through 2009. Cohort 1 (a closed cohort) included all RA patients who had continuous fee-for-service coverage (enrolled in Medicare Part A and Part B and not in a Medicare Advantage Plan) with at least 2 RA diagnosis codes from a rheumatologist in 2006 and who also had Medicare Part D pharmacy coverage beginning January 1, 2007. The start of followup for all patients in this cohort was January 1, 2007, and patients were allowed to remain in the cohort until they lost fee-for-service or Medicare Part D coverage, until they died, or until December 31, 2009. No new patients were allowed to enter the cohort. The purpose of this closed cohort was to select a fixed and relatively stable group of RA patients (with respect to Medicare coverage) to estimate the prevalence of biologic agent use. The requirement for RA diagnoses to come from rheumatologists was intended both to improve the certainty of the RA diagnosis () as well as to ensure that patients who might be in need of a biologic agent were being treated by physicians who likely would be comfortable prescribing these agents.
Patients in cohort 2 (an open or dynamic cohort) were required to have continuous Medicare fee-for-service and Medicare Part D coverage for at least 12 consecutive months during the period from 2006 through 2009 and 2 International Classification of Diseases, Ninth Revision, Clinical Modification RA diagnosis codes (714.x) from rheumatologist outpatient visits that occurred between 7 and 365 days apart. The followup period could start only after both eligibility criteria were satisfied and ended when the beneficiary lost fee-for-service or Medicare Part D coverage, when the beneficiary died, or on December 31, 2009. Furthermore, cohort 2 consisted of patients who initiated a biologic agent during followup and whose initiation date of therapy with biologic agents was immediately preceded by at least 12 months free of exposure to that specific medication, which is a new-user design (). These new users were further categorized as biologic free (received no biologic agent of any kind in the previous 12 months) or switchers (received at least 1 different biologic agent in the previous 12 months). A treatment episode began on the date of first use of each specific biologic medication, defined as the index date. Cohort 2 allowed the examination of patterns of biologic agent use among new users rather than prevalent users of particular biologic agents. Patients in cohort 2 were allowed to contribute >1 treatment episodes to the analysis examining the use of concomitant nonbiologic DMARDs. The analysis examining factors associated with the choice between injection or infusion biologic agents was further limited to new users who initiated one of the 3 most commonly used anti-TNF therapies (etanercept, infliximab, or adalimumab). The analysis was also limited to the first episode of new use of anti-TNF therapy during the study period.
Exposure to RA medications
For each day during followup, exposure to biologic agents and nonbiologic DMARDs was determined based on the records of filled prescriptions identified using National Drug Codes or records of infusions received identified using Healthcare Common Procedure Coding System (HCPCS) codes (for pharmacy-filled medications). Administration of new infusion biologic agents typically is recorded in Medicare claims using a nonspecific HCPCS code (e.g., J3490, J3590) for ∼1 year after licensure. Therefore, we identified administration of abatacept and certolizumab administered in 2006 and early 2007 (abatacept) and 2009 (certolizumab) using the nonspecific HCPCS codes coupled with an RA diagnosis code on the claim, an appropriate number of units dispensed (e.g., 500, 750, or 1,000 for abatacept or 200 or 400 for certolizumab), and the submitted or allowable per-unit costs within the appropriate range.
Prevalence of biologic DMARD utilization over calendar time
Beginning on January 1, 2007, among patients in cohort 1, we examined the proportion of patients exposed to each biologic agent in every 3-month period of followup. Patients were considered exposed to biologic agents if they had at least 1 infusion or filled prescription during each 3-month period. Patients contributed data to multiple 3-month intervals as long as they were under followup throughout each 3-month period. In addition, we examined the prevalence of biologic agent use among patients treated by physicians practicing at offices with infusion suites, defined as physicians who billed for infusion under their own provider numbers in the prior year.
Prevalence of concomitant DMARD utilization among new biologic agent users
The prevalence of concomitant DMARD use was assessed among new biologic agent users in cohort 2, stratified by which biologic agent was initiated and by whether the biologic agent users were biologic free or switchers. Concomitant DMARD use was defined as ≥1 filled prescription for a nonbiologic DMARD within 121 days after initiation of therapy with biologic agents. The following hierarchy was applied to classify nonbiologic DMARD use: any oral methotrexate (MTX); any subcutaneous MTX; any use of sulfasalazine, hydroxychloroquine, or leflunomide; any other nonbiologic DMARDs (e.g., azathioprine and minocycline); or none of these medications. In addition, we performed a sensitivity analysis using a 180-day window to examine the use of concomitant DMARDs.
Factors associated with choice of injection or infusion biologic agents among new anti-TNF users
Among the biologic-free RA patients in cohort 2, we identified those who initiated the most commonly used self-injectable, subcutaneously administered biologic agents (etanercept or adalimumab) or an infusion biologic agent (infliximab). The use of other injectable anti-TNF agents (i.e., certolizumab or golimumab) as first-line agents was too rare to examine. We evaluated the factors possibly associated with the receipt of etanercept/adalimumab versus infliximab, including demographics, concomitant glucocorticoids and nonbiologic DMARDs, calendar year, physician preference measured in the 6 months preceding each initiation, and whether the patient received any state-related financial assistance to help pay their Medicare Part B premiums in the previous 6 months, which is a proxy for low-income patients and government cost-sharing subsidies that make Medicare Part D medications more affordable for such patients. In addition, we examined the amount of the patients' out-of-pocket copayments for injectable biologic agents by comparing patients receiving state-related financial assistance to those who did not receive such assistance to help pay for their Medicare premium.
Because physician preference and many of the patient demographics could be strongly associated with the type of insurance, and because in a prior study stratification by type of insurance led to mitigation of the associations observed between patient characteristics and the physician prescribing preference (), we performed a stratified analysis by whether the patient received any state financial assistance for payment of Medicare Part B premiums. In addition, we assessed the direct association between physicians billing for infusion biologic agents and the choice of infusion versus injectable biologic agents among all and in both subgroups of patients by their receipt of state assistance for Medicare Part B premiums.
Physician preference was quantified for each physician in 6-month calendar blocks from 2006–2009 as the percentage of their RA patients treated with any infusion biologic agent for RA divided by the number of RA patients treated with a biologic agent in each physician's practice during that 6-month interval. For example, a physician who treated 8 patients with infliximab (or another infused biologic agent including abatacept or rituximab) of 20 total patients treated with biologic agents in that 6-month period would be represented as having 40% of their patients receiving an infused biologic agent. The prescribing physician was identified as the most recent physician visit with a diagnosis code for RA within 180 days prior to the initiation of therapy with biologic agents. If there was >1 such visit, preference was given to rheumatology providers as identified by the specialty field on each Medicare claim. Using this approach, physician visits with a rheumatologist and/or with an RA diagnosis could be found for 97% of all anti-TNF initiations. These percentages describing the physician preference for infused biologic agents were updated at 6-month intervals and grouped into time-varying quartiles. Because estimating physician preference for intravenously administered biologic agents might be unstable if physicians had very few RA patients in each 6-month period, we required physicians to have at least 5 RA patients in each interval to group their preference into quartiles. If the physicians did not meet this requirement, their preference was categorized separately rather than in a quartile. This time-varying quantification of physician preference allowed for the preference to change over time. The preference was lagged by one 6-month calendar interval so that a physician's preference to use intravenous or injectable therapy with biologic agents in 1 period was used as the predictor variable for the next 6-month period.
In addition, we examined the association between physician preference (dependent variable) for infused biologic agents and whether the physician billed for infusions under his or her own provider number (rather than sending the patient to an outpatient infusion center). This was used to correlate with financial incentives that could in part motivate a physician to select infliximab as the first-line anti-TNF agent for an RA patient.
Logistic regression was used to determine patient and physician characteristics associated with the receipt of injection versus infusion biologic agents among patients initiating anti-TNF therapy who were biologic free in the 12 months prior to anti-TNF initiation. Within-physician clustering was not adjusted for in the model because physician preference was of interest and adjusting for physician-level clustering would have made it difficult to describe the effect of physician preference. All analyses were conducted using SAS, version 9.3. The study protocol was approved by the University of Alabama at Birmingham Institutional Review Board. Data Use Agreements from the Centers for Medicare and Medicaid Services governed the use of the data.
We identified 79,883 RA patients who satisfied both RA and Medicare coverage criteria in 2006 for inclusion in cohort 1. The characteristics of these patients are shown in Table 1. The mean ± SD age was 70.1 ± 11.5 years, and ∼80% were women. In total, 29% of patients received some form of state assistance to help subsidize Medicare premiums.
Table 1. Baseline characteristics of rheumatoid arthritis patients in cohort 1*
Total (n = 79,883)
Values are the number (percentage) unless otherwise indicated. Baseline characteristics ascertained in the preceding 365 days.
Age, mean ± SD years
70.1 ± 11.5
Receiving state assistance to pay Medicare premiums
Charlson comorbidity score
Chronic pulmonary disease
Ischemic heart disease
Health services utilization
No. of physician visits, mean ± SD
19.2 ± 14.4
Through 2009, the proportions of prevalent users of any biologic agents (∼27%), self-injectable biologic agents (∼11%), and infusion biologic agents (∼16%) remained essentially unchanged among RA patients (Figure 1A), although use of some specific agents changed over time. Among patients treated by physicians practicing at offices with infusion suites, the prevalence of infusion biologic agent use was higher (19–20%) than that observed among all RA patients throughout the study period, whereas the prevalence of self-injectable biologic agents was slightly lower (9–10%). Among infusion biologic agents, the prevalence of infliximab use declined from 13.8% in the first quarter of 2007 to 11.9% in the fourth quarter of 2009 (Figure 1B). In contrast, the prevalence of abatacept increased from 1.5–3.7% during this time.
We identified a total of 31,270 treatment episodes in cohort 2; 17,662 free of any biologic agent use in the previous year and 13,608 switched from a different biologic agent. After limiting to etanercept, infliximab, and adalimumab, we identified 12,360 treatment episodes (70%), and after further limiting the data to the first treatment episode for each patient, we identified 11,966 unique RA patients (Table 2). The factors significantly associated with the use of infliximab rather than etanercept or adalimumab included a stronger physician preference for intravenous therapies of biologic agents (odds ratio [OR] for the highest to the lowest quartile of physician preference for infused biologic agents 7.31, 95% confidence interval [95% CI] 6.41–8.33). The patients who received state government financial assistance were less likely to use infliximab (OR 0.41, 95% CI 0.37–0.45). Older age and use of concomitant MTX were significantly associated with a greater likelihood to use infliximab rather than subcutaneous anti-TNF therapy. Among patients who used oral MTX prior to initiating anti-TNF therapy, the median dosage of oral MTX for the last filled MTX prescription was 17.5 mg per week (interquartile range 12.5–20.0).
Table 2. Patient characteristics associated with receipt of etanercept or adalimumab versus infliximab among biologic-free new users (cohort 2; n = 11,966 RA patients)*
Etanercept or adalimumab (n = 6,657)
Infliximab (n = 5,309)
Odds ratio (95% confidence interval) for receipt of infliximab
Unadjusted (n = 11,966)
Multivariable adjusted (n = 11,966)
With state assistance (n = 4,901)
Without state assistance (n = 7,065)
All characteristics were assessed in the 12 months prior to the start of the new anti–tumor necrosis factor agent, unless otherwise indicated. RA = rheumatoid arthritis; N/A = not applicable; DMARD = disease-modifying antirheumatic drug.
aEstimated in the 6 months prior to the index date.
Physician preference for infusion biologic agent, %
Quartile 1 (ref.)
Calendar year, %
Across the lowest (first) to highest (fourth) quartile of physician preference for intravenously infused biologic agents, the proportions of physicians who personally billed for infusions under their own provider number were 34%, 59%, 80%, and 88%, respectively. The average out-of-pocket patient payment for each filled prescription for etanercept or adalimumab among patients with government assistance was $4.10 and was $424.63 among those who did not receive state financial assistance. A subgroup analysis by receipt of government assistance did not alter the association we observed between physician preference and the choice of infusion versus injection biologic agents (Table 2). A direct assessment of the association between billing for infusion and choice of infusion versus injection biologic agents found that physicians who billed for infusion were more likely to initiate infusion biologic agents among all patients (OR 1.3, P > 0.001), including those who received state assistance (OR 1.2, P = 0.01) and those who did not (OR 1.3, P > 0.001).
Approximately 32% of biologic-free new users received monotherapy with biologic agents (i.e., no DMARDs such as MTX); the proportion ranged from 24% (infliximab) to 51% (certolizumab) (Figure 2A). Infliximab users had the highest proportion of patients receiving concomitant oral MTX (53%) compared with users of other biologic agents. A similar but slightly greater proportion of biologic agent switchers received biologic agents as monotherapy (36%), ranging from a low of 30% (infliximab) to a high of 44% (certolizumab) (Figure 2B). Results from the sensitivity analysis utilizing a 180-day window to identify monotherapy with biologic agents were consistent with those in the primary analysis; 29% and 32% of biologic-free new users and biologic agent switchers received monotherapy with biologic agents, respectively.
The findings from our study answer a number of important questions related to the use of and access to biologic agents for Medicare beneficiaries with RA. First and foremost, our data suggest that because even among beneficiaries with Medicare Part D, infusion biologic agents were still preferred over self-injectable biologic agents. Using 100% of Medicare data for RA patients, the prevalence of injection (10–11%) and infusion (16–17%) biologic agents remained essentially unchanged from 2007–2009; anti-TNF biologic agents (etanercept, adalimumab, and infliximab) were the most commonly used biologic agents for the treatment of RA. Among the various anti-TNF agents, infliximab has been found to be the least cost effective for the treatment of RA ([12, 13]). However, the preferential coverage of infused therapies under Medicare Part B has resulted in greater use of infliximab among Medicare beneficiaries compared to those enrolled in commercial insurance plans (). Infliximab also accounts for a very large expenditure to the Medicare program. In 2009, it ranked fourth among all Medicare Part B–covered medications in terms of percent share of national Medicare expenditures, accounting for 4% of $11.1 billion ().
Cost sharing has been found to play an important role in the use of health services, including prescription medications among RA patients (). Our data suggest that high out-of-pocket patient copayment was an important financial barrier and limited the impact of the Medicare Part D program on increasing the accessibility of self-injectable, subcutaneously administered biologic agents. We found that patients receiving state financial assistance were more likely to use self-injectable biologic agents, which is consistent with the greatly reduced out-of-pocket copayment accompanying a subcutaneous anti-TNF prescription ($4.10 versus $424.63 for those not receiving financial assistance).
In addition to patient copayments, we found that physician preference was significantly associated with the receipt of infusion biologic agents after adjusting for a number of other patient characteristics. Physician preference for infusion therapies was also strongly related to financial considerations; 88% of physicians in the highest quartile of preference for intravenous RA drugs were billing for infusion therapies, whereas only 34% of physicians in the lowest quartile of preference for intravenous therapies billed themselves for administration of these agents. To address the concern that this result may be partially explained by physicians having somewhat different case mixes of RA patients in their practice, we performed a subgroup analysis by receipt of state assistance, which was the patient characteristic most strongly associated with the outcome. In each subgroup of patients, we observed the same association between physician preference and choice of infusion versus self-injectable biologic agents. This suggests that financial and reimbursement considerations play an important role in selecting the first-line anti-TNF agent given to patients. These findings also suggest that physicians and patients are responsive to the coverage benefits provided by the Medicare program. In the future, alteration of Medicare benefits for intravenous and injectable biologic agents may meaningfully impact the choice of which drugs are given in what sequence to RA patients.
We found that approximately one-third of RA patients initiating or switching to biologic agents were not receiving MTX or other nonbiologic DMARDs and were receiving monotherapy with biologic agents. The rates were similar to the rate reported in a prior study (30%) conducted among RA patients enrolled in private and public plans between 1999 and 2004 (). Among MTX-naive patients, data from randomized clinical trials have shown that monotherapy with anti-TNF biologic agents was not superior to MTX monotherapy for improvement in RA signs and symptoms ([16-18]). Moreover, infliximab, rituximab, and golimumab are only approved for use with concomitant MTX by the US Food and Drug Administration. Recent trials have shown that newer agents administered as monotherapy may be superior to both MTX monotherapy or even anti-TNF monotherapy ([19-21]). The large number of RA patients receiving monotherapy with biologic agents suggests a need for additional research focusing on optimizing care for these RA patients. Finally, among people initiating therapy with biologic agents who were recently using MTX, the median dosage of oral MTX used was only 17.5 mg per week, suggesting that physicians may not be maximizing MTX before adding a biologic agent, or that MTX is not tolerated at higher doses in this patient population with a mean age of 70 years.
Our study is among the few that have examined temporal trends in the use of biologic agents after the implementation of Medicare Part D for prescription medications. Our findings were derived from and thus applicable to all RA patients with Medicare fee-for-service and Medicare Part D coverage receiving care from a rheumatologist. Medicare beneficiaries with fee-for-service coverage compose 82% of all Medicare beneficiaries in the US (). Our study does have a number of limitations. First, we did not have data on the use of biologic agents prior to the introduction of Medicare Part D in 2006, which prevented us from performing a before and after analysis to assess the prevalence of use of infusion and injection biologic agents before and after the introduction of Medicare Part D. Second, we assigned the treating physician to each RA patient based on a recent visit for RA. As a result, the physicians we identified as prescribing biologic agents may have been misclassified. However, if such misclassification had occurred, our results evaluating the effect of physician preference on the choice of infliximab versus etanercept/adalimumab as first-line treatment should have been biased toward the null (finding no association). Finally, the algorithms we used to identify biologic agents that were initially coded with nonspecific HCPCS codes were not validated, although we are currently in the process of validating them (). Regardless, any misclassification of these codes would have had a trivial effect on our results because these codes were used for a relatively short period of time since the permanent code for abatacept became available January 1, 2007 and only a small number of certolizumab users received this agent through their Medicare Part B medical benefit in the latter half of 2009. Additionally, it is possible that patients with a preference for infusion biologic agents could seek out rheumatologist practices with infusion suites on their premises. This seems unlikely because patients generally would not know whether their doctor has on-site infusion capabilities when selecting a physician. Finally, because the numbers of biologic-free RA patients initiating certolizumab and golimumab were small (Figure 2A), the estimated proportions (40–50%) of patients receiving monotherapy with biologic agents were less stable compared with the corresponding estimates of other biologic agents.
In summary, our study examined the use of biologic agents among Medicare beneficiaries longitudinally from 2007 through 2009, a period marked by the establishment of the Medicare Part D program. We found that while prescribing patterns of biologic agents were multifactorial, the treatment decisions were strongly associated with preferential coverage and reimbursement for infusion biologic agents, which may have created financial pressures from the perspectives of both the patients and physicians to make certain therapies more or less accessible. Despite the availability of Medicare Part D, high out-of-pocket payments likely limited the impact of this insurance program on increasing patient access to self-injectable biologic agents. The proportion of patients receiving monotherapy with biologic agents was high and demonstrates the importance of recognizing this large subgroup of patients in terms of optimizing RA treatment strategies when concomitant therapy with nonbiologic agents is not used.
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. Curtis 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. Zhang, Xie, Delzell, Chen, Curtis.
Acquisition of data. Kilgore, Curtis.
Analysis and interpretation of data. Zhang, Xie, Delzell, Chen, Kilgore, Yun, Saag, Lewis, Curtis.