Abstract
Objective. Baricitinib is an oral, once-daily selective Janus kinase (JAK1/JAK2) inhibitor for adults with moderately to severely active rheumatoid arthritis (RA). We evaluated baricitinib’s safety profile through 288 weeks (up to September 1, 2016) with an integrated database [8 phase III/II/Ib trials, 1 longterm extension (LTE)].
Methods. The “all-bari-RA” group included patients who received any baricitinib dose. Placebo comparison was based on the 6 studies with 4 mg and placebo up to Week 24 (“placebo-4 mg” dataset). Dose response assessment was based on 4 studies with 2 mg and 4 mg including LTE data (“2 mg-4 mg–extended”). The uncommon events description used the non-controlled all-bari-RA.
Results. There were 3492 patients who received baricitinib for 6637 total patient-years (PY) of exposure (median 2.1 yrs, maximum 5.5 yrs). No differences in rates of death, adverse events leading to drug discontinuation, malignancies, major adverse cardiovascular event (MACE), or serious infections were seen for 4 mg versus placebo or for 4 mg versus 2 mg. Infections including herpes zoster were significantly more frequent for 4 mg versus placebo. Deep vein thrombosis/pulmonary embolism were reported with 4 mg but not placebo [all-bari-RA incidence rate (IR) 0.5/100 PY]; the IR did not differ between doses (0.5 vs 0.6/100 PY, 2 mg vs 4 mg, respectively) or compared to published RA rates. All-bari-RA had 6 cases of lymphoma (IR 0.09/100 PY), 3 gastrointestinal perforations (0.05/100 PY), 10 cases of tuberculosis (all in endemic areas; 0.15/100 PY), and 22 all-cause deaths (0.33/100 PY). IR for malignancies (0.8/100 PY) and MACE (0.5/100 PY) were low and did not increase with prolonged exposure.
Conclusion. In this integrated analysis of patients with moderate to severe active RA with exposure up to 5.5 years, baricitinib has an acceptable safety profile in the context of demonstrated efficacy. Trial registration numbers: NCT01185353, NCT00902486, NCT01469013, NCT01710358, NCT01721044, NCT01721057, NCT01711359, and NCT01885078 at clinicaltrials.gov.
Rheumatoid arthritis (RA) treatment goals include controlling synovitis, improving and preserving physical function, and preventing joint damage and disability1. Despite availability of conventional synthetic and biological disease-modifying antirheumatic drugs (bDMARD), many patients do not achieve remission/low disease activity, lose response over time, or have safety or tolerability issues, including infections2,3.
Janus kinase (JAK) inhibitors, which target cytokine signaling pathways implicated in RA pathogenesis4,5,6, offer an alternative treatment option. Baricitinib, an oral selective JAK1 and JAK2 inhibitor, demonstrated clinical efficacy with acceptable safety in phase III trials7,8,9,10.
Comprehensive evaluation across trials with a longer timeframe is necessary to fully understand a drug’s safety profile. Here we characterize the safety profile of baricitinib by pooling data from available RA clinical trials, including a longterm extension (LTE) study.
MATERIALS AND METHODS
Study designs and patients
Patient-level data from 8 randomized clinical trials (4 phase III, 3 phase II, 1 phase Ib) and 1 ongoing LTE trial were included; data went up to September 1, 2016 (Table 17–13). Eligible phase II/III trial patients were aged ≥ 18 years with moderately to severely active RA. Exclusions included current or recent (< 30 days prior to study entry) clinically serious infection requiring antimicrobial treatment [including active or untreated latent tuberculosis (TB)] and selected laboratory abnormalities (hepatic/renal function tests, selected hematology, markers of infection). Evaluated baricitinib doses ranged from 1 to 15 mg daily, with 2 and 4 mg in phase III and the LTE. Trials were conducted in accordance with the Declaration of Helsinki and Good Clinical Practice guidelines, and were approved by each center’s institutional review board (Supplementary Table 1, available with the online version of this article). All patients provided written informed consent.
Patient populations.
Patients completing phase III trials were eligible for the LTE study. Patients randomized to 2 mg and not rescued in the originating study continued taking 2 mg in the LTE; all other patients received 4 mg. Patients receiving 4 mg for at least 15 months without rescue and achieving sustained low disease activity [Clinical Disease Activity Index (CDAI) score ≤ 10] or remission (CDAI score ≤ 2.814) were blindly re-randomized to 4 mg or tapered down to 2 mg. Phase II NCT01185353 patients were also eligible for the LTE and were treated with 4 mg.
Analysis sets were organized as follows: (1) “Placebo-4 mg” was 6 studies with patients randomized to placebo or 4 mg through 24 weeks of treatment. Data censored at rescue or end of placebo-controlled period (“as-treated” analysis); (2) “Placebo-2 mg-4 mg” was 4 studies with patients randomized to placebo, 2 mg, or 4 mg up to 24 weeks of treatment. Data censored at rescue or end of placebo-controlled period; (3) “2 mg-4 mg-extended” was patients from placebo-2 mg-4 mg plus data from extension periods. Data censored at rescue or dose change; (4) “All-bari-RA” was all patients who received at least 1 dose of baricitinib; includes all available data without censoring for rescue or dose change.
Analysis sets 1–3 (“as-treated”) allowed randomized comparison between treatment groups. Primary analyses were based on placebo-4 mg, the largest placebo-controlled set. Dose response evaluation for laboratory abnormalities was based on placebo-2 mg-4 mg, which allows analysis in the short term while avoiding potentially biased dose-response inferences from combining trials with and without a 2-mg arm. Dose response for adverse events (AE) was based on 2 mg-4 mg-extended, which maximizes randomized dose comparison information. Because of the longterm latency period, malignancy [excluding nonmelanoma skin cancer (NMSC)] was also evaluated without censoring for rescue or dose change (“as-randomized” analysis)15. Uncommon events were primarily evaluated using all-bari-RA. Although uncontrolled, data were not censored at dose change or rescue, which provides the largest patient-years (PY) of exposure. Lastly, to evaluate the longterm laboratory profile for DMARD-inadequate responders initially randomized to and maintained on 4 mg (in RA-BEAM, RA-BEACON, RA-BUILD, NCT01185353, and the LTE), a “4-mg longterm cohort” was extracted from all-bari-RA.
Safety data for active comparators through 52 weeks [adalimumab (ADA; RA-BEAM); methotrexate (MTX; RA-BEGIN)] were previously reported9,10, and except for deaths, are not included here.
Assessments
Safety assessments included treatment-emergent AE (TEAE), AE leading to temporary interruption or permanent discontinuation of study drug, serious AE (SAE), AE of special interest, and deaths. An SAE was any event meeting International Conference on Harmonisation E2A seriousness criteria16.
Placebo and each baricitinib group were compared using the Cochran-Mantel-Haenszel test stratified by trial. Differences between doses were evaluated by Mantel-Haenszel incidence rate difference adjusting for trial. Exposure-adjusted incidence rates (EAIR) were calculated as the number of patients with an event per 100 PY of overall exposure time. For AE of special interest, IR were calculated as the number of patients with an event per 100 PY of observation time including any postdrug followup time, with observation time censored at event date.
RESULTS
Patients
Patient demographics including baseline age, steroid use, and disease severity were generally similar across treatment groups within analysis sets (Table 2). Most patients were female (∼80%), with mean age ∼53 years, ∼9 years since RA diagnosis, and moderate or high baseline disease activity (CDAI > 10.0 to 22.0, or > 22.0, respectively)17,18.
Baseline demographics and measures of disease activity.
In all-bari-RA, 3492 patients received ≥ 1 dose of baricitinib for a total of 6636.7 PY exposure. A total of 78% of patients had ≥ 1 year and 54% had ≥ 2 years of treatment with a maximum exposure of 5.5 years (Table 3). Patients were treated for about 400 PY in a placebo-controlled period, and about 187 PY in placebo-2 mg-4 mg. In 2 mg-4 mg-extended, 4 mg had greater PY than 2 mg, and more 4 mg patients had ≥ 1 years of exposure with similar ≥ 2 years (Table 3).
Safety variables of special interest.
AE, including SAE
The SAE EAIR was similar between groups in placebo-4 mg (12.7 vs 12.9) and numerically higher for 4 mg compared to 2 mg in 2 mg-4 mg–extended (13.2 vs 10.1; Table 3). No single system organ class appeared to account for this difference.
Thirty-one deaths occurred in the baricitinib RA program. Two deaths (myocardial infarction) occurred during screening and were not considered related to study procedures. Three deaths occurred among placebo-treated (PY = 393.8, IR 0.76) and 4 deaths among active-comparator–treated patients (PY = 447.4, IR 0.89). Twenty-two deaths (causes of death, Table 3, footnote c) occurred among all baricitinib-treated patients (PY = 6636.7, IR 0.33 (95% CI 0.2–0.5).
The incidence for temporary interruption or permanent discontinuation of study drug due to an AE was numerically higher for 4 mg compared to placebo (Table 3), most frequently due to infections (Table 4). The most common infection leading to discontinuation was herpes zoster, for which all phase III — but not LTE — protocols required discontinuation. Some laboratory changes were reported by the investigator as AE and are included in Table 4; Table 519,20 shows details on select laboratory changes based on central laboratory testing.
Adverse events (AE) detail.
Changes in selected laboratory values and clinical chemistry (weeks 0–24).a
AE of special interest
Infection was the most common TEAE in placebo-4 mg and was more frequent in 4 mg (EAIR 88.4 vs 75.9, 36.3% vs 27.9% of patients). The higher infection EAIR for 4 mg compared to placebo (Table 3) is attributed to a higher incidence of upper respiratory tract infections (EAIR 44.7 vs 39.4, 18.4% vs 14.5% of patients), herpes zoster (4.3 vs 1.0, 1.8% vs 0.4% of patients), and herpes simplex (5.4 vs 2.5, 2.2% vs 0.9% of patients). In 2 mg-4 mg–extended, the TE infection EAIR was comparable between doses. The serious infection IR was similar between treatment groups across sets. The placebo-4 mg IR was 4.2 vs 3.8 (Table 3). The all-bari-RA IR was 2.9 (95% CI 2.5–3.4) and the most common serious infections were pneumonia (n = 36, EAIR 0.5), herpes zoster (n = 26, EAIR 0.4), urinary tract infection (n = 18, EAIR 0.3), and cellulitis (n = 10, EAIR 0.1). Serious infection rates in patients who were taking glucocorticoids at baseline were about twice those seen in those not taking glucocorticoids; importantly, however, this was seen not only for patients receiving baricitinib but also with placebo (data not shown).
Ten TB cases were reported in all-bari-RA (IR 0.15, 95% CI 0.07–0.27; Table 3). All occurred in endemic areas (Argentina, India, Russian Federation, South Africa, South Korea, and Taiwan). One case occurred in the placebo-controlled period, 1 following rescue from placebo to 4 mg, and 8 after LTE entry.
The herpes zoster IR was significantly higher for 4 mg compared to placebo (4.3 vs 1.0) and for 4 mg compared to 2 mg in 2 mg-4 mg–extended (Table 3). The all-bari-RA IR was 3.2 (95% CI 2.8–3.7). All cases were cutaneous. Nearly 9% (18/212) had multidermatomal distribution (> 3 contiguous or ≥ 2 noncontiguous dermatomes), none had visceral involvement, and incidence did not increase over time (data not shown).
Malignancies
In the placebo versus 4 mg analysis, the malignancy (excluding NMSC) IR was comparable between groups. In the 2 mg-4 mg–extended analysis, the 4 mg IR was numerically higher compared to 2 mg with the “as-treated” analysis, while the “as-randomized” analysis showed a similar IR between 2 mg and 4 mg (Table 3). The all-bari-RA IR was 0.8 (95% CI 0.6–1.0), with no increased incidence over time (Figure 1A). The age- and sex-adjusted standardized incidence ratio (SIR) based on Surveillance, Epidemiology, and End Results Program data21 was 1.04 (95% CI 0.79–1.36).
Malignancy-related events over time for all-bari-RA analysis set (IR). A. Cumulative incidence rate of malignancy (excluding NMSC) by time period for all-bari-RA patients (IR per 100 PY; 95% CI). B. Cumulative incidence rate of NMSC by time period of report for all-bari-RA patients (IR per 100 PY; 95% CI). RA: rheumatoid arthritis; IR: incidence rate; NMSC: nonmelanoma skin cancer; PY: patient-years.
To further quantify malignancy risk in all-bari-RA, we assessed patients who initiated and maintained 4 mg (n = 2658, 4645 PY). In this all-bari-RA 4 mg subcohort, the malignancy (excluding NMSC) IR was 0.8 (95% CI 0.6–1.08).
The NMSC IR was higher for 4 mg compared to 2 mg in 2 mg-4 mg–extended (Table 3), driven by a higher incidence diagnosed within 24 weeks of starting 4 mg (Supplementary Table 2, available with the online version of this article). The all-bari-RA IR was 0.4 (95% CI 0.2–0.5) and did not increase over time (Figure 1B).
No lymphoma cases were reported in the controlled periods, and 6 were reported during the LTE (all-bari-RA IR 0.09, 95% CI 0.03–0.19; Table 3). Four patients initially received placebo and switched/rescued to 4 mg; one initially received ADA and switched to 4 mg, and 1 initiated on 4 mg; patients received baricitinib for an average of 485 days (range 342–670 days) prior to lymphoma event. Lymphoma types included a gastric mucosa-associated lymphoid tissue B cell lesion (n = 1, successfully treated with Helicobacter pylori eradication alone) and B cell (n = 4) and T cell (n = 1) lymphoma. Five of the 6 cases were taking background MTX, and the other was taking concomitant tacrolimus. Two cases of lymphoproliferative disorder were reported. One patient began taking ADA and was diagnosed 112 days after rescue to 4 mg. The other initiated with 4 mg and was diagnosed after 259 days of treatment. As of last followup, corrective lymphoma treatment was not administered in either case.
Gastrointestinal (GI) perforation
Three cases were reported in all-bari-RA (IR 0.05, 95% CI 0.01–0.13; Table 3): a perforated appendix, a perforated diverticulum, and a proximal intestinal perforation after knee surgery. All patients were taking background MTX and nonsteroidal antiinflammatory drugs, and 2 were taking prednisone.
Major adverse cardiovascular event (MACE)
The all-bari-RA MACE IR was 0.5 (95% CI 0.4–0.7; Table 3). The incidence of MACE and the individual components were comparable between groups.
Deep vein thrombosis (DVT)/pulmonary embolism (PE)
In placebo-4 mg, there were no cases of DVT/PE with placebo and 5 cases with 4 mg. One case occurred after discontinuing baricitinib, and 1 case resolved with continued administration of baricitinib and without anticoagulant treatment. In the remaining 3 cases, baricitinib was either continued or temporarily interrupted, then restarted without worsening or recurrence. Two of the 5 events were serious. All patients had multiple DVT/PE risk factors. After the September 2016 data cutoff, a followup medical review identified a sixth event of DVT (termed thrombophlebitis) in the baricitinib 4-mg group during the placebo-controlled period in a patient taking oral contraceptives. One fatal PE was reported with MTX monotherapy during the controlled period (Table 3, footnote c). In 2 mg-4 mg–extended, the DVT/PE incidence was comparable between doses (IR 0.5 vs 0.6, 2 mg vs 4 mg, respectively). Thirty-one patients reported DVT/PE in all-bari-RA (IR 0.5, 95% CI 0.3–0.7) and the IR was stable over time (Supplementary Figure 1, available with the online version of this article). Independent association with VTE incidence was seen for increased age, increased body mass index, history of DVT/PE, and use of selective cyclooxygenase-2 inhibitors, but not for other factors such as baseline disease activity or corticosteroid use. To further quantify any potential DVT/PE risk, the 928 patients starting 4 mg after placebo were assessed; 1 DVT was reported through 24 weeks of baricitinib treatment.
Laboratory and chemistry changes
In placebo-4 mg and placebo-2 mg-4 mg, no significant differences between treatment groups were observed for hemoglobin shifts to less than the lower limit of normal (LLN), < 10 g/dl, or < 8 g/dl (Table 5). In the DMARD-inadequate responder 4 mg longterm cohort, a small mean decrease from baseline in hemoglobin through Week 20 was observed, and returned to baseline or higher with continued treatment (Supplementary Figure 2A, available with the online version of this article). In placebo-4 mg and placebo-2 mg-4 mg, no significant differences between treatment groups were observed for shifts in lymphocytes to < 500 cells/mm3, neutrophils to < 1000 cells/mm3, or platelets to < LLN (Table 5). Mean lymphocyte and neutrophil count changes associated with longterm baricitinib treatment are shown in Supplementary Figure 2B–C. Lymphocyte counts initially increased, then declined to baseline. Neutrophil counts initially decreased within normality, then remained stable. Baricitinib 4 mg was associated with a modest increase in mean platelet counts that peaked at 2 weeks, returned toward baseline and remained stable (Supplementary Figure 2D). A small proportion of patients had a platelet count > 600,000/mm3 (Table 5). No association was observed between platelet count increase ≥ 400,000/mm3 and DVT/PE (25.8% vs 36.1% for patients with and without a DVT/PE).
Significant increases from baseline in low-density lipoprotein (LDL) to ≥ 130 mg/dl and high-density lipoprotein (HDL) to ≥ 60 mg/dl were observed with both doses (Table 5); however, LDL/HDL ratio did not change (mean: 0.01, −0.04, and −0.01; placebo, 2 mg, 4 mg, respectively). In general, LDL and HDL increased in the first 12 weeks and stabilized thereafter (data not shown). In baricitinib-treated patients, dose-dependent, largely asymptomatic increases in creatine phosphokinase (CPK) were observed. Discontinuation owing to increased CPK or muscle symptoms was uncommon (0.2%), and the proportion of patients experiencing muscle symptoms did not differ across treatment groups.
Small mean increases (< 0.1 mg/dl) in serum creatinine were observed with baricitinib within the initial weeks of treatment (Supplementary Figure 2E, available with the online version of this article). Elevations in alanine aminotransferase (ALT) > upper limit of normal (ULN) were more frequent for baricitinib 4 mg compared to placebo in placebo-4 mg and placebo–2 mg-4 mg (Table 5; Supplementary Figure 2F). ALT elevations to ≥ 3 × ULN were comparable across treatment groups while non–dose-dependent elevations ≥ 5 × ULN were more frequent for baricitinib compared to placebo. In all-bari-RA, permanent discontinuations due to hepatobiliary AE were uncommon (EAIR 0.3; Table 4). No case met Hy’s Law criteria for drug-induced liver injury22.
DISCUSSION
We report an assessment of baricitinib safety in patients with RA through 5.5 years of treatment. Our data suggest that the incidence of death, SAE including infections, and malignancy are similar to those observed for other therapeutic trials. Few patients discontinued as a result of adverse events.
The all-bari-RA mortality rate was similar to other RA therapeutic programs23,24,25,26,27,28. The clinical conditions contributing to death were common to patients with RA and none predominated (Table 3, footnote c).
Patients with RA are at elevated risk of infection from their disease29,30 and its therapies2,3,31. While a higher overall TE infection rate was observed with baricitinib 4 mg vs placebo, serious infection rates were similar. The all-bari-RA serious infection rate was similar to other RA therapeutic trials23,25,26,27,28,29,30,32. Infections leading to death were uncommon (IR 0.07) in baricitinib-treated patients. TB was observed in baricitinib-treated patients. While tumor necrosis factor (TNF) inhibitors have been shown to increase TB risk33,34, it is unclear whether JAK inhibitors have similar effects. In tofacitinib trials, most TB cases (81%) occurred in endemic countries, and rates varied according to the general population’s background TB rate35. This increased risk versus the general population is similar to that observed in real-world studies with TNF inhibitors33,36. TB screening prior to therapy should reduce risk17.
Although there was no observed increased risk for serious infections with baricitinib, there was an increased risk of herpes zoster (no visceral case), similar to that seen with other JAK inhibitors27,35,37,38 and higher than that observed with bDMARD39. Increased zoster risk with JAK inhibitors may result from inhibition of Type 1 interferons, which signal through a JAK1/tyrosine kinase 2 heterodimer. Dose-dependent increases in zoster risk without an increased risk for serious infections have been seen with Type 1 antiinterferon antibodies in patients with systemic lupus erythematosus40,41. This suggests an on-target mechanism, rather than generalized immune suppression, for increased zoster risk. Zoster vaccination for selected patients prior to initiating JAK inhibitors may be a therapeutic option36,42,43.
There were 52 malignancy cases (excluding NMSC). With short-term treatment, the IR between placebo and 4 mg was similar. In 2 mg-4 mg extended, the 4 mg IR was numerically higher compared to 2 mg. However, with “as-randomized” analysis, the IR between doses was similar. To account for long latency, as-randomized analysis does not censor data at rescue or dose change. Additionally, screening effects may lower the risk of uncommon/rare events in earlier phases of a trial15. As-randomized analysis reduces the bias introduced by postbaseline, unidirectional switching of patients from the lower to higher baricitinib dose (due to rescue). Lastly, the IR for both all-bari-RA and all-bari-RA 4 mg subcohort of patients who initiated and maintained on 4 mg was 0.8/100 PY. The latter may represent a more accurate estimate for 4 mg than that observed in 2 mg-4 mg–extended because it is based on over 7-times more PY of exposure (4645 vs 604). Exposure time within baricitinib trials is relatively limited, and longer time periods will be needed to further evaluate malignancy risk. At present, the malignancy (excluding NMSC) IR appears to be similar to those reported in other RA therapeutic programs24,25,28,44,45 and remained stable over time. The SIR of 1.04 (95% CI 0.79–1.36) indicates that the cancer IR in the baricitinib RA program is not increased compared to a similar United States population sample21.
Patients with RA are at an increased risk of DVT/PE46,47. In placebo-4 mg, DVT/PE were observed with 4 mg but not placebo (5 vs 0 patients). Although a causal relationship to baricitinib cannot be excluded, traditional risk factors were present in each case. Additionally, for patients converting from placebo to 4 mg, only 1 DVT case was reported in over 900 patients through 24 weeks of treatment. No dose or temporal dependency in DVT/PE risk was observed with prolonged administration. In all-bari-RA, the DVT/PE IR of 0.5/100 PY was comparable with background rates of 0.3 to 0.8/100 PY in real-world studies of the RA population46,47,48. Further study will be necessary to evaluate this potential risk.
Regarding other AE of special interest, the incidence of MACE was low. There were 3 GI perforations [IR of 0.05/100 PY (0.5/1000 PY)], and the observed IR appears to be lower than that observed for tofacitinib, tocilizumab, and other bDMARD in real-world data (IR range 0.73 to 1.55/1000 PY)49.
Fewer than 1% of patients stopped baricitinib because of laboratory changes, and very few patients experienced Common Terminology Criteria for AE (CTCAE) grade ≥ 3 changes in specific laboratory variables. While JAK selectivity of baricitinib differs from tofacitinib, the existing overlap likely explains the similarity in laboratory changes including increases in LDL, HDL, CPK, liver tests, and creatinine39. Additionally, with both compounds neutrophils drop after drug start, although this has not been correlated with an increased infection risk26. However, a similar small percentage of patients developed CTCAE grade 3 and 4 changes in lymphocytes between placebo and baricitinib arms, and overall, mean lymphocyte levels did not decrease over time with baricitinib use. While patients taking baricitinib had a small transient mean decrease in hemoglobin, the proportions of patients with abnormally low hemoglobin did not differ significantly between baricitinib and placebo, and very few experienced CTCAE grade ≥ 3 changes (< 8 mg/dl).
Limitations of this analysis include the short placebo-controlled period, which diminishes AE assessment of baricitinib versus the underlying disease, particularly for uncommon events (e.g., malignancy, MACE, DVT/PE)50. The IR of these uncommon events compared to published background rates provides context. Baricitinib 2-mg exposure was limited, although the 4-mg safety profile should inform that of 2 mg. Limitations of LTE studies include lack of a control arm, modification of background therapy according to clinicians’ discretion, and variability of dosing such as rescue or taper. However, these factors more closely reflect usual clinical care; thus, the results could be applicable to real-world use.
We have evaluated the safety of baricitinib in over 3400 patients with RA treated for up to 5.5 years. Throughout its development, baricitinib was generally well tolerated. Infection risk, particularly for herpes zoster, is elevated as with other JAK inhibitors, and clinicians should take steps to prevent and monitor for such infections. Longterm risks of malignancy need further study, but currently there is no signal suggesting an increased risk. The potential risk for DVT/PE warrants further characterization, including in the postmarketing setting. Overall, in the context of demonstrated efficacy7,8,9,10 in patients with active RA, baricitinib 4 mg and 2 mg once daily had an acceptable safety profile through up to 5.5 years of longterm exposure.
ONLINE SUPPLEMENT
Supplementary material accompanies the online version of this article.
Acknowledgment
We thank the patients, investigators, and study staff who were involved in these studies. We also thank Cate Jones, PhD, Eli Lilly and Co., for assistance with tables, manuscript preparation, and process support, and Julie Sherman, AAS, Eli Lilly and Co., for assistance with figures.
Footnotes
Full Release Article. For details see Reprints and Permissions at jrheum.org
- Accepted for publication July 1, 2018.
Free online via JRheum Full Release option
