Abstract
Objective. To investigate the distribution of whole-body magnetic resonance imaging (WB-MRI) inflammatory lesions of peripheral joints and entheses, and their response to adalimumab (ADA) treatment and agreement with clinical measures of disease activity in patients with axial spondyloarthritis (axSpA).
Methods. Explorative analysis of an investigator-initiated randomized controlled trial of ADA. WB-MRI was performed at weeks 0, 6, 24, and 48. Detailed analyses of WB-MRI lesions in peripheral joints and entheses were performed, including agreement with clinical measures of disease activity.
Results. WB-MRI inflammatory lesions were most frequently observed in the acromioclavicular, metatarsophalangeal, and wrist joints (> 10% of joints), and at the greater trochanter, calcaneal insertion of the Achilles tendon, and ischial tuberosity (> 15% of entheses). Inflammation resolved in ≥ 2/3 of involved sternoclavicular, metacarpophalangeal, first carpometacarpal, hip, and tarsometatarsal joints, and pubic symphyses and medial femoral condyles. In contrast, inflammation resolved in ≤ 1/6 of involved acromioclavicular joints, knee joints, and supraspinatus tendon insertions at humerus. Tenderness of joints and entheses agreed poorly with WB-MRI inflammation (κ < 0.40). Joint tenderness resolved more frequently in MRI-positive than MRI-negative joints (8/13, 62% vs 9/34, 26%) after 6 weeks of active treatment.
Conclusion. Inflammatory lesions of peripheral joints and entheses in patients with predominantly axSpA, and changes therein, can be mapped using WB-MRI, and it may contribute to differentiate between inflammatory and noninflammatory joint tenderness. (Trial registration: ClinicalTrials NCT01029847).
Spondyloarthritis (SpA) is a group of diseases with shared genetics and clinical manifestations1. Peripheral arthritis and enthesitis are frequent disease manifestations and are included in the Assessment of Spondyloarthritis International Society (ASAS) classification criteria for axial and peripheral SpA2,3. Classical clinical signs of inflammation at peripheral joint and entheses, e.g., swelling, may be apparent, but the patient-reported pain and tenderness often does not correlate with clinical signs of inflammation. Tenderness in patients with SpA may be of inflammatory or noninflammatory origins4,5.
Whole-body magnetic resonance imaging (WB-MRI) is a new modality in musculoskeletal (MSK) imaging that may visualize inflammatory changes in multiple peripheral joints and entheses at the same time. WB-MRI has gained interest for the detection of arthritis and enthesitis, and for objectively monitoring changes during treatment6,7,8,9,10. In previous WB-MRI studies of patients with SpA, these sites have frequently been found to be involved: joints of the anterior chest wall, the calcaneal insertion of the Achilles tendon and plantar fascia, iliac crest, greater femoral trochanter, and ischial tuberosity6,7,8,11,12,13. These studies had rather small sample sizes, were cross-sectional, or included only patients with ankylosing spondylitis (AS), and the distribution of peripheral MRI manifestations in a broader group of patients with axial SpA (axSpA) is, therefore, still largely unknown. It is of interest how inflammatory lesions change during an effective antiinflammatory therapy, because this provides information on the validity of the MRI findings. Moreover, WB-MRI may also be of value to distinguish pain associated with inflammation versus pain not associated with inflammation in joints and entheses.
The objective of our study was to investigate patients with axSpA for the distribution of inflammatory lesions of peripheral joints and entheses, and their response to adalimumab (ADA) treatment by WB-MRI. Moreover, we aimed to investigate the agreement between WB-MRI findings and clinical joint tenderness and swelling, entheseal tenderness, and other clinical measures of disease activity.
MATERIALS AND METHODS
Study design
The ASIM study (Adalimumab in Axial Spondyloarthritis) consisted of 49 patients who fulfilled the ASAS criteria for axSpA, had sacroiliitis as assessed by radiography and/or MRI, and a Bath Ankylosing Spondylitis Disease Activity Index (BASDAI) score ≥ 4.0 despite treatment with nonsteroidal antiinflammatory drugs. The study was a randomized, double-blind, placebo-controlled trial of ADA 40 mg or placebo subcutaneously (SC) every other week (eow) for 6 weeks, followed by ADA 40 mg SC eow from Week 6 to Week 48 in both treatment groups. At Week 24, clinical responders (decrease in BASDAI of 50% or 2.0) continued ADA while nonresponders were allowed treatment with other drugs following local treatment guidelines. The inclusion and exclusion criteria, study procedures, and main results for clinical outcomes and WB-MRI indices of inflammation have been reported previously14. The study was approved by the Regional Committee on Health Research Ethics, Region Hovedstaden, Denmark, approval number: H1-2013-118. Trial registration: ClinicalTrials, NCT01029847. All participants provided written consent.
WB-MRI acquisition
WB-MRI was performed at baseline and at weeks 6, 24, and 48 on a Philips 3.0 Tesla scanner using 6 separate imaging stations with a whole-body quadrature coil: (1) coronal and sagittal images of cervical spine/shoulders including the sternoclavicular joints; (2) coronal images of thoracic spine; (3) coronal images of lumbar spine, sacroiliac joints, and pelvis including the hips; (4) coronal images of hips and hands; (5) coronal images of knees; and (6) coronal and axial images of ankles and feet14. Short-tau inversion recovery (STIR), T1-weighted spin-echo (T1W), and T1W post-gadolinium (post-Gd-T1W) sequences were obtained14.
WB-MRI scoring
An experienced MSK radiologist (I.E.) evaluated all images in known chronology blinded to radiography and clinical data. Reading images with known chronology was done to increase precision and feasibility, given the complexity of assessing WB-MRI images at 4 timepoints in unknown order, knowing that it may introduce a risk of bias in changes over time. Osteitis was assessed on STIR sequences with post-Gd-T1W sequences used for reference only, while synovitis and entheseal soft tissue inflammation were assessed based on both these sequences.
Fifty-six peripheral joints [glenohumeral, acromioclavicular, sternoclavicular, wrist, carpometacarpal, metacarpophalangeal (MCP), hand interphalangeal (IP) joints, hip, knee, ankle, tarsometatarsal, and metatarsophalangeal (MTP) joints] were scored separately for synovitis and osteitis on a semiquantitative scale (0 = absent, 1 = mild to moderate, 2 = severe). Joints were also assessed for the presence/absence of erosion.
Fifteen peripheral entheseal sites (supraspinatus tendon insertion at humerus, iliac crest, ischial tuberosity, pubic symphysis, greater femoral trochanter, insertion of the collateral ligament at the medial femoral condyle, calcaneal Achilles tendon insertion, and the fifth lumbar spinous process) were scored separately for entheseal osteitis and soft tissue inflammation (0 = absent, 1 = mild to moderate, 2 = severe). The ischial tuberosity consists of the area where the hamstring muscles insert, and the greater femoral trochanters consists of the area where the gluteus medius inserts. However, adjacent structures when inflamed might also cause increased signal in the bone marrow and soft tissue at these sites.
A previously developed WB-MRI Enthesis Inflammation Index (range 0–60), which sums osteitis and soft-tissue inflammation scores of entheses into a patient-level index, and a WB-MRI Peripheral Joint Inflammation Index (range 0–184), which sums osteitis and synovitis scores of all individual joints, were applied for correlation and cluster analysis with other variables14.
Baseline and Week 24 images of 8 patients were scored twice to assess intrarater reliability. The costosternal joints and IP joints of the feet were only assessable in < 10% of the patients and were excluded from the analysis.
Clinical examination
Patients had swollen joint count-68 joints (SJC68) and tender joint count-70 joints (TJC70) performed, and the conventional SJC44 and TJC44 were also derived15. All entheseal sites in the Maastricht Ankylosing Spondylitis Enthesitis Score (MASES), Spondyloarthritis Research Consortium of Canada (SPARCC) Enthesitis Index, and Leeds Enthesitis Index (LEI) were assessed for tenderness16,17,18. A tender enthesis count of 33 entheses (TEC-33) based on all the entheses included in the MASES, SPARCC, and LEI was constructed14.
Statistical analyses
For the main analyses, WB-MRI data at the lesion level were handled as binary (osteitis present/absent, synovitis present/absent, soft tissue inflammation present/absent) because most of the positive scores, 722 of 765 (94%), were scored as 1 (mild or moderate) and only a few, 43 (6%), were scored as 2 (severe). As an additional analysis, MRI-positive joints were defined as any signs of osteitis and/or synovitis (i.e., score ≥ 1), while MRI-positive entheses were defined as any signs of osteitis and/or soft tissue inflammation (i.e., score ≥ 1).
Agreement between WB-MRI and tenderness, as well as the intrarater reliability of scoring WB-MRI, were assessed using Cohen’s κ, with values > 0.75 representing excellent agreement, values between 0.4 and 0.75 as fair to good agreement, and values < 0.4 as poor agreement19. In a posthoc analysis to further analyze the disease characteristics of patients with discrepancies between WB-MRI findings and clinical tenderness, 2 subgroups were created: 1 subgroup of patients with ≥ 5 MRI-positive joints or entheses that were not tender (n = 11), and 1 subgroup of patients with ≥ 5 MRI-negative joints or entheses that were tender (n = 12). These cutoffs were chosen posthoc to include about 1/5 of the patients in each subgroup.
Hierarchical cluster analysis of the WB-MRI peripheral joint inflammation index and the WB-MRI enthesis inflammation index along with other disease activity measures was performed. The SPARCC Spine/Sacroiliac joint (SIJ) Inflammation Index was derived by adding the SPARCC Spine Inflammation Index and the SPARCC SIJ Inflammation Index. C-reactive protein (CRP), TEC-33, TJC70, and SPARCC Spine/SIJ Inflammation Index were log transformed to be about normally distributed; log(x + 1) was used to allow for 0 values. Thereafter, all variables were scaled to mean 0 and SD 1. Distances between measures were calculated using the Manhattan distance, where the pairwise discrepancies in 2 measures across all patients are summed. Hierarchical cluster analysis was performed using “average” as aggregation method, where all values are simultaneously taken into account. To assess the uncertainty of the clustering procedure, the bootstrap probability for each cluster was computed using 10,000 bootstrap samples20.
RESULTS
Distribution of peripheral arthritis as assessed by WB-MRI and palpational tenderness
At baseline, 114 joints out of 2174 (5%) were MRI-positive (osteitis and/or synovitis), 27 joints (1%) had osteitis, and 100 joints (4%) had synovitis (Table 1, Figure 1). By clinical examination, 131 joints out of 3430 (4%) were tender and 8 (0.2%) were swollen. Five patients (10%) had ≥ 1 swollen joint, 26 patients (53%) had ≥ 1 tender joint, and 35 patients (71%) had ≥ 1 MRI-positive joint.
Percentage of joints and entheses that were MRI-positive at baseline. A: supraspinatus tendon insertion at humerus; B: fifth lumbar spinous process; C: iliac crest; D: pubic symphysis; E: ischial tuberosity; F: greater femoral trochanter; G: medial femoral condyle; H: calcaneal Achilles tendon insertion; MRI: magnetic resonance imaging.
Percentage of joints with clinical tenderness and MRI lesions at baseline and the resolution and development of MRI-positive joints during followup.
The most frequent MRI-positive joints were the first MTP (38%), acromioclavicular (16%), and wrist joints (12%). The most frequent tender joints were the sternoclavicular (18%), first to fourth MTP (8–12%), and shoulder joints (9%). Erosions were most frequently observed in the acromioclavicular (9%) and first MTP joints (7%). Image examples are shown in Supplementary Figure 1 (available with the online version of this article).
Distribution of enthesitis as assessed by WB-MRI and tenderness
At baseline, 597 entheses were assessed by WB-MRI, 28 entheses (5%) were scored positive for osteitis, 45 entheses (8%) were scored positive for soft tissue inflammation, and in total 59 entheses (10%) were MRI-positive (osteitis and/or soft tissue inflammation; Table 2, Figure 1). In comparison, 327 out of 1617 entheses (20%) were tender on palpation. Forty patients (82%) had ≥ 1 tender enthesis and 28 patients (57%) had ≥ 1 enthesis with inflammation on MRI. Overall, 41 patients (84%) had ≥ 1 tender joint or enthesis, and 41 patients (84%) had ≥ 1 MRI-positive joint or enthesis.
Percentage of entheses with clinical tenderness and MRI lesions at baseline and the resolution and development of MRI-positive entheses during followup.
The most frequent MRI-positive entheses were greater femoral trochanter (21%), calcaneal Achilles tendon insertion (17%), and ischial tuberosity (16%), whereas the most frequent tender entheseal sites were at the first and seventh costosternal junctions (35–43%), posterior superior iliac spine (38%), and greater femoral trochanter (35%). Image examples are shown in Supplementary Figure 1 (available with the online version of this article).
Clinical indices of enthesitis, tender, and swollen joints
At Week 6, TJC70 decreased significantly more in the ADA group compared to the placebo group. No significant changes in SJC68 were observed over time. The MASES, LEI, and SPARCC enthesitis indices and TEC-33 all decreased numerically during treatment, but the between-group difference in change at Week 6 did not reach statistical significance (Supplementary Table 1, available with the online version of this article).
Agreement between WB-MRI arthritis/enthesitis and clinical tenderness
The agreement between WB-MRI arthritis and tenderness of the individual joints was poor, with κ values site-by-site < 0.4, except for the third to fifth MTP joints (κ for these joints ranged 0.6–1.0, but the majority were judged not assessable by MRI). Among 2139 joints in total at baseline, 17/1965 were concordantly positive/negative, 60 were tender but MRI-negative, and 97 were MRI-positive but not tender (κ 0.14). In addition, regarding the entheses, agreement between WB-MRI and tenderness was poor, with κ values site-by-site < 0.4. Among 597 entheses in total at baseline, 19/388 were concordantly positive/negative, 104 were tender but MRI-negative, and 37 were MRI-positive but not tender (κ 0.08).
When limiting the analysis to sites that were scored as “severe” by MRI at baseline, none of the 6 joints scored as “severe” by MRI were clinically tender or swollen, and only 2 of 9 entheses scored as “severe” by MRI (the left and right ischial tuberosity in a single patient) were clinically tender.
Intrarater reliability of WB-MRI at the joint and enthesis level
Intrarater agreement at the joint level was fair to good, with κ 0.46 and percentage exact agreement 96%. Similar results were found for entheses, with κ 0.59 and percentage exact agreement 93%.
Patients with major discordance between MRI inflammation and clinical tenderness
Patients with ≥ 5 MRI-negative sites with tenderness were more often women, had shorter disease duration and more frequently nonradiographic axSpA (nr-axSpA), with less spinal inflammation and new bone formation, and tended to have more fatigue (Table 3). In contrast, patients with ≥ 5 MRI-positive sites with no tenderness tended to be men, had longer disease duration, and more frequently radiographic axSpA, i.e., AS, more spinal inflammation and new bone formation, and tended to have less fatigue.
Baseline characteristics of all patients and subgroups of patients with many discordant tender joints/entheses and many discordant MRI-positive joints/entheses.
Correlation between MRI inflammation and other disease activity measures
The WB-MRI Enthesis Inflammation Index and WB-MRI Peripheral Joint Inflammation Index correlated with each other (κ = 0.37, p = 0.008). The WB-MRI Enthesis Inflammation Index correlated with the SPARCC Spine MRI Inflammation Index (κ = 0.32, p = 0.03). The WB-MRI Enthesis Inflammation Index tended to correlate weakly with CRP (κ = 0.20, p = 0.15), while the WB-MRI Peripheral Joint Inflammation Index did not correlate with CRP (κ = 0.00, p = 0.99). There was no correlation between the WB-MRI Enthesis Inflammation Index and TEC-33 (κ = 0.03, p = 0.81), and no correlation between the WB-MRI Peripheral Joint Inflammation Index and TJC70 (κ = −0.09, p = 0.56). Similar results were found when the analysis was limited to the 13 entheses and 56 joints that were assessed both clinically and by MRI. TEC-33 correlated most closely with TJC70 (κ = 0.64, p < 0.001), fatigue (κ = 0.36, p = 0.01), and pain (κ = 0.32, p = 0.02).
Hierarchical cluster analysis of disease activity measures
Using hierarchical cluster analysis in an investigative posthoc analysis showed that objective measures of disease activity, i.e., WB-MRI indices of inflammation of peripheral joints and entheses, and inflammation of the axial skeleton (SPARCC Spine/SIJ) on conventional MRI, CRP, and SJC68, formed 1 cluster, while the patient-reported measures of disease activity, including pain, fatigue, and palpational tenderness, seemed to form another separate cluster (Figure 2).
Clustering of different measures of disease activity in patients with axial spondyloarthritis. Numbers at clustering points indicate how strongly each cluster is supported by the data (bootstrap probability — the frequency with which a cluster appears in 10,000 bootstrap replicates). MRI: magnetic resonance imaging; WB-MRI: whole-body MRI; TEC-33: tender enthesis count of 33 entheses; TJC-70: tender joint count-70 joints; SJC-68: swollen joint count-68 joints; WB-MRI enth. infl.: WB-MRI entheseal inflammation index; WB-MRI joint infl.: WB-MRI joint inflammation index; CRP: C-reactive protein; SPARCC Spine/SIJ: Spondyloarthritis Research Consortium of Canada MRI Spine and Sacroiliac Joint Index.
Resolution of tenderness and WB-MRI lesions during treatment
The TJC70, TJC-44, MASES, SPARCC, LEI, and TEC-33 improved numerically over time in both groups, but the between-group difference in change at Week 6, i.e., at the end of the placebo period, did not reach significance, except for TJC70. However, this Week 6 between-group difference in change in TJC70 was driven by an observed worsening in the placebo group more than actual improvement in the ADA group (Supplementary Table 1, available with the online version of this article).
At the lesion level, clinical tenderness of the individual joints/entheses resolved at Week 6, with similar frequencies in the ADA group (36%/42%) and the placebo group (52%/50%), whereas MRI inflammation of joints/entheses tended to resolve more often in the ADA group (31%/26%) than in the placebo group (10%/17%). In the ADA group, clinical tenderness of joints and entheses that were positive for baseline MRI inflammation tended to resolve more frequently, while in the placebo group, clinical tenderness of joints and entheses that were negative for baseline MRI inflammation tended to resolve more frequently (Table 4). At Week 6, tenderness of joints resolved more frequently when positive for inflammation by MRI (Fisher’s exact test, p = 0.041), while no difference in the resolution of tenderness of entheses in relation to the presence or absence of MRI inflammation was observed.
Disappearance of clinical tenderness from baseline to Week 6, stratified by the presence or absence of baseline MRI inflammation.
DISCUSSION
In our study of patients with predominantly axSpA, WB-MRI documented inflammation of peripheral joints and/or entheses in most patients. Inflammatory lesions as assessed by WB-MRI tended to disappear during treatment with ADA, which supports the validity of WB-MRI as a potential method of measuring inflammation during followup of patients with SpA. Further, a posthoc subanalysis showed that tender joints with MRI inflammation improved clinically during tumor necrosis factor inhibitor (TNFi) therapy more frequently than tender joints without MRI inflammation, which suggests that MRI may be able to differentiate inflammatory from other noninflammatory causes of tenderness and pain in patients with SpA. To our knowledge, no studies have previously investigated whether WB-MRI inflammation of peripheral joints and entheses predicts treatment response, while in axSpA, MRI inflammation of the sacroiliac joints is known to predict treatment response to TNFi therapy21.
MRI inflammation present at baseline resolved most frequently (≥ 50% of lesions) in sternoclavicular, MCP, first carpometacarpal, hip, and tarsometatarsal joints, and at the pubic symphysis, medial femoral condyle, and the calcaneal insertion of the Achilles tendon. Thus, MRI inflammation detected at these sites is likely to be truly related to an inflammatory disease such as SpA. In contrast, inflammation resolved in only a few of the acromioclavicular joints and first MTP joints; these sites are known to be prone to osteoarthritis (OA) and the persisting inflammation that was observed may be unrelated to SpA. Only a few of the osteitis lesions at the supraspinatus tendon insertions at humerus (14%) resolved, and because of the limited image resolution and only coronal images of the shoulder, this may be hard to discriminate from fluid in bone erosions at this site of the shoulder in patients with axSpA22; discrimination between supraspinatus and infraspinatus tendon insertions may also be very difficult. A frequent involvement of the acromioclavicular and first MTP joints was not found in 2 other studies, but whether this reflects genuine differences between the patient cohorts or different image resolution or reader sensitivity is unclear11,12.
If WB-MRI were to be used as an objective measure of SpA remission, sites prone to degenerative changes may need to be disregarded when assessing the peripheral inflammatory activity related to SpA. An option would be to assess each joint for OA and disregard scores for inflammation of joints where OA is judged present. However, WB-MRI has poorer resolution compared to conventional-dedicated MRI and is therefore not currently sufficient for assessing damage (except for gross pathology) in small structures, e.g., finger or toe joints; newer MRI units and improved sequence types can provide better resolution.
The agreement between tenderness and objective signs of inflammation of peripheral joints and entheses as assessed by WB-MRI was poor overall. Subtle areas with inflammation might be missed by MRI because of the rather coarse image resolution, and small joints of the hands and feet were at the threshold of resolution, with just 1 or 2 slices depicting a joint, and partial volume averaging was a challenge. This may have limited the intrarater agreement in our study and may contribute to the poor correlation between clinical and imaging findings. However, a poor correlation between MRI findings and clinical assessment of entheses has been found in several studies7,8,13,23. Therefore, we tried to look at possible explanations. In posthoc analyses, we observed that the subgroup of patients with many tender joints and entheses were more frequently women with nr-axSpA, while the subgroup of patients with many MRI-positive sites were more frequently men fulfilling the radiographical criteria for AS. Similar differences between men and women have been reported in a large French cohort, where women had worse patient-reported outcomes despite less radiographic sacroiliitis and MRI inflammation of sacroiliac joints and spine24. Women tend to have nonradiographic disease and more often fulfill the diagnostic criteria for fibromyalgia4,25, and the lack of good correlation between MRI inflammation and other measures of disease activity may thus in part reflect variations across the spectrum of axSpA26.
A disconnect between patient-reported and objective measures also emerged when disease measures were analyzed in hierarchical cluster analysis. We identified a cluster of patient-reported measures, including palpational tenderness of joints and entheses, and a separate cluster of objective measures, including MRI, CRP, and SJC. Joint and entheseal tenderness may or may not be related to active SpA at the tender sites, but the positive predictive value of tenderness was low when using MRI inflammation as standard reference (22% of joints with tenderness were MRI-positive, while 15% of entheses with tenderness were MRI-positive).
The relationship of WB-MRI inflammation to other objective measures of disease activity and the better clinical response of joints when positive by MRI suggest that the method has credibility, despite the concerns mentioned above. However, the clinical relevance of subclinical MRI inflammation for the diagnosis and management of patients with axSpA is unknown, and the data on the prevalence of WB-MRI–positive joints and entheses in healthy subjects and subjects with degenerative changes across different age groups are also very limited. Thus, at this stage, WB-MRI is not recommended for use in routine clinical practice in patients with axSpA but is of high interest for research purposes. The Outcome Measures in Rheumatology (OMERACT) MRI in Arthritis Working Group has recommended which scan planes to use for WB-MRI image acquisition and assessment27. In our present study, it is a limitation that sagittal MRI images of ankles and knees were not performed, which meant that certain structures, e.g., the plantar fascia and patellar ligament, could not be scored and therefore our study does not provide data on these sites. WB-MRI was scored by 1 experienced MSK radiologist and we acknowledge that an additional reader might have improved the overall robustness of the conclusions. An initiative regarding the refinement of a semiquantitative scoring system of WB-MRI and testing its reliability among several readers at varying levels of experience is currently ongoing within the OMERACT MRI in Arthritis Working Group.
The distribution of inflammatory lesions was mapped using WB-MRI of peripheral joints and entheses in 49 patients with predominantly axSpA. Joint and entheseal tenderness agreed poorly with WB-MRI inflammation. Patients with many tender joints or entheses that were MRI-negative were frequently women with nr-axSpA, while patients with many MRI-positive joints or entheses that were not tender were frequently men with AS. WB-MRI seems to be a promising objective tool for assessing the distribution and changes over time in inflammation of peripheral joints and entheses, and may separate inflammation of joints and entheses from tenderness of noninflammatory origin.
ONLINE SUPPLEMENT
Supplementary material accompanies the online version of this article.
Acknowledgment
Thanks go to all the patients who participated in this trial.
Footnotes
Dr. Pedersen and Dr. Østergaard are co-senior authors.
The investigator-initiated study was supported by AbbVie, which manufactures adalimumab. S.K. received grants from The Danish Rheumatism Association and Rigshospitalet. S.J.P. has received speaker fees from MSD, Pfizer, AbbVie, Novartis, and UCB; has been an advisory board member for AbbVie, and received research support from AbbVie and MSD. M.Ø. has received research support and/or consultancy/speaker fees from AbbVie, BMS, Boehringer-Ingelheim, Celgene, Eli-Lilly, Centocor, GSK, Hospira, Janssen, Merck, Mundipharma, Novartis, Novo, Orion, Pfizer, Regeneron, Schering-Plough, Roche, Takeda, UCB, and Wyeth.
- Accepted for publication March 1, 2019.
