Objective The aim of this study was to assess the correlation between serum and intestinal anti-tumour necrosis factor (TNF) levels, and their relationship to endoscopic disease activity and levels of TNF.
Design Cross-sectional study of 30 patients receiving treatment with infliximab or adalimumab for Crohn's disease or UC. For each patient, a sample of serum was matched to tissue biopsies. Endoscopic and histological disease activity was recorded for each tissue sample.
Results There was a significant positive correlation between anti-TNF in serum and tissue (r=0.3920, p=0.002), especially in uninflamed tissue (r=0.50, p<0.001), but not with those samples that had inflammation (r=0.19, p=0.54). Anti-TNF concentration in tissue correlated with degree of endoscopic inflammation, except for tissue with severe inflammation in which anti-TNF levels were again lower (mean normalised anti-TNF in tissue: uninflamed=0.93, mild=2.17, moderate=13.71, severe=2.2 inflammation (p=0.0042)). The ratio of anti-TNF-to-TNF in tissue was highest in uninflamed areas and lowest in severely inflamed areas. Patients with active mucosal disease had a higher rate of serum to tissue drug level mismatch when compared to those in remission (73.3% vs 33.3%, respectively; p=0.03).
Conclusions Our data suggest that local tissue inflammation characterised by high levels of TNF serves as a sink for anti-TNF. We further postulate that some patients with high serum anti-TNF levels have active disease because tissue levels of anti-TNF are insufficient to neutralise local TNF production.
- CROHN'S DISEASE
- ULCERATIVE COLITIS
- INFLAMMATORY BOWEL DISEASE
- INTESTINAL EPITHELIUM
Statistics from Altmetric.com
Significance of this study
What is already known on this subject?
Anti-tumour necrosis factor (TNF) biologics are effective in the treatment of Crohn's disease and ulcerative colitis.
Serum levels of anti-TNF drugs correlate with efficacy of the drug.
Some patients with detectable levels of anti-TNF in serum do not respond to anti-TNFs.
What are the new findings?
Anti-TNF concentration (infliximab and adalimumab) can be reproducibly measured in tissue and correlates with serum levels.
Inflamed tissue has significantly higher TNF and anti-TNF levels compared to matched uninflamed samples.
In areas of ongoing inflammation, the ratio of tissue TNF to anti-TNF is elevated compared to uninflamed tissue.
How might it impact on clinical practice in the foreseeable future?
The finding that local tissue TNF is greater than the levels of anti-TNF in inflamed, especially severely inflamed tissue, provides a mechanistic explanation for patients with active disease in spite of therapeutic serum levels of anti-TNF.
Patients with active disease in spite of detectable levels of anti-TNF in the serum may need higher doses of anti-TNFs or a different class of medication to neutralise tissue inflammation.
Inflammatory bowel diseases (IBD), Crohn's disease (CD) and ulcerative colitis (UC) are characterised by chronic intestinal inflammation in the absence of a pathogen. Studies have found an increased concentration of tumour necrosis factor (TNF) in the intestinal mucosa of patients with CD and UC when compared to controls.1–3 TNF in tissue has also been found to correlate with clinical and endoscopic disease activity.4 These findings along with animal models of IBD prompted the use of anti-TNF antibodies in the treatment of UC and CD.5 Therapy with anti-TNF agents has been shown to induce and maintain clinical remission in both diseases as well as induce mucosal healing.6–12
Infliximab (IFX) and adalimumab are the two most commonly used biologics in the treatment of IBD. Both are monoclonal immunoglobulin G1 (IgG1) antibodies that bind with high affinity and specificity to human TNF. Unfortunately, not all patients achieve remission with anti-TNFs. Several explanations could account for a lack of response to anti-TNFs or a loss of response. First, there may be inadequate levels of the drug.8 ,12 Patients may develop antibodies to the biological agent inhibiting its action and increasing its clearance. A recent study demonstrates that response to anti-TNF therapy correlates with the number of TNF-expressing cells in the mucosa.13 Finally, some patients may not respond because the mechanism driving inflammation is not dependent on TNF.
While several studies have found a correlation between serum anti-TNF levels and clinical outcomes, no studies have evaluated the intestinal tissue drug level and its correlation with serum concentrations and disease activity. There has been a notion that inflamed tissue serves as a sump for anti-TNF but that is unproven. Alternatively, anti-TNF concentrations may be evenly distributed to tissues in proportion to its serum levels. These concepts are important, as some patients have active disease despite having what is considered a therapeutic drug level. Patients with detectable or higher serum levels of the anti-TNF but ongoing inflammation have been deemed as anti-TNF mechanistic failures and some have suggested changing to a drug with a different mechanism of action.14
The aim of this study was to determine if the tissue concentrations of the anti-TNFs (IFX, adalimumab) could be accurately measured and whether tissue levels had a relationship with serum concentrations of the anti-TNF. We also wished to determine the relationship between anti-TNF levels in tissue and systemic and mucosal inflammation. Given the role of TNF in disease pathogenesis, we asked whether TNF levels, either in tissue or systemically, helped determine efficacy of the anti-TNF therapy. Our results are directed at edifying a mechanistic explanation for lack of response to anti-TNFs as well as improving therapeutic drug monitoring (TDM).
Subjects and setting
The study of Anti-TNF Tissue Level and Antibodies in Serum (ATLAS) was a cross-sectional study of patients with CD or UC enrolled in a prospective tissue and serum repository at the Crohn's and Colitis Center of the University of Miami (Miami, Florida). The study was approved by the University of Miami Miller School of Medicine Institutional Review Board. Enrolment was done between October of 2010 and October of 2012 and included patients on maintenance treatment with IFX or adalimumab who had serum and tissue samples drawn concomitantly with colonoscopy to assess endoscopic disease activity.
All patients were on maintenance therapy and had completed induction with IFX (5 mg/kg at weeks 0, 2 and 6) or adalimumab (160 mg at week 0 and 80 mg at week 2). The maintenance doses of IFX varied between 5 and 10 mg/kg and between every 6 and 8 weeks. For adalimumab maintenance, 40 mg was given every other week or every week. Dosing of the biologics was determined by the treating gastroenterologist. The study samples were drawn at the time of colonoscopy.
Tissue and serum samples
For each patient, samples were obtained from inflamed and normal or uninflamed tissue from the ileum and the colon. In patients with no active disease, one sample was taken from the ileum and a second sample was taken from the colon (ascending for patients with CD, and ascending and sigmoid for patients with UC). For those patients found to have inflammation, one sample was taken from the inflamed area and another was taken from normal or uninflamed tissue. All tissue samples were obtained with a biopsy forceps and placed in RNA later (Qiagen). We performed a subanalysis that found that flash frozen samples at −80°C and those in RNA later yielded similar results with respect to protein levels (data not shown). We, therefore, continued using samples frozen (−80°C) in RNA later. Serum samples were taken at the time the colonoscopy was done. De-identified samples were sent to Prometheus laboratories (San Diego, California, USA) for measurement of anti-TNF and TNF levels in serum and tissue (described below).
Evaluations and predictive variables
Independent variables considered were demographics, IBD phenotype, smoking status, body weight and body mass index, previous exposure to biologics and concomitant medications used for the treatment of IBD.
Disease phenotype was classified according to the Montreal classification.15 CD was categorised as ileal, colonic or ileo-colonic, with or without upper gastrointestinal tract involvement and stricturing or fistulising disease. UC was classified as proctitis, left-sided disease or extensive involvement (pan-colitis).
Five aminosalicylates were considered if taken for 30 or more consecutive days prior to the index date. Corticosteroid-dependent was defined as worsening of symptoms with dose reduction or discontinuation as per the treating physician. No patient was on budesonide. Rectally administered topical steroids (eg, enemas or suppositories) were not considered in the analysis. Immunomodulators (azathioprine, mercaptopurine and methotrexate) were considered if taken for ≥60 days at any dose prior to the day the samples were drawn. Previous exposure to biologics was recorded and stratified by drug (IFX, adalimumab, certolizumab pegol and/or natalizumab) and time of exposure (more or less than 1 year).
TNF and anti-TNF level measurement
TNF and anti-TNF drug levels were measured in serum as well as intestinal biopsies. TNF and anti-TNF levels in tissue were reported as raw values and also normalised level by adjusting for epithelial cell quantity in the intestinal tissue as determined by the human epidermal growth factor receptor 2 (HER2) level. Intestinal inflamed tissues in patients with IBD are comprised of epithelial cells, leucocytes and other cells, including infiltrating bacteria.16 HER2 reflects the epithelial content and therefore, can serve to normalise for epithelial cells in the setting of variable inflammation.17 Tissue lysates from flash frozen biopsies were prepared by cryofracture technique using a Covaris CryoPrep CP02 followed by lysis. The tissue particles were lysed in buffer containing Trizma pH 7.4, NaCl, EDTA, Triton X-100, Highly Accelerated Life Test Protease and Sodium Orthovanadate. For tissues stored in RNA later, the tissue was removed from the solution and then 50–150 µL of lysis buffer added, followed by vortex at maximum speed for 5 s. The sample was then vortexed every 5 min during a 20 min incubation on ice prior to aliquotting. The protein content of the tissue lysates was calculated using the BCS assay and then normalised to 1 mg/mL. TNF and HER2 were measured using a proprietary, modified ELISA assay platform. The lowest limit of quantitation for TNF in serum and tissue were 0.056 and 0.064 pg/mL, respectively.18 ,19
The homogeneous mobility shift assay (HMSA) was utilised to determine serum and tissue levels of adalimumab and IFX as well as their respective antidrug antibodies.20 In brief, the ATA-HMSA serum samples were first acid dissociated and then neutralised in the presence of IFX and adalimumab-AlexaFluor-488. The plate was incubated for 1 h at reverse transcriptase on an orbital shaker to complete the formation of the immune complexes. The equilibrated samples were filtered through a filter plate equipped with a Durapore membrane (0.22 µm; EMD Millipore, Billerica, Massachusetts, USA) into a 96-well receiver plate (Nunc, Thermo Fisher Scientific, Waltham, Massachusetts, USA). The recovered solutions were individually loaded into an high-performance liquid chromatography (HPLC) system (Agilent Technologies 1200 series HPLC system, Santa Clara, California, USA) equipped with a BioSep SEC-3000 column (Phenomenex, Torrance, California, USA). The chromatography was run at the flow rate of 1 mL/min with 1× phosphate buffered saline (pH 7.3) as the mobile phase for a total of 20 min and was monitored with a fluorescence detector at excitation and emission wavelengths of 494 and 525 nm, respectively. ChemStation Software (Agilent Technologies, Santa Clara, California, USA) was used to setup and collect data from the runs automatically and continuously. Data analysis was performed with the use of a proprietary-automated programme run on R software (R Development Core Team, Vienna, Austria). To obtain the actual drug and antidrug antibodies concentrations in the serum, the interpolated results from the standard curves were multiplied by the dilution factor. The lower limits of quantitation for adalimumab and antibodies to adalimumab were 1.6 µg/mL and 1.7 U/mL, respectively. The lower limits of quantitation for IFX and antibodies to IFX were 1.0 µg/mL and 3.1 U/mL, respectively. Values as low as 0.55 U/mL could be detected for antibodies to adalimumab and were used in the calculations of patient positivity. The upper limits of quantitation for adalimumab and antibodies to adalimumab were 50 µg/mL and 50 U/mL, respectively. High-sensitivity versions of the assays were used for tissue analysis with limits of quantitation for drug levels of 50 ng/mL. The investigators conducting the measurements were blinded to the demographic, clinical and histological data of the patient.
The primary outcome was the presence of mucosal inflammation seen during colonoscopy. Secondary outcomes were microscopic inflammatory disease activity and C reactive protein (CRP) levels. The endoscopist reported whether the biopsied mucosal tissue was inflamed, previously involved or never involved, and graded the endoscopic degree of inflamed mucosa as 0 (uninflamed), 1 (mild inflammation), 2 (moderate inflammation) or 3 (severe inflammation). For UC, the numerical score is based on the Mayo criteria.21 For Crohn’s disease, mild inflammation was classified as erythema, loss of vascular markings with or without scattered aphthae; moderate inflammation as diffuse aphthous ulcerations or larger shallow ulcerations; and severe inflammation as deep ulcerations with or without narrowing. In the case of no inflammation, the endoscopist distinguishes between never involved versus previously involved. This analysis is done in the endoscopy suite as the biopsies are taken and recorded by the research coordinator prospectively. The pathologists also independently rate the inflammation as mild, moderate or severe, which is recorded in our database. Endoscopic remission was defined as the lack of any inflammatory findings in the intestinal mucosa (erosions, ulcers, granularity or friability). Biopsies were analysed by the pathologist as part of the patient’s standard care and stratified as active or absent inflammation (defined as absence of a microscopic inflammatory infiltrate (neutrophils in the lamina propria, basal plasmocytosis and basal lymphoid aggregates)). Samples must have had no endoscopic or microscopic inflammation in order to be considered ‘uninflamed’. CRP levels were measured using a standard ELISA assay from the same serum sample drawn for the TNF/anti-TNF level analysis.
Descriptive statistics were used to examine the baseline characteristics of the study population. Continuous variables were compared using Student t test or the Mann-Whitney U test (for non-parametric variables). The χ2 test was used to evaluate distributions of categorical variables. Correlation between serum and tissue TNF and drug levels were performed using Spearman's rank correlation test. For analytical purposes, quartile analysis was used to stratify patients based on multiple variables. Paired t tests were performed to evaluate the intraindividual variability of anti-TNF tissue levels. Analysis of variance was used to analyse the differences between three or more means.
Anti-TNF levels can be accurately measured in tissue samples
As a proof of principle, we sought to determine the feasibility of measuring levels of anti-TNFs from frozen tissue or tissue that has been stored in RNA preservation solution (RNA later). We found that both flash frozen samples and those stored in RNA later yielded similar results in terms of levels of anti-TNF as well as TNF protein levels (data not shown). We analysed 30 patients for whom we had matched serum and tissue from a colonoscopy. Twenty-four (80%) had CD and 6 (20%) had UC. Twelve patients (40%) were receiving maintenance treatment with IFX and 18 (60%) were on adalimumab. The baseline characteristics of the study group are shown in table 1. For each patient, we analysed one serum sample and tissue samples from two distinct sites (60 tissue samples). We had 6 inflamed and 21 uninflamed colon samples, and 11 inflamed and 22 uninflamed ileal samples (figure 1). All uninflamed samples were taken from areas that had no previous involvement as determined by the endoscopist at the time of the procedure and were normal histologically (see Methods section).
In addition to measuring anti-TNF levels, we measured concentrations of the epithelial marker HER2 to normalise for the amount of cellular protein in samples. Epithelial marker-based normalisation method has been used in other published studies to normalise for protein levels.17 We also looked at cytokeratin and obtained similar results as HER2 normalisation. We refer to normalised TNF and anti-TNF levels when corrected for HER2 content.
To examine the accuracy of our measurements, we investigated the intraindividual variability of tissue drug levels in samples in patients with two uninflamed biopsies from different areas of the gastrointestinal tract. We also looked at the intraindividual variability of inflamed and uninflamed samples. In those patients with two uninflamed mucosal biopsies, there was no statistically significant difference between the paired samples (p=0.34; figure 2A). When comparing paired inflamed and uninflamed samples from a single patient, we found that the samples with inflammation had a significantly higher drug level when compared to their paired uninflamed sample (p=0.03; figure 2B).
Correlation of anti-TNF in serum and tissue
Given that we can accurately detect anti-TNF in tissue and our observation that there was a difference in anti-TNF levels in uninflamed versus inflamed tissue, we asked whether the tissue levels correlated with serum levels of the anti-TNF and whether the relationship between the serum level was the same in uninflamed versus inflamed tissue. There was a significant positive correlation between the drug in serum and tissue (table 2; r=0.3920, p=0.002). This was true even when restricting the analysis to specific tissue locations, that is, a significant positive correlation was seen between serum and tissue drug levels in ileum (r=0.3444, p=0.0497) and colon (r=0.4364, p=0.0258). When samples were stratified by the type of anti-TNF (ie, IFX vs adalimumab), there was a significant correlation in serum and tissue drug levels for those patients on IFX (r=0.5052, p=0.0165), but not for patients receiving adalimumab (r=0.2347, p=0.1747). Interestingly, when looking at the correlation by inflammatory state of the tissue (inflamed vs uninflamed), there was a stronger correlation between serum anti-TNF levels and uninflamed tissue (r=0.50, p<0.001), but no correlation with those samples that had inflammation (r=0.19, p=0.54). These data lead us to conclude that serum levels and tissue levels of anti-TNF are correlated, but serum anti-TNF levels better predict anti-TNF levels in normal (uninflamed) tissue rather than inflamed tissue.
Tissue anti-TNF levels correlate with the presence of increased TNF in the intestinal mucosa
We next wanted to understand why anti-TNF levels may vary in tissue from patients for a given serum concentration of anti-TNF. We hypothesised that local TNF production may retain more anti-TNF in tissue. We stratified TNF tissue levels based on degree of inflammation. The mean normalised TNF in tissue with none, mild, moderate and severe inflammation was 0.36, 1.18, 6.88 and 5.3, respectively (p=0.0042; figure 3A). There was a positive correlation between grade of mucosal inflammation and normalised TNF levels in tissue (r=0.4, p=0.003). A separate analysis was performed in order to evaluate if the severity of tissue inflammation was correlated with anti-TNF tissue levels. The mean anti-TNF/HER2 ratios in tissue with none, mild, moderate and severe inflammation were 0.93, 2.17, 13.71 and 2.2, respectively (p=0.01 for the trend; figure 3B). These data support that TNF and anti-TNF levels correlate with severity of inflammation. In more severe inflammation, tissue levels of anti-TNF are actually lower suggesting more rapid clearance.
Given that inflamed tissue has not only high levels of TNF but also higher levels of anti-TNF, we hypothesised that anti-TNF tissue levels may be low when adjusted for tissue levels of TNF. We found that normalised anti-TNF to TNF ratios in uninflamed tissue was numerically higher than in the inflamed tissue, but the difference did not reach statistical significance (17.8 vs 3.7, p=0.11). Furthermore, we found that the ratio of anti-TNF to TNF was inversely proportional to the degree of inflammation (17.8 for no inflammation, 4.2 for mild, 3.8 for moderate and 2.0 for severe). Even though there was a numerical trend, the difference did not reach statistical significance (p=0.88; figure 4). These data suggest that in patients with active inflammation, the ratio of anti-TNF in tissue to local TNF levels is decreased.
Serum to tissue anti-TNF drug discordance in patients with active disease
In clinical practice, we currently assess serum levels of the anti-TNFs. Studies have varied with respect to the serum level of anti-TNF that is considered ‘therapeutic’.22–24 Most of the patients in our cohort had levels above 3 μg/mL, which has been used as a ‘therapeutic’ cut-off in some studies in patients on IFX. Unfortunately, some patients still present with active disease, despite achieving ‘therapeutic’ levels. We hypothesised that some patients with active inflammation had a discordance between the serum levels of anti-TNF and the tissue level of anti-TNF. To perform this analysis, we divided patients into quartiles of serum anti-TNF levels and quartiles of tissue anti-TNF levels. We looked at patients in the top three quartiles of serum anti-TNF (levels of anti-TNF ranged from 6.1 to 33.1) and examined those patients in this group who had the lowest quartile of normalised anti-TNF tissue levels (0.17–0.37). We labelled these as the ‘mismatch group’. We found that 11 of the 15 (73.3%) patients with active endoscopic disease had a serum-tissue drug level mismatch while only 5 of the 15 patients (33.3%) in endoscopic remission had this finding (p=0.03). These data suggest that patients with active disease may have relatively low levels of tissue anti-TNF in spite of elevated levels of serum anti-TNF.
Systemic inflammation and anti-TNF levels
We next examined the relationship of systemic markers of inflammation and anti-TNF levels. Previous studies have found that anti-TNF levels correlate inversely with CRP levels. Our study also found a similar result. There was a significant inverse correlation between serum CRP and serum anti-TNF (r=−0.3994, p=0.0107; figure 5).
We asked whether the ratio of TNF in serum versus anti-TNF in serum correlated with systemic markers of inflammation. We divided patients into those with elevated CRP >5 mg/dL versus low CRP samples (≤5 mg/dL). We found that the ratio of TNF to anti-TNF was significantly higher in the high CRP group (mean serum TNF/serum drug 1.66 vs 0.53 in the low CRP group, p=0.013; figure 6). Normalised tissue TNF levels were also numerically higher in patients with high CRP levels as compared to low CRP samples, but did not reach statistical significance (mean normalised tissue TNF: 2.5 in CRP high vs 0.5 in CRP low, p=0.15).
One of the issues in non-invasive monitoring of patients with IBD is the sensitivity of CRP for mucosal inflammation. We had only a small number of patients who had normal CRP but active mucosal disease (5/30). We asked whether elevated TNF levels in serum could have identified these patients. We found, however, that the difference in serum TNF levels was not different between those with normal CRP who had inflammation and those with normal CRP with no inflammation, but the sample size was limited (data not shown). We did find, however, that patients with mucosal inflammation but normal CRP had significantly higher tissue TNF levels than those with no inflammation (mean normalised tissue TNF: 1.7 in normal CRP with inflammation vs 0.3 in normal CRP without inflammation, p=0.0061). Thus, neither CRP nor serum TNF could have helped to identify these patients with mucosal inflammation.
TDM is evolving as a critical part of the algorithm of managing patients with IBD who are on biologics. Studies are ongoing to determine if active monitoring and adjusting levels will ensure better outcomes for patients. However, as we move ahead incorporating TDM into clinical practice, we lack some basic understanding of the pharmacokinetics of these biologics especially in tissue. Ultimately, the goal is for patients to achieve mucosal healing and even microscopic healing. We would like, however, to have ways to predict which levels of biologics correlate best with mucosal healing.
In the current study, we examined the tissue levels of anti-TNF and how they correlate with serum drug levels, inflammatory markers and TNF in the tissue. Studies have suggested that anti-TNF is lost in the stool from the inflamed colon of patients with severe UC.25 It is also known that inflamed tissue has increased levels of TNF.4 Olsen et al4 ,26 showed that TNF is highly expressed in the inflamed intestinal mucosa of patients with UC and that IFX decreases its level. Studies have not, however, simultaneously examined tissue concentrations of anti-TNF and compared this to tissue TNF or serum levels of anti-TNF. We looked at the relationship between TNF and anti-TNF levels in serum and tissue, and their correlation with biochemical and endoscopic disease activity.
We found that inflamed tissue had significantly higher anti-TNF levels compared to matched uninflamed samples. We also established that the anti-TNF levels correlated with increased local TNF in inflamed tissues and with the severity of inflammation. Inflamed tissue also had a numerically higher level of TNF when compared to normal mucosa. Importantly, however, in moderate to severely inflamed tissue, the anti-TNF to TNF ratio was lower implying that there was insufficient anti-TNF to neutralise the TNF. These data suggest that local inflammation characterised by high levels of TNF serves as a sink for anti-TNF. Interestingly, those samples with severe inflammation (grade 3) had a lower drug level than the ones with moderate inflammation. A possible explanation for this finding is that at a certain level of mucosal inflammation and injury, there will be increased protein clearance, including losses in stool.
While there was an overall correlation between serum and tissue anti-TNF levels, we also found that patients with active IBD had a higher rate of ‘serum to tissue anti-TNF mismatch’ (patients with serum drug levels in the higher three quartiles that had adjusted tissue levels in the lowest quartile). Although our study had a limited number of patients, the data suggest that in patients with active disease, ‘high’ serum drug levels may not always translate into high tissue levels. This discordance may be due to other factors, such as rapid clearance of the anti-TNF in inflamed tissue. Thus, it is difficult to define these patients as anti-TNF failures since tissue concentrations may be insufficient for that patient. However, in certain patients it may not be practical to further increase the anti-TNF dose. This further questions the lack of a specific target serum drug level that predicts remission in all patients as shown in a previous study.24 While a ‘therapeutic tissue drug level’ is not defined, we postulate that the drug-to-TNF ratio in tissue may be more important than the drug level itself.
It is important to mention that this study presents several limitations. First, the study included a limited number of patients and this needs to be considered when interpreting the results and extrapolating them to clinical practice. Also, while the grade of inflammation in each tissue sample was graded using a predefined scale, this is not a universally validated tool.
The data presented in this study add to the weight of evidence that TDM offers the ability to guide therapy. An understanding of the variable presence of anti-TNF in tissue and the finding that, in spite of seemingly therapeutic levels of anti-TNF in the serum, patients continue to have a high ratio of TNF to anti-TNF in tissue provides a mechanistic explanation for persistent inflammation in certain patients. Larger, prospective studies should address whether TNF levels, either in serum or tissue, can help predict dosing of anti-TNFs in an individual patient.
Contributors AJY contributed to study concept and design, acquisition of the data, analysis and interpretation of the data, drafting of the manuscript, statistical analysis, and approved the final draft submitted. AJ contributed to analysis and interpretation of the data, drafting of the manuscript, critical revision of the manuscript for important intellectual content, and approved the final draft submitted. DAS contributed to study concept and design, analysis and interpretation of the data, critical revision of the manuscript for important intellectual content, and approved the final draft submitted. JSB and FP contributed to study concept and design, analysis and interpretation of the data, critical revision of the manuscript for important intellectual content, study supervision, and approved the final draft submitted. MAQ contributed to acquisition of samples and data, technical and administrative support, critical revision of the manuscript for important intellectual content, and approved the final draft submitted. RK contributed to statistical analysis and interpretation of the data, critical revision of the manuscript for important intellectual content, and approved the final draft submitted. ARD contributed to study concept and design, critical revision of the manuscript for important intellectual content, and approved the final draft submitted. SS contributed to study concept and design, critical revision of the manuscript for important intellectual content, study supervision, technical support, and approved the final draft submitted. MTA contributed to study concept and design, critical revision of the manuscript for important intellectual content, study supervision, and approved the final draft submitted.
Competing interests AJ, RK, FP and SS were employees of Prometheus laboratories at the time the study was conducted. MTA has done consultations for Abbvie Laboratories, Janssen, Prometheus laboratories and UCB.
Ethics approval University of Miami Human Subject Research Office.
Provenance and peer review Not commissioned; externally peer reviewed.
If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.