Introduction Psoriatic arthritis (PsA) is a chronic, inflammatory, musculoskeletal disease that affects up to 30% of patients with psoriasis. Current challenges in clinical care and research include personalised treatment, understanding the divergence of therapy response and unravelling the multifactorial pathophysiology of this complex disease. Moreover, there is an urgent clinical need to predict, assess and understand the cellular and molecular pathways underlying the response to disease-modifying antirheumatic drugs (DMARDs). The TOFA-PREDICT clinical trial addresses this need. Our primary objective is to determine key immunological factors predicting tofacitinib efficacy and drug-free remission in PsA.
Methods and analysis In this investigator-initiated, phase III, multicentre, open-label, four-arm randomised controlled trial, we plan to integrate clinical, molecular and imaging parameters of 160 patients with PsA. DMARD-naïve patients are randomised to methotrexate or tofacitinib. Additionally, patients who are non-responsive to conventional synthetic (cs)DMARDs continue their current csDMARD and are randomised to etanercept or tofacitinib. This results in four arms each with 40 patients. Patients are followed for 1 year. Treatment response is defined as minimal disease activity at week 16. Clinical data, biosamples and images are collected at baseline, 4 weeks and 16 weeks; at treatment failure (treatment switch) and 52 weeks. For the first 80 patients, we will use a systems medicine approach to assess multiomics biomarkers and develop a prediction model for treatment response. Subsequently, data from the second 80 patients will be used for validation.
Ethics and dissemination The study was approved by the Medical Research Ethics Committee in Utrecht, Netherlands, is registered in the European Clinical Trials Database and is carried out in accordance with the Declaration of Helsinki. The study’s progress is monitored by Julius Clinical, a science-driven contract research organisation.
Trial registration number EudraCT: 2017-003900-28.
- Magnetic Resonance Imaging
This is an open access article distributed in accordance with the Creative Commons Attribution Non Commercial (CC BY-NC 4.0) license, which permits others to distribute, remix, adapt, build upon this work non-commercially, and license their derivative works on different terms, provided the original work is properly cited, appropriate credit is given, any changes made indicated, and the use is non-commercial. See: http://creativecommons.org/licenses/by-nc/4.0/.
Statistics from Altmetric.com
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.
Strengths and limitations of this study
Our multiomics systems medicine approach integrates molecular, imaging and clinical data, which facilitates identification of pretreatment profiles that are associated with disease-modifying antirheumatic drug response in psoriatic arthritis (PsA).
We use a two-step data analysis approach to both discover and validate predictive profiles.
Sensitive imaging techniques are used to evaluate treatment response at multiple time points, enabling comparison with conventional response measures.
Although the TOFA-PREDICT includes therapies with three different mechanisms of action (methotrexate, a tumour necrosis factor inhibitor and a Janus kinase inhibitor), this does not cover the full therapeutic armamentarium available for PsA.
The two-step approach with discovery and validation bisects the cohort, leading to reduced sample size per treatment group.
Psoriatic arthritis (PsA) is a chronic, autoinflammatory and autoimmune musculoskeletal disease that affects up to 30% of patients with psoriasis.1 It is considered a heterogeneous disease, as patients have a variable disease course and clinical phenotype.1–4 The hallmarks of PsA include cutaneous psoriasis, nail dystrophy, peripheral arthritis, axial spondyloarthritis, dactylitis and enthesitis.1–3 PsA may also feature extramusculoskeletal manifestations and comorbidities that impact overall morbidity and mortality, including anxiety, depression, uveitis, inflammatory bowel disease, metabolic syndrome and cardiovascular events.5–11
PsA can cause severe joint damage early in the disease course, contribute to functional disability and chronic pain and, as such, negatively impact quality of life.2 4 12–14 Delayed treatment initiation is associated with progression of joint erosions, decreased long-term physical function and reduced risk of medication-free remission.13–16 A delayed diagnosis of 6 months may already negatively impact physical function and joint erosions.14 These data highlight the necessity of timely initiation of effective treatment with disease-modifying antirheumatic drugs (DMARDs).17 18
Challenges in treatment and assessing response to therapy
The care for patients with PsA faces several challenges.19 The first challenge arises in unravelling the mechanisms that underlie pathogenesis. Although over the past 15 years many researchers have studied its complex aetiology, the exact molecular mechanisms underpinning PsA pathogenesis remain unknown.3 20 It is important to improve our understanding of the genetic, environmental and immune-mediated factors that initiate and maintain the disease, as discoveries about dysregulated immunological pathways can facilitate the development of new therapies. For example, identification of the implications of the tumour necrosis factor (TNF) alpha and interleukin (IL)-23/IL-17 pathways have led to rapid development of effective therapeutic agents.1 3 Moreover, stratification of patients with inflammatory arthritis by immunological phenotype for selection of therapy has shown promise. For example, favourable treatment response in patients with PsA who were stratified based on circulating T helper cell profiles has been reported.21 In rheumatoid arthritis, a machine learning (ML) model based on divergent transcriptional signatures in peripheral blood mononuclear cells (PBMCs), monocytes and CD4+ T cells was reported to predict treatment response in adalimumab or etanercept (ETN)-treated patients.22 These examples underline how unravelling disease pathogenesis may improve clinical practice.
The second challenge comprises a lack of methods to select the optimal treatment for each patient.4 12 23 Evidence-based treatment strategies for PsA were developed by the European Alliance of Associations for Rheumatology (EULAR) and the Group for Research and Assessment of Psoriasis and Psoriatic Arthritis (GRAPPA). However, treatment response rates are disappointing.24 25 Up to 40% of patients respond insufficiently to a first DMARD, and strongly divergent drug responses are observed.3 4 12 Although conventional synthetic (cs)DMARDs are frequently used as first-line therapy, there is limited evidence available on their effectiveness in PsA.26–28 Moreover, the number of csDMARDs, biological (b)DMARDs and targeted synthetic (ts)DMARDs is rapidly increasing and head-to-head trials are scarce.23 29–31 Hence, clinicians have no tools at their disposal to predict which DMARD will be effective for an individual patient.23 This lack of precision medicine is a clinically relevant problem for a potentially aggressive disease that may impact quality of life, affect multiple organ systems, has an economic burden on the healthcare system and demands costly treatment that potentially causes adverse events.12–19 21
The third challenge comprises the wide array of novel imaging modalities and the growing number of analytical methods that have become available for the evaluation of therapy response in PsA. Conventional radiography lacks sensitivity, especially in patients with early disease in whom little radiographic abnormalities are observed.32 Furthermore, the visual interpretation of medical images is time consuming, bound with interobserver variation and limited to semiquantitative outcomes that may be insensitive to detect small changes over time. On the contrary, computer-based medical image analysis can generate uniform, quantitative results in a (semi)automatic manner. Adding these techniques in trials and in clinical practice may add to unravelling mechanisms as well as improvement of treatment.
Overall, there is an urgent clinical need to assess and understand the cellular and molecular pathways underlying DMARD treatment response in PsA. To this end, the TOFA-PREDICT trial was designed. In this investor-initiated, phase III, multicentre, four-arm randomised trial, a multiomics systems medicine approach is used to integrate pretreatment clinical, transcriptomic, metabolomic, proteomic, flow cytometric and imaging data to discover profiles of patients with PsA that predict response to tofacitinib (TOF), as compared with methotrexate (MTX) and ETN. By expanding our knowledge of the underlying mechanisms, course and treatment response, the TOFA-PREDICT study also aims to identify novel biomarkers for diagnosis and disease monitoring.3 19
In the TOFA-PREDICT trial, sensitive imaging techniques, including MRI and fluorine-18-fluorodeoxyglucose positron emission tomography/CT (18F-FDG PET/CT), are applied to monitor disease activity. The current trial can deliver important data on the value of these more advanced imaging methods. With the use of ankle MRI scans early, possibly reversible and inflammatory features of PsA can be visualised at the heel, which is the most frequently affected site for enthesitis in PsA.33 34 Moreover, 18F-FDG PET/CT might aid in the measurement of local and systemic inflammation in PsA, including (peri)articular and vascular inflammation.
Identify pretreatment profiles with integrated clinical, transcriptomic, metabolomic, proteomic, flow cytometric and imaging data that predict response to treatment with TOF in DMARD-naïve and DMARD-non-responsive patients with PsA.
Compare clinical efficacy of treatment with TOF, MTX and ETN in DMARD-naïve and DMARD-non-responsive patients with active PsA.
Compare structural response to treatment of active PsA with TOF, MTX and ETN using (semi)quantitative ankle MRI outcomes, radiographic outcomes and 18F-FDG PET/CT outcomes.
Determine (medication-specific) molecular mechanisms predicting and underlying clinical response to TOF, in comparison to MTX, and ETN in active PsA.
Methods and analysis
TOFA-PREDICT is a multicentre (seven), investigator-initiated, phase III, open-label, four-arm randomised controlled study conducted in the Netherlands. A total of 160 patients with PsA who fulfil the Classification Criteria for Psoriatic Arthritis will be included in two groups, each with two treatment arms.35 The first group consists of DMARD-naïve patients, who are randomised to MTX (arm 1) or TOF (arm 2). The second group consists of DMARD-non-responsive patients, who continue csDMARD background therapy and are randomised to addition of ETN (arm 3) or TOF (arm 4).
Eligibility criteria are displayed in table 1. The TOFA-PREDICT trial started on 4 April 2018 and the scheduled end date is 1 July 2025. By the end of 2022, inclusion of the first cohort of 80 patients is completed. The evaluation of the first cohort will be initiated early 2023.
The first group of patients are DMARD naïve and have active PsA. Typically, these patients are at an early stage of PsA. Patients are randomised to receive either MTX monotherapy 25 mg once a week, subcutaneously (standard of care therapy, arm 1), or TOF monotherapy 5 mg two times per day, orally (investigational therapy, arm 2). Randomisation is performed per site in computer-generated random blocks. Patients will be assessed according to a predefined schedule of regular study visits (table 2). In case of treatment failure (see the ‘Treatment failure’ section), combination therapy will be initiated: patients randomised to MTX will also start TOF and vice versa. If drug intolerance warrants discontinuation of the drug, a switch will be made to the alternate drug as monotherapy (TOF to MTX and vice versa).
The second group of patients are non-responders to previous treatment with either MTX, leflunomide or sulfasalazine, or to previous treatment with combination therapy of a csDMARD and one previous bDMARD. A history of one bDMARD prior to inclusion is allowed, except for prior use of ETN. Prior use of a tsDMARD (Janus kinase inhibitor, abatacept) is also not allowed. Only patients who have had secondary treatment failure to a TNF inhibitor (TNFi), defined as initial good response, but diminished clinical efficacy over time, are eligible to participate in the study.36 These DMARD non-responders continue background therapy with csDMARD and are randomised to receive the addition of either ETN 50 mg once a week, subcutaneously (arm 3), or TOF 5 mg two times per day, orally (arm 4). ETN was chosen as it was reimbursed and no preference for a specific TNFi is mentioned in current EULAR and GRAPPA international guidelines for the treatment of PsA.24 25 In the event of treatment failure or drug intolerance (see the ‘Treatment failure’ section), a switch from ETN to TOF or vice versa will be made (figure 1).
Study visits are performed at baseline and weeks 4, 16, 26, 39 and 52. Each study visit comprises multiple study assessments (a schematic overview is depicted in table 2). From week 16 onwards, the American College of Rheumatology (ACR)50 score is calculated every study visit to determine treatment failure.37 The ACR50 score is described in the Outcomes section. Patients are evaluated additionally to the above-described visits according to regular clinical practice, including blood sampling for safety measurements according to regular practice. During all visits, adverse events and serious adverse events (SAE) are documented with respect to safety.
Treatment failure is defined as failing to achieve an ACR50 response on two consecutive visits from week 16 onwards. If a patient does not attain the ACR50 response at a regular study visit, an additional study visit is scheduled 4 weeks later. At this ‘treatment failure’ visit the ACR50 response is reassessed. In the event that the ACR50 response is again not attained, ‘treatment failure’ is confirmed and a crossover to the alternate treatment protocol within that study group takes place (figure 1). A minimum washout of 1 week will be applied to patients switching from TOF to ETN (or vice versa). If the ACR50 response is attained at the ’treatment failure’ visit, regular 12-week visit intervals will continue and the patient will not switch therapy. In addition, drug intolerability that warrants discontinuation (eg, side effects, laboratory abnormalities) is defined as treatment failure at any time point. In the case of MTX, dosage lowering is the first step in case of drug intolerability. For ETN and TOF, dosage changes are not possible and drug intolerability indicates treatment failure. Crossover will not take place in the last 3 months of follow-up.
MTX dosage adjustments
MTX is initiated in the DMARD-naïve arm at a dosage of 15 mg/week subcutaneously. The dosage is increased to 25 mg/week after 4 weeks, unless the ACR50 response is attained or side effects prevent safe dosage escalation. By increasing the dosage to 25 mg/week at week 4, the primary end point of the study can be compared between MTX and TOF at week 16 (ie, 12 weeks of administering the maximal dosage of MTX). MTX dosage may be reduced during follow-up if ACR50 has been attained and/or if side effects occur, in accordance with standard clinical care.
In accordance with standard clinical care, the following escape therapies are allowed: non-steroidal anti-inflammatory drugs, intra-articular corticosteroid injections and, from week 24 onwards, topical corticosteroids.
End of study
After 52 weeks of follow-up, all patients will resume regular clinical care while continuing the DMARD therapy that was initiated during the study. Treatment in regular care will also be resumed by patients who discontinue trial medication due to (serious) adverse events, treatment failure after crossover or other reasons. From these patients, we will only collect a selection of data after 52 weeks of follow-up (table 2: footnote §).
Data collection and samples
All collected clinical data are entered in an online database (research online; Julius Center University Medical Center (UMC) Utrecht) designed for the TOFA-PREDICT trial. Blood samples for the multiomics analyses are collected at several time points throughout the study (figure 1). In addition, blood samples are taken to monitor drug safety after the start of MTX, TOF or ETN. Blood samples for the multiomics analyses are collected at seven different study sites. After protocolised transport, all blood samples are processed in a standardised way in the UMC Utrecht. The samples are pseudoanonymised and after magnetic-activated cell sorting, PBMC subsets are stored. Additionally, serum, plasma and PBMC subset lysates are stored. All blood samples for multiomics analyses are registered with Quaero Systems. The multiomics analyses of the stored samples are performed in batches at a later stage, taking confounders such as treatment arm, visit number and demographics into account. All data are integrated at the Data Research Environment (anDREa). The omics data will be made available in public databases after primary analyses and publication.
Patient and public involvement
Patients were not involved in the development of the research question, the design and conduct of the study, choice of outcome measures nor recruitment.
Systems medicine approach
The primary objective is to discover and validate pretreatment clinical, transcriptomic, metabolomic, proteomic, flow cytometric and imaging profiles that predict treatment response. Response and non-response are defined as attaining or not attaining minimal disease activity (MDA), respectively, after 16 weeks of treatment. To define these profiles, a multiomics systems medicine approach will be used for which transcriptomic, metabolomic, proteomic and flow cytometric data are collected. Transcriptomic and flow cytometric analysis will be performed on PBMC (subset)s. Proteomic and metabolomic analyses will be performed on serum and/or plasma samples. These molecular and cellular data will be added to the clinical, structural and imaging data (ankle MRIs, whole-body 18F-FDG PET/CT and radiographs of the hands and feet). Systems medicine data analyses will be used to combine the different omics layers in our attempt to identify profiles that predict treatment response.
Clinical efficacy measures
We use MDA at week 16 as the primary outcome for the identification of molecular and cellular profiles that predict treatment response. MDA is a validated, PsA-specific composite measure that includes evaluation of arthritis (tender and swollen joint count), skin disease (Psoriasis Area and Severity Index and body surface area), enthesitis and patient-reported outcomes (Health Assessment Questionnaire (HAQ), visual analogue scale (VAS) for pain and VAS for patient global assessment).38 39 The clinical relevance of composite measures that include multiple disease domains has become increasingly evident over recent years.38–40 To define treatment failure, we use the ACR50 response because treatment effect during follow-up is most commonly detected as a change from baseline. ACR50 is a composite measure defined as 50% improvement in the number of both swollen and tender joints, next to 50% improvement in at least three of the following outcomes: HAQ, acute phase reactant (we use C-reactive protein), VAS for patient global assessment, VAS for physician global assessment and VAS for pain.37 41 42 We calculate the ACR50 every 12 weeks starting from week 16. Moreover, we assess dactylitis, blood pressure, body mass index, laboratory parameters and additional patient-reported outcomes, and calculate additional PsA-specific composite indices.43
At baseline, weeks 4, 16, 26, 39, 52 and at treatment failure visits, patients fill out online questionnaires to monitor disease activity and their mental and physical health. TOFA-PREDICT employs the following questionnaires: Assessment of SpondyloArthritis International Society Health Index, Dermatology Life Quality Index, EuroQol-5 Dimension Scale, HAQ, Self-Administered Psoriasis Area and Severity Index, the Work Productivity and Activity Impairment Questionnaire and two VAS scores to assess pain and the patients’ global assessment.44–49
Three imaging techniques are applied in the TOFA-PREDICT study: MRI scans of both ankles, whole-body 18F-FDG PET/CT and conventional radiography of the hands and feet. At baseline, week 16 and week 52, MRI scans of both ankles are obtained. MRI scans are performed using MR equipment with a field strength of 1.5 or 3 T. The ankles are scanned separately using an extremity coil. The MRI protocol was developed in accordance with the European Society of Musculoskeletal Radiology recommendations and contains the following sequences: 3D proton density with fat suppression (FS), transversal T1 turbo spin echo and 3D T1 FS before and after intravenous gadolinium injection.50 The estimated total time in the MRI room is <60 min per patient per visit. Ankle MRIs are visually evaluated using Psoriatic Arthritis Magnetic Resonance Imaging Scoring System (PsAMRIS), adapted for the heel, and (Heel Enthesitis Magnetic Resonance Imaging Scoring System (HEMRIS) measures.33 51 Using deep learning, quantitative outcome measures for ankle MRIs will be developed aiming to quantify (peri)articular inflammatory joint changes such as synovitis, bone marrow oedema and enthesitis.
At baseline and week 52, whole-body 18F-FDG PET/CT scans are obtained. ¹⁸F-FDG is administered intravenously after an overnight fast. Dosing of ¹⁸F-FDG depends on local guidelines. After administration of ¹⁸F-FDG, the 18F-FDG PET/CT is performed 1 hour later. A non-contrast-enhanced low-dose CT is performed for attenuation correction. In this multicentre trial, all PET/CT reconstructions are compliant to European Association of Nuclear Medicine Research Ltd guidelines in order to achieve comparable quantitative outcome parameters, such as standardised uptake values.52 The main 18F-FDG PET/CT outcome measures are vascular and (peri)articular inflammation.
At baseline and at week 52, radiographs of hands and feet are acquired. Radiographs of hands and feet are evaluated using the PsA-modified Sharp-van der Heijde score.53 MRI, 18F-FDG PET/CT and radiography observers are blinded to diagnosis and treatment.
Sample size calculation
The primary objective of TOFA-PREDICT is to predict the treatment response (attaining or not attaining MDA after 16 weeks of treatment in active PsA) using the multiomics analysis of pretreatment omics data. To evaluate the sample size needed to detect differentially expressed genes/proteins (DEGPs) between responders and non-responders we simulated several scenarios. These scenarios used a range of number of prognostic genes (50–500), dispersion (0.1–0.5) and false discovery rates (FDR; 0.01–0.1) within each scenario assuming a minimum fold change in DEGPs of 2, 80% power and testing of a total of 20 000 genes with a mean expression (read count) of 50. Separate analyses were performed for an equal distribution between responders and non-responders (50:50) and for unequal distributions of responders and non-responders (40:60 and 25:75). Results in the scenario assuming 400 differentially expressed genes, an FDR of 0.05 and an unequal distribution between responders and non-responders (40:60) assuming dispersion values as found in previous RNA-seq data from our group (eg, CD14+ monocytes, dispersion value 0.11) resulted in a sample size of 20 patients per arm. Therefore, we assumed a sample size of 80 (20 patients per arm) to be sufficient to detect relevant expression signatures. Sample size was calculated using the R package ‘RnaSeqSampleSize’ (V.3.6.1).54 For other omics platforms, required sample sizes are considered smaller based on the smaller number of markers (eg, proteins up to 180 and metabolites up to 800). To enable external validation, a similar cohort will follow the first 80 patients up to a total of 160 included patients.
Systems medicine approach
Different layers of baseline omics data will be analysed separately and will be integrated with clinical parameters (eg, gender, disease duration, etc), patient-reported parameters and imaging data for the discovery and validation of molecular and cellular signatures that serve as biomarkers to predict treatment response after 16 weeks of treatment (primary end point). Furthermore, molecular signatures will be computed using omics data collected at weeks 4 and 52 (or treatment failure) in addition to baseline data. We will explore the molecular signatures using bioinformatic approaches. The observations made during the exploration of the data will guide the choice of tools and algorithms for the next step of the data analysis.55 For each analysis step, we will perform permutation analysis and k-fold cross-validation to test the reliability of the molecular signature. Moreover, we will integrate multiomics data to discover molecular signatures that are supported by different layers of data, strengthening the reliability of the discovered signature. For prediction at baseline, the expression (ie, fold change) of the separate omics layers will be analysed. Thereafter, using resulting relevant expression signatures in addition to established clinical and imaging predictors as features, we will build integrated and internally validated ML models to predict response to TOF and separately response to MTX and ETN. A final statistical analysis plan (SAP) will be defined prior to database lock using the optimal techniques for analysing expression profiles and optimal ML models to use. Genes or gene modules from these signatures and models will bring forth new hypotheses that can be verified experimentally, contributing towards a better understanding of the disease mechanisms and a predictive model for disease outcome and therapy response.
After inclusion of the first 80 patients (~20 patients per group), the first step of the predictive multiomics analysis will be performed. Of all the available multiomics data, predictive biomarkers are identified as either relevant (statistically significant), irrelevant (statistically insignificant) or promising (based on clinical and scientific reasons without formal statistical significance). For each omics platform, an optimal predictive assay for treatment response will be developed. Also, all relevant biomarkers will be integrated in multiomics approaches and added to clinical data and structural imaging data to develop an exploratory prediction model for treatment response. To externally validate the identified biomarkers, we implement a second step in the analysis. Both the relevant and promising biomarkers will be analysed in the subsequent cohort of 80 patients to replicate the results from the first phase. The proposed omics assays from the first cohort will be validated in the second cohort. Finally, the combined relevant and promising biomarkers of all 160 patients will be integrated in multiomics approaches and added to structural imaging data and clinical data to develop a final and clinically applicable prediction model using pretreatment markers. In this phase, the added predictive value of omics markers over known, easily available (clinical) baseline predictors will also be assessed.
Clinical efficacy and structural response
Efficacy of treatment and imaging outcomes will be compared between different treatment arms using logistic or linear regression analyses taking into account established prognostic indicators (such as structural damage, elevated acute phase reactants and polyarthritis, to be finalised in the SAP) and centre (as the stratification factor used in randomisation). The significance level (α) will be set at 0.05, with p values less than or equal to α considered statistically significant.
Missing data and SAEs
Cases that are lost to follow-up and other missing data will be presented descriptively. If the percentage of missing data exceeds 5%, multiple imputation will be performed, based on data type and quantity of the missing data. For binary secondary drug efficacy outcomes, missing data will be defined as non-response to prevent overestimation of the effect. SAEs and suspected unexpected serious adverse reactions will be reported descriptively.
Ethics and dissemination
Ethics approval and informed consent
The study was approved by the Medical Research Ethics Committee in Utrecht, Netherlands (MREC reference number: NL63439.041.17), and is carried out in accordance with the Declaration of Helsinki. The trial is registered in the European Clinical Trials (EudraCT) Database (reference number: 2017-003900-28). All participants provided written informed consent. The study progress is monitored by a science-driven contract research organisation (Julius Clinical).
The results of the primary and secondary objectives of the study will be published in international peer-reviewed journals and on national and international scientific conferences.
Patient consent for publication
We would like to thank Timothy Radstake for initiating the study. Furthermore, we thank Anneloes van Loo, Karin Schrijvers, Ria Boot and Khalid Laajab for their valuable contributions to study logistics and study assessments. Also, we thank Anne Karien Marijnissen and Marieke Vianen for their help in legal and ethical affairs, for data management, for management of finances and for study coordination. We thank Michel Olde Nordkamp, Cornelis Bekker, Rina Kommer-Wichers and Arno Concepcion, research technicians, for their supervision and dedicated work in the laboratory. Lastly, we want to acknowledge all rheumatologists (in training), radiologists and nuclear medicine physicians (in training), research nurses, pharmacists, pharmacy assistants and laboratory technicians from the participating centres that have made the conduction of this study possible.
NJK, FTP, JNP and NLAV contributed equally.
Collaborators TOFA-PREDICT author group: Astrid van Tubergen (Department of Medicine, Division of Rheumatology, Maastricht University Medical Center, Maastricht, the Netherlands; Care and Public Health Research Institute (CAPHRI), Maastricht University, Maastricht, the Netherlands), Kasper Hermans (Department of Medicine, Division of Rheumatology, Maastricht University Medical Center, Maastricht, the Netherlands; Care and Public Health Research Institute (CAPHRI), Maastricht University, Maastricht, the Netherlands), Harold Vonkeman (Department of Rheumatology and Clinical Immunology, Medisch Spectrum Twente Hospital, Enschede, The Netherlands), Marthe Hettema (Department of Rheumatology and Clinical Immunology, Medisch Spectrum Twente Hospital, Enschede, The Netherlands), Marc Kok (Department of Rheumatology and Clinical Immunology, Maasstad Hospital, Rotterdam, The Netherlands), Radjesh Bisoendial (Department of Rheumatology and Clinical Immunology, Maasstad Hospital, Rotterdam, The Netherlands), Arno van Kuijk (Amsterdam Rheumatology and Immunology Center/Reade, Amsterdam, The Netherlands; Academic Medical Center, University of Amsterdam, Amsterdam, The Netherlands), Jim Wiegel (Amsterdam Rheumatology and Immunology Center/Reade, Amsterdam, The Netherlands; Academic Medical Center, University of Amsterdam, Amsterdam, The Netherlands), Tim Jansen (Department of Rheumatology and Clinical Immunology, VieCuri MC, Venlo, The Netherlands), Antoaneta Comarniceanu (Department of Rheumatology and Clinical Immunology, VieCuri MC, Venlo, The Netherlands), Lenny Geurts – van Bon (Department of Rheumatology and Clinical Immunology, ZGT, Almelo, The Netherlands), Lydia Schipper (Department of Rheumatology and Clinical Immunology, Elisabeth-TweeSteden Hospital, Tilburg, The Netherlands), Sandra van Bijnen (Department of Rheumatology and Clinical Immunology, Elisabeth-TweeSteden Hospital, Tilburg, The Netherlands), Siska Wijngaarden (Department of Rheumatology and Clinical Immunology, Gelre Hospitals, Apeldoorn, The Netherlands).
Contributors All authors substantially contributed to discussion, reviewing, revising and improvement of the protocol and writing of the manuscript before submission. FTP, NJK, JNP and NLAV are the lead investigators. FL and MH are the principal instigators. JT and JS provided clinical input and supervision of research clinicians. EFAL contributed to study design. PW is responsible for methodology and statistics in the design and analysis of the study. MH, PW, FL and SAYH contributed to the analysis plan and processing of blood samples. WF, MPJ, SA, FL and PAdJ are responsible for imaging in design and completion of the study. In each participating site, a lead investigator (rheumatologist) is responsible for identification, recruitment, data collection and completion of CRFs, follow-up of study patients, adherence to study protocol and investigators’ brochure and participation in feedback sessions during the conduct of this study. Lead investigators and other contributing colleagues in the participating sites are listed in the TOFA-PREDICT author group.
Funding This work was supported by Pfizer (New York, New York, USA). The collaboration project is cofunded by the PPP Allowance made available by Health~Holland, Top Sector Life Sciences & Health, to stimulate public–private partnerships (grant number: LSHM17074).
Competing interests None declared.
Patient and public involvement Patients and/or the public were not involved in the design, or conduct, or reporting, or dissemination plans of this research.
Provenance and peer review Not commissioned; externally peer reviewed.