Article Text


Extended report
Adalimumab in patients with hand osteoarthritis refractory to analgesics and NSAIDs: a randomised, multicentre, double-blind, placebo-controlled trial
  1. X Chevalier1,
  2. P Ravaud2,
  3. E Maheu3,
  4. G Baron2,
  5. A Rialland4,
  6. P Vergnaud5,
  7. C Roux6,
  8. Y Maugars7,
  9. D Mulleman8,
  10. C Lukas9,
  11. D Wendling10,
  12. P Lafforgue11,
  13. D Loeuille12,
  14. V Foltz13,
  15. P Richette14
  16. On behalf of the French section of osteoarthritis
  1. 1Department of Rheumatology, Henri Mondor Hospital, Creteil, France
  2. 2INSERM, AP-HP (Assistance Publique des Hôpitaux de Paris), Hôpital Hôtel Dieu, Centre d'Epidémiologie Clinique, Univ. Paris Descartes, Sorbonne Paris Cité, Faculté de Médecine, Paris, France
  3. 3Department of Rheumatology, Hopital Saint Antoine University Paris VI, Paris, France
  4. 4Unité de Recherche Clinique, Hopital Henri Mondor, UPEC Paris XII
  5. 5Laboratoire Synarc, Lyon, France
  6. 6Department of Rheumatology, Hopital de l'Archet, University Nice, France
  7. 7Department of Rheumatology, University Nantes, France
  8. 8Department of Rheumatology, Hopital Trousseau, University Tours, France
  9. 9Department of Rheumatology, Hopital Lapeyronnie, University Montpellier, France
  10. 10Department of Rheumatology, Centre Hospitalier Universitaire de Besançon, Université de Franche-Comté, Besançon, University of Besancon, France
  11. 11Department of Rheumatology, Hopital de la Conception, University of Marseille, France
  12. 12Department of Rheumatology, Hopital Brabois, University of Nancy, France
  13. 13Department of Rheumatology, Hopital La Pitié, University Paris VI, France
  14. 14Université Paris Diderot, Assistance Publique-Hôpitaux de Paris, Hôpital Lariboisière, Fédération de Rhumatologie, Paris, France
  1. Correspondence to Professor Xavier Chevalier, Department of Rheumatology, Henri Mondor Hospital, Creteil 94010, France; xavier.chevalier{at}


Aim To test the efficiency of tumour necrosis factor blockers (adalimumab) in patients with painful refractory (non-responders to analgesics and non-steroidal anti-inflammatory drugs (NSAIDs)) hand osteoarthritis (OA).

Methods We performed a randomised, double-blind, placebo-controlled, parallel group, multicentre study. Patients were randomised to: 1/1 adalimumab 40 mg for two subcutaneous injections at a 15-day interval or placebo and monitored for 6 months. The primary outcome was the percentage of patients with an improvement of more than 50% in global pain (Visual Analogue Scale) between week 0 (W0) and week 6 (W6). Secondary outcomes included the number of painful joints, swollen joints, morning stiffness duration, patient and practitioner global assessments, functional indexes for hand OA, and consumption of analgesics. Analysis on the mean primary outcome measure was done on patients who received at least one injection.

Results 99 patients were recruited and 85 patients were randomised. Among them, 37 patients in the placebo group and 41 in the adalimumab group received at least one injection and were evaluated at W6 (n=78) on the main efficacy outcome. Mean age was 62 years, 85% were women, and mean level of pain was 62 mm at W0. At W6, 35.1% in the adalimumab group versus 27.3% in the placebo group had a pain reduction ≥50% (RR 1.12 (95% CI 0.82 to 1.54; p=0.48). There were no statistical differences for all secondary end points. The rate of adverse events was similar in the two groups.

Conclusions Adalimumab was not superior to placebo to alleviate pain in patients with hand OA not responding to analgesics and NSAIDs.

Trials registration number NCT00597623.

  • Hand Osteoarthritis
  • Anti-TNF
  • Treatment

Statistics from


Digital hand osteoarthritis (OA) involving the proximal and distal interphalangeal (IP) joints is the most frequent form of OA, affecting middle-aged women and the elderly (reaching a radiological prevalence of 80%).1 ,2 Symptomatic hand OA in subjects aged more than 70 years affects 13% of men and 26% of women.3 Among symptomatic patients, there is a subset of difficult-to-treat patients, frequently women, who present a high level of pain and disability despite an optimal combination of non-pharmacological and pharmacological treatments, as recommended.4–6 This clinical setting may be associated with the presence on X-rays of central subchondral erosions in IP joints and is named erosive hand OA.4

In this setting, several disease-modifying drugs used in inflammatory arthritis such as methotrexate, sulfasalazine and hydroxychloroquine have been tried, but without any proven efficacy.7 ,8

Therefore, there is a real need for an emerging effective therapy in hand OA where the impact on the patient's daily quality of life could be as severe as in rheumatoid arthritis.9 ,10 Patients suffering from such a severe form of hand OA may develop further the feeling of a neglected disease.11 ,12

Tumour necrosis factor (TNF) α is a well-known pro-inflammatory cytokine produced by the synovial cells and chondrocytes in OA which, together with interleukin (IL)-1 b, plays a pivotal role in cartilage matrix degradation.13–18 Beyond its pro-inflammatory effects, TNF a further activates sensory neurons through PGE2 release, and increases the expression of the neurotrophin nerve growth factor, which is a key mediator of pain in OA, present in synovial fluid and inflamed synovial membrane.19 ,20

In addition, because patients with severe hand OA may present signs of inflammation in the IP joints that can mimic peripheral psoriatic arthritis (PsA), with synovitis seen in ultra-sonography or by MRI, it has been hypothesised that TNF α may play an important role in this disease.21–23

In vitro, blocking TNF action decreases production of enzymes and pro-inflammatory mediators in OA cartilage explants.24 Targeting synovial membrane inflammation by TNF α blockers has shown dramatic effect on symptoms and structure in patients with PsA and rheumatoid arthritis.25 By contrast, there are very little data on the effect of anti-TNFα in patients with hand OA and no well conducted randomised controlled trials (RCTs) focused on evolution of symptoms.26–28

The objective of this study was to test the hypothesis that TNF α blockers have a symptomatic efficacy in patients suffering from hand OA with a high level of pain and no response to analgesic and non-steroidal anti-inflammatory drugs (NSAIDs), in order to face a pragmatic situation where the practitioners miss any efficient therapeutic issues.


Trial design

This is a randomised, double-blind, parallel group, placebo-controlled, 6-month, multicentre trial (16 active sites among 20) testing the efficacy of a TNF α blocker adalimumab (two subcutaneous injections of 40 mg each at a 2-week interval) in patients with hand OA who do not respond to analgesics and NSAIDs (the so-called refractory hand OA), according to international guidelines. The protocol and amendments were approved by the Independent Ethics Committees from Henri Mondor Hospital (Creteil, France). As required by legislation, the research was authorised by the French Regulatory Agency. The study was conducted in accordance with the International Conference on Harmonization Guideline for Good Clinical Practice and the Declaration of Helsinki. This trial has been registered at (n° NCT00597623).

This is an academic study sponsored by the Assistance Publique-Hopitaux de Paris.

Eligibility of patients

Eligible patients were adults with painful hand OA meeting the classification criteria of the American College of Rheumatology for hand OA and not responding to analgesics and NSAIDs.29 Inclusion criteria included: aged between 40 and 80, at least three IP joints symptomatic for more than 3 months, a level of pain over 40 mm (global pain in the last 24 h), non-responding to analgesics level 1 or 2, and to NSAIDs for at least 10 days in the past 3 months. X-rays of the hand should have shown at least three IP joints with OA features (Kellgren and Lawrence superior or equal to grade 2). Women of child-bearing age should use an effective contraceptive method. Patients were recruited by referral and gave their informed consent to participate after receiving comprehensive information.

Exclusion criteria

They included women of child-bearing age who are not using contraception; previous therapies by TNF α blockers; secondary hand OA (to previous inflammatory diseases) and any painful syndrome of the upper limb which may interfere with evaluation of hand pain; inflammatory diseases and contraindications to anti-TNF α therapies such as acute or chronic infectious states, latent tuberculosis meaning either history of a primo infection (positive tuberculin test=10 mm following 5UI as stated by the French security agency) or risk of reactivation of silent tuberculosis (previous history of untreated tuberculosis, recent contact with a person suspected of having active tuberculosis, phlyctenular or positive tuberculin test, pulmonary sequel of tuberculosis); infection with HIV or hepatitis viruses (C and B); conditions with increased risk factors for infectious disease such as chronic cutaneous ulcers, urinary catheters and previous infection from any medical devices; cancers or malignant blood diseases except previous cancer history if cured for at least 5 years, baso cellular cutaneous cancer; any precancerous conditions (intestinal polyps, cervical dysplasia, the start of monoclonal gammapathy proliferation, myelodysplasia syndromes); demyelinating disease of the central nervous system, severe cardiac insufficiency and high level of antinuclear antibodies (<1/800); leucopenia; kidney insufficiency (creatine clearance <50 mL: min); hepatic cytolysis; skin disease which contraindicated subcutaneous injections and cutaneous psoriasis; anticoagulant therapies, scheduled surgery for hand OA or any surgical procedure anticipated within the next 6 months; intra-articular injections of corticosteroids in the past month or hyaluronic acid (HA) intra-articular injection in IP joints in the 3 months prior to joining the study; slow-acting drugs for OA (soybean and avocado extracts, glucosamine, chondroitin, diarcerhein) started less than 6 months prior to the study; other treatments for hand OA such as methotrexate, hydroxychloroquine and salazopyrine; colchicine; oral corticosteroids; psychiatric disorders; unstable diabetes; patients who are not registered with social security; and history of alcohol abuse.

Study setting

The study was conducted in 16 rheumatological departments from tertiary care hospitals centres in France.


During the preinclusion period, after screening and before the first injection in the following 7–28 days, eligibility criteria were checked. At day 0 (named W0), the patient received the first injection. The second injection was performed 15 days later (named W2) and the following assessment visits were done at weeks 6, 10, 14 and 26. At the W0 visit, each patient underwent a physical examination and vital signs were recorded. At each visit, items representing secondary efficacy end points were recorded (listed in the Clinical variables section).

Prohibited/authorised treatments

For all visits, analgesics and NSAIDs should be stopped 72 h before. Authorised treatments included only acetaminophen (up to 4 g/day). NSAIDs were not authorised until W6 (corresponding to 4 weeks after the second injection).


Safety was evaluated at each visit and the patient was advised to inform the medical staff of any emerging medical problems (for definition of adverse events see online supplementary file).

Blood and urine samples were collected at W0 and at W6 for biomarker studies (see online supplementary file, including dosage in the serum of cartilage oligomeric matrix protein (COMP), HA, high-sensitivity C reactive protein (hsCRP), PIIANP (type IIA collagen N-propeptide), IL-1 β, IL-6 and TNF α, and in the urine, level of CTX II (degradation products of C-terminal telopeptides of type II collagen).

Outcome measures and follow-up

Clinical variables

Our primary efficacy end point was the percentage of patients achieving a pain improvement of at least 50% on a 0–100 mm Visual Analogue Scale (VAS) between W0 and W6. The question asked was: ‘What is the global level of pain in your hands in the past 24 h?’

Secondary efficacy outcomes were mean change in pain level assessed on a 0–100 mm VAS between W0 and W6 and between W0 and W26. Other secondary efficacy parameters were evaluated at W6 and W26: pain (on a 0–100 mm VAS), duration of morning stiffness (minutes), patient global assessment on a 0–100 mm VAS, practitioner global assessment on a 0–100 mm VAS, number of painful IP joints (0–30), number of painful IP joints under digital pressure (0–30), number of swollen IP joints (0–30), evolution of FIHOA (0–30), evolution of Cochin Hand Functional index (0–90), and consumption of analgesics between the two groups.30 ,31 Data for all efficacy parameters were collected at weeks 2, 4, 6, 10, 14 and 26.

Concerning rescue medications, patients were classified as non-users, or occasional and daily users, and their consumption was analysed.

Biological variables

Cytokines and biochemical markers of joint metabolism were measured with commercially available kits, using assay procedures specified by the manufacturers (see online supplementary file). Serum samples and fasting second morning void urine were collected before injection, at W0 and at W6.

Sample size calculation

We hypothesised a 60% rate of responders at 6 weeks (pain improvement of at least 50%) in the adalimumab group and 30% in the placebo group, with an a risk of 5% and a power of 80%. The sample size had to be 42 patients per group. Anticipating 20% of patients prematurely withdrawing from the trial, 110 patients (55 per group) had to be included.


Participants were randomised with a 1:1 ratio to adalimumab (Abbott) at 40 mg for two subcutaneous injections, at a 15-day interval, or placebo using a computer-generated central randomisation schedule.

The list of randomisation was created by Abbott, using Clinicocopia statistical software, using a random block size of eight and independently of investigators involved in the trial. After the investigator had obtained the patient's informed consent and had verified the inclusion/exclusion criteria, a fax was sent to the URC Mondor which subsequently assigned the allocated treatment. An independent pharmacist dispensed the allocated treatment.

The randomisation was centralised and not stratified.

Each syringe was masked with an identical volume of 0.8 mL, conditioned in a block of two syringes, stored at 2–4°C. They were prepackaged and consecutively numbered for each patient according to the randomisation schedule. Each patient was assigned an order number.


Patients and physicians were blinded to treatments. Each subcutaneous injection was performed by a nurse, independently of the medical investigators and of the outcome measures. All investigators, staff and participants were kept unaware of the outcome measurements and trial results.

Statistical analysis

All statistical analyses were performed at an independent centre (Hotel-Dieu Hospital, Paris, France), using SAS software, V.9.3 (SAS Institute Inc). A two-sided p<0.05 was retained for each analysis. Means and SDs or median with first and third quartiles were used to summarise distributions and means with a 95% CI were used to summarise group differences. Efficacy analyses were conducted on all randomised patients having received at least one injection. A post hoc analysis on the primary outcome was performed on the subpopulation with at least three swollen joints at baseline. Selection of patients and follow-up have been described according to the consort statement.32

The percentage of patients with more than a 50% of improvement in pain VAS in each group was compared using a log-binomial regression (with a log link and a binomial distribution). Clustering due to centre was taken into account using generalised estimating equations with an independence or exchangeable working correlation structure. A relative risk with a 95% CI was computed. To impute missing data, we assumed that all missing participants did not experience the 50% improvement. As a sensitivity analysis, we assumed that all participants experienced the 50% improvement.

For quantitative secondary outcomes, we used the mixed effects model for repeated measurements to estimate and compare adjusted mean changes from baseline to 6 weeks and from baseline to 26 weeks. The likelihood-based model uses all available data and provides valid results under the assumption that missing data are missing at random. Fixed effects included the baseline value of the outcome, centre, treatment, time and treatment by time interaction. An unstructured correlation matrix was used to model the within-subject error correlation structure. Changes between groups at W6 in joint biomarkers, hsCRP and cytokine levels were compared according to a Mann–Whitney test. Correlations between change in TNF levels and changes in clinical outcomes at W6 in patients treated with adalimumab were assessed by a Spearman correlation coefficient.


Flow chart

Figure 1 shows the study flow chart of participants. Of the 99 recruited patients from 16 centres who gave their informed consent after receiving extensive information about the trial, 14 were excluded before randomisation. A total of 85 patients were enrolled and randomly assigned (42 in the adalimumab group and 43 in the placebo group). Of these, two patients (one in each group) were further excluded from all statistical analyses because it was impossible to determine which treatment was finally received after allocation (mistake in treatment allocation due to a computer glitch). In the placebo group, five patients who did not receive the allocated intervention (two excluded for impossibility of a normal follow-up, one for refusing informed consent, one for contraindication for TNF therapy and one for a misunderstanding), while blinding was maintained, were also excluded for comparative analyses. Moreover, these five patients have baseline data for demographic characteristics but not for efficacy outcomes (baseline or follow-up). Finally, 41 in the adalimumab group received at least one injection and 37 patients in the placebo group made up the population for the main efficacy analysis.

Figure 1

Flow chart of the study.


The recruitment period began in January 2008 and ended in October 2009. Participants attended clinical visits at weeks 0, 2, 4, 6 (main outcome), 10, 14 and 26.

Patient characteristics

The adalimumab and the placebo groups had similar characteristics, except the baseline level of pain which was slightly higher in the adalimumab group: 67.1±13.1 versus 63.6±12.7 mm in the placebo group (table 1). For the whole population, the mean age of patients (85.8% women) was 62.5±6.9 years with a mean Body Mass Index of 24.9±4.07 kg/m2. Mean disease duration of symptomatic hand OA was identical in both groups: 13.5 years. The majority of patients were taking an ongoing treatment for their hand OA (83%). The most painful hand OA was the right hand in 54% of cases.

Table 1

Baseline characteristics of patients with hand OA

Primary outcome efficacy analysis

In the mITT set (defined as patients who received at least one injection), the primary outcome at W6 was missing in 4/37 patients in the placebo group and in 4/41 patients in the adalimumab group (table 2). With respect to the percentage of patients with pain improvement of more than 50% at W6, there was no difference between the groups: 35.1% of adalimumab-treated patients were responders compared with 27.3% of placebo-treated patients (RR=1.12; 95% CI 0.82 to 1.55; p=0.47). Considering missing values as non-responders in both groups and taking into account the centre effect, the relative risk was RR=1.10 (95% CI 0.82 to 1.49; p=0.51). Considering missing values as responders, results were the same: RR 1.11, 95% CI (0.82 to 1.52), p=0.50.

Table 2

Results for efficacy outcome: percentage of patients with at least 50% improvement in pain from baseline to week 6 (W6) (primary end point) and week 26 (W26)

Secondary outcome measures

Results are reported for all secondary outcome measures in table 3. The difference in the mean change in pain score on the VAS (0–100 mm) over 6 weeks between adalimumab and placebo was: −2.5 mm (95% CI −14.0 to 9.0), p=0.67 (table 2).33 Similarly, the mean difference in pain score on VAS (0–100 mm) at W26 was not different between the groups: −1.4 (95% CI −13.0 to 10.1; p=0.80). The course of the changes in pain score in the two groups is given in figure 2.

Table 3

Results for secondary efficacy outcome: estimate of least-squares mean change in outcomes from baseline to 6 or 26 weeks

Figure 2

Pain score changes over time.

There was no difference in any secondary end points between the two groups either at W6 or W26, except for the decrease in the number of swollen joints between W0 and W26 which favoured the adalimumab group: mean between-group difference: −1.9 (95% CI −3.2 to −0.6; p=0.006). The percentage of patients with analgesic intake did not differ between the two groups (70% in the placebo group and 68% in the adalimumab group; p=0.85).

A post hoc analysis on the evolution of pain in patients with at least three swollen joints at baseline was performed and found no difference between the two groups in this subset (n=30 in the adalimumab group and n=27 in the placebo group). After considering missing values as non-responders in both groups and taking into account the centre effect, RR was 1.08 (95% CI 0.84 to 1.38; p=0.56).

Analysis of biomarkers

Baseline levels of the different biomarkers in the two groups are summarised in table 1. Changes at W6 in joint biomarkers, CRP levels and cytokine levels after treatment with placebo or adalimumab are summarised in table 4. There was no difference in the changes for any biomarkers between the two groups. Finally, we assessed whether changes in the TNF a level were linked to the clinical outcome at W6 in the group of patients treated with adalimumab but there were no significant correlation (all p>0.05).

Table 4

Changes at week 6 in joint biomarkers, hsCRP and cytokines levels after treatment of hand osteoarthritis patients with placebo or adalimumab

Adverse events

The rate of adverse events was similar between the two groups. Overall adverse event was 73.0% (27/37) in the placebo group and 75.6% (31/41) in the adalimumab group. Severe adverse events were reported in 5.4% (2/37) in the placebo group and 9.8% (4/41) in the adalimumab group. There was no serious event related to subcutaneous injection of adalimumab, in particular no serious infection in those non-immune-compromised patients. Pain related to subcutaneous injection was observed in six subcutaneous injections: three in the placebo group and in three adalimumab subjects, and were described as burning, pain or pruritus of minor or mild intensity.


This RCT failed to demonstrate a greater efficacy of adalimumab (two injections at a 2-week interval) over placebo to alleviate pain in subjects with painful digital hand OA refractory to both analgesics and NSAIDs. The main clinical outcome (% of patients with at least a 50% improvement in pain from baseline to 6 weeks) was not different between the adalimumab and placebo groups (p=0.48). The level of difference in pain change on a VAS was −2.5 mm at W6, far below the level of clinical relevance.33 ,34 Furthermore, there was no difference between the groups at any time point as shown in figure 2. In addition, none of the secondary outcomes showed any significant difference between the groups. Analgesic intake did not differ between groups. All these data strongly suggest that pain and disability in hand OA are not TNF α-dependant.

There was no epidemiological difference in the patient profile between the two groups at baseline that may explain this lack of efficacy of TNF α blockers, with the exception of a slightly higher pain level in the adalimumab group (67 vs 63 mm in the placebo group). However, all analyses were adjusted for this baseline level which therefore cannot explain the negative result of this trial. Lack of difference between the two groups cannot be explained by a low disease activity at baseline (pain was around 60 mm in both groups) or by an unexpected high level of response in the placebo group (27% of responders corresponded closely well to our initial hypothesis of 30% of responders). The placebo effect in patients with hand OA was previously reported to be very high, with an effect size of 0.8 on pain.35 Otherwise, anti-TNF α therapy (limited to two injections) was perfectly well tolerated.

The failure of TNF α blockers in painful hand OA may indicate that TNF α is not the right target in order to improve symptoms in hand OA. Similarly, a small, open-label trial (12 patients) with inflammatory hand OA who received six subcutaneous injections of adalimumab every other week (12 weeks) failed to demonstrate any pain improvement.26

However, it could be assumed that TNF α blockers may be more active in joints that are painful and swollen in keeping with the pro-inflammatory role of TNF α in the synovial membrane.13 In a 1-year RCT, adalimumab reduced the number of new erosions only in the subset of patients with clinically swollen IP joints at baseline.27 But, in our study, post hoc analysis of the subgroup of patients with more than three swollen joints at baseline did not predict any positive response to TNF a blockers on pain.

Another hypothesis could be that TNF α blockers may exert a differential effect on pain and structure. In patients with rheumatoid arthritis, infliximab therapy compared with other DMARDs reduced both the radiological progression of prevalent hand OA and the incidence of hand OA.28 In contrast, in a recent 1-year placebo-controlled trial, adalimumab (40 mg every two weeks) failed to demonstrate a structural effect.27 Similarly, we did not observe any difference in the evolution of biomarkers of cartilage turnover (CTX-II, PIIANP, HA and COMP) or in serum levels of pro-inflammatory cytokines and in hsCRP.

Our study has several limitations. Only two subcutaneous injections of adalimumab could be regarded as insufficient to observe any clinical benefit. However, administration of anti-TNF in inflammatory arthritis such as PsA is marked by a rapid response (starting after only two injections).24 In addition, long-term administration of adalimumab in hand OA failed to demonstrate any benefit and thus does not support a prolonged treatment with anti-TNF α.27 The choice of high response (more than 50% of pain improvement) as the main criterion might show a milder clinical response. This hypothesis is not confirmed regarding the equivalence of pain improvement in both groups.

Though the current study is negative and does not support our initial expectation, we still need therapeutic solutions for these patients. A recent survey showed that participants described a lack of help and information with regard to the management of their hand OA and described a lack of comprehension from practitioners of the impact of hand OA in their life.11 ,12 This reaction is certainly intensified by the lack of any efficient treatment.

Looking for new targets such as IL-1 or IL-6 should certainly be investigated in the future before closing the door.36 One brief report on three patients indicated some beneficial results with IL-1 Ra in patients with painful hand OA.37

In a situation without gold-standard therapy, it is important to continue to explore new therapies with a good benefit–risk balance. The subgroup of patients with ‘inflammatory’ hand OA and swollen joints may represent the best population to target in future clinical trials of future therapies.


We are grateful to the patients who agreed to be included in this trial, and we thank all the co-investigators: Créteil: Fevre Claire, Farrenq Valérie; Paris (Saint Antoine): Champey Julien; Lyon: Piperno Muriel; Nice: Allam Yacine, Euller Ziegler Liana; Nancy: Rat Anne-Christine, Sommier Jean-Phillipe; Nantes, Maugars yves; Tours: Mammou Saloua; Dijon: Piroth Christine; Marseille: Legré Virginie; Lille: Pouyol Francois, Flipo René-Marc, Montpellier: Gérard-Dran Delphine; Paris (Pitié-Salpêtrière): Banneville Béatrice, Inaoui Rachida; Paris (Bichat): Meyer Olivier, Ballard Magali; Poitiers: Solau-Gervais Elisabeth, Besancon: Wendling Daniel; Paris (Lariboisière): Peyr Olivier.


View Abstract

Supplementary materials

  • Supplementary Data

    This web only file has been produced by the BMJ Publishing Group from an electronic file supplied by the author(s) and has not been edited for content.

    Files in this Data Supplement:

  • Lay summary

    This web only file has been produced by the BMJ Publishing Group from an electronic file supplied by the author(s) and has not been edited for content.


  • Handling editor Tore K Kvien

  • Contributors The study was designed by the academic investigators. Representatives of TAP collected the data, and statisticians at TAP conducted all statistical analyses. All authors had access to the data and vouch for the veracity and completeness of the data and the data analysis. The manuscript was written in its entirety by the authors.

  • Funding Sponsored by Hopitaux de Paris direction de la Recherche Clinique (DRC) (Hospitals of Paris, Clinical Research Department), St Louis, Paris (Assistance Publique-Hopitaux de Paris (APHP)) and supported by the Inserm Pro A. Abbot grant which was given with the agreement of the sponsor the DRC APHP, URC Mondor, but Abbot was not involved in the design, implementation and results of the study, which was performed independently of the firm.

  • Patient consent Obtained.

  • Ethics approval Independent Ethics Committees from Henri Mondor Hospital (Creteil, France).

  • Provenance and peer review Not commissioned; externally peer reviewed.

Request Permissions

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.