Background & Aims: An open study reported that patients with Crohn's disease in remission who have taken azathioprine for longer than 3.5 years are at low risk of relapse when azathioprine is discontinued. To confirm this observation, we performed a multicenter, double-blind, noninferiority withdrawal study. Methods: Patients who were in clinical remission on azathioprine for ≥42 months were randomized to continue azathioprine or to receive an equivalent placebo for 18 months. The primary end point was clinical relapse at 18 months. Results: Forty patients were randomly assigned to receive azathioprine and 43 to receive placebo. Characteristics of patients at entry were similar in the 2 study groups. At 18 months, 3 patients had a relapse in the azathioprine group, and 9 had a relapse in the placebo group. Kaplan-Meier estimates of the relapse rate at 18 months were 8% ± 4% and 21% ± 6%, respectively. The hypothesis that placebo was inferior to azathioprine was not rejected (P = .195). Among the baseline variables, C-reactive protein level >20 mg/L, time without steroids <50 months, and hemoglobin level <12 g/dL were found to be predictive of relapse in the multivariate analysis. Conclusions: This study shows that azathioprine withdrawal is not equivalent to continued therapy with azathioprine for maintenance of remission in patients with Crohn's disease who have been in remission on azathioprine for ≥3.5 years. Thus, azathioprine maintenance therapy should be continued beyond 3.5 years. Background & Aims: An open study reported that patients with Crohn's disease in remission who have taken azathioprine for longer than 3.5 years are at low risk of relapse when azathioprine is discontinued. To confirm this observation, we performed a multicenter, double-blind, noninferiority withdrawal study. Methods: Patients who were in clinical remission on azathioprine for ≥42 months were randomized to continue azathioprine or to receive an equivalent placebo for 18 months. The primary end point was clinical relapse at 18 months. Results: Forty patients were randomly assigned to receive azathioprine and 43 to receive placebo. Characteristics of patients at entry were similar in the 2 study groups. At 18 months, 3 patients had a relapse in the azathioprine group, and 9 had a relapse in the placebo group. Kaplan-Meier estimates of the relapse rate at 18 months were 8% ± 4% and 21% ± 6%, respectively. The hypothesis that placebo was inferior to azathioprine was not rejected (P = .195). Among the baseline variables, C-reactive protein level >20 mg/L, time without steroids <50 months, and hemoglobin level <12 g/dL were found to be predictive of relapse in the multivariate analysis. Conclusions: This study shows that azathioprine withdrawal is not equivalent to continued therapy with azathioprine for maintenance of remission in patients with Crohn's disease who have been in remission on azathioprine for ≥3.5 years. Thus, azathioprine maintenance therapy should be continued beyond 3.5 years. Crohn's disease is a chronic inflammatory disorder of the gastrointestinal tract that can lead to disabling symptoms during disease flares and can result in complications such as fistulas, abdominal abscesses, intestinal obstructions, and, occasionally, cancer. The immunosuppressive agent azathioprine (2.5–3.0 mg/kg) and a related prodrug, 6-mercaptopurine (1.0–1.5 mg/kg), are effective for maintenance of steroid-induced remission over 12–15 months in patients with active Crohn's disease. 1O'Donoghue D.P. Dawson A.M. Powel-Tuck K. Bown R.L. Double-blind withdrawal of azathioprine as maintenance treatment for Crohn's disease.Lancet. 1978; 2: 944-946Google Scholar, 2Candy S. Wright J.P. Gerber M. Adams G. Gerig M. Goodman R. A double blind controlled study of azathioprine in the treatment and maintenance of remission in CD.Gut. 1995; 37: 674-678Google Scholar, 3Present D.H. Korelitz B.I. Wisch N. Glass J.L. Sachar D.B. Pasternack B.S. Treatment of Crohn's disease with 6-mercaptopurine. A long-term, randomized, double-blind study.N Engl J Med. 1980; 302: 981-987Google Scholar, 4Markowitz J. Grancher K. Kohn N. Lesser M. Daum F. A multicenter trial of 6-mercaptopurine and prednisone in children with newly diagnosed Crohn's disease.Gastroenterology. 2000; 119: 895-902Abstract Full Text Full Text PDF Scopus (685) Google Scholar Several long-term follow-up studies 5Bouhnik Y. Lémann M. Mary J.Y. et al.Long-term follow-up of patients with Crohn's disease treated with azathioprine or 6-mercaptopurine.Lancet. 1996; 347: 215-219Abstract Google Scholar, 6Fraser A.G. Orchard T.R. Jewell D.P. The efficacy of azathioprine for the treatment of inflammatory bowel disease a 30 year review.Gut. 2002; 50: 485-489Google Scholar, 7Kim P.S. Zlatanic J. Korelitz B.I. Gleim G.W. Optimum duration of treatment with 6-mercaptopurine for Crohn's disease.Am J Gastroenterol. 1999; 94: 3254-3257Google Scholar have shown that the benefit of azathioprine may be sustained over many years in most patients, but controlled data are lacking. The question of whether treatment with azathioprine can be safely interrupted after a period of prolonged remission is of great interest to patients and physicians because most patients affected by Crohn's disease are young, have a long life expectancy, and often have concerns about the long-term safety of the drug. Despite increasing evidence of safety from several studies, 5Bouhnik Y. Lémann M. Mary J.Y. et al.Long-term follow-up of patients with Crohn's disease treated with azathioprine or 6-mercaptopurine.Lancet. 1996; 347: 215-219Abstract Google Scholar, 6Fraser A.G. Orchard T.R. Jewell D.P. The efficacy of azathioprine for the treatment of inflammatory bowel disease a 30 year review.Gut. 2002; 50: 485-489Google Scholar, 7Kim P.S. Zlatanic J. Korelitz B.I. Gleim G.W. Optimum duration of treatment with 6-mercaptopurine for Crohn's disease.Am J Gastroenterol. 1999; 94: 3254-3257Google Scholar, 8Fraser A.G. Orchard T.R. Robinson E.M. Jewell D.P. Long-term risk of malignancy after treatment of inflammatory bowel disease with azathioprine.Aliment Pharmacol Ther. 2002; 16: 1225-1232Google Scholar, 9Connell W.R. Kamm M.A. Dickson M. Balkwill A.M. Ritchie J.K. Lennard-Jones J.E. Long-term neoplasia risk after azathioprine treatment in inflammatory bowel disease.Lancet. 1994; 343: 1249-1252Google Scholar, 10Present D.H. Meltzer S.J. Krumholz M.P. Wolke A. Korelitz B.I. 6-Mercaptopurine in the management of inflammatory bowel disease short and long term toxicity.Ann Intern Med. 1989; 111: 641-649Google Scholar a small increase in the frequency of malignancy, especially lymphoma, cannot be excluded. 11Korelitz B.I. Mirsky F.J. Fleisher M.R. Warman J.I. Wisch N. Gleim G.W. Malignant neoplasms subsequent to treatment of inflammatory bowel disease with 6-mercaptopurine.Am J Gastroenterol. 1999; 94: 3248-3253Google Scholar, 12Kumar S. Fend F. Quintanilla-Martinez L. et al.Epstein-Barr virus-positive primary gastrointestinal Hodgkin's disease association with inflammatory bowel disease and immunosuppression.Am J Surg Pathol. 2000; 24: 66-73Google Scholar, 13Larvol L. Soule J.C. Le Tourneau A. Reversible lymphoma in the setting of azathioprine therapy for Crohn's disease.N Engl J Med. 1994; 331: 883-884Google Scholar, 14Farrell R.J. Ang Y. Kileen P. O'Briain D.S. Kelleher D. Keeling P.W. Weir D.G. Increased incidence of non-Hodgkin's lymphoma in inflammatory bowel disease patients on immunosuppressive therapy but overall risk is low.Gut. 2000; 47: 514-519Google Scholar, 15Ahmed E. Fraser A. Korelitz B. Brensinger C. Lewis J. Azathioprine and 6-mercaptopurine use for inflammatory bowel disease is associated with an increased risk of lymphoma (abstr).Gastroenterology. 2004; 126: A135Google Scholar Immunosuppressive therapy can favor the emergence of occasionally severe opportunistic infection. 16Posthuma E.F.M. Westendorp R.G.J. van der Sluys Veer A. Kluin-Nelemans J.C. Kluin P.M. Lamers C.B.H.W. Fatal infectious mononucleosis a severe complication in the treatment of Crohn's disease with azathioprine.Gut. 1995; 36: 311-313Google Scholar, 17Khatchatourian M. Seaton T.L. An unusual complication of immunosuppressive therapy in inflammatory bowel disease.Am J Gastroenterol. 1997; 92: 1558-1560Google Scholar Regular monitoring of blood counts and liver tests is mandatory to avoid delayed bone marrow and liver toxicities. 18Connell W.R. Kamm M.A. Ritchie J.K. Lennard-Jones J.E. Bone marrow toxicity caused by azathioprine in inflammatory bowel disease 27 years of experience.Gut. 1993; 34: 1081-1085Google Scholar A retrospective study published by some of our group 5Bouhnik Y. Lémann M. Mary J.Y. et al.Long-term follow-up of patients with Crohn's disease treated with azathioprine or 6-mercaptopurine.Lancet. 1996; 347: 215-219Abstract Google Scholar suggested that withdrawal of azathioprine without relapse might be possible in patients who have been in complete remission without steroids for longer than 3.5 years, because the 2-year relapse rate seemed similar whether the treatment was continued or stopped after this time. This uncontrolled observation on a small subset of patients required confirmation by a prospective controlled trial. Therefore, we decided to perform a multicenter, randomized, double-blind, noninferiority withdrawal trial in patients who were in clinical remission induced by azathioprine for ≥42 months. The study was performed at 11 sites in France and 1 site in Belgium, all of which are members of the Groupe d'Etude Thérapeutique des Affections Inflammatoires du tube Digestif (GETAID). Recruitment of patients took place from October 1995 to November 1999. Eligible patients were at least 18 years old and had Crohn's disease according to established clinical, endoscopic, radiological, and histological criteria. Criteria for inclusion in the study were (1) continuous treatment with azathioprine for at least 42 months; (2) no flare-up; no treatment with oral prednisone (>10 mg/day), budesonide, artificial nutrition, or other immunosuppressive or biological agents; and no surgery (with the exception of limited perianal surgery) during the preceding 42 months; and (3) no treatment with rectal steroids, aminosalicylates, metronidazole, or ciprofloxacin during the preceding 6 months. Patients were excluded if they had active disease (defined by a Crohn's Disease Activity Index 19Best W.R. Becktel J.M. Singleton J.W. Rederived values of the eight coefficients of the Crohn's Disease Activity Index (CDAI).Gastroenterology. 1979; 77: 843-846Scopus (519) Google Scholar [CDAI] score >150 at entry), if they had Crohn's disease limited to the perianal area, and if they were treated with azathioprine for the indication of prevention of postoperative recurrence after a curative surgical resection. At study entry, patients were randomized in a 1:1 ratio to 18 months of double-blind therapy with oral azathioprine once daily at the dose they were taking before enrollment in the trial or to placebo. The randomization was performed centrally by using permutation tables of size 2 or 4 after stratification by center according to the number of patients anticipated to be enrolled at each center. Placebo and azathioprine (provided by GlaxoSmithKline, Marly-le-Roi, France) were formulated in 25-mg tablets to be indistinguishable from each another in appearance and taste. Clinical and biological monitoring, performed at baseline and then every 3 months, included calculation of the CDAI score and biological tests (blood counts, sedimentation rate, C-reactive protein, alanine aminotransferase, and γ-glutamyltransferase). The CDAI score 19Best W.R. Becktel J.M. Singleton J.W. Rederived values of the eight coefficients of the Crohn's Disease Activity Index (CDAI).Gastroenterology. 1979; 77: 843-846Scopus (519) Google Scholar incorporates 8 items: the number of liquid or very soft stools, the intensity of abdominal pain, and general well-being in the 7 days preceding the assessment; the presence or absence of an abdominal mass or extraintestinal manifestation; the use of opiates as antidiarrheal medication; hematocrit; and body weight. The score can range from 0 to 600; a score of ≤150 is considered to indicate clinical remission. The patients were asked to fill out a daily diary card for 7 consecutive days before each visit or in case of exacerbation of symptoms. To maintain blinding for the principal investigators at each site, the results of biological tests were reviewed by coinvestigators who had no contact with the patients, and these results were reported in a separate case report form. If leukopenia (<1500 neutrophils per milliliter) or mild transaminase increases (<2 times the upper limit of normal) occurred, then the dose of the study drug could be decreased by 50%. For patients with involvement of the terminal ileum or of the colon, an ileocolonoscopy was offered to the patient at study entry, but refusal was not regarded as a reason for exclusion. Immediately after the procedure, the endoscopist calculated the Crohn's Disease Endoscopic Index of Severity (CDEIS) score. 20Mary J.Y. Modigliani R. Development and validation of an endoscopic index of the severity for Crohn's disease a prospective multicentre study. Groupe d'Etudes Therapeutiques des Affections Inflammatoires du tube Digestif (GETAID).Gut. 1989; 30: 983-989Google Scholar This score is based on the surface of the lesions and of ulcerations, the presence of deep or superficial ulcerations, and the presence of ulcerated or nonulcerated stenosis in the terminal ileum and 4 segments of the colon. The CDEIS score can range from 0, indicating a complete absence of lesions, to approximately 40. The study was approved by the ethics committee for each participating center. Written informed consent was obtained from all patients. The primary efficacy criterion was the proportion of patients who experienced a clinical relapse over the 18-month period. Relapse was defined as a CDAI score >250, a CDAI score between 150 and 250 on 3 consecutive weeks with an increase of at least 75 points above the baseline value, or the need for surgery for Crohn's disease (with the exception of limited perianal surgery). This definition was proposed to eliminate small and transient increases of the CDAI score, which can be related to causes other than relapse (for instance, intestinal infection or irritable bowel syndrome) and do not need specific treatment adaptation. The control treatment (continuing azathioprine) is effective; we therefore tested for noninferior efficacy of azathioprine withdrawal (placebo). 21Blackwelder W.C. "Proving the null hypothesis" in clinical trials.Control Clin Trials. 1982; 3: 345-353Google Scholar We considered that an increase of 20% in the relapse rate at 18 months in the placebo group would be an upper limit at which the clinical benefit of azathioprine withdrawal would be acceptable, taking into account the potential risks of azathioprine continuation. With this criterion for noninferiority, we calculated that 40 subjects were required in each group for the study to provide at least 80% power, with a 5% one-sided type I error, assuming a 10% relapse rate in the azathioprine group. 5Bouhnik Y. Lémann M. Mary J.Y. et al.Long-term follow-up of patients with Crohn's disease treated with azathioprine or 6-mercaptopurine.Lancet. 1996; 347: 215-219Abstract Google Scholar Comparison of the distributions of clinical and biological variables at baseline between the 2 treatment groups was performed with the χ2 test or the Mann-Whitney test. The primary analysis was performed according to the intention-to-treat analysis principle including all patients randomly assigned to treatment; follow-up of patients who discontinued the study for reasons other than relapse was censored at the time of discontinuation. A per-protocol analysis was also performed. Probabilities of relapse were estimated by the Kaplan-Meier method by using the Greenwood formula for standard error estimates. Noninferiority of azathioprine withdrawal as compared with azathioprine continuation was tested by comparing the upper limit of the 95% confidence interval (CI) of the difference between relapse rates with the limit 0.20, the maximum tolerable limit chosen a priori to define noninferiority. This method, universally used in noninferiority trials, has been shown to give results very similar to those with the more complicated comparison of relapse rates, in which the difference distribution is centered at 0.20 under the null hypothesis. 22Dunnett C.W. Gent M. Significance testing to establish equivalence between treatments, with special reference to data in the form of 2 × 2 tables.Biometrics. 1977; 33: 593-602Google Scholar The z statistic was used to determine the upper limit of the one-sided 95% CI of the increase in relapse rate. 21Blackwelder W.C. "Proving the null hypothesis" in clinical trials.Control Clin Trials. 1982; 3: 345-353Google Scholar The influence of concomitant variables on time to relapse was also examined with the Cox proportional hazard model adjusted on treatment group according to the following baseline variables: sex, age, smoking habits, disease site, duration of disease, duration of remission, duration of azathioprine administration, azathioprine daily dose, previous treatment, time without steroids, CDAI, C-reactive protein concentration, erythrocyte sedimentation rate, leukocyte counts, hemoglobin level and mean corpuscular volume, platelet count, CDEIS, and presence of ulceration at colonoscopy. Continuous variables were categorized into 2 or 3 groups according to a previously reported method. 23Peto R. Pike M.C. Armitage P. et al.Design and analysis of randomized clinical trials requiring prolonged observation of each patient. II. Analysis and examples.Br J Cancer. 1977; 35: 1-39Crossref Scopus (7166) Google Scholar, 24Byar D.P. Identification of prognostic factors.in: Buyse M.E. Staquet M.J. Sylverster R.J. Cancer clinical trials methods and practice. Oxford Medical Publications, Oxford1988: 423Google Scholar Each variable was first divided into 3 categories at approximately the 33rd and 67th percentiles. If the relative relapse rates were not substantially different in 2 adjacent categories, then these 2 categories were grouped together. If no clear pattern was observed, then the median was taken as a cut point. Classic cutoff values, when available, were also used. The proportional hazards model was used first to study the influence of each factor on the time to relapse and then to identify independent prognostic factors after adjustment on treatment group. For the latter, a multivariate analysis was applied by using a stepwise procedure to variables with a P value of <.20 in the univariate analysis. In the multivariate analysis, P < .05 was considered as the level of significance. Clinical and biological variables are presented as mean and standard deviation. Relapse rates are presented as estimates with standard error; relative risks are presented as estimates with 95% CIs. Between October 1995 and November 1999, a total of 83 patients were enrolled; 43 patients were randomly assigned to receive placebo, and 40 were randomly assigned to receive azathioprine. Demographic and other baseline characteristics were similar in the 2 study groups (Table 1). One patient in the azathioprine group never took study medication. Five patients discontinued the study for reasons other than relapse. Three patients withdrew their consent, 2 at month 9 (1 patient in each treatment group) for personal reasons and 1 at month 12 (placebo group) after he had experienced a perianal abscess leading to a limited operation, but without symptoms of intestinal relapse. Two other patients (1 in each treatment group) stopped the study drug because of adverse events. A trial profile is shown in Figure 1.Table 1Baseline Characteristics of the Study PatientsCharacteristicAzathioprine (n = 40)Placebo (n = 43)Age, y (mean ± SD)40 ± 1436 ± 11Male sex, n (%)19 (47)18 (42)Smoker, n (%)17 (42)18 (42)Disease site, n (%) Small bowel only5 (13)4 (9) Colon only19 (48)17 (40) Both16 (40)22 (51)Perianal lesions17 (43)19 (44)Previous segmental resection, n (%)14 (35)17 (40)Duration of disease, y (mean ± SD)11 ± 611 ± 5Duration of remission, mo (mean ± SD)57 ± 1762 ± 26Duration of azathioprine, mo (mean ± SD)64 ± 2167 ± 27Azathioprine dose, mg/kg body weight per day (mean ± SD)1.7 ± 0.41.7 ± 0.5Steroids <10 mg/day, n (%)3 (8)2 (5)CDAI (mean ± SD)41 ± 4339 ± 47C-reactive protein, mg/L (mean ± SD)5.3 ± 6.9 (n = 36)6.9 ± 8.2 (n = 40)Erythrocyte sedimentation rate, mm (mean ± SD)11 ± 6 (n = 34)14 ± 9 (n = 41)CDEIS (mean ± SD)2.5 ± 2.8 (n = 24)2.4 ± 3.5 (n = 21)CDEIS = 0, n (%)6 (25) (n = 24)10 (48) (n = 21)Presence of ulceration at colonoscopy, n (%)13 (54) (n = 24)8 (38) (n = 21) Open table in a new tab At 18 months, 3 of 40 patients in the azathioprine group had relapsed (mean ± SE of the 18-month relapse rate: 7.9% ± 4.4%), and 9 of 43 in the placebo group had relapsed (21.3% ± 6.3%). The difference in relapse rates between the 2 treatment groups was 13.4% ± 7.7%. The upper limit of the one-sided 95% CI for the difference was 26.0%. This was >20.0%, the value set as the upper limit for noninferiority when the study was designed. Therefore, the hypothesis that azathioprine continuation is superior to azathioprine withdrawal could not be rejected (P = .195). If the case of perianal abscess was considered as a relapse, then the upper limit of the one-sided 95% CI for the difference was 28.3% (P = .278), and the result was not changed. By using the per-protocol analysis, the upper limit was 26.7%, leading to the same conclusion (P = .211) as the intention-to-treat analysis. The Kaplan-Meier estimates of the time to relapse in the azathioprine group and the placebo group are shown in Figure 2. The duration of remission (mean ± SE) was 17.3 ± 0.5 months in the azathioprine group and 15.9 ± 0.9 months in the placebo group. In the Cox proportional hazards model, the baseline variables associated with a higher risk of relapse are given in Table 2. Two biological markers, C-reactive protein ≥20 mg/L and hemoglobin <12 g/dL, and time without steroids ≥50 months were found to be independent factors associated with a higher relapse rate in the 18 following months. Taking into account these prognostic factors, the superiority hypothesis still could not be rejected (P = .36).Table 2Influence of Baseline Variables on Time to Relapse Through the Cox Proportional Hazard Model Adjusted on Treatment GroupVariableRelative risk of relapse (95% CI)P valueUnivariate analysis Age (n = 83) (>40 y vs 30–40 y vs ≤30 y)2.5 (1.1–5.7).024 Time without steroids (n = 83) (≥50 mo vs <50 mo)4.0 (1.2–13.3).016 Hemoglobin (n = 81) (≥12 g/dL vs <12 g/dL)4.6 (0.9–22.7).043 Neutrophils (n = 79) (<4000/mL vs ≥4000/mL)7.9 (1.0–60.8).019 C-reactive protein (n = 76) (<20 mg/L vs ≥20 mg/L)12.3 (2.3–64.3).0002Multivariate analysis (n = 75) C-reactive protein (<20 mg/L vs ≥20 mg/L)16.9 (2.7–104.3)<.0001 Time without steroids (≥50 mo vs <50 mo)5.2 (1.5–18.1).004 Hemoglobin (≥12 g/dL vs <12 g/dL)8.7 (1.6–48.8).0034NOTE. A higher relative risk means a shorter time to relapse. For each variable, categories are presented by increased risk of relapse. Open table in a new tab NOTE. A higher relative risk means a shorter time to relapse. For each variable, categories are presented by increased risk of relapse. The mean CDEIS level was low in the 45 patients who had a colonoscopy at baseline (Table 1), but only 36% of them had complete mucosal healing (CDEIS was 0: no lesion at all). Ulcerations were still present in 47% of patients with endoscopic lesions. The presence of endoscopic lesions (CDEIS > 0) or of ulcerations was not predictive of relapse. Adverse events occurred in 2 patients: 1 patient from the placebo group had a facial rash at month 12 and stopped the treatment, and in another patient, from the azathioprine group, a myelodysplastic syndrome with bone marrow karyotype abnormalities in chromosome 7 was diagnosed at month 6, and the patient died 6 months later. This 32-year-old man had a 12-year history of Crohn's disease and was treated with azathioprine (2 mg · kg−1 · day−1) for 68 months. Another patient had mild leukopenia at month 12, and this led to a reduction in azathioprine dosage. Controlled trials, 1O'Donoghue D.P. Dawson A.M. Powel-Tuck K. Bown R.L. Double-blind withdrawal of azathioprine as maintenance treatment for Crohn's disease.Lancet. 1978; 2: 944-946Google Scholar, 2Candy S. Wright J.P. Gerber M. Adams G. Gerig M. Goodman R. A double blind controlled study of azathioprine in the treatment and maintenance of remission in CD.Gut. 1995; 37: 674-678Google Scholar, 3Present D.H. Korelitz B.I. Wisch N. Glass J.L. Sachar D.B. Pasternack B.S. Treatment of Crohn's disease with 6-mercaptopurine. A long-term, randomized, double-blind study.N Engl J Med. 1980; 302: 981-987Google Scholar a meta-analysis, 25Pearson D.C. May G.R. Fick G.H. Sutherland L.R. Azathioprine and 6-mercaptopurine in Crohn's disease. A meta-analysis.Ann Intern Med. 1995; 122: 132-134Google Scholar and a Cochrane review 26Pearson D.C. May G.R. Fick G. Sutherland L.R. Azathioprine for maintaining remission of Crohn's disease.Cochrane Database Syst Rev. 2000; (CD000067)Google Scholar have shown that azathioprine and 6-mercaptopurine are effective in Crohn's disease, both for maintaining remission and as steroid-sparing agents in chronic, active, steroid-dependent cases. However, there has been no consensus about the duration of the treatment once remission has been obtained. Because thiopurines only suppress disease activity, the issue of when to stop the drug is a difficult clinical decision and is often a matter of concern for patients. This multicenter, double-blind, noninferiority withdrawal trial shows that even after a long period of disease quiescence on azathioprine (minimum, 42 months; median duration of azathioprine therapy on entering the trial, 55 months), azathioprine withdrawal is not equivalent to continuing azathioprine for the prevention of clinical relapse over the ensuing 18 months. The upper limit of the one-sided 95% CI of the difference between relapse rates (26.0%) was superior to that considered, before initiating the trial, as the maximal difference to accept noninferiority (20%). The mean relapse rate in patients who stopped the drug (placebo group) was nearly 3 times that observed in those who were maintained on treatment (azathioprine group): 21.3% and 7.9%, respectively. The noninferiority design used in our study was justified by our initial hypothesis. The classic design used in clinical trials is intended to show a difference of efficacy between a new experimental treatment and the control treatment (or placebo), with a null hypothesis (H0) of equality between treatment efficacies. In such trials, the risk of wrongly concluding a difference (type I error) is fixed at 5%, and the risk of missing an advantage of the new treatment is fixed at 10%–20%, corresponding to a power of 80%–90%. The comparison between treatment groups is usually performed by a two-sided test. In recent years, randomized trials designed to establish the noninferiority of a new treatment as compared with a standard effective treatment have been increasingly used. 27Lallemant M. Jourdain G. Le Coeur S. et al.A trial of shortened zidovudine regimens to prevent mother-to-child transmission of human immunodeficiency virus type 1.N Engl J Med. 2000; 343: 982-991Google Scholar, 28The Columbus InvestigatorsLow-molecular-weight heparin in the treatment of patients with venous thromboembolism.N Engl J Med. 1997; 337: 657-662Google Scholar This design is used to establish that an experimental treatment has efficacy similar to a reference treatment, taking into account that the experimental treatment has an advantage over the reference treatment on some important secondary end point (for instance, toxicity). 21Blackwelder W.C. "Proving the null hypothesis" in clinical trials.Control Clin Trials. 1982; 3: 345-353Google Scholar The null hypothesis H0 is that the reference treatment is superior to the experimental treatment by at least a prespecified amount (a one-sided hypothesis). When the test comparing treatment efficacies is statistically significant, H0 is rejected, and the conclusion associated with type I error level of 5% is the noninferiority of the experimental treatment relative to the control treatment. When the test does not reach statistical significance, the conclusion is that the study fails to show the noninferiority of the experimental treatment compared with the control treatment (or to reject the superiority of the control treatment compared with the experimental one). In our study, the reference treatment was azathioprine continuation, a treatment of established efficacy, and we tested for noninferiority of azathioprine withdrawal (placebo). Our initial assumption that the relapse rate in the azathioprine group would be approximately 10% was confirmed by the study (7.9%). The limit beyond which the equivalence would be rejected was fixed a priori at 20.0%, a percentage that could be regarded as high. This limit was, however, considered clinically relevant at the time of the conception of the trial on the basis of the benefit/risk ratio for patients receiving long-term treatment with thiopurines, taking into account the consequences of relapse and the possibility of re-treating the patient with azathioprine in case of relapse on the one hand and the potential risks of long-term treatment on the other hand. Choosing a limit lower than 20% would have led to the same conclusion, bu