There is a new debate on the role of thiopurines in Crohn's disease. This viewpoint discusses the current evidence and balances thiopurines against other treatment options in the therapeutic algorithm of Crohn's disease.
Azathioprine, and in the United States also 6-mercaptopurine, have been a mainstay of drug therapy in Crohn's disease for decades. Their use was evidence based by a variety of controlled trials and their combined metaanalysis.1 However, recent re-analysis discarded the notion that thiopurines induce remission2 which is consistent with their slow onset of action. In the two trials on maintenance when combined the relative risk of relapse was 0.64 but failed to reach significance. Three additional withdrawal trials indicated that continued medication with azathioprine did prevent relapse (RR = 0.39) and the difference to controls was statistically significant.
More recently, the SONIC trial compared the early use of azathioprine, infliximab and the combination of both.3 The results have been used to advocate the early top-down use of infliximab with or without azathioprine due to significantly better results as compared to azathioprine alone. It is important to realize that the study design was remarkably unfair for the azathioprine only arm and so was the interpretation. It is evident that azathioprine works best when combined with an effective agent for remission induction, may it be steroids4,5 or infliximab.3 The main message of SONIC is that a Slow acting agent Only is Not effectIve in Crohn's disease. At any rate, following SONIC the use of thiopurines in Crohn's disease was under debate.
With interest we have therefore read the head-to-head publication of two new azathioprine trials focusing on recently diagnosed Crohn's disease, accompanied by an editorial.6–8 Unfortunately, both investigator driven studies were underpowered and seriously flawed for various reasons. According to the available data9,10 about 40% of patients with newly diagnosed Crohn's disease will experience an intermittent and uncomplicated course of disease without an early need for immunosuppression. Nevertheless, Panes et al. randomized “all” patients with newly diagnosed Crohn's disease to receive either azathioprine or placebo. Roughly half of the patients in both groups discontinued the study prematurely. Not surprisingly the study overall failed to show a superiority of azathioprine in this unselected patient cohort. If a cohort or a major proportion receives an unnecessary medication the negative impact of side effects will not be balanced by its benefits. The positive message of this trial is that the use of azathioprine significantly reduced the rate of more severe disease flares (11.8% vs. 30.2% in the placebo group, p = 0.01), i.e. in those patients that apparently had a need for immunosuppression. Since this was only a post-hoc analysis the negative overall message in the title was that early azathioprine is no more effective than placebo.
The second trial by Cosnes et al. avoided this problem by selecting patients with risk factors for a disabling course of disease. Unfortunately, they used an open label design, so that both patients and physicians were aware of the medication given. Even more difficult for interpretation was the similar study medication in both arms: around 60% of patients in the conventional arm received azathioprine in the second and third year. If only 40% of the control group were true controls, what do you expect? This is reminiscent of the classical top-down trial by D'Haens et al.11 where treatment in the control arm was also liberalized so that in the end clinical outcome was similar. Despite this handicap in the current study design there was a trend in the early azathioprine group to experience less flares, significantly less active perianal lesions and less perianal surgery, in line with other data.12 Finally, also comparable to the D'Haens et al. trial,11 there was a remarkably low requirement for infliximab (31 and 25% in both groups, respectively).
There is a definite risk that based on these at first sight negative trials thiopurines will be discarded from the Crohn's treatment portfolio in general. In addition to the specific trial limitations mentioned above, there are more general problems with the extrapolation of study results which likely apply also to these trials in early Crohn's disease. As reported recently, the patients recruited into trials are very different from the more general Crohn's population treated in an outpatient practice.13 Even though this particular study focused on moderate to severe IBD most of their patients would not have met relevant inclusion/exclusion criteria and the authors rightly question whether trial results indeed prove efficacy beyond the rather specific trial population. There may be no way around this problem but most physicians are not even aware of this limitation.
Another common error is the uncritical comparison of trial results. There is little point in the editorialists'6 effort to adjudicate a percentage level of remission to a given drug. These numbers largely depend on the specific cohort studied and not just on the medication. Unfortunately, head to head long-term comparisons of anti-TNF and azathioprine on “normal” Crohn's patients are unavailable, and probably never will be. SONIC also focused on early Crohn's as did the trials discussed above.
If azathioprine indeed fell into general disrepute, are TNF-antibodies really the better alternative? Unfortunately, especially in large trials using biologicals it has become customary to focus on patients responding to a specific treatment at a given time during the trial and to refer to the responder population as a “new” 100%. This is a classical selection bias prone to overestimate effect sizes. Thus, after a year the remission rate normalized to the starter cohort is only in the range of 20%. If, in a similar but retrospective way, the focus is on azathioprine responders,14 these have a smaller proportion of patient years with active disease and are less likely to be hospitalized or undergo intestinal surgery, than patients who did not receive an immunosuppressant.
Finally, although the full picture of side effects is beyond the scope of this discussion, it should be noted that the anti-TNF strategy suggested as an alternative to a first line of classical immunosuppression is accompanied by an increased risk of opportunistic infections,15 melanoma,16 psoriasis,17 probably also postoperative anastomotic complications18 and lymphoma.19 Clearly, thiopurines also carry significant and maybe comparable adverse event risk but it is always wise to keep both risk profiles in mind.
The logical conclusion to be drawn is not only that the early use of azathioprine should be restricted to patients in need of immunosuppression but also that we currently fail to identify this cohort in advance.20 Therefore it seems reasonable to advocate an accelerated step up approach during the first 12 months with (ex post) identification of around 60% of patients in need for immunosuppressive therapy. The currently best evidence for thiopurines is in patients who have achieved remission using steroids or anti-TNF agents. A substantial proportion of these patients will be able to significantly reduce or even wean off steroids21 or infliximab: about half of the patients on combined anti-TNF and azathioprine do not require anti-TNF forever but do well on the thiopurine alone.22 The postoperative use of azathioprine is less clear but some data indeed support this indication.23
Due to the limited treatment options in Crohn's disease the most effective (and aggressive) therapies (i.e. anti-TNF antibodies) should be reserved for the severe and refractory cases. Otherwise no treatment options are left once first line therapy has failed or if the patient relapses. This more balanced view is in disagreement with the editorialists who essentially support anti-TNF use in most Crohn's patients, if it was not for the expense.6
It is a tell-tale story that 24 IBD centers in France recruited only 132 recently diagnosed Crohn's patients over a 5 year period and Spain randomized only 131 over 3 years in 31 hospitals. In comparison, industry-sponsored trials nowadays easily recruit more than 1100 patients in a similar period.24 Both the GETAID and the GETECCU groups should be lauded to at least try to maintain a culture of industry-independent academic clinical trials but both studies show how difficult this effort is in reality. Otherwise we would have to depend solely on pharmaceutical company sponsored trials with their known bias, which is rarely against the company's drug.25
Conflict of interest statement
Klaus R. Herrlinger and Eduard F. Stange declare no conflict of interest.