Reply

To the Editor

We thank Drs. Akkoc and Khan for their comments on the 2015 American College of Rheumatology/Spondylitis Association of America/Spondyloarthritis Research and Treatment Network recommendations for the treatment of ankylosing spondylitis and nonradiographic SpA. They question whether the conditional recommendation supporting the use of TNFi in the treatment of patients with nonradiographic axial SpA should be modified to exclude specifically patients with a normal serum CRP level or lack of evidence of active sacroiliitis on MRI at baseline. They base their assertion on evidence from controlled trials showing weaker effects of etanercept and adalimumab in subgroups of patients who did not have an elevated CRP level or sacroiliitis on MRI, with the effects being, in some studies, significantly different from those in patients with an elevated CRP level or sacroiliitis on MRI.

Subgroup analyses of clinical trials can be useful for understanding potential modifiers of treatment responses, but these analyses are susceptible to several methodologic pitfalls, including lack of testing for interactions, limited power, Type I statistical errors due to multiple testing, and post hoc selection of attributes to investigate for modifying effects (1). Because of these issues, many findings from subgroup analyses have not been confirmed when tested a priori in prospective trials, and statisticians are generally wary of giving too much credence to findings from post hoc subgroup analyses (2). A common interpretative error is to compare treatment effects within subgroups without comparing differential responses between subgroups using tests of interactions. In the Ability-1 trial cited by Akkoc and Khan, tests of interactions demonstrated significant differences in responses to adalimumab in patients with an elevated baseline CRP level versus a normal baseline CRP level, but also demonstrated differences according to age and duration of SpA, with significantly better responses in younger patients and patients with early disease; no significant treatment interaction with MRI-assessed sacroiliitis or with 8 other clinical factors was found (3). In the study by Dougados et al, there were no significant treatment interactions with either baseline CRP level or MRI-assessed sacroiliitis (4). The reports by Landewé and colleagues and Haibel and colleagues did not include subgroup analyses for the controlled portions of their trials (5,6), and we hesitate to assume that similar findings apply to these trials in the absence of data. In the trial reported by Barkham et al, only subjects who did have sacroiliitis on MRI were enrolled, meaning testing for differential responses was not possible (7). In the recent golimumab trial cited by Akkoc and Khan that was not included in our review, tests of interactions were not reported (8).

While there is some evidence to suggest that responses differ between patients with and those without elevated CRP levels, these findings are not yet sufficiently established to form a basis for excluding all patients without an elevated CRP level from receiving TNFi therapy. CRP levels are elevated in only a minority of patients, so requiring that the level be elevated prior to initiation of this treatment would eliminate many patients. There is also some, although less strong, evidence supporting differential treatment responses between patients with and those without sacroiliitis on MRI. Insurance restrictions and cost may limit the availability of MRI.

As reflected in the conditional recommendation, these factors should be considered in the decision-making process when potential benefits and risks of treatment are considered by patients and clinicians. However, the committee does not believe the current evidence supports a proscription against the use of TNFi in patients with nonradiographic axial SpA who do not have an elevated CRP level or sacroiliitis seen on MRI. Additional research is needed to clarify whether treatment with TNFi can appropriately be targeted more specifically in these patients.