ObjectivesTo determine whether C-reactive protein (CRP) level is predictive of response to tumor necrosis factor-α blocker treatment in patients with ankylosing spondylitis (AS) and whether there is an optimal CRP range for treatment initiation. MethodsIn this post hoc analysis, data on etanercept-treated patients with AS were pooled from four randomized trials. Week 12 responses (ASAS20, ASAS50, ASDAS-CRP < 1.3, and ASDAS-CRP ∆ ≤ 1.1) were evaluated in relationship to baseline CRP levels (normal, defined as ≤ upper limit of normal [≤ ULN]; elevated, > ULN; high, > ULN and ≤ 3xULN; and very high, > 3xULN), baseline levels of patient-reported outcomes (PROs), and CRP levels at weeks 2, 4, and 8, using univariate and stepwise predictor analyses. In addition, relationships between baseline CRP and other baseline predictors were analyzed using stepwise models of response. ResultsAmong 867 patients, baseline CRP levels were normal in 371 (43%) patients, high in 299 (34%), and very high in 197 (23%). Very high baseline CRP was a significant predictor for all four week-12 outcomes, compared with normal CRP. Conversely, normal CRP at weeks 2, 4, and 8 was a stronger predictor of week 12 response than elevated CRP. PROs were less consistent predictors of response. In addition, there was a significant association between higher baseline CRP and lower age of disease onset (< 40 years) and between normal CRP and lower disease burden. ConclusionsIn patients with AS, both baseline and post-baseline CRP levels can be predictive of response to treatment at week 12, more consistently than PROs. Clinical trialsNCT00421915, NCT00247962, NCT00418548, NCT00356356