I read with interest the article by Reinisch and colleagues on the ability of C-reactive protein (CRP) levels in Crohn's disease patients responding to infliximab to predict future relapse. The article reconfirms the finding of ACCENT I that a low baseline CRP can predict nonresponse, and one might interpret this as showing that patients without active inflammation are unlikely to respond to an anti-inflammatory drug.
That CRP levels predict relapse in Crohn's is, of course, also not a novel finding, having been demonstrated on a number of occasions and neither is the ability to predict prolonged response to anti-TNF. It is, however, notable that the positive predictive values of 3-month CRP for predicting response at 1 year are similar to those found previously. It is these predictive values, and their confirmation that point to the limitations of CRP in predicting disease course.
The target as the authors conclude is to permit treatment optimisation, and ideally we would do this without awaiting a relapse, but rather by predicting it. The PPV for maintained remission at 54 weeks of 51.8% means that a clinician could as well predict the 1-year response in a patient by tossing a coin as by performing a CRP. Of course most of us would test CRP more frequently and so perhaps the ability of the assay to predict 22-week response from week 14 is more relevant. Here 25% of those predicted to remain in remission will relapse, and only 22% of those predicted to relapse will.
This article therefore brings us to an understanding both that CRP can allow us to better inform our patients of their prognosis but equally that it is far too inaccurate to be a reliable guide to dose intensification. As the authors so rightly conclude, more work is needed.