This study shows no significant advantage of Brewerton or Nørgaard radiographs in detection of erosions at metacarpophalangeal joints in patients with polyarthritis over the conventional posteroanterior view. The observer agreement for detection of exact number of erosions/pair of hands is uniformly low. Therefore, erosion index cannot be reliably used in epidemiologic surveys or therapeutic trials as an outcome measure unless steps are taken to improve the observer agreement.