Study was conceptually and statistically biased

EDITOR,—Lyn March and colleagues' study comparing a non-steroidal anti-inflammatory drug with paracetamol in osteoarthritis displays conceptual and statistical bias towards paracetamol.1 The conceptual bias—that non-steroidal anti-inflammatory drugs are illogical when there is little inflammation—would not arise if these drugs were called potent pain killers. After all, paracetamol has antipyretic activity but is used in the absence of fever. Statistical bias arose by the authors' designation of paracetamol as the drug of choice if differences from diclofenac did not reach significance. It was impossible, however, for diclofenac to be significantly better on a global rating: when all preferences …