Adjusting for a variable (the collider – nonsteroidal antiinflammatory drug [NSAID]) that is influenced by an unmeasured cause of the exposure and an unmeasured cause of the outcome, creates a spurious noncausal pathway (A). The spurious pathway created (red) is M-shaped if variables are arranged in this pattern, and so this is sometimes referred to as M-bias. The unmeasured cause of propofol exposure might be clinician preferences (U₁) and the unmeasured cause of mortality might be frailty or renal impairment (U₂) which relatively contraindicates NSAID. If NSAID were thought to be a cause of mortality, then controlling for NSAID will simultaneously block a confounding pathway (B) and may be of net benefit.