Two treatments have similar estimated probabilities of adverse effects, but one effect is brief and reversible while the other may cause permanent impairment. Why can this difference matter?