The importance of event rates for clinical site effects in multicenter trials