How can we test/compare survival curves after multiple imputations?