borderline significantP = 0.057
Random-effects multiple regression showed that pre-intervention house indices were reduced significantly at focal cycles 1 (from 12.2% to 3.5%, P = 0.001) and 4 (from 9.7% to 3.0%, P = 0.047), with borderline significant effects at cycle 5 (P = 0.057, from 5.5% to 2.7%).