Data are means ± 90% confidence limit. Effects are shown in units standardized by dividing by the baseline between-subject SD averaged over all studies. Magnitudes are based on the following scale: <0.20, trivial; 0.20–0.60, small; 0.60–1.20, moderate; and >1.20, large. Nontrivial magnitudes are beneficial, unless stated otherwise. Shaded cells indicate clear beneficial effects. Cells in boldface indicate clear harmful effects. Italics indicate estimates from a fixed-effects meta-analysis. *Random-effects analysis failed to produce estimates, even for the simplest model (mode of exercise as the only fixed effect). †Insufficient studies to include study characteristics in random- or fixed-effects analyses.