be small by the Cohen convention.66 These resultssupport hypothesis 1; that is, therapeutic exercise has asignificant, positive effect on habitual walking speed. Interms of the clinical relevance of these findings, therewas a 57% success rate for exercise training to improvehabitual gait speed in the study population.61 Stateddifferently, exercise training was 1.83 times more likelyto improve habitual gait speed when compared with thecontrol. Furthermore, exercise training resulted in anoverall gait speed change of 0.01 m/s in the 24 studies.Finally, the fail-safe analysis indicated that it would takean additional 109 studies with no significant findings forthe P value to no longer be significant (P.05 level).These findings support hypothesis 1.The factors that may have produced these results (ie, theblocking of studies by type of therapeutic exercise[strength training and combination training], intensityof exercise [high, moderate, and low], and exercisedosage [high and low]) were examined next. As indicatedin Table 5, when the effect sizes of the studies ineach block were combined, the combined effect size ofthe studies that used moderate- and low-intensity exercise(14 studies, Ncombined953) and low-dosage exercise(10 studies, Ncombined703) had Z-scores that were notsignificant (P.05). Thus, further analysis was not indicated.61,62 The combined effect sizes for the remainingblocks of studies were significant and are presentedbelow (Tab. 5).Exercise ModeCombining the 8 strength training studies (n280)resulted in an overall effect size of rcombined.145(P.017, 95% CI0.028–0.258), which was significantat P.05 (Tab. 5 and Fig. 3). Although the combinedeffect size was small,66 the success rate for strengthtraining to improve habitual gait speed was 56%.Strength training was 1.61 more likely to improve gaitspeed as compared with the control. The gait speedchange from strength training was 0.02 m/s. The fail-safenumber was only 8 studies.The overall effect size of the 16 combination trainingstudies (n1,022) was rcombined.176 (P.002, 95%CI0.116–0.235) and significant at the P.05 level. Thisagain is considered to be a small effect size,66 but thesuccess rate was 58% and the odds ratio was 1.95,indicating that combination training would be morelikely to improve habitual gait speed than the control(Tab. 5 and Fig. 4). The gait speed change from combinationtraining was 0.01 m/s. For this group of studies,the fail-safe number was 40 studies. These findingssupport hypothesis 1.Exercise IntensityCombined data from the 10 studies (n349) that used ahigh-intensity exercise program resulted in a combinedeffect size of rcombined.184 (P.001, 95% CI0.082–0.285), which is significant at the P.05 level and supportshypothesis 3. However, the gait speed change was0.02 m/s, which is considered small.66 The success ratewas 59% and the odds ratio was 2.12, showing that highintensity exercise training would be more likely toimprove habitual gait speed compared with the control.The fail-safe number was 32 studies (Tab. 5 and Fig. 5).Exercise DosageFourteen studies (n599) included in the high-dosageexercise training group yielded an overall small effectsize66 of rcombined.190 (P.001, 95% CI0.112– 0.266)and, therefore, also supported hypothesis 3. As withprevious analyses, the use of a high-dosage exerciseprogram would be 58% more successful or 1.89 timesmore likely to improve habitual gait speed as comparedwith a control (Tab. 5 and Fig. 6). The 14 studies yieldedan overall gait speed change of 0.02 m/s. The fail-safenumber was 42 studies.Effect of Therapeutic Exercise on Fast Gait SpeedThe 18 studies (n752) included in the analysis for fastgait speed were found to be homogeneous (223.28,df17, P.14); therefore, the data were combined acrossstudies. Again the funnel plot indicated that publicationbias was not a major problem for these studies. Thecombination of effect sizes from these studies yieldedan overall and statistically insignificant effect size ofrcombined.055 (P.173, 95% CI0.017– 0.126)(Tab. 5 and Fig. 7). Hypothesis 2 was not supported, andno further analysis was performed on the fast gait speedstudies