Statistical Analyses

The mean difference (MD) in HDL-C, LDL-C, TC, TG, and BMI was examined for each trial. MD was defined as follows: (mean value at post-intervention in the exercise group − mean value at baseline in the exercise group) − (mean value at post-intervention in the control group − mean value at baseline in the control group). The weighted MD (WMD), i.e., overall MD, was weighted by the inverse variance of differences from baseline to final assessment in each trial and calculated with a random-effects model using the DerSimonian –Laird method²²⁾. The correlation coefficient between the baseline and the final assessment was assumed to be 0.50²³⁾. Cochran Q statistics were calculated, and the heterogeneity of WMD was examined among trials. The I² statistic represented the level of heterogeneity. An I² of 25% or lower indicated a low risk, an I² of 25%–75% indicated a moderate risk, and an I² greater than 75% indicated a high risk²⁴⁾.

Subgroup analyses of WMD in the HDL-C, LDL-C, TC, and TG levels were performed by classifying trials. The study group was stratified into 14 categories: dyslipidemia, no dyslipidemia, no medication, overweight, not overweight, significant decrease in the MD in BMI, no significant decrease in the MD in BMI, walking, jogging, bicycle ergometer, moderate-intensity exercise, vigorous-intensity exercise, exercise volume ≥ 150 min/week, and exercise volume < 150 min/week. Dyslipidemia was defined based on the Japan Atherosclerosis Society guidelines²⁵⁾, and trials with a mean HDL-C < 40 mg/dL, mean LDL-C ≥ 140 mg/dL, or mean TG ≥ 150 mg/dL of subjects were designated as having dyslipidemic subjects. If none of the subjects in a trial was taking medication, then a trial was categorized as no medication. Overweight was defined in accordance with WHO²⁶⁾, and trials involving subjects with a mean BMI ≥ 25.0 kg/m² were designated as having overweight subjects. On the basis of the American College of Sports Medicine (ACSM) guidelines²⁷⁾, moderate-intensity exercise was defined as 55%–69% of the maximum heart rate (HR_max). Trials involving exercise at 70% or greater of HR_max were categorized as involving vigorous-intensity exercise. If a trial indicated exercise intensity as a percentage of maximum oxygen uptake, the value was converted to a percentage of HR_max using the equation of Londeree and Ames²⁸⁾. Exercise volume was based on the exercise volume as recommended in the American Heart Association (AHA) and the ACSM guidelines²⁹⁾. In addition, a sensitivity analysis was used to evaluate the influence of a risk of bias according to the Cochrane Risk of Bias tool²¹⁾. Trials falling into one or more domains of a high risk of bias were excluded, and the WMD in HDL-C, LDL-C, TC, and TG was then calculated.

Publication bias was evaluated by assessing the symmetry of funnel plots produced by the MD in serum lipids or lipoproteins (x-axis) and the inverse of the standard error (y-axis). First, Egger's regression test was used to evaluate the asymmetry of funnel plots³⁰⁾. Second, the trim and fill method of Duval and Tweedie was used to estimate the number of missing trials³¹⁾. If the results suggested that trials were missing, then the WMD in lipids and lipoproteins was adjusted in light of the effect of these trials.

The results for baseline variables were expressed as the mean ± SD weighted by the number of subjects. In all the statistical tests, a P value < 0.05 was considered to be statistically significant. The results of MD and WMD were expressed as the 95% confidence intervals (CI). The Comprehensive Meta-Analysis soft program (Version 2.2; Biostat, Inc., Englewood, NJ, USA) was used to perform the meta-analysis.