To our knowledge, this is the first systematic review investigating the pooled effects of different exercise interventions on anthropometric outcomes, blood lipids and cardiorespiratory fitness. The main findings of this meta-analysis suggest that in subjects with a BMI ≥25 kg/m2, AET is more efficient in reducing BW, WC and FM as well as in increasing VO2max uptake when compared to RT, respectively. However, RT turned out to be more suitable when it comes to an improvement of lean body mass. Furthermore, the present results provide evidence that a combined intervention seems to be the most promising tool for management of overweight and obesity. CT was more powerful in reducing anthropometric risk factors like BW, WC or FM when compared to RT, and more effective in raising LBM when compared to AET. Pooled direct and indirect evidence on these three exercise interventions showed that CT was the most efficacious to reduce anthropometric outcomes such BW, WC and FM (with the respective ranking probabilities, following Bayesian network meta-analysis: 63%, 63% and 90%).
Since waist circumference correlates with abdominal fat mass and is considered to be an independent predictor of CDV, it can be used as a surrogate marker of abdominal fat mass 
._ENREF_52 De Koning et al. reported a 2%-increase in CVD risk for each 1cm-gain in WC 
. By transferring these findings to the results of the present meta-analyses, AET was associated with a reduction in CVD risk that was approximately 2% stronger as in the respective RT counterparts. Moreover, CT resulted in a decline in CVD risk that was by 4% more distinct as compared to the effects of an RT intervention. Aerobic exercise is known to increase the sympathetic tone, and the subsequent release of adrenergic transmitters leads to an increased lipolysis especially in abdominal fat 
Regarding lean body mass, the results of the present meta-analyses show that both RT and CT are more effective in raising LBM when compared to AET, respectively. An increase in LBM contributes to the maintenance or may even reflect an increase in resting metabolic rate 
. Apparently, RT triggers the preservation and buildup of body protein thereby altering the relationship between LBM and FM 
. In a previous study, it was shown that, if performed twice a week, RT facilitated an increase in LBM by 1−2 kg in the course of 6 months and could prevent age-associated loss of LBM 
Results suggest that exercise interventions containing aerobic sessions (whether isolated or as part of a combination training) improve cardiorespiratory fitness when compared to RT as a single training modality. A gain in cardiorespiratory fitness is known to be associated with reduced cardiovascular mortality and cancer incidence in men and women 
. A pooled analysis by Kodama et al. 
_ENREF_66 investigating the impact of cardiorespiratory fitness on all-cause mortality and cardiovascular events revealed that an increase in VO2
max in the amount of one metabolic equivalent correlated with a 13%-reduction of all-cause mortality as well as with a 15%-decrease in CHD/CVD risk, respectively. The authors suggested that the +1-MET-improving effects on VO2
max are comparable to corresponding influences of a decrease in WC (-7 cm), SBP (-5 mmHg), TG (-88 mg/dl), and FG (−18 mg/dl) as well as to increases in HDL-C (+7.72 mg/dl), respectively. When applying these findings to the results of the present meta-analysis, AET outperformed RT as a single training modality with a further 7.5%-risk reduction in all-cause mortality and a further 8.5%-risk reduction in CHD/CVD, respectively.
The present systematic review has several strengths and weaknesses. The meta-analysis were conducted following a stringent protocol, i.e. in all trials, participants were randomly assigned to the intervention groups, and only supervised training protocols were included. Randomized controlled trials are considered to be the gold standard for evaluating the effects of an intervention and are subject to fewer biases as compared to observational studies. The network meta-analysis included all individuals for each outcome. Moreover the present meta-analysis had a substantial sample size (range: 323 to 664) volunteers, thus providing the power to detect statistically significant mean differences as well as to assess publication bias. Network meta-analysis methods were used to obtain coherent estimates of all treatments relative to each other, using all available evidence and adequately accounting for evidence from 3-arm trials (i.e. avoiding the repeated use of data from such trials in different comparisons). This is of particular importance in this application where there were several trials simultaneously comparing all the interventions. Overall, the estimated between-studies heterogeneity parameters were small for all networks, and there was no evidence of inconsistency, which further strengthens the conclusions. Trial characteristics suggest the consistency/similarity assumption is satisfied, which is confirmed by the statistical analysis.
Limitations of the present review include the limited number of studies and the heterogeneity of the study designs. The trials covered in the meta-analyses showed variations in population characteristics (e.g. overweight, obese, age, number and ratio of male and female participants).
A considerable confounder could be the volume of exercise (min/week) prescribed. Two studies reported exercise intensity in the CT group to be twice as high as compared to their respective RT and/or AET counterparts 
. However, a sensitivity analysis excluding these studies confirmed the results of the primary analysis. Other potential confounders included differences in dietary intake and activity performed outside the monitoring and supervision by the investigators. Most studies reported the method of randomization as well as other data required for risk of bias assessment, which might be due to the fact that the trials were performed within the previous 20 years (between 1994 and 2012). However, another major limitation is the size of the study population, i.e. 11 of the 15 trials had a sample size of less than 60 participants, demanding a conservative interpretation of the results. With respect to publication bias, funnel plots for this systematic review showed low to moderate asymmetry suggesting that e.g. lack of published trials with inconclusive results cannot be completely excluded as a confounder of the present meta-analysis (Figure S15-16 in File S1
). According to the results of the Begg's and Egger's linear regression tests, there is evidence for a potential publication bias for BW following pairwise comparison of AET vs. RT and VO2
max following direct comparison of CT vs. AET. Therefore, these results should be interpreted with caution. Future trials should focus on high-quality methodological assessment (allocation concealment, blinding of outcome assessment, and intention-to-treat analysis), long-term effects (≥12 months), and larger sample size.
In conclusion, the present systematic review and meta-analysis focused on RCTs mutually comparing AET, RT, and CT. Anthropometrical as well as cardiorespiratory fitness parameters turned out to be significantly more improved following AET or CT protocols as compared to their respective RT counterparts. With respect to the limitations of the present systematic review, a conservative interpretation of the data is required. The primary objective in obesity management is the reduction of body fat. According to the results of the pairwise meta-analysis, reduction of fat mass was significantly more pronounced following AET, and CT as compared to RT. However, addition of RT to AET strategies may prevent loss of LBM, which is a common problem in the course of weight loss in obesity management programs. Evidence from the network meta-analysis suggests that CT is the most efficacious exercise modality in the prevention and treatment of overweight, and obesity and should therefore recommended whenever possible.