Summary of the meta-analyses included in the review
| Reference | Included studies | Number of included studies (sample size) | Performance test(s) | Effect size (95% CI) and p value* | 95% PI | Per cent change | I2 (95% CI) |
| Christensen et al 4 | Single or double-blind crossover study designs. | Nine studies (n=97). | Speed during running, cycling or rowing.† | 0.41 (0.15 to 0.68); p=0.002. | 0.41 (0.09 to 0.73). | ~2 | 0% (0% to 35%). |
| Conger et al 5 | Crossover study designs. | Carbohydrate versus caffeine+carbohydrate: 21 studies (n=333). Caffeine versus placebo: 36 studies (n=352). | Any form of aerobic exercise if it was 10 min or longer in duration. | Carbohydrate versus caffeine+carbohydrate: 0.26 (0.15 to 0.38); p<0.001. Caffeine versus placebo: 0.51 (0.41 to 0.62); p<0.001. | Carbohydrate versus caffeine+carbohydrate: 0.26 (−0.18 to 0.70). Caffeine versus placebo: 0.51 (−0.06 to 1.08). | Carbohydrate versus caffeine+carbohydrate: +6%. Caffeine versus placebo: +16%. | Carbohydrate versus caffeine+carbohydrate: 7% (0% to 42%). Caffeine versus placebo: 24% (0% to 50%). |
| Doherty and Smith6 | Double-blind crossover study designs. | 24 studies (n=217) for aerobic exercise, 6 studies for graded exercise tests (n=62) and 12 studies for short-term high-intensity exercise (n=127). | Exercise testing divided to aerobic exercise, graded exercise tests and short-term high-intensity exercise. | Aerobic exercise: 0.63 (0.50 to 0.77). Graded exercise tests: 0.17 (−0.02 to 0.36). Short-term high-intensity exercise: 0.16 (0.01 to 0.31).‡ | Aerobic exercise: 0.63 (0.06 to 1.20). Graded exercise tests: 0.17 (−0.09 to 0.44). Short-term high-intensity exercise: 0.16 (−0.18 to 0.50). | +12% across all exercise tests. | Aerobic exercise: 4% (0% to 52%). Graded exercise tests: 0% (0% to 42%). Short-term high-intensity exercise: 0% (0% to 24%). |
| Gonçalves Ribeiro et al 8 | Crossover study designs. | Seven studies (n=91) for time trial duration, four studies (n=52) for power and two studies (n=31) for running distance. | Time trial duration, power and running distance. | Time trial duration: 0.40 (0.11 to 0.70); p=0.007. Power: 0.18 (−0.21 to 0.56); p=0.366. Running distance: 0.38 (−0.13 to 0.88); p=0.142. | Time trial duration: 0.40 (0.01 to 0.79). Power: 0.18 (−0.65, 1.01). Running distance: unable to determine. | Time trial duration: +2%. Power: +4%. Running distance: +11%. | Time trial duration: 0% (0% to 19%). Power: 0% (0% to 43%). Running distance: 0% (unable to determine). |
| Grgic9 | Crossover study designs. | 16 studies (n=246). | Peak and mean power in the 30 s Wingate test. | Peak power: 0.27 (0.08 to 0.47); p=0.006. Mean power: 0.18 (0.05 to 0.31); p=0.005. | Peak power: 0.27 (−0.35 to 0.89). Mean power: 0.18 (0.04 to 0.32). | Peak power: +4%. Mean power: +3%. | Peak power: 7% (0% to 44%). Mean power: 0% (0% to 28%). |
| Grgic and Pickering10 | Crossover study designs. | 10 studies (n=133). | Peak torque in an isokinetic strength assessment. | 0.16 (0.06 to 0.26); p=0.003. | 0.16 (−0.17 to 0.49). | +5%. | 15% (0% to 57%). |
| Grgic et al 11 | Single or double-blind crossover study designs. | 10 studies (n=149) for strength and 10 studies (n=145) for vertical jump. | Strength in the 1RM test and height in a vertical jump test. | 1RM: 0.20 (0.03 to 0.36); p=0.023. Vertical jump: 0.17 (0.00 to 0.34); p=0.047. | 1RM: 0.20 (0.02 to 0.39). Vertical jump: 0.17 (−0.03, 0.37). | 1RM: +3%. Vertical jump: +3%. | 1RM: 0% (0% to 37%). Vertical jump: 0% (0% to 47%). |
| Polito et al 12 | Double-blind crossover study designs. | 16 studies (n=239) for muscular endurance and three studies (n=46) for the 1RM test. | Muscular endurance (assessed by repetitions to fatigue) and strength in the 1RM test. | Muscular endurance: 0.38 (0.29 to 0.48); p<0.001. 1RM: 0.09 (−0.07 to 0.25); p=0.25. | Muscular endurance: 0.38 (0.02 to 0.74). 1RM: 0.09 (−0.09 to 0.27). | Muscular endurance: +6%. 1RM: +2%. | Muscular endurance: 24% (0 to 56%). 1RM: 0% (0% to 43%). |
| Shen et al 13 | Crossover study designs. | 40 studies (n=582). | Any form of aerobic exercise if it was 5 min or longer in duration. | 0.33 (0.21 to 0.45).‡ | 0.33 (0.21 to 0.45). | +3%. | 0% (0% to 14%). |
| Southward et al 14 | Crossover study designs. | 44 studies (n=639 for time trial duration and n=350 for time trial power). | Duration of the time trial or power during a time trial. | Time trial duration: 0.28 (0.17 to 0.40); p<0.0001. Time trial power: 0.22 (0.07 to 0.37); p=0.004. | Time trial duration: 0.28 (0.17 to 0.40). Time trial power: 0.22 (0.06 to 0.38). | Time trial duration: +2%. Time trial power: +3%. | Time trial duration: 0% (0% to −56%). Time trial power: 0% (0% to 14%). |
| Warren et al 15 | Crossover and between group study designs. | 27 studies (n=576) for MVC and 23 studies (n=388) for muscular endurance. | MVC and muscular endurance. | MVC: 0.19 (0.09 to 0.29); p<0.001. Muscular endurance: 0.28 (0.14 to 0.42); p<0.001. | MVC: 0.19 (−0.18 to 0.56). Muscular endurance: 0.28 (−0.29 to 0.85). | MVC: +4%. Muscular endurance: +14%. | MVC: 44% (13% to 65%). Muscular endurance: 12% (0% to 46%). |
*Positive effect sizes and percentages show favouring of caffeine over placebo.
†Defined as the maximal achieved speed in exercise performance tests lasting from 45 s to 8 min that had either a fixed duration or a fixed distance.
‡P values were not provided.
1RM, one repetition maximum; MVC, maximal voluntary contraction; PI, prediction interval.