Data abstraction and statistical analysis
Two independent reviewers abstracted data, and a third reviewer resolved differences. For studies investigating multiple sprints, data from the first sprint only were included in statistical analysis because the first sprint is when peak power achievement is expected. A weighted mean difference (WMD) between creatine and placebo groups was calculated for each outcome using Review Manager 4.1 software (developed by The Cochrane Collaboration). A fixed effects model was used unless statistical heterogeneity was significant (P < .05), in which case a random effects model was used. Subanalyses were planned on several factors that were anticipated to be sources for variation, including (1) dose and duration of creatine administration, (2) concomitant resistance training, (3) different baseline level of physical training, (4) age, and (5) sex.
Results
Description of studies
After reviewing titles and available abstracts of more than 500 articles, we retrieved 66 potentially relevant studies, 16 of which met inclusion criteria for the analysis.17,20-34 Characteristics of these studies are summarized in the Table. Included studies represented 20 discrete samples and 414 subjects. Two studies20,21 evaluated creatine supplementation in men older than 60 years, whereas all the others studied younger subjects (range, 18–36 years). Only 1 study included women.17 Creatine dosages were similar across included studies (typically 20 g/d for the first 4–7 days of supplementation and 5 g/d thereafter). Studies that evaluated maximal weight lifting performance were more likely to include adjuvant resistance training programs in their protocols than those that evaluated cycle ergometry sprint or isokinetic dynamometer performance. None included cycle ergometry training.
TABLE
Characteristics of included studies
Reference | No. subjects (sex) | Dose per day and duration | Training level | Weight training during study? | Outcome measurement | Quality score (out of 10) | Comparability of creatine & placebo groups at baseline* |
---|---|---|---|---|---|---|---|
Barnett 1996 | 17 (M) | 280 mg/kg ×4 d | Active | No | CP | 2.5 | + |
Cooke 1995 | 12 (M) | 20 g ×5 d | Untrained | No | CP | 2.5 | + |
Cooke 1997† | 80 (M) | 20 g ×5 d | Trained or active | No | CP | 2 | +++ |
Dawson 1995‡ | 18 (M), 22(M) | 20 g ×5 d | Active | No | CP | 3 | +++ |
Jones 1999 | 16 (M) | 20 g ×5 d then 5 g ×10 wk | Trained | Yes | CM | 3 | +++ |
Stone 1999 | 20 (M) | 0.22 g/kg ×35 d | Trained | Yes | CM, BP, S | 4.5 | +++ |
Kelly 1998 | 18 (M) | 20 g ×5 d then 5 g ×26 d | Trained | Yes | 3BP | 2 | – |
Noonan 1998 | 39 (M) | 20 g ×5 d then 300 mg/kg ×8 wk | Trained | Yes | BP | 5.5 | +++ |
Peeters 1999 | 35 (M) | 20 g ×3 d then 10 g ×6 wk | Trained | Yes | BP | 3 | +++ |
Vandenberghe 1997 | 19 (F) | 20 g ×4 d then 5 g ×10 wk | Untrained | Yes | BP, S | 5 | +++ |
Pearson 1999 | 16 (M) | 5 g ×10 wk | Trained | Yes | BP, S, PT | 3 | +++ |
Volek 1999 | 19 (M) | 25 g ×7 d then 5 g ×12 wk | Trained | Yes | BP, S | 4.5 | +++ |
Gilliam 2000 | 23 (M) | 20 g ×5 d | Active but untrained | No | PT | 2.5 | + |
Rawson 1999§ | 20 (M) | 20 g ×10 d then 4 g ×20 d | Untrained | No | AF, PT | 4.5 | +++ |
Rawson 2000§ | 17 (M) | 20 g ×5 d | Untrained | No | AF | 3.5 | +++ |
Becque 2000 | 23 (M) | 20 g ×5 d then 2 g ×6 wk | Trained | Yes | AF | 5 | + |
*Comparability between groups was assessed for age, anthropomorphic measurements, and strength outcomes. +++ = similar for all 3 characteristics; + = similar for strength outcome measurements; – = not comparable at baseline for strength outcome. | |||||||
†Four protocols with 20 subjects each evaluating the same strength outcome measurement reported in Cooke 1997. | |||||||
‡Two separate experiments reported in Dawson 1995. | |||||||
§Included subjects > 60 years old; in all others subjects were < 36 years old. | |||||||
AF, 1 repetition maximum arm flexor strength; BP, 1 repetition maximum bench press strength; 3BP, 3 repetition maximum bench press strength; CM, cycle ergometer mean peak power; CP, cycle ergometer peak power; PT, isokinetic leg flexion/extension peak torque; S, 1 repetition maximum squat strength. |
Methodological quality of included studies
The methodological quality of studies was generally low (Table). The mean quality score was 3.5 ± 1.2 (mean ± SD) out of a possible 10 (range, 2–5.5). None of the studies identified the method of randomization used or specifically reported an intention-to-treat analysis. None specifically reported masking of outcome assessment. In general, these significant flaws in study design would tend to result in overestimation of the benefit of creatine supplementation.
Absolute strength
When 1- to 3-repetition maximum bench press strength measurements were statistically combined (they were homogeneous), the creatine supplementation group showed an absolute strength increase of 6.85 kg (95% confidence interval [CI], 5.24–8.47; n = 143) lifted per repetition greater than that seen with placebo alone (Figure 1). There was no additional advantage in strength performance after 9 to 12 weeks of supplementation (WMD = 6.6 kg; 95% CI, 3.5–9.5) compared with 4 to 8 weeks of supplementation (WMD = 6.6 kg; 95% CI, 4.8–8.4). Subanalysis for an interaction with resistance training, previous training level, age, or sex was not possible because all studies measuring bench press strength except one17 investigated creatine supple mentation in previously trained young men who continued resistance training during supplementation. The 1 study in previously sedentary young women17 did find a trend toward increased bench press strength, although independently this change was not statistically significant.