The Effects of Creatine Supplementation Combined with Resistance Training on Regional Measures of Muscle Hypertrophy: A Systematic Review with Meta-Analysis

2.1. Literature Search

We preregistered our methods for both the systematic review and the meta-analysis on the Open Science Framework website (https://osf.io/c7bez and https://osf.io/hdqzb (accessed on 15 January 2023), respectively). We searched the PubMed/MEDLINE, Scopus, and Web of Science databases from inception to February 2023 to locate relevant studies. The search syntax was performed using the following combination of terms: (“creatine”) AND (“supplement*”) AND (“resistance training” OR “resistance exercise” OR “weight lifting” OR “weight-lifting” OR “weightlifting” OR “strength exercise” OR “strength training” OR “strengthening” OR “resistive exercise” OR “resistive training”) AND (“muscle hypertrophy” OR “muscular hypertrophy” OR “muscle mass” OR “muscle size” OR “muscle thickness” OR “cross-sectional area” OR “cross sectional area” OR “muscle volume”). The other articles included in our systematic review were either known by the authors or identified by manually searching the bibliographies of the retrieved articles. Two researchers (RB and AM) screened the retrieved abstracts and reviewed the full texts for studies that conceivably met the inclusion criteria. Inclusion required agreement between both researchers; in cases where a disagreement arose, a third researcher (BJS) resolved the dispute. The methods followed the guidelines set forth by the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) [20].

2.2. Inclusion/Exclusion Criteria

We included studies that: (1) investigated the longitudinal effects of creatine supplementation (in any form) combined with RT vs. RT without creatine supplementation; (2) had a duration of ≥6 weeks; (3) included adults (≥18 years of age); (4) were published in English-language peer-reviewed journals; and (5) reported pre–post study changes in site-specific hypertrophy, employing a validated imaging modality including magnetic resonance imaging (MRI), computed tomography (CT), or ultrasound.

Studies were excluded if: (1) the participants had pre-existing musculoskeletal disorders, cardiovascular diseases, or any other condition that could be considered detrimental to the RT performance; (2) other potentially anabolic ingredients were included in the supplementation formula (since protein is a food source, we allowed for supplementation with this macronutrient, provided its provision was equated between conditions); (3) blood flow restriction was incorporated into the RT protocol; or (4) there was insufficient numerical or graphical data provided to assess the differences between the conditions.

2.3. Data Coding and Analysis

We extracted the data from the included studies and coded them in an Excel spreadsheet (Microsoft Corporation, Redmond, Washington), which was performed by 2 authors (FA and MC) using the following classifications: (1) study characteristics (author, year of publication, and sample size); (2) participant demographics (age, sex, and RT status); (3) training methods (sets, exercises, frequency, duration, and repetitions); (4) supplementation methods (dose, timing, blinding, placebo, and protein supplementation); and (5) pre- and post-training means and standard deviations of hypertrophy. In cases where the studies lacked sufficient information regarding pre–post changes, we contacted the authors to request the missing data. If we were unable to acquire this data from authors, we extracted values from the figures using the WebPlotDigitizer online software (https://apps.automeris.io/wpd/, (accessed on 15 March 2023)) where applicable. To account for the possibility of coder drift, a third researcher (MS) re-coded 30% of the studies, all of which were randomly selected for assessment [21]. The per case agreement was determined by dividing the number of variables that were coded the same by the total number of variables. Acceptance required a mean agreement of ≥90%. Any discrepancies in the extracted data were resolved through discussion and the mutual consensus of the coders.

2.4. Methodological Quality

As previously described [22], we assessed the methodological quality of the included studies via the Downs and Black assessment tool [23], which is a 27-item checklist that addresses the following aspects of a study’s design: reporting (items 1–10), external validity (items 11–13), internal validity (items 14–26), and statistical power (item 27). Consistent with previous systematic reviews of exercise interventions, we modified the checklist by adding two items relating to participant adherence (item 28) and training supervision (item 29) [24,25,26]. Each item in the checklist was scored with a “1” if the criterion was satisfied or with “0” if the criterion was not satisfied. Based on the summary scores, the studies were classified as follows: “good quality” (21–29 points); “moderate quality” (11–20 points); or “poor quality” (less than 11 points) [25,26]. Three reviewers (AP, FA, and AM) independently rated each study; any disagreements in the study ratings were resolved by a majority consensus.

2.5. Statistics

A Bayesian framework was chosen over a frequentist approach as it can provide more flexible modeling, enabling the results to be presented intuitively through the reporting of subjective probabilities [27]. Where sufficient data were available, the comparative effects comparing RT with and without creatine supplementation were quantified using controlled absolute mean differences effect sizes to facilitate interpretations. Where there was a need to combine the results on different scales, controlled standardized mean differences effect sizes were used. Three-level random-effect Bayesian hierarchical models were used to pool the effect sizes and model the average effects, variance within the studies, variance between the studies, and covariance of multiple outcomes reported in the same study (e.g., hypertrophy of different sites and/or outcomes reported at multiple time points following the baseline). Within-study variance is influenced by pre–post correlations [28] that are generally not reported. Rather than specifying a single correlation value, this was estimated but constrained using informative prior distributions. We intended to use these informative prior distributions for the comparative effect sizes based on previous meta-analysis data [29]. However, due to limited information regarding regional changes in muscle mass, we ultimately employed default weakly informative prior distributions.

The inconsistency in the models was described by comparing the variances across the three levels. Inferences from all the analyses were made from posterior samples generated using the Hamiltonian Markov Chain Monte Carlo method and via the use of credible intervals (CrI), and the probabilities were calculated. The interpretations were based on the range of values within the CrI and the calculations of the probability that the magnitude of the average effect size exceeded the qualitative thresholds (i.e., small, medium, and large) that were specific to the strength and conditioning interventions [29]. Meta-regression or subgroup analyses were performed when there were sufficient data, including a minimum of 4 data points per category level or 10 data points for the continuous variables [30]. The small-study effects (publication bias, etc.) were visually inspected with funnel plots and quantified with a multi-level extension of Egger’s regression intercept test [31]. The analyses were performed using the R wrapper package brms interfaced with Stan to perform the sampling [32].

3.1. Descriptive Data

A total of 10 articles met the inclusion criteria (see Figure 1 and Table 1). The duration of the studies ranged from 6 to 52 weeks. Four studies included young adults (aged 21–26 years) [33,34,35,36] and 6 studies included older adults (aged 57–72 years) [37,38,39,40]. Four studies included only males [35,37,38,39], one study included only females [40], and five studies included both males and females [33,34,36,41,42]. Two studies employed resistance-trained participants [34,36] and the others employed untrained participants. All the studies incorporated a parallel group design and all the RT sessions were performed two–five times per week. One study solely focused on training the elbow flexors [35]; all the other studies implemented total-body training protocols. One study involved a creatine “loading” phase, which involved the consumption of 20 g/day of creatine for 5 consecutive days prior to consuming 5 g/day for the remaining duration of the study [35]; one study involved the supplementation of either 6 g/day of creatine or 6 g/day of creatine in combination with 30 g of whey protein [41]; and all the other studies implemented dosing protocols of either 0.1 or 0.15 g/kg/day of creatine. Three studies involved the ingestion of creatine two–four times per week [33,36,39] and all the other studies involved the ingestion of creatine five–seven times per week. One study measured the muscle CSA of the lower leg and forearm using peripheral quantitative computed tomography [42], one study measured the CSA of the vastus lateralis using ultrasonography [41], and the remaining studies measured the muscle thickness of the upper and lower extremities using ultrasonography.

3.2. Univariate Analysis of Standardized Differences

An analysis of the controlled standardized mean differences was conducted by pooling all the outcomes and measurement scales. A total of 44 outcomes were included across the 10 studies meeting the inclusion criteria (elbow flexors: nine; elbow extensors: eight; knee flexors: eight; knee extensors: eight; ankle plantarflexors: four; ankle dorsiflexors: three; vastus lateralis: two; forearm: one; and lower leg: one). A forest plot of the meta-analysis is presented in Figure 2, with a pooled mean estimate of 0.11 (95%CrI: −0.02 to 0.25) providing evidence for a very small effect favoring creatine supplementation (P(>0) = 0.961, P(>0.1) = 0.588, P(>0.2) = 0.089). Substantial heterogeneity was identified ($\tau$ = 0.10 (75%CrI: 0.03 to 0.20)), with central estimates indicating a low covariance between multiple outcomes reported from the same study (ICC = 0.17 (75%CrI: 0.01 to 0.79)). Egger’s regression intercept test produced wide intervals and a visual inspection of the funnel plot (Figure 3) did not identify any small-study-related issues.

These distributions represent “shrunken estimates” based on all the effect sizes included and the random effects model fitted and borrowed information across the studies to reduce uncertainty. The black circles and connected intervals represent the median values and 95% credible intervals for the shrunken estimates. The white circles and intervals represent the raw estimates and sampling variances calculated directly from the study data.

3.3. Multivariate Analysis of Absolute Differences

The initial analyses were conducted using the controlled absolute changes in the muscle hypertrophy measured by an ultrasound in cm. The multivariable analyses were conducted and split into upper-body (elbow extensors: seven studies [33,34,36,37,38,39,40]; elbow flexors: eight studies [33,34,35,36,37,38,39,40]) and lower-body (knee extensors and knee flexors: seven studies [33,34,36,37,38,39,40]).

The meta-analysis results for the upper body are illustrated in Figure 4, providing evidence favoring creatine supplementation. The marginal pooled controlled mean difference estimates were 0.16 cm (95%CrI: −0.10 to 0.39) and 0.10 cm (95%CrI: −0.13 to 0.32) for the elbow extensors and flexors, respectively. Substantive heterogeneity was identified (elbow extensors: $\tau$ = 0.15 cm (75%CrI: 0.04 to 0.31); elbow flexors: $\tau$ = 0.13 cm (75%CrI: 0.04 to 0.27)) with a positive but uncertain correlation ($\rho$ = 0.38 (75%CrI: −0.11 to 0.84)). The probability that the controlled absolute mean differences favored creatine supplementation for both the extensors and flexors was $P\left(>0\right)=$ 0.801.

The meta-analysis results for the lower body are illustrated in Figure 5 and also provide evidence favoring creatine supplementation. The marginal pooled controlled mean difference estimates were 0.13 cm (95%CrI: −0.07 to 0.37) and 0.11 cm (95%CrI: −0.06 to 0.31) for the knee extensors and flexors, respectively. Substantive heterogeneity was identified (knee extensors: $\tau$ = 0.11 cm (75%CrI: 0.03 to 0.24); knee flexors: $\tau$ = 0.07 cm (75%CrI: 0.02 to 0.17)) with a positive but uncertain correlation ($\rho$ = 0.18 (75%CrI: −0.32 to 0.66)). The probability that the controlled absolute mean difference favored creatine supplementation for both the extensors and flexors was $P\left(>0\right)=$ 0.850.

3.4. Analysis of Moderating Effects

The potential moderating effects of age and the duration of the intervention were investigated using controlled standardized mean differences across all the outcomes. Six studies comprising 26 outcomes were conducted with older adults (mean age: 61.6 years) and four studies comprising 18 outcomes were conducted with younger participants (mean age: 23.5 years). Evidence was obtained that the controlled standardized mean differences were greater for younger participants (${\beta }_{Younger}=$ −0.17 (95%CrI: −0.45 to 0.09); P(Younger > Older) = 0.910; Figure 6).

Seven studies comprising 30 outcomes were conducted with short interventions (mean duration: 8.6 weeks, range: 6–16 weeks) and three studies comprising 14 outcomes were conducted with long interventions (52 weeks). A substantive overlap was obtained for the controlled standardized mean difference distributions, with limited evidence of greater values for a short intervention (${\beta }_{Short:Long}=$ −0.09 (95%CrI: −0.37 to 0.19); P(Short > Long) = 0.767; Figure 6).

3.5. Study Quality

A qualitative assessment of the studies via the Downs and Black checklist indicated a median score of 20 (range: 17 to 25 points). Four studies were deemed to be of good quality [35,37,38,40], six studies were classified as being of moderate quality [33,34,36,39,41,42], and no studies were found to be of poor quality.