Chocolate milk is one of the few drinks in sports nutrition to have a pretty good reputation. Not only is it delicious, but it’s also rich in calcium, carbohydrates, flavonoids, and electrolytes. These are nutrients that seem to aid sports performance.
The conclusions from a few studies over the years have supported its use for improving sports performance. However, there are also several contradictory findings that didn’t find it to be superior to other sports drinks that contain carbohydrates and electrolytes.
A recent systematic review and meta-analysis published in Nature tried to gather all the literature on this topic and assess how effective chocolate milk was for improving several measures which include:
Ratings of perceived exhaustion (RPE)
Time to exhaustion (TTE)
Serum creatine kinase
This list of primary outcomes in the paper is nearly identical to what they registered on PROSPERO, which is great because it gives the reader some confidence that the authors didn’t run additional analyses to find significant results for their main outcomes… or did they (more on that below). The authors only included studies in their review that met the following criteria,
“Original articles utilizing a controlled trial design on trained participants or athletes and considered the effect of post-exercise CM consumption on subsequent exercise performance or recovery indices were included.”
After sifting through a mountain of studies, a total of 12 studies were included in the systematic review and 11 were included in the meta-analysis.
Let’s take a look at the quality of the randomized trials that were included in this paper because systematic reviews and meta-analyses are only as good as the studies included. The authors assessed the quality of the studies using Cochrane’s risk of bias tool. I’ve recreated their graphic to show what they’d rated the studies as.
Red = High Risk of Bias
Yellow = Unclear
Green = Low Risk of Bias
So, it seems most of the trials seemed to be of “fair quality.” That doesn’t inspire much confidence, but it’s not horrible. Now we can get into the results for each of the outcomes.
TTE: MD = 1.46 min, 95% (CI): −0.31, 3.22
RPE: MD = −0.04, 95% CI: −0.52, 0.44
Lactate: MD = −0.67 mmol/L, 95% CI: −1.49, 0.16
Creatine Kinase: MD = −70.61 U/L, 95% CI: −169.26, 28.04
Heart Rate: MD = −0.24 bpm, 95% CI: −4.22, 3.75
None of the outcomes above were statistically significant, but many of the intervals lean toward an effect. They’re also incredibly wide for most of the outcomes, indicating a great deal of random error.
Also seemed to be no evidence of publication bias for any of the outcomes based on funnel plots and Begg’s and Egger’s tests. So, based on the planned analyses by the authors, they didn’t find that chocolate milk consumption had a difference on any of the outcomes.
It is worth noting that there was significant between-study heterogeneity, as measured by the I2 statistic and Cochran’s Q, which can be thought of as measures of real differences in effect sizes not attributed to random error.
Typically, authors do subgroup analyses to explore these true sources of differences, and when the authors for this paper ran subgroup analyses for several of the outcomes, they did find statistically significant differences for some of them. However, statistically significant or not, none of the subgroup analyses were planned, so some of the significant results could simply be false positives. Again, this is also why preregistration is so important!
“We will conduct subgroup analysis and meta-regression where possible based on study design and supplementation dose if there was any evidence of heterogeneity between included studies.”
So while they did have a genuine reason to explore the sources of dispersion, these results are pretty much exploratory and shouldn’t be taken as evidence that chocolate milk is a performance enhancing drink.Yet, that’s what the media did, as shown below.
So does chocolate milk improve sports performance? Maybe. I don’t know. However, I do know that the results of this study don’t provide any strong evidence for the claim that it does or that it doesn’t. There may be some reason to follow up, based on some of the findings of the subgroup analyses, but if we want more robust evidence for either a meaningful effect or a nonmeaningful effect, we’ll need larger well-designed randomized trials.
But again, this isn’t new. I’ve covered several instances on my blog where science journalists simply got it wrong, some of which are listed below.
The content from this post was also featured on the following video:
That’s all for today folks.