I got something interesting in my inbox the other day. Something that I assume was a friend’s email address getting hacked – although it’s the least offensive (apparent) hack I’ve ever seen (he says as the viruses circulate around his computer’s RAM).
It was a nearly blank email with a link to a Dr. Oz clip about the weight-loss promoting effects of green coffee extract, which contains high concentrations of chlorogenic acids. These molecules are said to promote weight loss through increasing metabolism.
Being a scientist means being a skeptic. In this case, because I already feel like it must be BS due to its connection with Dr. Oz (an Oprah-elevated proponent of many untested, ‘alternative’ therapies), the challenge for me is to admit the possibility that this stuff may work. So, rather than looking through the data to see if there’s anything to deny the claim, I’m really trying hard to look at the data to see any glimmer of possibility.
Here’s a link to the Dr. Oz article online. The article was published in the January 2012 Diabetes, Metabolic Syndrome and Obesity, and happily the entire article is available free of charge. So let’s look at the data…
The article examines a “22-week crossover study was conducted to examine the efficacy and safety of a commercial green coffee extract product GCA™ at reducing weight and body mass in 16 overweight adults.” Half of the participants were male and half female – a typical study setup (although I do worry about how data is handled when looking at both sexes together, so let’s pay attention to that.)
Dr. Oz’s website indicates that “The subjects (taking the supplement) lost an average of almost 18 pounds – this was 10% of their overall body weight and 4.4% of their overall body fat.” These are pretty hefty claims, but I could use losing 18lbs, so let’s see where this goes.
The study followed those eight men and eight women for 22 weeks. At the beginning of the study, the average body mass index (BMI) at the start of the study was 28.22 ± 0.91 kg/m2 . Determine your own BMI here.
Note that BMI < 18.5 is underweight
18.5 – 25 healthy weight
25 – 30 overweight
This puts the study participants at the high end of overweight, but ‘preobese’.
Dosages of the green coffee extract and placebo were as follows:
“This study utilized two dosage levels of GCA, as well as a placebo. The high-dose condition was 350 mg of GCA taken orally three times daily. The low-dose condition was 350 mg of GCA taken orally twice daily. The placebo condition consisted of a 350 mg inert capsule of an inactive substance taken orally three times daily.”
I don’t think I’m the first one to point out that it’s hard to have a double blind trial when the dosages are distinguishable (two times vs three times daily). At least the placebo should be indistinguishable from the high dose.
One early eye-catching piece of data is from Table I, that summarizes the data of all precipitants as
BMI (kg/m2) pre study:28.22 ± 0.91 post study:25.25 ± 1.19 change-2.92 ± 0.85**, -10.3%
On average, all subjects lost weight during the study. But this really tells us nothing because we could see a 10% drop in BMI if the test arm lost 20% and then placebo arm stayed the same, or we could see the same thing if the weight loss occurred during ALL arms of the study.
Perhaps this reporting of data is justified by the next statement that participants all rotated through being on high dose, lose dose or placebo with intervening washout periods. Presumably, this makes the most of a small sampling of people, but I do find it harder to be confident about the data. Then again, I have never been involved in any human trial of this kind.
here’s the data:
High Dose arm:
start BMI (kg/m2) 26.78 ± 1.55 –> end 26.03 ± 1.36
Low Dose arm:
start BMI (kg/m2) 26.25 ± 1.37 –> end 25.66 ± 1.20
start BMI (kg/m2) 25.66 ± 1.20 –> and 26.67 ± 1.72
At first glance this might appear to be pretty good. But let’s graph it out:
the data continue to look great.
Now, with error bars:
Huh. Not so hot anymore.
Also, I’m not how sure this was done, but they get p values for HD p = 0.002, LD p = 0.003, placebo p = 0.384. These stats mean that the HD and LD groups are showing very significant differences, while the placebo group is not. You should be able to see this in the graph with error bars (as an approximation of significance). Unfortunately, I see a whole lot of no nothing. But, perhaps BMI is not the appropriate way to observe weight change (we are, after all not seeing specific weight changes, but changes within a group, i.e. diversity)
Another way to try to see what’s going on is to take a look at the weight data:
The data were presented in a number of other ways, but each of these was confusing and didn’t illustrate any clear conclusion (my interpretation).If the individuals’ data were visualized as a scatter plot, this might show us something – or data for each individuals change while in each group… As it is, we see unclear data with spectacular statistics, but we don’t get to see enough to be convinced of the changes.
Rather than go on and get more and more skeptical, let’s say, although we don’t see a lot here, the data,as reported, would make us want to see a larger study with some revisions for control of diet, exercise monitoring and a change in the way osage is administered so as to maintain the ‘blindness’ of the study.