Weighing up body fat scales (part 1)
I've always wondered just how accurate body fat scales actually are, and after quite a lot of research and no actual answers, I decided to try and work it out for myself. Or, at least get a better idea by collecting some real world data. Cue body fat scale experiment!
What I did
I booked a DEXA scan and I bought some body fat scales.
If you haven't heard of a DEXA scan (I hadn't until Brian told me about it), DEXA stands for Dual-energy X-ray absorptiometry (in case you're interested) and it's basically a low dose X-ray that gives you a really good visual image of your bone density, muscle mass and fat percentage.
On the morning of my scan, I weighed myself on each of the body fat scales I'd bought, then headed off for the scan. I used the scan results as my control and could then compare each of the scales against that. It was only one point in time, but the idea was to see how they compared against 'the correct answer' on that specific day.
The scales I compared against the DEXA scan output:
FitBit Aria
Kamtron cs20m
Renpho es-cs20m
Tanita bc-602
Withings body+
All of these devices except the Tanita connected to a smartphone application. I installed these on my test Android device (first generation Pixel). I should note that just by taking the Renpho and Kamtron scales out of the packaging and connecting them to their relevant mobile applications, I could tell that these are the same scales, just in different outer shells. The apps were identical just with different names and the scales behaved in exactly the same way and even had the same model number. I therefore expected these two devices to give me identical results.
The results
Just taking into account body weight and body fat percentage measurements, the most accurate scales by a fair margin were the FitBit Aria scales. Bottom of the pile were the Kamtron, which underperformed the Renpho all the way through, against all of my expectations.
In order, from most to least accurate on the day when compared to the DEXA scan output:
1. FitBit Aria
Body weight: -0.04kg
Body fat percentage: +0.1%
The FitBit Aria scales only measure these two things, so further comparison wasn't possible.
2. Renpho es-cs20m
Body weight: -0.36kg
Body fat percentage: +1.7%
Muscle mass: -1.49kg
Bone mass: +0.24kg
BMR: +22kcal
3. Withings body+
Body weight: -0.36kg
Body fat percentage: -0.74%
Muscle mass: +0.28kg
Bone mass: -0.12kg
The Withings Body+ scales didn't have a BMR feature.
4. Tanita bc-602
Body weight: -0.4kg
Body fat percentage: -2.4%
Muscle mass: +1.21kg
Bone mass: -0.1kg
BMR: +1148kcal
Because the Tanita scales also measured segmented muscle mass and fat percentage, I also compared these to the DEXA output:
Segmented muscle mass:
Left arm: -0.17kg
Right arm: -0.17kg
Left leg: +0.02kg
Right leg: -0.06kg
Trunk & head: +1.59kg
Segmented fat percentage:
Left arm: -5.1%
Right arm: -5.1%
Left leg: +0.4%
Right leg: +1.2%
Trunk & head: -2.8%
5. Kamtron cs20m
Body weight: -0.45kg
Body fat percentage: -3.7%
Muscle mass: +1.59kg
Bone mass: +0.42kg
BMR: +91kcal
Conclusions
The body fat scales actually performed much better than I expected they would, on the whole. Probably in part due to the fact that my body fat percentage in the DEXA scan was higher than I was expecting; apparently everyone thinks that when they get their results, so that made me feel a little better!
The FitBit Aria scales came out on top as the most accurate, but they only measure body weight and body fat percentage, so if you want more measurements like muscle mass and bone mass, the Renpho was the next most accurate all round. The BMR measurements were a bit erratic, particularly from the Tanita scales. So, if you do have scales that measure this, I'd be inclined to automatically assume your true BMR is lower.
This was also only one point in time, and I didn't think it was entirely fair to judge the scales solely based on this, so onto Part 2 of this experiment, where I used all of the scales for a few days straight so I could collect a few measurements over time and compare the scales with each other...