If you are in a hurry, read this first
Here’s a summary of what we found, using some new Mergeflow functionalities, and clinical trials:
- Quite a number of clinical trials look at connections between lifestyle and medical conditions such as “obesity”, “diabetes”, and “cardiovascular conditions”.
- Besides “mainstream” nutritional supplements, such as vitamin D or omega-3 fatty acids, some less-well-known supplements are investigated as well. Anthocyans, for example, which are claimed to have antioxidant effects.
- Wearables do not (yet) play a role in clinical trials that specifically investigate nutritional supplements.
- Beyond nutritional supplements, Fitbit seems to be the wearable of choice for monitoring patients’ activity patterns in clinical trials related to obesity, diabetes, and cardiovascular conditions.
If you are not in a hurry, some background first
For this analysis, we used Grid Search. With Grid Search, you can explore entire topic landscapes, rather than run individual searches. I will come back to this later. We also used Grid Search to explore how you can use machine learning to address climate change, or how you can discover company strategies in additive manufacturing.
As a data set, we used clinical trials here. Clinical trials are reports on clinical research, for example in the context of drug development (cf. https://en.wikipedia.org/wiki/Clinical_trial for more details). Every week, Mergeflow collects 600 new clinical trials on average, and our data set goes back ca. 20 years.
Now, what is Grid Search really, and what does it have to do with clinical trials?
Let’s say you want to explore a whole range of topics. For example, you might be interested in connections between various nutritional supplements (vitamins, minerals, etc.) and various medical conditions (e.g. diabetes, obesity, cardiovascular). In order to discover these connections, you could then search for information on all possible combinations. But given the combinatorics, this is a non-starter. It would take you days or more to do this. And then only imagine if you also wanted to continuously monitor new developments. No way.
Grid Search does all these combinatorics for you. Once you have defined your queries (the nutritional supplements and the medical conditions), Mergeflow Grid does the rest. The result is a matrix that shows you the connections between your topics (yes, we will have one below).
By the way, another way of exploring topic intersections is Mergeflow Teams. It enables collaborative tech discovery across topics, without the pains of information management and manual data labeling.
Why clinical trials? Because as far as research goes, clinical trials really matter. They are very close to how medicine is or will actually be practiced. They are subject to approval by national regulatory authorities such as the NIH (https://www.nih.gov/). Perhaps you could say that they are about “turning discovery into health”, which is the motto of NIH. This means that they are a very good place for discovering game-changing or breakthrough innovations.
What did we want to know?
Before we start, let me emphasize that none of us here at Mergeflow has any medical expertise to speak of. We are computer scientists, economists, neuroscientists; but as far as medical research goes, we are simply a bunch of curious people. So please bear this in mind when you read this post. If you have medical expertise, we’d love to hear from you, learn from you, and discuss the kinds of questions that you would like to ask (and how we might be able to help you do that).
We wanted to know if there are any clinical trials that investigate how various nutritional supplements relate to medical conditions such as diabetes, obesity, and cardiovascular diseases. And we also wanted to know if and how wearable devices play a role here, given that some of them are starting to be used for ECGs and other diagnostics. Last but not least, we were interested in connections with lifestyle (e.g. physical exercise).
How did we look?
We set up a number of Mergeflow queries for nutritional supplements, wearable diagnostics, lifestyle, obesity, diabetes, and cardiovascular.
Nutritional supplements: We used parts of a list from the NIH, their “dietary supplements fact sheets”: https://ods.od.nih.gov/factsheets/list-all/.
Wearable diagnostics: Besides searching for “noninvasive diagnostics” generally, we also searched for specific providers, e.g. Fitbit, Garmin, Misfit, Withings, Apple.
Lifestyle: We included physical exercise, sleep, insomnia, sedentary lifestyle.
Obesity: We looked e.g. for obesity in general, body mass index, metabolic syndrome, body fat, endocrine disease / function.
Diabetes: We included e.g. diabetes, glucose, Kussmaul breathing.
Cardiovascular: Here, we included e.g. atrial fibrillation, coronary heart disease, high blood pressure, myocarditis.
For defining our queries, we relied heavily on Wikipedia.
And what did we find?
We took all our queries in the groups described above, and ran them through a Mergeflow Grid in clinical trials of the last ten years. This produced a rather big matrix. In order to see the whole thing, click on it. It might be a bit overwhelming at first, but don’t worry, we will walk you through it.
First, here are some basics of how the matrix works:
- It is symmetrical, i.e. rows and columns are identical; each row and each column represents one of our queries.
- The boldface labels are our topic groups (nutritional supplements, lifestyle, wearables, obesity, diabetes, cardiovascular).
- Red fields = total number of clinical trials per single query (e.g. all clinical trials for “Vitamin C / ascorbic acid”).
- Green fields = number of clinical trials per query combination (e.g. number of clinical trials for “Vitamin C and blood sugar”).
OK, now let’s get to work.
If you look at the overall patterns across the matrix, you will notice that there are strong correlations among our medical conditions (diabetes, obesity, cardiovascular). This is probably not very surprising and might be considered a “sanity check” for our method.
What is probably a bit more interesting is the strong connection between lifestyle and the medical conditions, but also to some extent the nutritional supplements, particularly e.g. to omega-3 fatty acids and vitamin D (again, click on the image to see it bigger):
Then there are some white spots. Of course, such white spots are most interesting if the topics by themselves have many hits. For example, there are many clincal trials on wearables and many trials on nutritional supplements. But the connections are basically empty. In other words, so far nobody seems to have run any clinical trials that look at nutritional supplements and tracked their effect on e.g. heart rate or calory burn with wearables (or used wearables to control for physical activity level differences among a trial test group).
More particular observations
Now let’s get into some more details, and look at some underlying data (our Mergeflow Grid matrix is interactive, zoomable, and clickable, so you can always get to the underlying data, which are clinical trials in our case). Our goal here is not to show you an exhaustive list of all things that might be interesting. Rather, we looked at a small selection of things we wanted to know more about. Our selection is certainly subjective, note also our disclaimer above, regarding our lack of medical expertise.
Anthocyans are naturally occurring pigments that, depending on a number of factors, can appear somewhere between red, blue, and black (https://en.wikipedia.org/wiki/Anthocyanin). We had heard on several occasions that anthocyans are supposed to have antioxidant effects. For examle, here is a blog post that talks about this:
We wanted to know if we see something on this topic in clinical trials. Yes, it turns out. There are a number of trials that have looked e.g. at connections between anthocyans and glucose or cardiovascular conditions (cf. red arrows in the screenshot below):
For these connections, anthocyans in the context of glucose and cardiovascular conditions, we zoomed in further and only considered clinical trials at phase 3. Phase 3 trials aim at “final confirmation of safety and efficacy” of a clinical regimen (cf. https://en.wikipedia.org/wiki/Clinical_trial). Here is a phase 3 trial on anthocyans in the context of glucose:
And here is one for the cardiovascular conditions context:
Next, we looked at vitamin K. We were interested in vitamin K because it was one of our stronger connections in our matrix (i.e. still colored when we set the threshold for connections quite high), as you can see in the screenshot below:
In case you have any medical expertise, this probably does not surprise you. The background here is that atrial fibrillation patients often take anticoagulants, such as warfarin, to avoid systemic clotting. Vitamin K has the opposite effect, in other words, warfarin etc. are vitamin K antagonists.
We found this connection interesting, given that popular opinion often is, in very general terms, that “vitamins are good for you”. What these results show is that the answer may rather be, “it depends”. Just because you can buy it over the counter, it might not mean that it is good, or even harmless in some cases.
Differences between wearables makers
Not surprisingly, wearables often occur in lifestyle contexts. But note that there are differences between wearables.
In many cases, Fitbit wearables are the device of choice for monitoring in a medical context.
Fitbit occurs a lot more in clinical trials than the other makers:
One strong context is “sleep”. Here, we zoomed in on phase 2 trials. Phase 2 trials aim to establish the efficacy of a drug (as compared to a placebo). For example, here is a trial that investigated a treatment for chronic fatigue syndrome:
A Fitbit device was used for monitoring patients’ activity levels.
Next, we looked at medical condition contexts. Again, Fitbit was dominant:
In most of these trials, wearables (mostly Fitbit) were used for activity level monitoring. We had seen this already, so we instead decided to zoom in on trials that investigated the wearable devices per se (i.e. their accuracy). For example, this study that started in 2017 investigates the Apple Watch (photoplethysmography = optical measurement of blood volume):
Another, more recent, study also investigates the accuracy of optical sensing technologies for heart rhythm monitoring. It uses Apple and Fitbit wearables:
This study also evaluates FibriCheck (https://fibricheck.com/), a software that detects atrial fibrillation from signals provided by wearables or smartphones (in which case the smartphone’s camera is used as optical sensing device).s) for these connections. Who knows, perhaps we will see something happen there soon.