In the first generation of our tests, we’ll be basing our reports on already published studies. For example, we look at spermidine (a polyamine), which is not available as a routine blood test and is a metabolite that could be acted upon. Epidemiological studies have shown that spermidine-rich diets are protective against cardiovascular disease and reduce the risk of cardiac death in humans. See references here:
Even though these studies have looked into metabolites in relation to chronic conditions, in the first generation of our reports, we’ll not be providing individuals with any diagnostic information and our tests right now are only intended for wellness purposes.
Regarding diagnostic predictive markers, I want to reiterate that we are not a diagnostics company at this stage and to quote from our post, “as our [longitudinal] metabolomics database grows [we will] look for new signatures of age-related diseases at earlier and earlier stages. (Such analysis will only be done on de-identified data, only with consent, and only for our work towards extending healthspan.) “.
With that being said, there are other groups that have done a great job of validating metabolite biomarkers that do provide relatively new predictive insights into chronic disease prediction and risk. One example is this paper where they looked at type 2 diabetes risk in individuals with *normal fasting glucose*
(https://link.springer.com/article/10.1007/s00125-018-4599-x):
Nineteen metabolites were selected repeatedly in the training dataset for type 2 diabetes incidence classification and were found to improve type 2 diabetes risk prediction beyond conventional type 2 diabetes risk factors (AUC was 0.81 for risk factors vs 0.90 for risk factors + metabolites, p = 1.1 × 10-4).
In adjusted Cox proportional hazard models, the type 2 diabetes risk per 1 SD increase in glycine, taurine and phenylalanine was 0.65 (95% CI 0.54, 0.78), 0.73 (95% CI 0.59, 0.9) and 1.35 (95% CI 1.11, 1.65), respectively. Mendelian randomisation demonstrated a similar relationship for type 2 diabetes risk per 1 SD genetically increased glycine (OR 0.89 [95% CI 0.8, 0.99]) and phenylalanine (OR 1.6 [95% CI 1.08, 2.4]).
Although we already measure these metabolites and others in our current panel and we are able to calculate these score, we will not be providing these score since we’re not a diagnostics company at our current stage.
Thanks for hanging around and answering questions. I too was wondering how the shipping affected the sample while it was in transit because part of what I read indicated that the sample was frozen upon arrival potentially (my impression here) to prevent degradation of the things that you hope to quantify.
Recognizing that the sample is not frozen before shipping and is therefore not "fresh" - 1) how much degradation occurs and 2) how have you quantified or measured the effects of a long strange trip through the postal services?
Thanks! We aren't working with metabolon but rather doing targeted metabolomics work. We do have some microbiome-related metabolites in our panel, and will definitely reach out once we want to expand it!
Appreciate the level of scrutiny, which is important for products like ours.
Re the affiliations, it is correct that these institutions aren’t directly endorsing iollo. Though both Jan and I are actively affiliated with said institutions and are translating what we have learned through our metabolomics research into iollo.
One thing to clarify here is that currently, we don’t treat patients and our tests are only intended to improve wellness. That being said, we do see biomarkers for specific food intake. Coffee is one example, where if someone has an unusually high levels of the biomarker for it and has difficulty sleeping during the night, reducing coffee intake may improve their sleep level and quality. We also have metabolites that associate with specific food, like beer, meat, vegetables and other that we can already use and that have been published. Also as our database grows, we’ll be able identify new food intake markers and will also publish and use them.
At our current stage, we don’t recommend any supplements. It is a part of our roadmap and once we do, we will only give supplement recommendations if we identify solid research that support their health benefits that can also be detected through metabolomics analysis.
Re the age-related diseases, what we mean on the website is that the genetic population attributable fraction for age-related diseases is around 15%, and that biological layers that captures the interactions between genetics and the environment which are related to these conditions is ~85%, which includes the metabolome. Source: https://journals.plos.org/plosone/article?id=10.1371/journal... We’ll work on further clarifying the language in a way that it would still be understandable from people outside of the research field.
Daniel, you are working in a field that has a low tolerance for doublespeak.
Parent has politely pointed out that the website needs revision. 85% of disease is related to the blood metabome? No. You will be detecting early disease signatures in a clincally actionable manner? No, you will almost certainly not be, and you know this.
The data isn't actually PHI. Stanford isn't actually a partner. AI in this field isn't actually real. Nature Medicine isn't actually Nature. Ect.
For historical reasons the benefit of doubt with regard to the scientific credibility of your startup is low. In my opinion, it would be better if you made an effort not to prove this assumption correct.
> Re the affiliations, it is correct that these institutions aren’t directly endorsing iollo.
You may want to talk to the relevant departments in the universities you list, their "innovations" or "spin-out" teams. Typically they have very specific terms under which you can use their logos and mention any form of association. You may well be in breach, and if you work in those universities this could land you in trouble.
> Coffee is one example, where if someone has an unusually high levels of the biomarker for it and has difficulty sleeping during the night, reducing coffee intake may improve their sleep level and quality.
This doesn't seem like something you need a blood test to figure out...
A quick counter point to the parent comment: I'd actually love to have supplement recommendations from a trusted source. I've heard on multiple occasions that many supplements found in large retailers don't have the ingredients they claim to have. So that would be highly valuable to me.
The problem of trust is never removed. But what must be trusted can be made explicit, the burden of trust can be shifted to more trustworthy parties, or to parties who have less problematic incentives.
What makes you think Iollo is more trustworthy or has better incentives?
They are using enough shady language that my trust level is pretty low, and I suspect one of the major incentives to recommend one supplement over another is financial.
Nothing. I was commenting generally about what is achievable in endeavors like this. Normally it's a parry for blockchain bros (who are stupidly enthusiastic) or cynics (who are often blind to vulnerabilities in existing trust relationships).
> Given the possibility that some findings might lead to the need for medication, are you going to have MDs or PAs or whoever on staff who can prescribe those?
Yes, for when we enter the diagnostics phase of our tests, we will be working with MDs and PAs who can prescribe medication.
> For metabolites found in very low concentrations where draw-to-draw variance is high, how do you deal with that when a person is only sending you one sample at a time?
We will have a reference cohort in place to which your measurements will be compared. Thus, even for a single measurement, we can make statements about values that are out of range. The most benefit will come from longitudinal sampling, where we can see how your blood metabolite levels move over time.
> Is there ever going to be any effort to have this payable by health insurance?
Yes, it’s on our roadmap. We’re also now working on having our tests HSA/FSA eligible.
> Are you doing anything with the feedback data? As in, someone sends you a sample, you tell them to make a behavioral change, does your advice change if future samples don't show a positive intervention effect? How do you know if patients comply with your recommendations?
We are. And you’re exactly on point in that if the interventions have 0 effects on a person, it will be omitted in subsequent reports. Though, we usually see effects on the metabolome with interventions and it becomes a matter of adjusting them. Which ties into our recommendations getting better and better for a person as they continue testing and acting on those recommendations. In terms of how we know if a person complies with recommendations, based on our database, we know of metabolic patterns that would indicate compliance,
> Are you going to offer genetic testing so, say, someone with high LDL or whatever can know if that's due to diet or they just lost the genetic lottery and nothing but statins can possibly help them?
Currently not, but one feature that we will be implementing soon would be the ability to upload genetic information and we’ll integrate that into our analysis.
> I think LDL is not a metabolite but since your "what's measured" page ends with "more published below" and then there is nothing below, I'm not sure what the full extent is of what you're testing for.
We were not sure where LDL was coming from here? Regarding our webpage, the “more published below” refers to the Nature Medicine paper, we cite further down on the page. https://www.nature.com/articles/s41591-021-01266-0
> If not LDL, presumably something you're testing for can have many sources, including genetic propensity, diet, and other environmental factors. How do you determine which of those is most causally relevant before prescribing an intervention?
That’s a good question and will be decided case by case. As an example, if your glucose levels are elevated and you have diabetes, it doesn't really matter why that happened, the consequences are the same. But indeed, there are special cases especially of genetic variants that need special attention. We will work on those and present the data accordingly in the reports.
1) You are right. We have repeatedly received this feedback now. We will make the data available for all plans, and we will add a one-kit purchase option.
2) Yes, this is the kind of report presentation we are working on and we will work with the community as much as possible to design these reports.
- We've seen that people want different levels of granularity of their metabolomic trends, and these are the test frequencies where we could accommodate for those granularity levels.
- Regarding the calls: In the beginning, we will offer guidance for every measurement, if they request it.
- For the wearables, we're including Fitbit, Garmin, Apple Watch, Whoop, and Aura, and for the diet tracking app it'll initially be myFitnessPal, and Cronometer.
- Right now we can only ship our tests inside the US, so anyone currently in and with an address in the US.
Regarding your first question on recommendations, and potential changes in established interventions (the egg example). In the earlier phases, we will mostly focus on established interventions that affect the metabolome and health and that have already been published by others, such as the DASH diet and exercise regimes. As you build your metabolomic trends over time, we'll then transition into more of our proprietary interventions.
Regarding your discussion of potential confounding factors due to changes in lifestyle parameters (swimming, sunlight etc.). That's an excellent question and important topic. For some metabolites, this does not matter. For example, if your glucose or Hba1c levels go above a certain value, you have diabetes, and it doesn't matter how it got there. For other metabolites, there might indeed be some external factors that influence the results. As you said, maybe you move somewhere cold, your metabolite levels suddenly switch, and the report says "warning". We have two answers for this: (1) For a lot of metabolites, these types of environmental factors and whether or not they play a role have been investigated in research studies and we thus know them. (2) Prior to each test, we will ask for as many lifestyle parameters as possible so we know that a certain change occurred and we can account for those in our analyses. Also over time, as we build our database, we will be able to automatically detect these changes for you and account for them (similar to Apple Watch's movement detector).
In fact, valid point on the data download. After the feedback we have received, we will change this and everyone will get their data. Regarding the basic tips, we are working on an example report so users can see what they are getting.
- https://www.nature.com/articles/nm.4222, Figure 5
- https://www.mdpi.com/2076-3271/9/2/22/htm
- https://www.ncbi.nlm.nih.gov/pmc/articles/PMC8750749/
Even though these studies have looked into metabolites in relation to chronic conditions, in the first generation of our reports, we’ll not be providing individuals with any diagnostic information and our tests right now are only intended for wellness purposes.
Regarding diagnostic predictive markers, I want to reiterate that we are not a diagnostics company at this stage and to quote from our post, “as our [longitudinal] metabolomics database grows [we will] look for new signatures of age-related diseases at earlier and earlier stages. (Such analysis will only be done on de-identified data, only with consent, and only for our work towards extending healthspan.) “.
With that being said, there are other groups that have done a great job of validating metabolite biomarkers that do provide relatively new predictive insights into chronic disease prediction and risk. One example is this paper where they looked at type 2 diabetes risk in individuals with *normal fasting glucose* (https://link.springer.com/article/10.1007/s00125-018-4599-x):
Nineteen metabolites were selected repeatedly in the training dataset for type 2 diabetes incidence classification and were found to improve type 2 diabetes risk prediction beyond conventional type 2 diabetes risk factors (AUC was 0.81 for risk factors vs 0.90 for risk factors + metabolites, p = 1.1 × 10-4).
In adjusted Cox proportional hazard models, the type 2 diabetes risk per 1 SD increase in glycine, taurine and phenylalanine was 0.65 (95% CI 0.54, 0.78), 0.73 (95% CI 0.59, 0.9) and 1.35 (95% CI 1.11, 1.65), respectively. Mendelian randomisation demonstrated a similar relationship for type 2 diabetes risk per 1 SD genetically increased glycine (OR 0.89 [95% CI 0.8, 0.99]) and phenylalanine (OR 1.6 [95% CI 1.08, 2.4]).
The same group also published on this topic before: https://www.nature.com/articles/nm.2307
Although we already measure these metabolites and others in our current panel and we are able to calculate these score, we will not be providing these score since we’re not a diagnostics company at our current stage.