Are there any blood tests that are particularly insightful when it coms to performance (or lack of)? I’ve been feeling a bit meh for a few weeks, and it wouldn’t surprise me if i was recovering from an illness etc. By chance my doctor has just done a load of blood tests (annual review), so I am hoping one of those might be useful.
100%!!! I run the following at the end and before the race season starts. Ideally you’d run in mid season and/or whenever you feel something is off.
Full CBC panel and add: (1) Sensitive Estradiol (LC/MS/MS), (2) Iron panel (iron, TIBC, TSAT, ferritin), and (3) Reticulocyte hemoglobin (CHr/Ret-He).
Then you can put that into ChatGPT with the prompt of all your metrics and tell it you’re trying to maximize cycling/endurance performance.
2 Likes
By chance I was having an annual health check which coincided with a meh week. I’ll work through the results and hopefully find they include some of the ones recommended above.
No
I don’t think there are any blood values that get better with training, as some sort of “performance markers”. However, training stress /overtraining might make some worse, so a blood test could catch that.
I’m skeptical. IMHO blood tests only make sense if they are interpreted by a medical professional, i. e. not ChatGPT. Things like iron deficiency should be something your GP has an eye on.
If (unlike here) the tests are handled by some lab that may or may not be accredited, I’d be extra skeptical.
1 Like
Performance is the best indicator of performance
1 Like
I agree a GP should be involved for ruling out disease and anything dangerous, that’s literally their job. What they’re not paid to do is squeeze the last 5–10% out of someone’s endurance performance.
For most primary care, “labs are in range” equals job done. For a cyclist, “in range” can still mean flat legs and crap recovery. But most primary care docs aren’t in the business of optimizing endurance performance. For them, “normal” iron or “normal” hormones is good enough. For a cyclist trying to race at a high level, ferritin 20 vs 60 or Hct 48 vs 52 can feel very different on the bike.
I use tools like ChatGPT and my own reading to understand my labs in a performance context, then take questions back to my doctor if need be. It’s not about replacing a medical professional, it’s about not limiting myself to “you’re in range, you’re fine” when I’m trying to push the edge of endurance, not just avoid illness.
2 Likes
You are right. But I’d still say you should (1) have a medical professional (with some background in sports) involved and (2) use a proper lab (to get trustworthy numbers).
For the basics like iron and vitamin D deficiency, a good GP should be able to handle that just fine. If your needs go beyond that, seek a doctor with experience in endurance sports.
But neither you or I are educated enough to interpret results. For example, in one of her short stories Sloane Crosley retold a story where she had a medical scare. Her iron levels were crazy and her GP thought she might have a congenital disease. Turns out it was a fluke, she had eaten a lot of spinach the night before.
I have to deal with ChatGPT on a regular basis (I work in higher ed and students use it to generate solutions to homework problems), and I wouldn’t trust its answers for anything remotely important. Just have a look at the training plans it creates (there are some threads here).
I get where you’re coming from about trusting medical professionals. I used to be 100% in that camp. The problem for me in the past was that, on paper, my labs were not just “normal” but excellent. Yet I was still falling apart. Exhausted, not recovering, headaches, sleep wrecked. All the feedback I kept getting was basically “you’re in range, you’re fine.” They were trying to push depression medication. That was clearly wrong for my n=1.
That’s when I stopped treating “GP says labs are normal” as the end of the discussion. Most primary care docs are good at spotting obvious disease and crisis situations; they are not set up, trained, or incentivized to optimise endurance performance or chase down why a supposedly “healthy” person feels terrible while training hard. In my case, it took digging deeper, reading, asking better questions, and pushing for more specific testing before anything actually changed.
I don’t see tools like ChatGPT as a replacement for a doctor, but I also don’t outsource all thinking to a doc anymore that spends less than 30 mins looking at me and my labs. Nobody knows me better than me, and if I don’t take charge of my health, fitness and performance than nobody will. I use AI, forums, and research to understand the numbers and generate a possible path forward. Then if needed I take that back to my physician and specialists. That’s very different from “let an ChatGPT manage MY health,” and frankly it’s the only reason I’m not still being told I’m perfectly healthy while barely functioning.
3 Likes
I’m not going to convice you and that’s not my goal. However, if your main exposure is students pasting homework in and handing in whatever comes out, you’re basically seeing worst case use. Garbage in, garbage out.
I don’t treat ChatGPT as a coach or a doctor, and I wouldn’t trust a 12 week training plan it made up from a one line prompt either. Used like that, it’s junk. Where it is useful is when you already know what you’re doing and use it as a tool, not an authority.
Give ChatGPT detailed context, very specific questions, then sanity check the answer against my own experience and other sources it becomes a great tool.
That’s no different than a library. Walk in blind, grab a random book, you might get a children’s novel when you needed a physiology text. The tool (library) isn’t the problem, how you use it is.
So yeah, ask AI for a training plan and follow it blindly is dumb. But never use it for anything remotely important is throwing out something that can be quite useful when the human on the other side actually understands the context and treats it as one input, not the gospel.
3 Likes
No, that’s not the only exposure I have. I’m a mathematical physicist working at a university. Colleagues and I are planning to use LLMs in education in positive ways and I have used it to e. g. help me find homework problems or problems on exams. Colleagues also use other machine learning methods in their research, and I am sometimes involved peripherally (e. g. as a committee member in a PhD defense committee). I have also applied ML to problems from the semiconductor industry when I did a detour in the industry.
No, it doesn’t. LLMs are very good at producing convincing-looking text on a topic, but are prone to hallucinations. In fact, it has been mathematically proven that LLMs will hallucinate.
When I asked questions related to research topics I know, it hallucinated papers and authors. (One was particularly funny, because it sort of understood that a co-author and very close friend of mine somehow was involved, but it didn’t get the first name right and referred to a paper that did not exist.)
Even with simpler things like asking it to solve homework problems (which I did try myself, too), very often its arguments are just plain wrong. But because the text “looks good” at first glance, you need expertise to separate the wheat from the chaff. For math and physics questions related to my domain of expertise, I have the necessary expertise. For most other things, I do not.
I would not advise you to use any LLM to get medical questions answered, especially questions for which you’d have trouble finding a human expert right away.
This issue could be due to a whole host of factors. E. g. your GP might not be equipped to deal with athletes or perhaps you really were in excellent health and simply overtrained. Or perhaps you had issues that could have been solved by a cycling coach. Or a combination of those. I have no idea and am not going to speculate, just saying that there could be so many reasons why it didn’t work out.
I am just cautioning against using LLMs like ChatGPT in lieu of experts. Even when asking seemingly simple questions, they will often produce convincing-looking BS.
3 Likes
Maybe a side question, have you used Perplexity? The reason I like it so far is it gives you citations to cross check.
You’re focused on “people using ChatGPT instead of experts.” I’m talking about “people using it alongside experts as a research/translation tool.”
Since you keep folding my use case back into the first category, there isn’t much point repeating myself. We clearly disagree on where the safe line is, so I’ll leave it there.
No, but other LLMs will also give you citations, real ones and fake ones. You could cross-check the answer, sure, but at a certain point you are doing research — and that requires expertise.
I’m cool with LLMs, but IMHO we still need to figure out how to best use them. E. g. a lot of people use them instead of a search engine, they ask questions and expect correct answers. In my experience, they work best when you combine your own expertise with an LLM. E. g. if you are a proficient copy editor, you can use LLMs to generate a good first draft, and your expertise flows into the request (e. g. you have to specify the tone). This way it can save you quite a bit of time. But you need to get there first.
Other colleagues use LLMs for tasks that you might have written “scripts” some years ago.
Since LLMs completely revamp education, and I am in higher ed, including in the education of those who want to become teachers, it’s really something that’s on my mind.
I understand how you are using it, you are using it to interpret blood test results and generate questions for your GP.
What I like about it - is it directly links to all the sources in-line. So while not perfect, it does speed up the research and cross checking if you were going to do it on your own anyways.
But - consider me a skeptic too. Trying not to be too much of a naysayer and use them more though, albeit critically
1 Like
As the thread drifted towards AI… I thought why not take a step back and not ask the AI to interpret the values, but moreover, ask the AI which values it would take… but take it with a grain of salt…
“if you were a cyclist and wanted to maximize your performance and could take three blood marks to guide your way along a good training what would those markers be and how should its values develop?”
As a hypothetical cyclist focused purely on performance optimization through blood markers, I would prioritize tracking markers related to endurance capacity, recovery/overtraining status, and fuel availability.
Here are the three blood markers I would choose and how their values should ideally change during effective training:
1. Hemoglobin (Hb) and Hematocrit (Hct)
Hemoglobin is the protein in red blood cells that carries oxygen from the lungs to the muscles. Hematocrit measures the proportion of red blood cells in your blood volume. These are fundamental metrics for aerobic capacity (VO2 max) [2].
-
Why: Higher Hb/Hct means greater oxygen-carrying capacity, which directly translates to improved endurance performance, especially in prolonged efforts or at altitude [2].
-
How values should develop: Values should be within a healthy physiological range, trending toward the higher end of normal as fitness improves and long as proper hydration and iron levels are maintained [2]. A significant sudden drop could indicate iron-deficiency anemia or overhydration, both detrimental to performance.
2. Ferritin
Ferritin is the protein that stores iron, essential for producing hemoglobin. Iron deficiency (even without full-blown anemia) is common among endurance athletes and can cause fatigue and impair performance [1].
-
Why: Adequate iron stores ensure the body can produce enough hemoglobin to meet the demands of high-volume training [1]. Tracking ferritin helps proactively prevent iron deficiency before it impacts performance or shows up in Hb/Hct levels.
-
How values should develop: Values should remain in an optimal range for athletes (often cited as >30-50 ng/mL, or higher depending on specific guidance). They should be stable or slightly increase during structured recovery periods and should not drop too low during intense training blocks [1].
3. Creatine Kinase (CK)
Creatine Kinase (CK) is an enzyme released into the bloodstream when muscle tissue is damaged. It serves as a general indicator of muscle stress, damage, and recovery status [3].
-
Why: While training necessarily causes some muscle damage, consistently elevated CK levels indicate insufficient recovery and potential overtraining [3]. Monitoring CK helps fine-tune the balance between training stimulus and rest.
-
How values should develop: CK levels will rise temporarily after hard workouts. The key is that these levels should return quickly to a stable, lower baseline after a few days of rest [3]. If CK remains persistently high despite recovery days, it’s a clear signal to reduce training load.
Disclaimer: These markers provide useful data points but should always be interpreted by a healthcare professional in the context of your overall health, training plan, and performance metrics.
1 Like