[aestetix]

I have two questions for you.

First, do you have any opinion on prescripting valciclovir? Apparently some ME/CFS presentations are linked to herpes, and taking valtrex significantly mitigates symptoms. However, in order for this to be effective, it needs to be taken in doses similar to truvada or viramune, not as a short term fix for a herpes outbreak.

Second, any opinion on Bhupesh Prusty's work, specifically regarding fibernectin and a potential biomarker being in bone marrow?

[lindsaywaterman]

Thanks for these - don't have an opinion on either but will look into. Broadly, I think immune dysfunction is a feature of most cfs so there is often a positive EBV test, or HSV test. I suspect this is more "downstream" than causative of CFS. However, it may be that there is a biomarker in the mitochondrial function, bone marrow etc...I hope one is found!

[smearth]

Why is the medical profession obsessed with biomarkers but not the next optimal step in patient treatment plans?

Root cause analysis of dysfunction in complex dynamic systems is a waste of energy.

Someone has the biomarker but they are a heroin addict which confers them a protective mechanism. What then?

If you have 5 heroin addicts with similar symptoms you can just experiment with likely beneficial next treatment steps. Then let machine learning track the treatment response patterns and reccomend next treatment steps.

Grouping health histories and symptons then recording responses to treatment plans then using machine learning to recommend next treatment steps is - surely better than searching for biomarkers so Drs can stick their hands up faster. Rockefeller gave us the Johns Hopkin Medical System , Can’t Bill Gates give us the complex chronic illness extension?

Why don’t Doctors have a 300 question questionaire for patients covering all aspects of the patients health? You could then group the patients into health dopplegangers. No need for biomarkers.

Combine the good part of alternative medicine ( carefully understanding the patients symptons) with machine learning.

There’s 8 billion humans. There are alot of sympton dopplegangers out there to experiment and record the effects of next treatment steps on.

Medical professionals are obsessed with being trivial pursuit champions and seem to lose sight of the fact that they’re often invited to play invisible bridge.

It is fantastic that they are great at trivial pursuit (The Johns Hopkin system is a modern marvel) and they do untold good above and beyond the measure of most citizens and far far more than myself but doing good doesn’t cure complex chronic illnesses. (I admire Doctors).

But it’s frustrating that we have to pretend that because Dr’s are great in one area (acute medicine) that the greatness translates to other areas. The halo effect is really strong.

Look up the research for recovery rates from cfs. Abysmal.

[nradov]

Machine learning doesn't actually work very well for that. Attempts to apply it as you suggest in medical research have mostly not produced clinically useful results. More basic research into root causes is the only approach that is likely to lead to truly effective treatments.

[smearth]

Could you please explain it’s limitations/constraints for this use case.

[nradov]

The data is crap. Patient reported data is notoriously unreliable. Even data reported by clinicians is highly inconsistent between organizations so it takes a huge cleansing and normalization effort to get anything usable for clinical research. Applying ML to it tends to find a bunch of spurious correlations that aren't clinically actionable.

[smearth]

Thank you.

I would use fitbits with daily step count, sleep duration and heart rate variability and breaths per minute and resting heartrate and body temperature to monitor patient progress.

I can ask the questions that would partition the useful dopplegangers. If you ask great questions you get the great answers. Then you just need to group the data intelligently.

What I can’t do is be a machine learning expert and database/sys admin and front end developer and negotiator with a public health system.

I’d let the patients vote on their next steps initially. I’d also have psych questionaires to measure their personalities. That would form part of the sympton /genetics doppleganger groupings. As would genetic profiles.

For CFS there aren’t that many logical next best steps. Most things have been tried, there is a sensible order to try things.

This is the case with most chronic illnesses.

However subsets within cfs are better off trying certain treatments before others.

The better a subset responds to a next treatment then the tighter the coupling of their doppleganger to that treatment. Then when any new people match that doppleganger you can reccomend they try that treatment. You’d shift there next best treatment orders around.

As you get more and more data the treatment recommendations get more and more granular and effective.

I could hire a naturopathic md to assign the treatment steps initially.

If anyone wants to do a start up - reach out, I can find funding. Just need a team. I’m going to do it.

I appreciate you sharing your knowledge. But I think you’re assessment of the situation is wrong. I think the general consensus is wrong and I have enough energy now to fafo. Given what I feel I know it feels immoral not to.

[InSteady]

If you gave me a 300 question health questionnaire on 15 occasions, spaced by several weeks each, you would likely get at least 3,000 unique answers.

[smearth]

If you ask great questions you get great answers.