The Kettlebell Physio

07/06/2026

Another week, another GPT for you. This one was a real b280 in the s7104/4 though!
I’ll explain.

Those are ICF codes for “pain in the neck” — remember the International Classification of Functioning, Disability and Health?

If your training was anything like mine, the ICF was presented as a model to be applied to every clinical case…
…and then was never seen or heard of again after graduation.

Perhaps your experience was different. Anyway, I recently came across a special issue led by Prof Todd Davenport (published in 2013), which made a case for physical therapists using the ICF when reporting case reports.

Before that, Todd and colleagues had published two worked examples using the ICF in case reports: DOIs: 10.2522/ptj.20080113 | 10.2519/jospt.2010.3264

Evidently, neither clinician-authored case reports nor routine use of the ICF have been widely adopted. I trust CLEAR will be encouraging that horse to take off, rather than flogging a dead one.

For anyone unfamiliar with the ICF, it is a World Health Organization framework for describing functioning, disability, and health; a shared language for clinical care and interdisciplinary communication, encouraging a BPS, person-centred approach.

Routine in principle, but not routine in practice. More commonly used in rehab settings, but even then the literature suggests its use has been limited (PMID: 24986707).

Still, I figured it might be a useful addition to the CLEAR project, for those keen to use it, and possibly for your case notes right now, to describe the person, not just a pathology. Given the coding system, I can see potential value for future reporting too.

Davenport presents two formats; a table and a figure. The TL;DR is this: ChatGPT is currently not “good enough” to recreate the figure cleanly without a lot of manual formatting pain, so, after many many hours of hair pulling, I shelved that idea.

This GPT now focuses on recreating the table. You enter the case details step by step (without identifiable patient information), and it produces an ICF-informed case profile in either a copy-editable table format or as an image, plus a corresponding case description.
Give it a go. Feedback welcome!

07/06/2026

To read the full protocol for CLEAR, follow the links.

07/06/2026

07/06/2026

Yes ✅
Yes ✅
Yes ✅
Yes ✅
Yes ✅
Yes ✅

We want to hear from you! Follow the link.

05/06/2026

Would you ask ChatGPT about vaccines?

Recent unpublished Australian research found that 1 in 5 adults reported disregarding their GP’s advice in favour of advice from ChatGPT.

That should make us pause.

AI is already shaping health decisions. At the same time, vaccine confidence remains fragile.

Australia’s HPV vaccine programme has been a world-leading success, contributing to a 90% reduction in HPV prevalence among vaccine-eligible people since the programme began. But recent Australian CDC reporting also shows that adolescent HPV vaccination coverage has fallen — from 85.7% in 2020 to 79.5% in 2024.

Many people have genuine questions about vaccines: safety, side effects, ingredients, timing, long-term effects, children, pregnancy, and trust.

But for those who feel unsure or resistant, getting reliable information can feel confusing, overwhelming, or judgemental.

That is why we created Ask About Vaccines.
It is a calm, non-judgemental GPT for people with vaccine questions, worries, doubts, or mixed feelings.

It does not lecture, shame, or pressure.
It listens first.
It helps people think through their concerns.
And when they want evidence, it points them to reliable sources.

Clear answers.
Respectful conversation.
No judgement.

Try it here: https://chatgpt.com/g/g-6a22485f63a88191911cbf8021ddebc1-ask-about-vaccines

30/05/2026

I recently saw a post on IG about this 2021 systematic review on ACL healing. The caption read:

“A 2021 systematic review of complete ACL tears found that every study reported healing — none found failure to heal.”

quoting the review’s conclusion: “The belief that the ACL cannot restore its fibre continuity might be outdated.”

It sounds compelling, but it's also easy to misread. The key problem is that “every study reported healing” does not mean “every ACL healed.” It means that every included study reported some cases where ACL fibre continuity appeared to return on MRI. That is very different from saying complete ACL tears reliably heal without surgery.

Several studies in the review also reported non-healing, incomplete healing, persistent laxity, delayed reconstruction, or re-rupture. A more accurate summary might be:

A 2021 systematic review found that MRI evidence of restored ACL fibre continuity was reported in some patients across all included studies, but healing rates varied widely and the evidence was low quality.

There is another important distinction; the review used “healing” to mean MRI fibre continuity, not full clinical recovery. That matters because a ligament that appears continuous on MRI is not automatically strong, stable, symptom-free, or ready for sport. MRI continuity does not necessarily mean the ACL is functioning normally.

For example, Van Meer et al. reported that ACL fibre continuity improved in 60% of patients, but Lachman test improvement occurred in only 32%, pivot shift improvement in 8%, and KT-1000 laxity improvement in 2%. In other words, the MRI could look better while functional stability remained limited.

The studies in the review were also highly varied. They included case series, retrospective studies, prospective cohorts, one questionable RCT, different bracing protocols, different rehabilitation approaches, variable follow-up periods, and different definitions of MRI “healing”.

They also included selected populations, not all ACL ruptures. For example,

- Fujimoto included patients with low athletic demands and a continuous ACL on MRI.
- Ahn selected patients with only mild instability.
- Jacobi excluded injuries older than three weeks and several associated injuries requiring surgery.
- Costa-Paz was a retrospective series of patients who had already shown spontaneous healing, so it cannot estimate the probability of healing.
- Razi included a large denominator, but only a small number healed.

These findings should not be overgeneralised to people with displaced tears, high-grade pivot shift, bucket-handle meniscal tears, multiligament injury, chronic ACL deficiency, or high-demand pivoting sport goals.

The phrase “none found failure to heal” is particularly misleading. The review meant that no study found zero cases of restored MRI continuity. It did not mean:

no patients failed non-operative care
no patients failed to regain fibre continuity
no patients had persistent instability
no patients required later ACL reconstruction
no patients re-injured

All of those occurred. Fujimoto reported 8/31 patients requiring ACL reconstruction due to instability. Jacobi reported 18/86 ACL-Jack patients requiring secondary reconstruction within 24 months. Costa-Paz reported two re-ruptures after apparent healing.

Bracing is another issue. The review included very different approaches: Kyuro brace, Nakamura extension-block brace, an unspecified brace allowing 20° flexion, ACL-Jack dynamic posterior drawer brace, simple immobilisation, bracing only for collateral ligament injury, and even no bracing or non-specific rehabilitation.

None of those is the same as saying “bracing heals ACLs”. The details probably matter: knee angle, duration, posterior drawer force, range limits, early weight-bearing, tissue coaptation, associated injuries, clinical supervision, and patient compliance.

So yes, this review challenges the old claim that a complete ACL rupture can never restore fibre continuity, but it does not prove that most complete ACL ruptures heal. It does not prove that bracing reliably causes healing. And it does not prove that MRI continuity equals full functional recovery.

For someone with a recent ACL rupture, they should have answers to these questions, to make an informed choice about what to do:

- Is the tear proximal, midsubstance, distal, displaced, or well coapted?
- Are there associated meniscal, chondral, MCL, PCL, or posterolateral injuries?
- Is there a high-grade pivot shift?
- Does the knee give way?
- Is the person able to comply with a specific bracing or rehabilitation protocol?
- Does follow-up MRI show continuity and good tissue morphology?
- Do clinical stability and functional testing support progression?
- Is the goal straight-line activity, recreational sport, or high-demand pivoting sport?

The conclusion is not that “complete ACL tears heal”, it is that some ACLs may restore MRI fibre continuity under certain conditions. The evidence is interesting, but still uncertain. Patients deserve a careful, individualised discussion about tear morphology, associated injuries, instability, goals, risks, and the limits of what MRI healing actually means.

When we share evidence online, it’s worth remembering that patients may be reading it while anxious, injured, and looking for direction. So the question isn’t just whether the statement is technically true, but whether it could be misunderstood in a way that shapes a real decision about care.

https://doi.org/10.1016/j.jbmt.2021.06.003

28/05/2026

The first three As in the “5 As of EBP” are:

ASK — the right clinical question
ACCESS — the available evidence
APPRAISE — the validity of that evidence

Great in theory. Harder in practice. I’m sure plenty of clinicians occasionally open PubMed or PEDro and enter a few search terms. But how many are routinely building PICO-based search strategies with Boolean operators? Too hard basket.

Accessing evidence is easier now. Tools like Litmaps, Consensus, Research Rabbit, Connected Papers — and good old social media — make it easier to find papers. The challenge is that the evidence we stumble across is often only one piece of a much larger puzzle.

Clinical decisions should be based on all the available evidence, where possible, not one convenient paper.

Then comes appraisal. If clinicians are routinely comparing papers against CASP checklists, I’ll take up running. And I really hate running. Again: too hard basket.

I recently shared a GPT to get over the ‘ASK’ hurdle - now I’m sharing one that does the APPRAISAL part. I think this will be a winner!

CLEAR – Evidence Search Builder
Helps turn a clinical question into a PubMed/PEDro search strategy.

CLEAR – Evidence Appraisal
Uses CASP checklists to help appraise research evidence, especially RCTs, systematic reviews, and meta-analyses.

These tools don’t replace clinical reasoning, but they may make the first steps of EBP a little more doable!

Please note: the same traffic-light rules apply when uploading PDFs.

28/05/2026

The first three As in the “5 As of EBP” are:

ASK — the right clinical question
ACCESS — the available evidence that answers the question
APPRAISE — the validity of the evidence

This is my opinion: great in theory, but not likely happening in practice.

I’m sure some people in community-based allied health open PubMed or PEDro once in a while and enter a few related search terms. But I’d be very surprised if anyone is constructing PICO-based search strategies with Boolean operators to find evidence. Too hard basket.

To liken EBP to baking a cake: if we switch out the first and most crucial ingredient, or leave it out entirely, that cake is not coming out anything like the recipe.

Accessing evidence is easier. Tools like Litmaps, Consensus, Research Rabbit, Connected Papers, and others make this step more achievable. Engaged and interested people also share research online all the time. For the interested clinician, finding evidence on almost any topic is now pretty quick and easy. Sometimes, we can even go straight to the source — the author. Yay for social media.

The challenge, in my opinion, is that the pieces of evidence we find are often only a few pieces from a much larger picture. Clinical decisions are supposed to be based on all of the available evidence, where possible, not one convenient piece of it. That would be highly problematic.

Finally: appraise. I’m confident enough to suggest that if there are clinicians out there routinely comparing papers against a CASP checklist, I’ll take up running (I f'ing hate running). Again, too hard basket.

If these steps are the foundation of EBP in practice, what might this suggest about how effectively EBP is actually being applied?

The CLEAR project seeks to address some of these barriers.

I recently shared a couple of GPTs designed to help. The first was the Evidence Search Builder, which supports the “ASK” step by helping construct a search strategy from a clinical question.
Back in 2023, I shared the CASP GPT.

Today, I’m sharing a much-improved version - CLEAR - Evidence Appraisal. It has been built with MSK-focused allied health clinicians in mind. It has access to the CASP checklists and can produce a clinically relevant appraisal of different study designs — ideally RCTs, systematic reviews, and meta-analyses - but it will appraise any for which there is a current checklist.

If you have a piece of research evidence and want to better understand its validity, and how useful it might be for the patient sitting in front of you looking for help, attach the PDF and let it do the work.

Neil

Please note: the same traffic-light rules apply.

https://chatgpt.com/g/g-6a178958e9a88191925ac33c3c90a9ce-clear-evidence-appraisal

27/05/2026

This is a graph view in Obsidian, linking concepts identified from seven recent systematic reviews and meta-analyses on exercise and bone health in post-menopausal women with osteoporosis. All the good stuff sits behind the pretty picture.

It was created in OpenAI’s Codex using a 'Skill' (these replace 'prompts') developed by Ilya Shabanov as part of The Effortless Academic. The same process can also be done in Claude.

I saw this just yesterday and downloaded the guide: https://x.com/Artifexx/status/2059062104888398102?s=20

The power of it is mind-boggling. If you're keen to get across lots of research evidence, I highly recommend it. Gone are the days of having to sift through individual papers, try to remember everything you've ever read on a topic, and try to work out how the heck everything is connected!

If you have a computer, you can do this too.

___

This was its response to a question from Greg Lehman (on another older post on another page) about findings from the Souza paper:

My read is that Souza’s “other factors” are mostly muscle-bone crosstalk factors, not non-loading in the sense of “bone adapts without stress.”

Other factors besides mechanical loading:
Muscle-derived biochemical signals
Souza explicitly points to myokines and muscle-derived factors as possible mediators. They name myostatin inhibition, IGF-1, irisin, IL-6, IL-7, IL-15, follistatin, TGF-beta, FGF-2, osteonectin, osteoglycin, and others as candidates influencing bone metabolism. The idea is that resistance training may affect osteoblast/osteoclast activity partly through endocrine/paracrine signaling from contracting muscle.

Muscle hypertrophy and muscle-bone interaction
Souza frames higher muscle mass as associated with lower prevalence of low BMD. So RT may help bone partly by increasing muscle size and force capacity over time, which then changes habitual skeletal loading, even if the training load itself is not very high.

Muscle contraction itself
They suggest that contraction-mediated effects may matter independently of absolute external load. This is why a low-load protocol could plausibly still matter if it creates enough muscular work and fatigue.

Baseline BMD status
In the vault analysis, Souza is linked to the finding that baseline BMD status influenced femoral neck BMD responses to resistance training: resistance training.md (line 16). Their interpretation was that people with lower initial BMD may be more responsive regardless of load.

Intervention duration and progression
Souza also found intervention duration influenced femoral neck responses: resistance training.md (line 17). Zhao 2025 adds a compatible point: resistance training significantly improved LS, FN, and TH BMD, while longer duration and frequency mattered for some sites: resistance training.md (line 18).

Why effort matters:
Effort is probably a proxy for the internal training stimulus, not just for “trying hard.” A 40% 1RM set taken close to failure can recruit more motor units, produce more cumulative muscle contraction, generate more metabolic stress, and potentially stimulate more muscle-derived signaling than an easy 40% 1RM set stopped far from fatigue.

So the key distinction is:
External load: the weight on the machine/bar, e.g. 40% vs 80% 1RM.
Internal effort: how close the working muscle gets to failure or high recruitment.
Biological stimulus: the combination of muscle tension, repeated contractions, fatigue, myokine release, hypertrophy signaling, and eventual progression.

Souza’s result that high- and low-load RT did not differ for femoral neck or lumbar spine BMD is captured in the vault here: low-load resistance training.md (line 15) and low-load resistance training.md (line 16). But they are not saying load is irrelevant. They are saying absolute load may not be the only route to an osteogenic stimulus.
One wrinkle: low-load training near failure can be uncomfortable, and Souza suggests that discomfort may prevent participants from actually reaching high effort or progressing well. So effort matters twice: biologically, because it may drive recruitment and muscle-bone signaling; practically, because if the protocol is too uncomfortable, people may not sustain the effort needed for adaptation.