Causal overstatements in modern physical activity research, 2024, Skarpsno

Midnattsol

Midnattsol

Moderator
Staff member
The challenge of causation in physical activity research
Although advancements such as access to large datasets with device-measured physical behaviour, and advances in statistics, have improved our understanding of the associations between physical activity (PA) and health outcomes, PA research often contains causal overstatements. The line between correlational and causal PA research is narrow, and confounding and reverse causation may lead to false conclusions. We contend that data must be able to answer a causal question before implications for ’24-hour’ PA guidelines and interventions are considered.

https://doi.org/10.1136/bjsports-2023-108031
 
Midnattsol said:
The challenge of causation in physical activity research
Although advancements such as access to large datasets with device-measured physical behaviour, and advances in statistics, have improved our understanding of the associations between physical activity (PA) and health outcomes, PA research often contains causal overstatements. The line between correlational and causal PA research is narrow, and confounding and reverse causation may lead to false conclusions. We contend that data must be able to answer a causal question before implications for ’24-hour’ PA guidelines and interventions are considered.

https://doi.org/10.1136/bjsports-2023-108031
Click to expand...
Well we can't read because paywall but I'd put a high note on the whole "PACE showed that 1 in 7 report some minor statistical improvement on a fatigue questionnaire after a year-long exercise training programs" means that exercise cures CFS. It basically agues that 1/7 of something like 10% improvement, so at best 0.01% = 100%. It's beyond absurd, we are literally 4 orders of magnitude off here, and the benefits are qualitative and never matched by objective benefits. And PACE was the best they could do, it even hides the massive harm their model causes.

It probably takes the top rank of widely misleading overstatements in history, certainly top 10. Like someone saying they can easily lift 10,000 kg but then only do it with 10 kg and call it a success. Even worse is the audience who sees it happening and cheers wildly.
 
A recent review [...] assessed how reallocating time between behaviours influenced different health outcomes. [...] the overall conclusion was that moderate-to vigorous PA, at the expense of light PA, sedentary behaviour or sleep, was associated with better health outcomes (eg, adiposity, biomarkers, mental health and chronic disease). However, since most of the included studies were cross-sectional, the authors expressed the obvious need for prospective and experimental study designs.
Click to expand...

Interestingly, the findings from the above-mentioned review coincide with a recent cross-sectional study published in a leading cardiology journal. This latter paper, which is undeniably well written and with comprehensive analyses based on data from five countries, concludes that there exist a clear hierarchy of behaviours and that moderate-to-vigorous PA demonstrated the strongest, most time-efficient protective associations with cardiometabolic outcomes. The study also implies an unfavourable association when sleep replaced any time spent active, for instance, standing. While the causal interpretation and consequently the causal language in the above-mentioned cross-sectional studies seems appealing, it is problematic because the impact of reverse causation is inevitable. The fact that the outcomes of interest are responsible for the variation in PA rather than the other way around is likely to lead to misinterpretation of the observed associations. It is impossible to estimate the causal effects of reallocating time between physical behaviours from studies of this nature.
Click to expand...

Larger data and more precise measurement of PA reduce measurement error but does not remove problems of confounding and reverse causation. Indeed, the causal structure of the data may become more complex when we have a myriad of opportunities to define PA behaviour. Dealing with ’24-hour’ observational data is challenging from a causal inference perspective and necessitates careful consideration of the joint effects of physical behaviours on health outcomes. We should, therefore, establish a robust causal framework on how we can estimate causal effects in the context of compositional 24-hour PA data. We should also triangulate results across other causal approaches (eg, instrumental variables analysis (with and without genetic instruments), within sibling comparisons, negative control) and conduct randomised experiments when feasible (eg, short-term effects of actually replacing time in different behaviours).
Click to expand...

It is time to stress the need to take the causal question seriously in order to create better empirical evidence that truly can support ‘24-hours’ PA guidelines. To achieve this, authors should call descriptive studies by their names, and journal editors should discourage the use of misleading causal language. This will lead to better scientific reporting and to a more successful encounter for the reader. Cross-sectional studies can provide useful descriptions of the distributions of physical behaviour in the population but should not provide basis for 24- hour recommendations. Because of methodological limitations, it is not possible for all research questions to be causally motivated, irrespective of study design, sample size, objective measures or biological plausibility of the association of interest.
Click to expand...
 
It is time to stress the need to take the causal question seriously...

I would go further: Determining causal relationships and pathways is the whole point of science, its sole purpose. Its success rises or falls entirely on how effectively it does that. Descriptive and correlation studies are useful only insofar as they contribute to that singular goal.

All else claiming to be science is just fluff and grift.
 
Sean said:
It is time to stress the need to take the causal question seriously...

I would go further: Determining causal relationships and pathways is the whole point of science, its sole purpose. Its success rises or falls entirely on how effectively it does that. Descriptive and correlation studies are useful only insofar as they contribute to that singular goal.

All else claiming to be science is just fluff and grift.
Click to expand...
This. A principle of scientific models (or theories) is that they have to be predictive, if not of all things, at least of one thing, and then very accurately. Models that can't predict accurately aren't very useful, while models that can't predict anything are functionally useless. They may have some intellectual value, but in the real world they aren't worth a damn.

Evidence-based pragmatic models can't ever predict anything, they are not scientific models, have no theory behind them. All the biopsychosocial/psychosomatic models can't predict a damn thing all put together, not one bit of useful prediction. That makes them useless, it's simply wrong to apply them in real life as anything but a "heh, maybe it could work for you, probably coincidentally, but I wouldn't personally spend $5 on this if I were you".

Which would be fine if they stuck to this, but as know, even though pragmatic trials aren't supposed to infer any causality, the entire premise of not just ME/CFS, but of all psychosomatic rehabilitation in general, is built entirely out of "those trials kinda show that it can be of help to some, therefore they are considered 100% safe and effective and that means that they are psychological disorders". This is the wrong part. I harp on it a lot, but even PACE's best results is some trivial 10% benefit for 1 in 7, and out of that the entire profession took away that it's 100% a cure and that it means that "chronic fatigue" is psychological/behavioral. This is the insane part, they know this is wrong and do it anyway, all because of permissive language that makes them get away with it.
 
You must log in or register to reply here.
Back
Top Bottom