2019 Supervised walking improves cardiorespiratory fitness, exercise tolerance, and fatigue in women with Sjögren’s syndrome; miyamoto et al

https://link.springer.com/article/10.1007/s00296-018-4213-z

interesting to see just how far graded activity / exercise is being trialled &/ or promoted for fatigue in other illness. This didnt report any problems you’d expect from an M.E graded exercise trial, no mention of PEM , adverse effects etc. One researcher I think is The man who led the (supposed) CFS sjogrens MRC study.

 

Conclusions
This supervised walking program was demonstrated to be feasible and safe with improvements...

This is a very clever way of implying that the program produced improvements without actually making the ungrounded claim. The program was demonstrated to be with improvements - absolutely true, but...

 

Anyone else feel suspicious whenever you read p = 0.049?

As usual, the authors are confusing VO2Max with VO2Peak... Similar to many ME/CFS studies, peak heart rates were below average for age, suggesting submaximal cardiovascular exertion during the test. Peak heartrate on the second test increased in both groups, but increased more in the intervention group, suggesting patients were working harder on the second CPET and thus true VO2Max was not achieved. (1st test: 150.7±19.3 (120.0; 181.0), 2nd test: 154.7±22.8 (117.0; 202.0) for the intervention group)

The fatigue questionnaire was this one:
https://www.facit.org/LiteratureRetrieve.aspx?ID=42365
(https://www.facit.org/facitorg/questionnaires)

The absolute difference between group means (on the FACIT-F) was ~5 points which has minimal clinical relevance and is also subject to a variety of uncontrolled biases (such as Hawthorne effect) due to lack of blinding...

 

Wouldn't that effect be present in petty much all studies - hard to get around it I suppose?

I have a statistics question because I don't quite understand how the authors got to their p-value of 0.016 for VO2max.

In the intervention group, it increased from 18.6 to 23 (+4.4), in the control group only from 21.4 to 22.1 (+0.7). Given the size of the standard deviation (around 10) and the range (around 30) this doesn't particularly large. There were only 22 patients in the control group. So how come the study was able to report this relatively small difference with a small sample size as statistically significant? They report that they did a Mann–Whitney test (probably because lack of normality in the data) but I thought this test didn't control for baseline values? And the end values at week 16 are pretty much the same so that couldn't have given the low p-value I suppose.

My statistics isn't that good, so It's likely that I'm overlooking something. Interested to hear if others can explain what's going on. Is it because the data was so non-normal that the descriptive numbers (mean and sd) are misleading?

 

Note that this study used supervised walking. Hence they're more likely to want to alter their behaviour, compared to unsupervised graded activity.

It's a key reason why controlling (blinded!) is desired - yes participants will alter their behaviour due to participating in a study, but they key is to make sure the bias is similar across all comparison groups.