Monday, August 8, 2011

Statistical vs. Clinical Significance: They Are Not the Same


MedPage Today featured an article about the beneficial effects of daytime wearing of compression stockings on obstructive sleep apnea. The premise was that increased edema in the neck could be caused by fluid coming from the legs when patients were in the supine position at night. Twelve patients who served as their own controls wore compression stockings for a week and then no stockings for a week alternating. The stockings lowered the amount of fluid in the neck by 60%, a statistically significant difference. So far, so good.

This resulted in another highly statistically significant finding, which was a 36% reduction in episodes of apnea [cessation of breathing] and hypopnea [inadequate breathing]. Sounds good, right? The problem is that the average number of episodes of apnea/hypopnea decreased from 48 per hour to 31 per hour. Patients experiencing more than 30 episodes of apnea/hypopnea per hour are classified as having severe obstructive sleep apnea. This means that the treatment only put the patients in the low range of severe obstructive sleep apnea. They still would require maximum therapy. Is a reduction in apnea/hypopnea episodes that does not move the patient out of the severe category really clinically significant? It does not seem so to me.

Although the MedPage piece did not address that issue, it did mention other limitations of the study including that study was small and brief, it included only otherwise healthy, non-obese patients and it did not look at whether the stockings had any impact of daytime sleepiness.

Another possible problem which I am familiar with concerning patients with varicose veins is that wearing compression stockings is uncomfortable, especially in the summer. They often do not fit well and are not particularly stylish. All of those factors lead to non-compliance.

Bottom line: Although there was a statistically significant improvement in the number of apnea/hypopnea episodes when compression stockings were worn, it does not appear that there was a clinically significant improvement in the course of obstructive sleep apnea.

24 comments:

m. s. said...

"Is a reduction in apnea/hypopnea episodes that does not move the patient out of the severe category really clinically significant? It does not seem so to me."

First of all, you're talking averages, while you should look at distributions. If the average has moved nearer to the 30 limit, it means that you can have more and more cases *below* 30 (even if always less than 50% of them, assuming that we talk of a Gaussian). So you can have what you call significant clinical improvement in a sizable number of cases. To know how many of them, I'd need to know the distribution, but I'd be surprised that it is have a sigma so narrow that you have none.

Second, I don't know much about sleep apnea, but how is something that brings you from 48 to 31 episodes not significant? Does it mean that someone with 31 episodes per hour is dramatically worse than someone with 29, but only slightly better than someone with 48? I would suppose it is a continuum -the fact that it doesn't fit an arbitrary definition limit doesn't mean it is worthless.

Skeptical Scalpel said...

Thanks for the comment. You make some good points.

You are correct about the average being lowered and some therefore are below 31. It's only 12 subjects. They are not typical OSA patients [not obese, no co-morbidities]. I too am not an expert on OSA, but I doubt that even a reduction to 25 A/H episodes per hour would change the treatment.

I never said it was worthless. Perhaps this will lead to some more dramatic results. I was simply using this study to illustrate the point that just because something is statistically significant, it is not necessarily clinically significant.

devicerandom said...

(same person as above post)

"I doubt that even a reduction to 25 A/H episodes per hour would change the treatment."

Yep, but does it mean that they aren't better?

"I never said it was worthless. [...] it is not necessarily clinically significant."

Then I do not know the distinction between "worthless" and "not clinically significant" :)

Eric said...

It clearly wouldn't work as a sole therapy, but it might reduce the CPAP pressure the patient would need, making it more comfortable and increasing compliance.

Ólafur Pálsson said...

Then someone else finds some magic treatment to reduce episodes by a flat 5 per hour, and that too would be clinically insignificant, even more than this article, but if both methods are used together you would see the biggest part of the curve move under the magic 30/hr (assuming Gaussian).

Anonymous said...

An AHI (apnea/hypopnea index - abnormal breathing events per hour) less than 30 is still moderately severe sleep apnea and is associated with significant morbidity and mortality without nasal-CPAP therapy ( The Lancet V365 p 1046). Small study - insignificant change. You are right...you know nothing about sleep apnea.

KonfusedKris said...

In epidemiology, even a 2% effect (p<0.05) size of a factor (e.g. effect of exercise on cognitive function in the elderly) would not be ignored because at a population level this 2% could be many real people. So although I do agree that reporting and discussing effect size, as well as significance are crucial, I would argue that lack of "clinical significance" doesn't negate the importance of a statisitically significant change.

Alethea said...

Meanwhile at the cellular level we find things that are barely statistically significant, but can be quite clinically significant. A very slight reduction in ion channel function (such as by a drug) can have extremely severe consequences on neuronal firing. So while statistical significance shouldn't be kicked to the curb, one does have to think about real life, too!

Han van der Heide said...

Yep, this seems like something worth looking into. Also because it is a relatively 'easy' way to help your apneu, patients can decide whether this works for them. I mean people with auto-immune diseases are willing to ingest worm eggs, what harm will compression stockings do compared to that?

Anonymous said...

Maybe something even more pressing regarding the study is the fact that only 12 patients took part. Is this anywhere near the number needed to have sufficient statistical power?

The very, very small number of patients in the study makes the statistical significance of the results questionable, let alone any clinical significance.

Skeptical Scalpel said...

Power is much more of a factor in studies that show no difference in outcomes. If the sample size is too small, a study showing no difference may simply be underpowered. This is also called a
Type II error." That is, assuming there is no difference when there really is a difference.

Anonymous said...

I think I understand... If I'm bitten by (for example) a rare and very poisonous snake, and the doctor says "If I don't treat you, you'll slip into a coma within an hour and die within 4 hours, but if I do treat you, you'll slip into a coma within an hour and die within 8 hours" then I would say that the treatment is statistically very significant (because my survival time is literally doubled) yet clinically insignificant (because either way, I'm in a coma in an hour, and don't live another full day.)

If sleep apnea is a slow killer, as other comments suggest (as opposed to the fast killer in the example above) then a treatment (wearing compression stockings) in which you are killed twice as slowly is statistically significant but clinically insignificant -- especially when other treatments (CPAP) are effective to eliminate the deleterious effects of severe apnea, or even severe apnea with compression stockings.

medzpro said...

most complain from my patient when wearing compression stockings is very-very uncomfortable, especially in Indonesia...

Skeptical Scalpel said...

Thanks for the comment. I agree. Compliance with wearing compressions stockings is poor, especially in hot weather.

Ellie K said...

Gosh, such a rough crowd you field here! As a statistician and former employee of a state department of health services, supporting non-infectious disease epidemiology, though never sleep apnea, I believe that a sample size of 12 is too small, almost a waste of effort. It isn't conclusive. Better to wait for funding for a larger study, and do it right.

I wear surgical compression stockings, 30-40mm/Hg. I live in Phoenix. It is hot. But I had my first pair fitted well, found a brand I liked, and listened to the fitting person. I don't mind wearing compression stockings because they make my ankles slender and the pain in my thighs go away (without surgical intervention). Toe-less thigh highs were recommended, which I find less restrictive. However, I have a strong incentive for compliance: Attractive legs, no more aching, even burning pain.

Initially, I giggled when I read this post, as it reminded me of a bump under a rug i.e. that compression stockings on one's legs would have a beneficial effect on the circulation in one's neck. Then I realized that systemic effects were plausible. They would need to be truly, observably significant, not merely statistically significant, to convince otherwise healthy, normal weight people to regularly wear compression hose, I suspect.

Skeptical Scalpel said...

Ellie K, you understand the importance of complying, unlike many patients who are no so committed.

klikharry said...

everything in your hospital is too crowded, same as here

info kesehatan said...

thanks for information..

Unknown said...

This is a huge problem ln internal medicine, poorly understood by
most. In pulmonary medicine there are always new inhalers
that result in. improvement in expiratory flow rates tht fail to make
a difference in the patient 's. ability to function. They do
cost more .though, good for the drug. companies.

Skeptical Scalpel said...

Unknown, thanks for commenting. I think this is true for many drugs. For example, are the newest PPIs really any better than the old ones?

RD said...

Dr Scalpel: This topic touches on a major flaw in statistical practice and in empirical methods. A lot of the use of statistical significance is a degraded philistine affair, applied mechanically and unreflectively to satisfy editors and referees. Computer software have gravely exacerbated this problem.

A whole monograph has been written about the abuse of statistical significance in economics. The following link is to the web site of one of the coauthors:
http://www.deirdremccloskey.com/articles/stats/preface_ziliak.php

Skeptical Scalpel said...

RD, thank you for the pertinent comments. You make a good point that software has made it easier for people to calculate p values.

Thanks also for the link. It was an interesting read.

Anonymous said...

Re: sleep apnea - I wish you did all know more about it. Each apneic event contributes to brain damage; any reduction in the AHI in my opinion is a good reduction, but best of all is all the way down to zero. I had never heard of compression hose doing anything for OSA, and will ask my own sleep medicine doc about it. But if they help - heck, I'll start wearing them.

Skeptical Scalpel said...

Anon, thanks for commenting. Do you have a link to a reference about your belief that any reduction in the AHI is beneficial for reducing brain damage? As I recall, the authors of the article did not mention that.

Post a Comment