Any thoughts on this paper?

[urge]

Perioperative Metoprolol and Risk of Stroke after Noncardiac Surgery
http://journals.lww.com/anesthesiol...e_Metoprolol_and_Risk_of_Stroke_after.20.aspx

Conclusions: Routine use of preoperative metoprolol, but not atenolol, is associated with stroke after noncardiac surgery, even after adjusting for comorbidities. Intraoperative metoprolol but not esmolol or labetalol, is associated with increased risk of perioperative stroke. Drugs other than metoprolol should be considered during the perioperative period if β blockade is required.

This month's Anesthesiology.

Makes you think twice.

---

Members don't see this ad.

 

[sevoflurane]

Perioperative Metoprolol and Risk of Stroke after Noncardiac Surgery
http://journals.lww.com/anesthesiol...e_Metoprolol_and_Risk_of_Stroke_after.20.aspx


This month's Anesthesiology.

Makes you think twice.

It's been a while urge, but isn't' that what the POISE trial showed?

 

[urge]

It's been a while urge, but isn't' that what the POISE trial showed?

Similar end point but different methodology. POISE got heavily criticized for using high doses of beta blockers, causing hypotension. This paper says metoprolol is the only bad beta blocker in the peri op period, not necessarily related to hypotension.

 

[Orin]

Larger studies have found no difference between strokes in folks using atenolol/metoprolol, outside of the OR. Like this one:
http://archinte.jamanetwork.com/article.aspx?articleid=1352792

Why would it be more likely to cause strokes in the periop period if it doesn't do it in the non-op period, assuming it's just based on the beta-blockade/hypotension effect?

I think their last figure kinda nails it that the big predictor of a perioperative stroke is A.fib and prior TIAs, not which beta blocker you happen to be on or which one you use intraop.

 

[sevoflurane]

Similar end point but different methodology. POISE got heavily criticized for using high doses of beta blockers, causing hypotension. This paper says metoprolol is the only bad beta blocker in the peri op period, not necessarily related to hypotension.

Kew. Thanks.

 

[Idiopathic]

not sure. the overall rate of stroke remains low (less than 1/1000) and plenty of risk factors that dont appear to be controlled/matched for are significant (BMI, renal insufifciency)

also while its concerning that of the 55 strokes, 6 got metoprolol (high incidence), 6 also good either esmolol or labetalol. i just dont know, this is the problem with regression tactics and case-matching, in my opinion - it assumes an equivalent provider response and management strategy. when you have a greater prevalence of your outcome variable (stroke), then maybe its easier to apply it to a variable such as this.

i mean, they matched some preoperative comorbidities but cant comment on length of surgery or emergent operation?

 

[deleted171991]

Just another "let's play with the database and use fancy statistics" study. Irrelevant for anything except medicolegal practice, like 90+% of what gets published.

This part is enough to make me vomit:

The incidence of perioperative stroke was 55 of 57,218 (0.09%). Preoperative metoprolol was associated with an approximately 4.2-fold increase in perioperative stroke (P < 0.001; 95% CI, 2.2–8.1). ...The use of intraoperative metoprolol was associated with a 3.3-fold increase in perioperative stroke (P = 0.003; 95% CI, 1.4–7.8)

Why did they waste time and money to study something that has an incidence of 0.09%? Who cares about the fact that a tiny number increased 3-4 times resulting in another tiny number? What kind of "confidence" should we have in a study that is based on 55 people, when there are so many factors in play in periop stroke? Why does this kind of crap even get published???

Then you have stuff like this (just reading the article at random):

The median estimated blood loss in patients who took metoprolol and had a postoperative stroke was 300 cc (interquartile range, 100–2,200 cc) compared with 125 cc (30–400 cc) in patients who took metoprolol and did not have a stroke (P = 0.18).

Seriously? How about increased blood loss is associated with increased risk of stroke? And why does one even mention something that has a p value of 0.18?

 

[urge]

Just another "let's play with the database and use fancy statistics" study. Irrelevant for anything except medicolegal practice, like 90+% of what gets published.

True. Every time you get these propensity matched groups, you have to ake it for granted. They don't release the hard data. lot of tricks could be played there.

Why did they waste time and money to study something that has an incidence of 0.09%? Who cares about the fact that a tiny number increased 3-4 times resulting in another tiny number? What kind of "confidence" should we have in a study that is based on 55 people, when there are so many factors in play in periop stroke? Why does this kind of crap even get published???

Consequences are severe. 1/1000 patients is relevant
for me.

Seriously? How about increased blood loss is associated with increased risk of stroke? And why does one even mention something that has a p value of 0.18?

Not sure you got what they meant. Anyone care to comment?

 

[deleted171991]

How can one draw valuable conclusions from a study that basically has 55 subjects out of which only 6 got metoprolol, especially if retrospective? This is bordering worthless speculation, no offense, I don't care what the math is. There are way too many variables at play to be able to draw valid conclusions in a lot of 55 (like the exact surgery they had). It's how one gets to conclusions that are proven wrong a decade later.

We should demolish this kind of science, not embrace it in a national journal. This is why academia is full of people who have nothing interesting to say/teach. This is their "science".

And of course increased blood loss will probably increase the risk of ischemic stroke. I don't need a study for that. And why does anybody bother to publish statistically irrelevant conclusions (p=0.18), if not to fill paper space? All of this mumbo-jumbo just gives ammunition to lawyers when presenting a case to a lay jury who has no idea of all the pitfalls in a study: "the patient had a stroke because the anesthesiologist gave him metoprolol".

Look at these "conclusions":

Conclusions: Routine use of preoperative metoprolol, but not atenolol, is associated with stroke after noncardiac surgery, even after adjusting for comorbidities. Intraoperative metoprolol but not esmolol or labetalol, is associated with increased risk of perioperative stroke. Drugs other than metoprolol should be considered during the perioperative period if β blockade is required.

There is no doubt, no "might", or "it seems that". Again, this is a paper approved for a national journal.

I am being impulsive here (I am neither a statistician nor a PhD/MPH), but I don't think I am wrong. It's just that the abstract reminds me of Dr. Ioannidis, the books about the improper use of (bio)statistics, and this (the last 20 seconds):

Consequences are severe. 1/1000 patients is relevant for me.

I respect that, but this kind of article leads us down the wrong path. We do so many unproven things already just because they are in the folklor, and thus "standard of care" (cricoid pressure for RSI comes to mind first). We really should stop paying attention to noise. And "Anesthesiology" should be 1/3rd as thick and 3 times more relevant.

 

[Ronin2258]

I will have to agree, I hate this data mining. There is too much in this that reads like they are trying to justify research positions, not make new knowledge.

I can only think of conclusions from data mining a this:

 

[fakin' the funk]

Consequences are severe. 1/1000 patients is relevant for me.

Yeah but is 1.5/1000 vs 1/1000 relevant? I think not. There are dozens of interventions we do with greater magnitude.

 

[Idiopathic]

so if there is an association, this is the only way to get at it. retrospective data is, by nature, dirty - matching pairs or groups helps to eliminate a lot of that dirt. so, when matching people by age and significant medical history (but leaving out a few things) they attempted to clean up the data, and showed (fairly specifically) that metoprolol use was a risk factor for perioperative stroke. yes, other factors are more likely to be important, and there is still unaccounted for variability, but, if true, this is information worth knowing, i.e. when you have other risk factors (blood loss, age, renal dysfunction, emergency case) maybe avoid that beta blocker entirely? i dont know, i happen to think these kinds of studies provide value on many levels.

 

[Idiopathic]

Yeah but is 1.5/1000 vs 1/1000 relevant? I think not. There are dozens of interventions we do with greater magnitude.

i would study and do something seemingly inconsequential for my patient to try and avoid a 1/2000 risk of stroke, wouldnt you? especially something as simple as giving labetalol over metoprolol

 

[deleted171991]

Anything is possible, but we should really stop spending money and time on things that are highly improbable. And we should definitely stop publishing almost anything without severe discernment, just because LWW needs stuff to fill the pages between the advertisements and the resume padders need more papers for their academic advancement.

Anesthesiology's response to this study should have been: please call back when you have real results. I am so sad that there isn't one real statistician, mathematician. MPH or PhD that would start dismembering these underpowered and overspeculating studies and shaming them into disappearance.

Just my impulsive 2 cents.

 

[deleted171991]

they attempted to clean up the data

I have very little respect for "cleaned up" data and its scientific significance. A lot of "noise" that people "clean up" is just signal that does not fit in their desired pattern.This is how one gets a "scientific" study that is beautifully disproved later, while harming who knows how many people in the meanwhile. We should really think "primum non nocere" in research, too. No academic advancement is worth it.

If one has too much noise in the data, one should design a study that has significantly more signal. One cannot just "amplify" the signal without knowing what it looks like; one might be amplifying just the parts one is biased for. This is why most retrospective studies are not worth altering our "standard of care", and why we worship the multicenter double-blinded randomized placebo-controlled trial god.

i dont know, i happen to think these kinds of studies provide value on many levels.

Yes, sure, they can give ideas that can be rechecked by other people, but not when you basically have 55 subjects, and the outcome that you are studying is so multifactorial.