Does perioperative beta blockade work?

Brady AR, Gibbs JS, Greenhalgh RM, Powell JT, Sydes MR; POBBLE trial investigators. Perioperative beta-blockade (POBBLE) for patients undergoing infrarenal vascular surgery: results of a randomized double-blind controlled trial. J Vasc Surg. 2005 Apr;41(4):602-9.

Brady et. al. J Vasc Surg. 2005 Apr;41(4):602-9 didn't see the effect in a 103 patient study.

This is a negative study. They found that beta blockers had no effect. What was the chance of seeing a difference in the groups in a 103 patient study? I calculated it at 0.29. So they have a 1 in 3 chance of seeing a 50% reduction in their event rate of 34%. It is not good design to assume a 50% reduction in such trials, 20% or 30% would be more reasonable. Designing it with a power of 0.29 is not a good idea. When the study comes up negative, the result is non contributory to the argument. It is just a study without a result. It has no meaning or value to the discussion. It is a non result.

The original Mangano study had 100 patients per group (twice the size of the Brady et. al. study and used a surrogate marker with a 40% event rate.) If perioperative beta blockers reduced the risk of their combined end point 90%, the Brady et. al. study would only have a power of 0.78. Poldermans study was similar in size to Brady et. al. but he had a 30% event rate on a hard outcome (death), and a 90% reduction. He had a very high risk study group with a huge effect size. Neither of these effects are good to rely on or assume in study design.

Mangano's study is interesting for several reasons. First it was quite small (200 patients) and it followed patients out for a number of years (5 were collected). The primary outcome variable was myocardial ischemia detected by Holter. Beta blockers reduced the risk of myocardial ischemia by 50%. We were stunned to see the differences in death rates in long term follow-up. We followed them out to 4 years before publishing to make sure there wasn't some problem with randomization or postoperative medication use. How would Mangano's study look if we included the MI's? The study would look very similar because you have half as many deaths in the atenolol group. When you pile up the death certificates on a table, the pile is half as tall. It is very convincing.

Does perioperative beta blockade increase risk ?

Lindenauer PK, Pekow P, Wang K, Mamidi DK, Gutierrez B, Benjamin EM: Perioperative beta-blocker therapy and mortality after major noncardiac surgery. N Engl J Med 2005; 353: 349-61

The first result of Lindenauer is that the number needed to treat to prevent one death is 33 in the high risk group. That is exactly the same NNT as Mangano et. al. Lindenauer found results that strongly support Mangano et. al. and Poldermans et. al. A NNT of 33 is really impressive. The drugs clearly work and clearly reduce the risk of death.

Now let's discuss the lack of a demonstrated outcome in the RCRI 0 and 1 patients. When you look at patients who are on preoperative medications, it is very common for medications to be removed in the perioperative period. We reviewed 5000 patients having CABG surgery and 85% had at least one anti-ischemic agent withdrawn. There were significant side effects of drug withdrawal. What fraction of Lindenauer's patients who were "on beta blockers" actually had them withdrawn in the perioperative period? He could not tell what day patients were operated on and did not follow drug use after day 2 or on discharge. From our previous work the number of patients withdrawn from beta blockers may be as high as 50%. Why are they withdrawn? It is hard to tell in epidemiologic trials. The most common reason seems to be failure to continue preoperative medications into the postoperative period.

So how does this effect Lindenauer's result? He found that low risk patients who are admitted to the hospital on a beta blocker do worse when compared to matched patients who are not on the drug. The event rate in this group is low, but withdrawal phenomena can occur in the 50% who discontinue the drug. It may be that the risks of beta blocker withdrawal overwhelm any benefit in the RCRI 0 groups.

Poldermans and Boersma agree. "Pending the availability of data from these trials (expected within four years), we believe it is appropriate to continue beta-blocker therapy in patients at low or intermediate risk, given the potential cardiac risks associated with the sudden interruption of beta-blocker therapy," they wrote.

It is hard to develop a hypothesis that explains why a drug that is beneficial in heart attacks, heart failure, hypertension, and in high risk surgical patients would be harmful in lower risk patients unless there is some other mechanism. Like, accidental withdrawal phenomena, or severe bradycardia. We have not seen the severe bradycardia phenomena but we clearly have demonstrated the adverse effects of withdrawal of beta blockers and alpha-2 agonists. Perioperative withdrawal of these agents is a bad idea.

Poise Trial:
If the issue is really clear cut, how did the IRB approve the POISE trial involving 10,000 patients in multiple countries? I have spoken to several of the investigators in the POISE trial. I have learned a lot from doing medical research, and from trying to change physician behavior. The first thing I learned was that it is really, really hard to change physicians behavior. If you have a therapy that clearly prevents death in several studies and has been adopted by the major medical parties involved as a standard of care, it takes on average 17 years to change physicians behavior. Why does it take so long? There are different types of physicians. Some take one good study that shows an effect and change practice (these physicians are in the minority). Others look at a series of studies and discuss changing practice. Others, take a series of studies and don't believe it no matter if the NNT to prevent a death is 33. Still others don't pay attention to the literature despite what it says.

Why is POISE doing a 10,000 patient study to see if perioperative beta blockers are beneficial? Well, they were not convinced by two studies with a total population of slightly over 300 patients. AHA and ACC weren't fully convinced either. They made the guidelines very restrictive. The POISE investigators were also interested in questions we didn't consider, such as "is perioperative beta blockade safe in patients with epidurals? Mangano et. al. was prior to the use of local anesthetics in postoperative epidurals. When I spoke at the CECANZ meeting the POISE investigators were very focused on the interaction of epidurals and beta blockers, an issue we had never even considered or thought about. ¾ of the questions I got at CECANZ were on the interaction of epidurals and beta blockers. I had never had such a question in 50 U.S. lectures on the topic. The POISE investigators view the issues differently from the average U.S. physician.

Let's look at RCRI for a second.
RCRI
1. High-risk type of surgery (intraperitoneal, intrathoracic, or suprainguinal vascular procedures)
2. Ischemic heart disease (history of myocardial infarction, history of a positive exercise test, current complaint of chest pain considered to be secondary to myocardial ischemia, use of nitrate therapy, or ECG with pathologic Q waves.
3. History of congestive heart failure (history of congestive heart failure, pulmonary edema, or paroxysmal noctural dyspnea, bilateral rales or S3 gallop, or chest radiography showing pulmonary vascular redistribution.
4. History of cerebral vascular disease. (History of transient ischemic attack or stroke).
5. Insulin therapy for diabetes.
6. Preoperative serum creatinin > 2.0 mg/dl.

The event rate in the RCRI 1 group is 1.4% in Lee's original paper. An event rate for RCRI 0 it is not listed but is likely to be still lower. To see a 50% reduction in a study with a 1.4% event rate you would need 4448 patients per group or 8896 total for two. So if the POISE trial were 100% RCRI 1 patients, you could see a 50% reduction with beta blockers. The likely real effect is most likely lower like 30%. With a 30% reduction in risk, which is what is seen in RCRI 3 and 4 patients, you need 13,993 patients per group or 27,986 patients. So, POISE has a high likelihood of not seeing any benefit in RCRI 0 or 1 patients. If there are withdrawal phenomena from inadvertent withdrawal of the drug, you may see adverse events in RCRI 0 and 1 patients.