Wednesday 30 November 2016

Intensive blood pressure lowering in patients with acute cerebral hemorrhage; don’t just do something... stand there!

I love me a good negative study. This is especially one that is well conducted in a high impact journal. The results are much more likely to be believed.

These authors of the ATACH-2 trial randomized patients with intraparenchymal hematoma to intensive systolic BP lowering (110 to 140 mm Hg) vs. standard treatment (140-180 mm Hg). They excluded patients with very low GCS and massive hematomas.

They mostly used nicardipine for BP reduction and this had to be started within 4.5 hours of onset of symptoms. The primary outcome was death or disability (mRS 4-6) at 3 months after randomization.


Patient enrolment was stopped because of futility at the 1000 mark.  The primary outcome was observed in 39% of the intensive-treatment group vs. 38% in standard treatment. As you would imagine, there were more adverse events with intensive treatment.

Of course there were some limitations. Externally validity might be an issue with the drugs used and patient demographics. Treatment didn’t start very early. We could argue over the primary outcome. Most patients weren’t that sick (85% had a GCS of 12-15) so perhaps we can’t extrapolate this to the sicker cohort etc.

Nevertheless, the results look pretty convincing.

They even conducted an ordinal shift analysis. These are all the rage in the stroke thrombolysis literature as they have a high degree of statistical efficiency to find small differences. But even this showed nothing.

So what are we to conclude?

Targeting a systolic BP between 140 to 180 is just fine. As with many things in medicine, we shouldn’t mess around too much. Don’t just do something... stand there!


Therapeutic hypothermia for in-hospital cardiac arrest? Who knows...?

There is some evidence to suggest that therapeutic hypothermia may (or may not) be efficacious for out-of-hospital cardiac arrest (OHCA). But what about cardiac arrest that happens in the hospital (IHCA)? She we bust out the esky and ice blocks?

These authors sought to answer this question by looking at a large registry of 26,183 patients with IHCA to see who did and didn’t get cooled. The primary outcome was survival to hospital discharge. To be clear, this was an observational study and patients were not randomized.

As you can imagine, there were quite substantial differences amongst the group of patients who were offered cooling and those that did not. So it’s a bit like comparing apples to oranges before we even start with therapeutic hypothermia.  

So how do we try to correct for all of this confounding? Throw in some fancy propensity matching!

After adjustment the therapeutic hypothermia was associated with lower in-hospital survival (27.4% vs. 29.2%). Cooling was also worse for most secondary outcomes.

So therapeutic hypothermia is harmful for IHCA?

Answer: we don’t know.

With all of the inherent limitations of this study design, the results are pretty irrelevant. Registry data is notoriously unreliable. Data is often missing, incomplete or just plain wrong. If you start with rotten eggs, you can’t make a very good Pavlova.

In addition, there is no amount of fancy statistics that can reliably account for all of the confounding. Lots of assumptions go into these “adjustments” and we can only correct for what we know and measure accurately. As Donald Rumsfeld said, “...there are also unknown unknowns.”

What are all the variables that make a patient sicker? Have we measured them accurately and precisely? There’s no way...

In the end, this study suggests there is clinical equipoise and it would be ethical to conduct a randomized clinical trial. Certainly this might give us a better answer... but I wouldn’t hold my breath.

For now, we can save our ice for something that really matters... margaritas anyone?


Steroids for refractory shock following cardiac arrest; such promising methods but such poor results

Should we give steroids to patients with OHCA and refractory shock? This is a decent research question. 

If one wants to know the truth, then the double-blind placebo controlled RCT is the absolute pinnacle of research.

The overall methods of this paper are great. Of course no papers are perfect and there were some issues but...

There was a fatal flaw.

It was only powered to include 50 patients! You have got to be kidding me.

It does not take a masters in biostatistics to know that such small numbers will only ever find huge changes in treatment effect. The likelihood of a type II error (false negative study) is massive.

(Just in case you were wondering, they found no difference if patients were given steroids or not)

This was going to be a negative study before they ever started. What were they thinking...? Sometimes I wish I had a time machine so I could go back and shake these people.  

Perhaps I shouldn’t be so critical. Feasibility of such a study is a major problem. These patients are not common and recruitment would have been very challenging. But most researchers anticipate this and choose other study designs and/or make other efforts & compromises. Why conduct research that you know will be flawed and a waste of time?

What should we conclude?

Absolutely nothing. Grumphfpfff...

But there is a good lesson to researchers and to those learning critical appraisal of the literature. Do your homework before you start. Choose a research question and study design that is feasible and has the potential for success.


One in Six Patients Hospitalized with Syncope has a PE? Danger... Danger...

The danger is taking this paper at face value! Never has such an important paper in the NEJM required the reader to exercise such caution with interpretation.

These authors looked to find the prevalence of PE among syncope patients who were admitted to hospital from 11 Emergency Departments in Italy. Che bello!

They performed a systematic workup to investigate PE in all of these 560 patients. Patients with low risk by Wells Score were excluded with a negative D-dimer. All others got mostly CTPA’s and a few VQ scans.

The prevalence of PE was 17.3%!

Holy merda! We need to do a lot more CTPA’s in patients with syncope!

Danger! Danger!

Remember, these were elderly patients with syncope admitted to hospital. If we include all patients with syncope (i.e. discharged from the ED) in the denominator the prevalence drops to 3.75%. But we’re still not done...

Of these admitted elderly patients, 40% had clinical signs of DVT, 20% had active cancer, and 10% had prolonged immobility. Is there any surprise that many of these patients might have a PE?

To be fair, there were 31 patients diagnosed with large PE’s that were admitted to hospital with an alternative diagnosis. This is about 5% of admitted patients or 1% of all patients that presented to the ED. Although this percentage is small, it does emphasize that we should continue to think, keep an open mind and avoid diagnosis momentum. But isn’t this what we should always be doing?

In addition, CTPA is not perfect. How many false positives did they have? And how many small PE’s were simple incidental findings that had absolutely nothing to do with their syncope? What is the number needed to kill by finding incidental “PE’s” and putting elderly patients on anticoagulation? What about other incidental findings, over-diagnosis, and contrast induced nephropathy?

Unfortunately none of the above issues were significantly highlighted in this paper. The authors, peer reviewers and journal editors should be ashamed. Fortunately, the NEJM is usually much better than this.

So what should we do?

Well, we should NOT perform indiscriminate CTPA’s on all patients with syncope. This will cause more harm than good. We should investigate PE if we think we should... keep an open mind; especially if they have risk factors and/or clinical signs & symptoms of DVT/PE.

But most of all, we should stai attento to the dangerous conclusions of this paper!


Prandoni P, Lensing A, Prins M, Sambuca Ilove, et al. Prevalence of Pulmonary Embolism among Patients Hospitalized for Syncope. N Engl J Med 2016;375:1524-31.