Wednesday 12 November 2014

The bias of systematic reviews and the neuraminidase inhibitors

6.9 billion dollars were spent around the world in 2009 stockpiling oseltamivir to combat pandemic influeza. So this drug must have magical powers. There’s little doubt it was magic for Roche. But does it actually benefit patients? Maybe we should look for the answer in systematic reviews of the neuraminidase inhibitors. These reviews and meta-analysis are meant to be about the highest form of evidence we have. But could they be subject to substantial bias? Could authors have financial conflict of interest that might sway their conclusions? Unfortunately the answer is an emphatic yes.

The author from the University of NSW Sydney found 26 systematic reviews about the neuraminidase inhibitors. Of these 13 examined prophylaxis and 24 examined treatment accounting for 37 distinct assessments. Two blinded reviewers looked at each of these to determine if the review was positive or negative. They then looked to see if there was a financial conflict of interest.

I think we can all predict the results but I was genuinely surprised at the extent of the results. Of the 8 reviews determined to have a conflict of interest, 7 (88%) were determined to be favourable. Of those without conflict of interest only 5 out of 29 (17%) were in favour. In addition, those without conflicts were much more likely to mention problems with publication bias, the quality of the studies and missing data.

So there is little doubt that systematic reviews can have substantial bias. Many assumptions go in to the selection of included studies, the grading of the evidence and all of this can be easily corrupted. Perhaps it is time we ditch the notion of systematic reviews being high quality evidence- especially if there are any conflicts of interest. Maybe we should bury them under the pile of 6.9 billion dollars worth of oseltamivir rotting in the ground.

Dunn AG, Arachi D, Hudgins J, et al. Financial Conflicts of Interest and Conclusions About Neuraminidase Inhibitors for Influenza. Ann Intern Med. 2014;161:513-518.

Tuesday 11 November 2014

The wonderful ARISE trial

Journal club in 2014 would certainly not be complete without a mention of the fantastic ARISE trial. I think most practicing emergency doctors now are well aware of this study and how it concluded no benefit of early goal directed therapy (EGDT) vs. usual care in early septic shock. Therefore I will only focus on a few key points.

This was a huge undertaking involving 51 centres mostly in Australia and New Zealand and enrolled 1600 patients. The authors are to be absolutely congratulated for such great effort. However, I’ll bet they are kicking themselves for not being the first cab off the rank of the new large RCT’s. ProCESS just beat them by six months. The last of these trials is the ProMISe trial coming out of the UK. Therefore we will be seeing lots more data and meta-analysis coming in the future.

In the ARISE trial, both groups had a mortality just shy of 20%. This seems to be the new norm regarding mortality rates in studies of sepsis now days. Much has been made about the gradual reduction in mortality over the years but I would imagine that much of this is probably due to early recognition of milder sepsis and the creation of a lower risk cohort rather than any major breakthrough. (Although some might consider this a breakthrough.)

Both arms in the ARISE trial got just about the same amount of fluids. Two and a half litres were given before enrollment and almost two more litres in the six hours after. (To be fair, there actually was a statistical difference with a p value containing lots of pretty zeros but this most certainly was clinically meaningless). More patients in the EGDT group got red-cell transfusions, central lines, vasopressors, dobutamine and went to ICU. So many more resources were used to no benefit. Could this be thought of harm of EGDT? I think so.

So once again the original study of EGDT which claimed a 15% absolute mortality reduction has been debunked. ARISE has taught us again to be skeptical of small studies showing large treatment effect. Where else have we seen this recently? Think hypothermia after cardiac arrest? Could NINDS be next?

The basic summary for the early care of septic shock in the ED? Early recognition, aggressive fluids, and early antibiotics. Who would have figured?


Peake SL, Delaney A, Bailey M, et al. Goal-Directed Resuscitation for Patients with Early Septic Shock. N Engl J Med 2014 Oct 16;371(16):1496-506.

Monday 10 November 2014

Should traumatic pneumothorax get a gentle pigtail or a massive painful chest tube?

Okay, so I think you have already figured out the answer so I’ll keep this brief.

There is quite a philosophic difference in how many clinicians manage spontaneous versus traumatic pneumothorax. Spontaneous usually gets nothing or a simple aspiration whereas traumatic often are managed with chest tubes. I have been teaching ATLS (EMST) for years and the current 9th edition still advocates a massive chest tube. Why such discord?

To be honest, the following paper is rather bad but at least addresses this important issue. So please ignore the small sample size, absolutely crazy power calculation (that must have been done in retrospect), lack of blinding and confounding factors.

40 adult patients with traumatic pneumothorax were randomized to either a pigtail catheter or a 28 French chest tube. They excluded anyone with haemothorax and those who needed emergent tubes.

The primary outcome was looking at pain on a numerical rating scale (NRS) at the insertion site of either the chest tube or pigtail. Secondary outcomes included success rate and complications.

The results confirmed the obvious. Pigtails hurt a lot less. Yes, a lot less. This was statistically significant even with such a small sample size. The duration of tube insertion and success rate were all similar. Safety can never be adequately addressed with the size of this study but one would imagine it should be fine.

Will this change what we do? The quality of this evidence shouldn't change anything. But it has caused me to look at recent studies (that I must have overlooked) advocating a smaller solution to uncomplicated traumatic pneumothorax. It may not sound kosher, but I’m going the pigtail from now on.


Kulvatunyuou N, Erickson L, Vijayasekaran, A, et al. Randomized clinical trial of pigtail catheter versus chest tube in injured patients with uncomplicated traumatic pneumothorax. Br J Surg 2014;101:17-22.

Do an ultrasound first for suspected kidney stones… if you’ve got the skills

You are seeing a patient in the ED with suspected renal colic. Should you do a point-of-care ultrasound first, a radiology ultrasound or just go straight to CT

Why not just CT everyone? After all, it is the gold standard test. Yes, but the authors correctly point out the radiation issues, cost and get kudos for mentioning morbidity from over-diagnosis and incidental findings. All of this coming from the lead author who is a San Francisco radiologist! (I wonder if she has had her keys to the radiology tea room revoked.)

Inclusion criteria were patients between 18-76 years old where the ER doctor decided to order imaging to establish or rule out a primary diagnosis of kidney stones.

Excluded patients included obesity, pregnancy, those at “high risk” for serious alternative diagnosis such as cholecystitis, appendicitis, AAA, bowel disorders and a few other things.

Patients were randomized in a 1 to 1 to 1 fashion to the first study being a point-of-care ultrasound, radiology ultrasound, or CT. After this was done, the ER doctor could do order whatever tests they felt appropriate.

There were three primary outcomes that all seem relevant but somewhat hard to measure.

The first was did the initial treatment strategy cause any harms? This was defined as high risk diagnosis with complications that could be related to a missed or delay in diagnosis. Three separate non-blinded reviewers determined this outcome and came to a consensus.

The other two outcomes were total radiation exposure and cost.

Results? 2759 patients were enrolled across 15 different ED’s. Most of these ED’s had a very active ultrasound program.40% of the patients had a prior diagnosis of kidney stones. Of the total, only 11 patients (0.4%) had a high risk diagnosis with complications in the first 30 days. There were no differences between the groups.

Of note, 60% of patients randomized to the point-of-care ultrasound never got a CT in the ED.

Not wanting to paraphrase an issue of such importance, the authors state the following:

“Our results do not suggest that patients should undergo only ultrasound imaging, but rather that ultrasonography should be used as the initial diagnostic imaging test…”

The obvious limitation to this study is external validity. I always seem to chuckle a little when I see a study of ultrasound that quotes test performance to three decimal points. The fact is that the sensitivity and specificity of the test will change each and every time a different person picks up the probe. This study clearly used doctors experienced in ultrasound. In addition remember they were also very good at excluding high risk patients on clinical grounds. Therefore caution is in order.

Will this change practice? As usual, the answer is it depends. It really matters who is holding the probe and making the clinical decisions. Nevertheless, this study gives strong support to those who might perform a point-of-care ultrasound without having to resort to CT if it was not felt to be indicated.

All of this from a radiologist…it must be a cold day in hell.


Smith-Bindman R, Aubin C, Bailitz J, et al. Ultrasound versus Computed Tomography for Suspected Nephrolithiasis. N Engl J Med 2014;371:1100-10.

Friday 7 November 2014

Can ethanol be neuroprotective?

Should we load head injured patients with a neuroprotective bolus of ethanol? Oh please let this be true!

Honestly, some papers are so good for journal club simply because they are so incredibly bad. The following is a great example. One might wonder if the peer reviewers were the head injured participants of this study.

These authors performed an un-blinded chart review with no methods to determine the physiologic values and outcomes of a cohort of 184 severely injured patients as defined by an ISS of >17. (Why 17 you might ask?)  Within this “matched” cohort, they compared those who had a blood alcohol level (BAL) greater than 0.05 g/dl (n=34) to those that were lower or had measurable alcohol (n=150).

Both groups had similar ISS. Those in the higher alcohol group had lower GCS scores (9.64 vs. 12) upon arrival. There was a trend towards higher values in the abbreviated injury score (AIS) for the head (3.29 vs. 2.81). But after two hours of admission the GCS scores “leveled out.”

Their conclusions are so great that I dare not paraphrase. “Severely injured patients with a raised BAL have higher incidence of severe traumatic brain injury...However the survival rate and in-hospital stay is not influenced. This supports the theory of a neuroprotective role of alcohol.”

Before you start hitting the bottle, please consider the following.

If someone is drunk, might they look to have a worse head injury? Yes you say? Then they would likely score higher on the head component of the AIS. Without getting too nerdy, this value is squared to become part of the ISS. Therefore almost half of the ISS in the drunk cohort came from their over appreciated “head injury.”

I think it is quite obvious that the drunk cohort was ultimately less injured but had the same outcomes. This would suggest a harmful effect of alcohol rather than a protective effect.

In the end, this is really a small and sloppy study that can’t tell us much of anything important. But its ultimate value may be for teaching critical appraisal of the literature. Cheers... and have one for me so long as I am not a trauma patient!

Covering: Scheyerer MJ, Dutschler J, Billeter A, et al. Effect of elevated serum alcohol level on the outcome of severely injured patients. Emerg Med J 2014;31:813-817