This might be the paper (or perhaps the first of many papers) that I’ve been waiting for. Like many colleagues in the #FOAMed world I’m a big fan of point of care ultrasound (POCUS) in the ED and I increasingly find it incredibly helpful in patients with respiratory associated symptoms.
Over the years I have often found patients with respiratory symptoms challenging to diagnose and I don’t think I’m alone. For a number of years I used to ask a fairly standard question in interviews that went something like this…..
‘So, tell me about a situation where you felt that you were in a significant diagnostic dilemma, a situation where you were not sure what the diagnosis was and/or what to do about it’
It’s hardly a Google question, but it was designed to explore how candidates handled uncertainty which I believe to be a key skill in medicine (and especially so in EM/Crit Care). There are no correct answers*, it’s a question designed to get candidates to talk about their own experiences, and amazingly I would estimate that about 90% of answers involved a ‘breathless patient’.
I must also admit to my own frailties and the evidence surrounding the imperfections of the chest X-ray in diagnosing significant pathology. For our sickest patients who often get portable oblique-o-darko-grainy grams in resus the X-ray is even more challenging (no criticism of our fab radiographers, they will freely admit that portables are not as good as departmental films). Even then we know that the standard chest X-ray misses conditions like pneumothorax and is often unable to identify conditions such as pulmonary embolus.
In summary. Our current experience and the current tools available to us at the bed side are imperfect.
This month the Lancet Respiratory Medicine journal publishes a randomised controlled trial on exactly this question. Can POCUS help emergency physicians get the answer, faster, safely and more accurately if ultrasound is incorporated into the initial assessment. As always I strongly advise you read the paper (which was surprising tricky to find as it is only available as an e-pub until it goes into print at some point in the future. so my thanks to @criticalcarereviews @cliffreid and Dr Laursen for helping me find it legally! ).
What kind of study is this?
Interesting, this is a randomised controlled trial of a diagnostic tool. I like this as we have argued before at St.Emlyn’s that diagnostic tools should be evaluated carefully and in many cases that means that we should be evaluating the impact of diagnostic tests on patients. Now this study does not go as far as that, we do not have much information on patient outcome (just crude mortality), but we can use this design to find out whether clinicians achieved a different diagnosis on real patients in real time. A typical diagnostic cohort study cannot do this.
Who was studied?
As an emergency physician I am most interested in whether the patients in this study are the same as those that I would like to use it on in my practice. Looking at the inclusion and exclusion criteria it’s similar but not quite the same. Streaming seems to take place in Denmark before patients are admitted to the medical emergency department with cardiac and suspected (presume low risk) PE patients being treated as out patients. This is significant as some of the diagnostically challenging patients are apparently initially cardiac, but turn out to be respiratory and vice versa. My group of patients in Virchester are less well differentiated as I might expect a broader cohort of patients.. The inclusion criteria (in the abstract above) are reasonable and seem to identify a group of reasonably unwell patients, so in respect of severity they are similar if not the same to the patients I want to use USS on in Virchester.
As an aside I really like the idea that the patient opened their own randomisation envelope in this study. I really like this a lot and I think we should see more of it. However, the enrollment was not consecutive, only taking place when the author was present. This can influence and bias a study as many of us believe that we see different types of breathless patients at different times of day, week and month (early morning cardiac failure anyone?).
What did they do?
Eligible patients were scanned within one hour of arrival. Scans of the lungs, heart and deep veins were performed by an emergency physician – but – it was always the same physician who did this. This clearly limits the generalisability of the findings as a wider range of physicians with perhaps less expertise and enthusiasm might result in rather different levels of accuracy.
The study is pretty small as we can see from the wide confidence intervals in the results and has been conducted as part of the lead authors PhD.
What about the gold standard?
Clearly in any diagnostic study we want to know what the patient actually had as that is what we want to compare it against. In this study the authors used a 4-hour determination of the patient’s likely diagnosis and then used a discharge diagnosis to determine a true gold standard. Such a gold standard is often used in clinical trials and although it has it’s problems from a practical perspective it is achievable and reasonable though subjective. The final diagnoses were agreed by two clinicians who were did not perform the ultrasound examinations.
What are the main results?
The headline figures here are that by 4 hours 88% of the patients in the USS group had a diagnosis that proved to be correct, whereas 63% of the non-USS group had the correct diagnosis. This may be because of the USS< but also perhaps as a result of more testing following USS. Perhaps as a result of USS findings, though this is unclear. The absolute risk difference then is 25% leading to a NND (number needed to diagnose) of 4. That’s a remarkably small number! I’m always a bit sceptical about such small NNTs and there are reasons described above as to why it may be so low, but it is certainly enough to raise interest.
In terms of patient outcome (which is what we are truly interested in) then there is no statistical difference between the two groups, though interestingly more patients died in the USS group as compared to the control group. However, the numbers are so small that we should not infer anything from this apart from the fact that we need a bigger trial.
What does this mean for us in the ED?
This is further evidence that POCUS of the chest may be of benefit in the ED. However, although this is an RCT there are some significant biases within it. Whilst I personally agree with the results and in all honesty I wish them to be true, there is not evidence here to firmly change practice. For me I would like to see more studies using a broader population base, multiple USS operators and larger numbers…., although when that is done I suspect it will show something similar (but not quite as dramatic).
I wanted this paper to be the game changer that I had hoped for, but it’s not that yet. However, there is enough here to make us stop and think about the technologies we currently use and to think hard about whether USS will be the answer. As with many studies we have managed to answer some questions, only to find ourselves with many more to ask.
I’d like to learn more
There is always #FOAMed to help.
1. SonoFlowCharts – The Lung Ultrasound Flowchart – A lines & B lines #FOAMed #FOAMus http://sonospot.com/
2. The ultrasound podcast Part 1, You’re patients deserve this.
3.Ultrasound for PNX with SCANCRIT
4. The ultrasound podcast Part 2 The wait is over
In the meantime may I suggest that you get some training on chest USS in the ED as you might need it sooner rather than later.
* There are no correct answers to the interview question, but there is an incorrect one. Any candidate who stated that they were never uncertain and that they were always right in their assessments would be thanked for coming but no job offer would be forthcoming.