This was a great paper to restart our journal club with and we brought it to the table after a succesful outing at our local trainee teaching day recently. I don’t know what your place is like but at our place we all do some POCUS, some of us do ore than others and some of us aren’t sure it adds very much. I’ve mentioned it before on previous posts but there have been some heated discussions had with Simon Carley questioning whether you can claim to be an up to date EP if you aren’t practising POCUS regularly.
Colours on the mast – I believe in POCUS as a tool in ED. I think it helps me and therefore my patients in terms of diagnostic accuracy and speed of diagnosis, and occasionally a surprise finding such as a big pericardial effusion in the heart failure patient. But thats just my ‘anecdata’ isn’t it?
Is there any evidence it makes a different to patient outcomes? Like mortality? Well that’s where this paper comes in….
Clinical Question: Does the use of POCUS in your ED, with an undifferentiated shocked patient, lead to an improved chance of survival for that patient?
Title of Paper: Does point of care ultrasonography improve patient outcomes in Emergency Department patients with undifferentiated hypotension? An international randomised controlled trial from the SHoC-ED Investigators.
Published: Annals of Emergency Medicine 2018
Population: Adults with a persistent BP <100 or a shock index >1 and undifferentiated shock.
Intervention: Protocolised POCUS
Comparitor: Standard care
Outcome: 30 day survival or survival to discharge.
So who were they looking at? These were patients either in the US or South Africa at a variety of centres ranging from tertiary referral centres, to the equivalent of a DGH. All centres had a dedicated POCUS program, and the researchers organised a trip to their counterparts hospitals prior to the trial to ensure that the standards accepted at the means of accreditation were similar in the US and RSA systems. This is a good thing.
They identified patients who didn’t have a clear cause for their ‘shock’ and screened them for enrolment. But this brings us onto an important part of the trial design. This was a convenience sample, in that patients could only be enrolled if certain clinicians were available to undertake the scans. it’s not clear who those clinicians were (as in were they the investigators or were they someone else). This has the risk of bias, in that the clinician may decide they were ‘too busy’ to enrol someone if they didn’t ‘like the look of them’ for the trial. Another area that risks introducing bias is the part of the inclusion criteria that insists upon only those with ‘undifferentiated’ shock being enrolled. We all have a different level of skill on different days of ‘sniffing out’ a diagnosis, and it may be that on some days i decide lots of patients clearly have pericardial effusions clinically, or that the ‘coffee ground’ vomitus is clearly an UGIB, again this process of ‘choice’ also could theoretically introduce bias.
Allocation concealment and randomisation protocols were undertaken, by site. Each site had blocks of 50 (control and POCUS), randomly (by a distant computer software) assigned to consecutively numbered sealed envelopes, which were opaque and the same size and weight. This, however isn’t gold standard for this process. What they could have done is had the randomisation schedule held offsite entirely with the computer software allocating the patient once they had been enrolled. The risks are that a clinician could (if devious enough) steam open the next envelope and find out whats next, or could work out what the next one is likely to be if they are aware of how many POCUS vs control patients there have been. I have been warned never to underestimate the lengths some will go to, in order to discover a randomisation schedule.
Gordon Fuller, one of our fab trainees and I’m sure future Prof, kindly gave me a brief tutorial and described a system of randomly permuted blocks by site that could make the likelihood of guessing the ‘next’ allocation nigh on impossible, whch again would have been better than a 50:50 split at each site.
The scans themselves were protocolised to look at lung bases, the pericardium, IVC, aorta, the hepato-renal and spleno-renal pouches, and some pelvic views in AP and longitudinal to look for fluid and asses the bladder. I think this approach is good and probably represents the way that most of us scan in such patients, a criticism that could be levied would be that the lung views weren’t complete which may have missed some pathologies such as anterior pneumothoracies and consolidation, I am also aware some of my colleagues undertake assessment of proximal vein compression in the legs to look for evidence of DVT in such patients. I guess with this you need to decide whether their scans are the scans you do, or if you feel burdened by the missing areas on a scan.
Results: Over 4 years they enrolled 273 patients across the 6 sites, with 135 in the control arm and 138 in POCUS. We thought this represented small numbers of patients. That’s less than 12 patients per site per year. I would estimate we have that number per month, if not per week. Again this may b a function of the convenience sample and the clinicians who were undertaking scanning, and possible limits this trials external validity.
Sepsis was the most common final diagnosis, with abnormalities of IVC size and collapsibility being the most common POCUS abnormality. They found no difference in mortality at 30 days or discharge across the groups. Even when sub-analysed by continent. Some of the secondary outcomes were also interesting – volume of fluid – no difference, CT scan rates – no difference, admission rates – no difference. Which brings me on to the admission rates…..around 82% of patients were admitted, and this doesn’t ‘feel right’ to me, when i think of the patients in this trial they are sick, maybe even big sick – they have undifferentiated shock, i don’t think i send home nearly 20% of these patients. Which makes me worry about the patients that were selected to be enrolled in the trial. I have read the paper over and over and I cant put a finger on where these patients came from.
So what do we think? What do you think? No difference in mortality, time to dust off the stethoscope and the ‘victorian parlour game’ (a turn of phrase from Francis Morris) of clinical examination?
We didn’t think so. Largely because mortality isn’t the thing that a test often has a direct effect on. Its not a treatment. I want to know the diagnostic accuracy of a test. Whether the test leads to a change in management, or a procedure that wouldn’t have been done otherwise, and if it does swing the management plan significantly in another direction then that’s great, neither diagnosis may have been fatal, but gosh sometimes its useful! I love that the authors have undertaken this work, though, mortality is a big patient orientated outcome but i’m just not sure it was the right one for this test.
Let me know what your conclusions are.
Keep fighting the good fight.
Be sure to check out the review done by the bottom line on this paper, they are an awesome resource.
I’m off to watch a Welshman ride a bike around France….. Allez-G