JC: Taking the PESIT – How Common is PE Among ED Patients with Syncope?

You might remember the tremendous stink around PE prevalence caused by the outcomes of the Pulmonary Embolism in Syncope Italian Trial (PESIT)¹ from when it was published back in 2016; we covered it here at St Emlyn’s², and there are other appraisals in the FOAM community – over at REBELEM^1,3, EMLitOfNot⁴e and EMCrit⁵ to name a few. The paper was pretty controversial in EM circles, not least because it seemed to conclude with a prevalence of PE of a huge 17% in patients who had presented with syncope – far higher than fits with the experiences of Emergency Physicians the world over, particularly when talking about our typical undifferentiated syncope patient population. There were other criticisms too – read the original paper here, then the appraisals above, and make your own mind up.

Other studies have since called these findings into question – Frizell et al⁶ performed a retrospective secondary analysis of patients presenting to the ED with syncope and found a far lower prevalence at 1.4% of the 348 patients enrolled. Ryan Radecki covered this paper over at EMLitOfNote⁷.

Last month, Annals of Emergency Medicine added an article in proof, which throws even more doubt at the original PESIT findings. Have a read of the article first⁸, then read on more below.

What is this paper about?

In this paper, the authors are interested in identifying the 30-day prevalence of PE among patients who attend the Emergency Department with a presenting complaint of syncope. This wasn’t a new data collection but an analysis of prospectively-collected data as part of a larger multi centre study looking at diagnoses in patients presenting with syncope. The authors have pooled data from two different prospective studies – one in Canada, one in the US – in order to get a better idea of how common PE is as a finding among these patients.

What do we even mean by syncope?

This is a really interesting question. It’s essential to define this term in order to collect data about it, and it seems as though this was done relatively sensibly. The authors tell us that patients were screened if they had a presenting complaint of “syncope, presyncope, fainting, blackout, loss of consciousness, fall, collapse, seizure, dizziness, or lightheadedness.” It’s easy to see from the outset that these entities are not the same, but patients rarely arrive at the ED describing their symptoms using completely accurate terminology – and in any case there’s a reasonable chance that two clinicians witnessing the same event might use different descriptors.

What matters here is initially capturing those attendances, then honing the group down to include only those actually presenting with syncope. Patients with prolonged loss of consciousness (more than five minutes) were excluded, along with those who had witnessed seizures or head injuries, or whose mental status [presumably this means conscious level] did not return to baseline after the event in question.

Were the original studies conducted in the same way?

Yes and no. Both studies looked at 30-day outcomes for patients with syncope, defined as loss of consciousness lasting less than 5 minutes (we know this because patients with prolonged loss of consciousness lasting more than 5 minutes were specifically excluded).

The Canadian study enrolled any adult patient (>16 years), while the US study enrolled only those >60 years; quite different populations.

Both studies tried hard to identify the diagnosis of PE within 30 days of the ED visit. They did this through chart analysis and telephone follow-up, with adjudication in both studies for ambiguous cases (a panel of three blinded clinicians in the Canadian study, a single independent clinician in the US study). The adjudication component is important; it may sound straightforward to ascertain a diagnosis from medical notes but in my experience that is not always the case – it very much depends on who is doing the documentation!

There were differences in data collection too; D-dimer usage was not recorded as part of the US study.

What did they find?

The headline results were as follows:

9091 patients were analysed between the two studies; 547 were investigated for PE as part of their workup and 55 had PE identified, plus one additional patient who was picked up during the 30-day telephone follow-up. This gave an overall prevalence of 0.6% (95% confidence interval 0.5%-0.8%).

The results are broken down further in the paper, which gives some more food for thought.

41 patients (0.5% of total) had their PE identified in ED, 8 (0.1% of the total) had PE identified during inpatient admission, and 7 (0.1% of total) had PE identified on 30-day follow-up after index visit and were not investigated for PE during the index visit.

11/56 patients with PE also had another non-PE serious outcome (eg arrhythmia, MI, aortic dissection.

4/56 patients diagnosed with PE died within 30 days of index visit, coded in the paper as “PE leading to death” in Table 1 – it’s unclear how much overlap there is with those with concomitant non-PE diagnoses but I think these are separate patients.

There are some other interesting things we can draw from the results, too.

In the Canadian cohort, 183 patients had a negative d-dimer (of the 278 who had non-age-adjusted d-dimer performed) and 18 of those patients were further investigated anyway; 1 with VQ scan, 17 with CTPA. I’m not completely sure what we are to make of that use of d-dimer…!

The authors also did some modelling around worst-case scenarios – they looked at what might happen among the patients who weren’t investigated for PE and those lost to follow-up/with deaths from an unknown cause, estimating a maximum total diagnoses of PE of 381 patients (4.1% of total; 95% confidence interval 3.7%-4.5%).

What does all this mean?

For me, this data is pretty reassuring, at least in part. The prevalence rates among patients with syncope – even in worst case scenarios – are lower than found in the PESIT trial, and this has at least face validity in my practice.

The authors conclude that the prevalence is likely lower than that reported in PESIT, thus reducing the emphasis on syncope as a primary presenting complaint suggesting PE and necessitating that syncope patients should have PE investigations. There are more patients in this combined cohort than in PESIT, which also sits in its favour.

What the study doesn’t tell us is how other risk factors or prediction tools – like PERC or Well’s scores – are affected by this data, because risk factors weren’t collected in the original investigation. Likewise, there is no data around location of the clots; we don’t know how many of these diagnosed PEs were subsegmental and thus of limited clinical relevance – so-called lung fluff or lung lint⁹ (with credit to Casey Parker¹⁰ and Seth Trueger⁹), exactly the kind of incidental findings where risks of treatment are likely to outweigh benefits, and which the PERC score is designed to help us avoid diagnosing in the first place.

I think for now we can return to where we were before PESIT kicked up an unpleasant smell in the ED. We can flush PESIT’s uncomfortably high prevalence from our minds and consider PE in conjunction with evidence-based diagnostic aids like PERC and Well’s scoring, rather than exposing everyone who faints to a big dose of radiation or the risks of anticoagulation.

vb

Natalie May

@_nmay

References