High Five for High PEITHO? Intermediate-high risk PE @ St Emlyns

Pulmonary Embolism is awful. You don’t need me to tell you. Just read the news.

However, the evidence-based fight to improve care continues and the Greek goddess rages on, with publication of PEITHO, PEITHO 2 and with PEITHO 3 in the wings for completion by June 2026. In addition, we now have a short interlude for HI PEITHO, published last month in the New England Journal of Medicine. These studies showcase the wonderful legacy of Guy Meyer, who has really got under the hood of acute PE with right ventricular strain and dedicated his life’s work to the study of risk informed therapy. Hats off. You can read more about Guy here.

HI PEITHO is the first large prospective randomised controlled trial of ultrasound facilitated, catheter directed fibrinolysis combined with anticoagulation in higher risk PE, compared to anticoagulation alone. You can read the abstract and the full paper in the NEJM, and as always, we would strongly recommend that you do. Critical Care reviews have also done a nice critical appraisal already, as has PulmCrit.

But having read it, here are some of our thoughts

Background

Most PE is eminently treatable with therapeutic dose anticoagulation, with an evidence base moving more and more towards total DOAC therapy, and apixaban in particular. However, some patients will be sick and may require more than anticoagulation alone. In patients with shock or significant haemodynamic compromise, national guidelines and current consensus support the use of systemic fibrinolysis to reduce early mortality. In patients without shock, but with evidence of right ventricular dysfunction, previous trials have shown that while thrombolysis may reduce the rate of cardiorespiratory deterioration, it unfortunately increases the risk of major haemorrhage and stroke.

Catheter directed fibrinolysis, supplemented with high frequency low power ultrasound could be just the right temperature porridge for these higher risk patients. This therapy facilitates lower doses of fibrinolysis and shorter infusion durations delivered direct to the thrombus site, with sonographic potentiation of fibrinolytic effects. It is costly however, dependent on expert operators, site specific and in dire need of supporting evidence to justify the logistical challenges of delivery.

Tell me about the design

HI PEITHO was an adaptive design open label superiority randomised trial with blinded outcome assessment, conducted across 59 sites in the USA and Europe. The study was sponsored by industry (Boston Scientific) but led scientifically by an academic institution and the PERT consortium. The PICO design was as follows:

Patients – Adults with acute PE confirmed on CTPA AND with an RV to LV ration of 1.0 or higher AND abnormal cardiac troponin levels AND evidence of cardiorespiratory distress (SBP <110, tachycardia >100 or tachypnoea >20 at any point within a 6 hour window preceding randomisation.

Intervention – Ultrasound facilitated catheter directed fibrinolysis with alteplase using the EkoSonic endovascular system plus anticoagulation. Dosing is described in the supplementary appendix and was dependent on thrombus burden, but appears to have amounted to 9mg total for unilateral disease and 18mg for bilateral disease.

Comparator – parenteral anticoagulation using LMWH or UFH as per standard guidelines.

Outcome – The primary outcome was a composite of PE related death, cardiorespiratory collapse or decompensation, non fatal symptomatic recurrence of PE within 7 days of randomisation. Safety outcomes included major bleeding as per ISTH definitions. Other secondary outcomes included change in RV:LV ration, all cause mortality and functional score.

Pretty comprehensive and logical on the whole. Dosing using the Ekos system was approximately 10-20% of the dose given with systemic therapy. Parenteral anticoagulation as a standardised control seems sensible, although the allowance for UFH may impact generalisabilty. A composite outcome was to be expected, but the inclusion of cardiorespiratory collapse was an interesting one. This was defined as one of; cardiac arrest; indication for CPR; signs of shock; placement on ECMO, endotracheal intubation, non invasive ventilation; or a NEWS score that rose to or persisted at 9 or higher. I’m not sure I hold these equivalent in my mind, so was interested to see how they broke down in the results. Symptomatic recurrence is also a woolly one, as only a very small proportion of PE events are likely to run into problems after starting therapeutic dose anticoagulation, usually as a result of an undiagnosed thrombophilia. As such, hard to imagine how one time CDT could be helpful in reducing these events. Finally, exclusion criteria of ‘persistent haemodynamic instability’ felt a bit subjective.

Any statistical nuggets?

The sample size calcs and analysis plan were interesting – the authors selected a 10% absolute risk reduction in the primary outcome as the proposed benefit with the intervention, which is within the realms of a moderate treatment effect. They estimated a sample size of 406 patients initially to identify this difference with 90% power, but also planned an interim analysis. When you look at the data this early, some would suggest that you are ‘spending’ some of the alpha (p value) early. As such, the authors selected a p value of 0.025 as the initial alpha, such that a future (second look) at the data would give still give them an estimated value of 0.05. This is tricky to get your hear round, but is good practice – if you are planning to look at the data early, you need to adapt your sample size to maintain your alpha and control of type 1 error. Following the interim analysis, the authors expanded their sample size. More on alpha spending here.

Randomisation was stratified by age and RV:LV ratio, which is prudent given the disparity in treatment effect within PEITHO. An Intention to treat analysis was used to ensure pragmatic application of results and preservation of balance from randomisation.

Who were the participants?

The authors screened 4313 patients over 4 years to obtain a sample of 544 patients in total. A 12.6% conversion rate from screening to recruitment is hard work, so the investigators clearly put the yards in. Most screen failures were through failure to meet inclusion/exclusion criteria but it is worth noting that approx. 480 patients did not consent or gave another/unknown reason not to participate. This leans towards selection bias a little, as I suppose an invitation to a complex intervention trial like this is all about how you phrase the offer.

Two patients were excluded due to not receiving the allocated intervention, leaving 542 patients total in the Intention to treat analysis. Baseline characteristics are reasonably balanced in the main, although it is worth noting that participants in the control arm had more chest pain, prolonged hypotension, tachypnoea, bilateral PEs and more history of congestive heart failure and TIA/Stroke. There was more cancer in the intervention group.

And the results?

In the ITT population, a primary-outcome event occurred in 11 patients (4.0%) in the intervention group and in 28 patients (10.3%) in the control group (relative risk, 0.39; 95% confidence interval [CI], 0.20 to 0.77; P=0.005). A significant result, and one that implies a large effect size (>50% relative risk reduction). The devil is in the detail though – what events actually made up this proportional reduction?

Interestingly, the efficacy benefit was overwhelmingly seen through a reduction in the rate of cardiorespiratory decompensation or collapse. There were actually more PE related deaths in the intervention group and equal recurrence between groups. In terms of what drove the rates of cardiorespiratory decompensation, this is in the supplement, but transparently presented and almost exclusively driven by difference in the event of NEWS >/= 9 (0.7% and 7% for the intervention and control groups, respectively) on two measurement, taken at least 15 minutes apart, and between 24h and 7 days from randomisation.

Bleeding events were higher in the intervention group, but did not reach statistical significance. There were no episodes of intracranial haemorrhage up to 30 days and only one ischaemic stroke in the intervention group. Death from any cause was higher in the intervention group at 7 and 30 days (RR 1.5 and 1.7, respectively), but with wide confidence intervals crossing 1 and therefore not statistically significant.

Any limitations to highlight?

The additional criteria of high heart rate, high respiratory rate and low blood pressure at inclusion are a bit of a curveball and will affect generalisability of these results. As such, the findings should not be extrapolated to patients not meeting such criteria, although there is a clear risk that they could be, as the definition of intermediate-high risk is confusing enough already to some. The authors laud the intervention being delivered within 2h of randomisation (and randomisation within 6h of diagnosis) but I worry that this again impacts generalisability – this window is rarely achievable outside direct presentation and exclusive management at a specialist centre. My hospital refers locally to our IR team via webpage, for MDT discussion the next day for example – does this mean these results do not apply to my patients? There are no long term follow up data available as yet, although the authors note this work is ongoing.

The open label nature of the intervention warrants discussion. I think the authors selected a reasonable design, as a sham procedure would convey significant risk and complexity. However, it does affect measurement of outcomes. Blinding the adjudication committee is all well and good when it comes to CTPA reports, and death related to PE, but if your key difference is in high NEWS scores, and your nurses/investigators know which intervention people have had, and have discretion over when to record the NEWS score, then I think you can’t get away from the potential for outcome bias.

The authors note the baseline imbalances and also highlight that only 10% of the patients were 75 or older, to their credit.

It’s not a limitation as such, but when you look at the exclusion criteria in the supplement, you will see they are very extensive. In particular, all patients who had received a single daily dose of LMWH, or a DOAC within 12h prior to potential randomisation, were excluded. This is presumably to reduce potential for bleeding complications by ensuring all patients use a BD regimen of LMWH during the intervention phase. However, this is a rarely used regimen in the UK and will significantly impact the generalisability of the study. NICE guidelines recommend treating prior to scan in the majority of patients. I can’t remember a patient with an intermediate-high risk PE I have seen recently, who has not had a full dose of LMWH prior to even considering CDT.

Finally, I can’t see any data on the distribution of contributing centres. 59 is a lot of sites, but if some only conducted 1-2 intervention procedures and others several hundred, there is a risk of systematic bias through cohorting skilled operators and experience.

Where does this all leave us?

First, congratulations to the authors and thank you to the participants. This is much needed data and must have been an absolutely monumental undertaking. Complex intervention trials are really difficult, and to deliver these results in a 7 year cycle is hugely impressive.

The authors conclude that the intervention in this trial led to a low risk of the composite of pulmonary embolism related death, cardiopulmonary decompensation or collapse, or symptomatic recurrence of PE within 7 days, when compared to anticoagulation alone. I think we can probably look at this data and be a little more nuanced, to say that the intervention essentially reduced the risk of having a high NEWS score over the next 7 days, or needing rescue intervention at a later stage (the majority of which was CDT in this study). The open label design of the study and limitations above raise some issues with internal validity, but the main issues are clearly with generalisability. I personally think it would be very hard to write any national guidelines based on these results mandating early referral for CDT.

More likely, I think this trial refines our scope for the intervention, provides further reassurance on bleeding risk and reminds us that PE patients with evidence of RV strain have a dynamic risk over the 7 days from diagnosis. Catheter directed therapy may well be a suitable option for some of these patients heading in the wrong direction. However, in a UK setting, I think we will remain confined to our current practice of discussion with a regional specialist and individualised decision making at the discretion of the operator. I personally don’t see anything in this paper that will make me advocate any harder than I do already for consideration of IR, and I think the cost of increasing availability would be enormous (not to mention tapered to training and appointment programmes). When NICE guidelines are updated I think we should push for a cost effectiveness and health economic analysis here, but I think we can infer from these results what it is likely to say – that CDT costs a lot, but current evidence suggests early QOL outcomes are not really impacted in a major way.

Final words.

All very interesting. There is a good editorial that’s worth a look as well. And some of you may have attended the RCEM conference this week to hear my talk on PE. If you did, and you wanted to look at the references in more detail, (as promised) they are listed below in section order.

Cheers

Dan

References:

Rationalising investigation and use of the PERC rule

Singh et al. EMJ 2013;30:701-706

Freund et al. JAMA 2018;319(6):559-566

nice.org.uk/guidance/ng158

choosingwiselycanada.org/recommendation/emergency-medicine/

Please Dimer Carefully

Stals et al. Ann Intern Med 2022; 175(2): 244-245

De Wit et al. Ann Emerg Med 2023; 81(5): 558-565

Roussel et al. Lancet Resp Med 2026; 14(1): 29-37