PulmCrit: Four fatal flaws in the new AHA/ACC risk-stratification system for PE

Prognostication is perhaps the most important aspect of treating PE, because it drives treatment intensity. Underestimating risk leads to inadequate treatment. Overestimating risk leads to excessive therapies, which carry their own risks.

Prognostication is difficult because PE patients exist along a continuous distribution of risk that changes over time. Nature doesn't neatly package PE patients into discrete bins. To communicate about risk and treatment, we strive to dichotomize this continuous risk distribution. Such dichotomization is inevitably flawed because our data are extremely limited (many studies examine individual risk factors, but not enough examine how these risk factors interact to synthesize risk).

ESC 2019 risk-stratification system

Let's start with the 2019 ESC (European Society of Cardiology) risk-stratification system. This is currently accepted and used by most published literature.

The ESC guidelines use the PESI (or sPESI) score to delineate a low-risk category of patients. (s)PESI is a widely validated tool to predict mortality in pulmonary embolism. But, there is a huge problem with (s)PESI that is frequently overlooked. The (s)PESI is a mortality prediction tool that doesn't account for medical interventions! Let's imagine that a patient presents to the hospital, has a cardiac arrest, is resuscitated, goes on ECMO, and eventually survives. According to the (s)PESI score that patient is a survivor! Because it fails to account for medical interventions, the (s)PESI score is weighted heavily towards comorbidities that cause unavoidable deaths (e.g., older age and cancer).

So the (s)PESI score has a blind spot in predicting which patients will deteriorate and can be saved with modern medicine (which is unfortunate, because this is precisely the group of patients we're trying to catch). To patch this blind spot, the ESC guidelines stipulate that, for the patient to be low-risk, they must also have no RV dilation and no troponin elevation (if troponin is ordered). This categorization system works adequately – patients who end up in the low-risk group are indeed low-risk.

At the other end of the spectrum, high-risk patients exhibit hypotension and shock. This group is similar to the previously designated “Massive PE.” These patients are obviously extremely ill and require aggressive resuscitation.

The intermediate-risk group contains… everyone else. This is a big, heterogeneous swath of patients. Here is where things get complicated.

Intermediate-low risk patients may have RV dilation or troponin elevation, but not both. These patients are unlikely to deteriorate. They don't require ICU admission or up-front advanced therapies (tPA, mechanical embolectomy). Fine.

Intermediate-high risk patients have both RV dilation and troponin elevation. This group is the most challenging. An intermediate-high risk designation alone isn't sufficient to indicate that a patient needs more aggressive treatment. Most intermediate-high patients have mild-moderate RV dilation (RV/LV ratio of ~1-1.5), a minimal troponin elevation, and may be doing pretty well. Such patients should be initially observed (with treatment escalation if they deteriorate). However, other intermediate-high risk patients have normotensive shock and are living uncomfortably close to the borderline between intermediate-high and high-risk. There are undoubtedly some patients in the intermediate-high group who benefit from immediate interventions such as tPA or mechanical thrombectomy (intermediate-actionable patients, as shown in the figure above). The challenge is finding them. There isn't sufficient data in the literature to parse out this group of patients clearly.

So that's the ESC risk-stratification system. It's not perfect, but it's pretty good. Low-risk patients are indeed low-risk. High-risk patients are indeed high-risk.

AHA/ACC 2026 risk-stratification system

The AHA/ACC risk-stratification is much worse. Risk group E is equivalent to the ESC's high-risk categorization, so there aren't any major differences there. The rest of this categorization system is a mess. Let's look at a few mistakes made by the AHA/ACC system.

error #1: failure to delineate low-risk PE

Sorting out low-risk versus elevated-risk PE might be the most critical decision in approaching a PE. The guidelines fail miserably at this.
The guidelines recommend that any of three different scores can be used to make this determination (PESI, S-PESI, or Bova). Recommending three different risk-stratification strategies is a formula for confusion. Different practitioners will use different scores to reach conflicting conclusions.
The ESC guidelines recognized that (s)PESI will fail to identify some acutely ill patients. Consequently, the ESC guidelines patched this hole by requiring that low-risk patients need both a low s(PESI) and also no RV dilation. This solution isn't perfect, but it gets the job done – it prevents sick patients from being miscategorized into the low-risk group. The AHA/ACC guidelines forgot about this deficit in the (s)PESI score. This will cause some patients who are actually quite ill to slip into their low-risk group (Category B), leading to under-treatment.
I like the Bova score. It's designed to predict hemodynamic deterioration or death. That's the ideal endpoint because it identifies patients who require intervention. The problem with the Bova score is that its performance isn't good enough to serve as a binary differentiator of lower vs. higher risk. The AHA/ACC guidelines recommend using a Bova score of ≦4 to define low-risk patients, but that will include some patients who actually have substantial risk of deterioration (patients with a Bova score of 3-4 points have an 18% risk of PE-related complications including death or hemodynamic deterioration).
So the bottom line is that using either the BOVA or PESI or sPESI alone as a differentiator between Category B and Category C is an absolute mess. Different practitioners may classify patients differently. Some patients with a moderately high risk will end up in Category B, leading to under-treatment.

error #2: vagueness about “biomarkers”

The AHA/ACC guidelines treat troponin and BNP as equivalent, grouping them together under the banner of “biomarkers.”
But BNP and troponin are quite different. Troponin is a more useful biomarker since it has higher specificity for a poor outcome. Troponin is more widely utilized to drive clinical decision-making.
BNP has lower specificity. If a CT scan of the RV and a troponin are already available, it's dubious whether BNP adds any independent information. Most risk-stratification tools don't utilize BNP.
By treating BNP as equivalent with troponin, the guidelines may cause patients with chronic heart failure (elevated BNP, normal troponin) to be triaged to an excessively high-risk category.

error #3: equating transient hypotension with incipient cardiopulmonary failure

The guidelines create an entire category of patients (D1) based on transient hypotension (defined as SBP <90 mm or a decrease of >40 mm Hg lasting <15 minutes or responding to IV fluids).
Transient hypotension is often an ominous sign that predicts poor outcomes. However, transient hypotension is not equivalent to incipient cardiopulmonary failure.
Lots of patients have chronically low SBP (e.g., chronic heart failure, cirrhosis, young women). For a patient with chronic systolic hypotension, a single SBP reading <90 mm may be insignificant.
Automated BP cuffs often produce inaccurate readings (e.g., if the patient is lying on their arm). If an automated BP cuff creates a single incorrect reading that is forwarded into the electronic medical record, this could derail the patient's entire risk-stratification and management plan.
Overall: transient hypotension should be taken seriously within the appropriate context (e.g., tachycardia, diaphoresis, dizziness, RV dilation). However, simply equating SBP<90 with “incipient cardiopulmonary failure” is incorrect.

error #4: nonsensical definition of normotensive shock

The guidelines define D2 as normotensive shock. That's a great concept. Unfortunately, they fail to define this correctly.
The guidelines define normotensive shock as: “Any: Lactate >2 mmol/L, acute kidney injury, urine output <0.5 mL/kg/hr, mental status change, cardiac index <2.2 L/min/m2, mean arterial pressure <60 mm Hg, increased shock score/stage (SCAI stage, CPES score).”
To start with the obvious: A MAP <60 mm indicates hypotension. Defining normotensive shock based on the presence of hypotension makes no sense.
The CPES (composite pulmonary embolism shock) score is a risk-stratification tool defined to predict which patients with normotensive PE may have low cardiac index. It's incorrect to equate a risk-stratification tool with a hemodynamic state.

conclusions

The ESC 2019 categorization system for PE is imperfect. Specifically, it fails to identify which patients in the intermediate-high group might benefit from immediate therapies. However, current evidence isn't sufficient to define this group of patients clearly, so this failure isn't really their fault.
The AHA/ACC categorization system for PE is dangerously flawed. If anyone did actually try to put this risk-stratification system into practice, it would likely cause substantial confusion and harm.
Hospitals and clinicians should continue using their current PE algorithms (rather than updating them based on the new guidelines).

Image credits: Heading photo by Loic Leray on Unsplash