4-Level Pulmonary Embolism Clinical Probability Score (4PEPS)

Clinical background

4PEPS was developed to bring several previously separate strategies for excluding pulmonary embolism together into a single instrument. The problem it seeks to solve is the overuse of computed tomography pulmonary angiography (CTPA): the symptoms of pulmonary embolism are non-specific, the D-dimer has low specificity, and imaging is readily available. The result is many false-positive D-dimer tests leading to CTPA that neither affects mortality nor is clinically necessary, but that carries a radiation risk and the risk of contrast nephropathy [1]. Several earlier strategies, such as the Wells score, the revised Geneva score, PERC and the YEARS algorithm, work well individually but rest on different ways of estimating clinical probability and on different D-dimer thresholds, which makes them hard to combine and increases the risk of misuse [1]. 4PEPS integrates them into a single four-level assessment with graded D-dimer thresholds.

Calculating the 4-level pulmonary embolism clinical probability score (4PEPS)

4PEPS is a weighted sum of 12 clinical variables:

$$\text{4PEPS} = P_{\text{age}} + P_{\text{chronic lung disease}} + P_{\text{HR}<80} + P_{\text{chest pain+dyspnoea}} + P_{\text{male sex}} + P_{\text{oestrogen}} + P_{\text{previous VTE}} + P_{\text{syncope}} + P_{\text{immobilisation}} + P_{\text{SpO}$$2<95%} + P{\text{calf pain/oedema}} + P_{\text{PE most likely}}

The points for each variable:

The score ranges theoretically from −4 to +22 (the negative weights sum to −4, the positive ones to +22). The highest values are not seen in practice, because some heavily weighted variables rarely occur together; in the validation cohorts the observed scores ran from −4 to +18. Some variables carry negative weights, reflecting the fact that variables strongly associated with the absence of pulmonary embolism (young age, chronic lung disease, low heart rate) lower the estimated probability.

The derivation cohort consisted of 5,588 patients (60% of a total of 11,114) from three pooled prospective emergency department databases from France, Belgium and the USA, recruited between 2003 and 2006 [1]. The mean age was 52 years (SD 18.5) and 61.8% were women. The prevalence of pulmonary embolism was 11% in the pooled cohort. Variables with more than 2% missing data were excluded, apart from those included in other established scoring systems. Variables were selected by univariate analysis followed by stepwise backward logistic regression, and points were assigned on the basis of the regression coefficients [1].

The thresholds for the four probability levels were defined a priori using a Bayesian approach and a safety threshold for a post-test probability below 2% [1]. Negative likelihood ratios for the D-dimer were estimated at 0.08 at a threshold of 1.0 μg/mL and 0.01 at an age-adjusted threshold, giving upper prevalence limits of 20% for low probability and 65% for moderate probability.

Interpretation in practice

The very low group (< 0) requires no further investigation provided the clinical assessment is correct. The low group (0–5) permits a generous D-dimer threshold of 1000 ng/mL, which is higher than the conventional threshold of 500 ng/mL and therefore yields fewer false-positive results. The moderate group (6–12) requires an age-adjusted D-dimer, which likewise reduces unnecessary imaging in older patients. The high group (≥ 13) proceeds directly to CTPA or V/Q scintigraphy, since at this probability the D-dimer cannot support a safe exclusion.

Validation and performance

In the derivation study, 4PEPS was validated in two external cohorts [1]. The first comprised 1,548 patients with a high prevalence of pulmonary embolism (21.5%) and the second 1,669 patients with a moderate prevalence (11.7%). The AUC was 0.79 (95% CI 0.76–0.82) and 0.78 (95% CI 0.74–0.81) respectively. False-negative test rates were 0.71% (95% CI 0.37–1.23) and 0.89% (95% CI 0.53–1.49), both below the safety threshold. The absolute reduction in imaging was 22% (95% CI −26 to −19) and 19% (95% CI −22 to −16) compared with the conventional strategy [1].

An independent external validation by Chiang et al. applied 4PEPS to a prospective European cohort of 734 patients with a high prevalence of pulmonary embolism (26%) [2]. The AUC was 0.85 (95% CI 0.82–0.88), and the overall false-negative rate was 1.2% (95% CI 0.59–2.23), below the ISTH-based safety threshold of 1.95% for this prevalence. 4PEPS reduced CTPA by 17 percentage points compared with the revised Geneva strategy (53% versus 70%). However, the prevalence of PE in the very low group (< 0) was 5.6% (5 of 90 patients, 95% CI 2.4–12.4), far exceeding the expected limit of below 2% [2]. Of the five false-negative patients, two had subsegmental and two segmental emboli and one a central embolus. In none of these cases had the assessor considered pulmonary embolism the most likely diagnosis, which suggests that the variable may behave differently when applied prospectively with consequences for further management than when coded retrospectively [2].

Another external validation was a post-hoc analysis of the YEARS study with 3,465 patients from Dutch emergency departments, of whom 14% had venous thromboembolism [3]. Discrimination was good, with an AUC of 0.82 (95% CI 0.80–0.84), and calibration was satisfactory. Efficiency, the proportion of patients in whom PE could be excluded without imaging, was 58% (95% CI 57–60). The false-negative rate was 1.3% (95% CI 0.86–1.9), somewhat higher than in the derivation study but within acceptable margins by ISTH criteria [3].

Limitations

The most important uncertainty concerns the very low group (< 0), where Chiang et al. found a PE prevalence of 5.6% in a high-prevalence cohort, far above the expected level of below 2% [2]. This raises questions about whether 4PEPS can safely exclude PE on clinical criteria alone in high-prevalence populations. In a related discussion, Freund et al. have argued that safety should be evaluated per probability category rather than as a pooled false-negative rate, since the latter masks risk in specific subgroups [2].

4PEPS lacks prospective outcome evidence. The derivation and all validations are retrospective analyses of previously collected data. A prospective cluster-randomised trial (SPEED&PEPS) with a planned enrolment of 2,560 patients at 20 emergency departments is being set up but has not yet published results [4].

The instrument was derived for outpatients and emergency patients with suspected pulmonary embolism. It does not apply to inpatients, patients on anticoagulation, or patients in whom pulmonary embolism has already been diagnosed. The variable "pulmonary embolism is the most likely diagnosis" is subjective and may vary between assessors, particularly in retrospective analyses where it carries no consequence for further management [2]. The instrument has not been incorporated into current major guidelines, as the 2019 ESC guidelines predate the publication of 4PEPS.

References

1. Roy PM et al. Derivation and Validation of a 4-Level Clinical Pretest Probability Score for Suspected Pulmonary Embolism to Safely Decrease Imaging Testing. JAMA Cardiol. 2021;6(6):669–677. PMID: 33656522
2. Chiang P et al. Pulmonary embolism risk stratification: external validation of the 4-level Clinical Pretest Probability Score (4PEPS). Res Pract Thromb Haemost. 2024;8(2):102348. PMID: 38444614
3. Stals MAM et al. Performance of the 4-Level Pulmonary Embolism Clinical Probability Score (4PEPS) in the diagnostic management of pulmonary embolism: An external validation study. Thromb Res. 2023;231:65–75. PMID: 37816274
4. Roy PM et al. Diagnostic Strategy for Suspected Pulmonary Embolism in Emergency Departments Based on the 4-Level Pulmonary Embolism Clinical Probability Score: Study Protocol of SPEED&PEPS Trial. Diagnostics (Basel). 2022;12(12):3101. PMID: 36553108