Safety of the pulmonary embolism rule‐out criteria rule: Findings from the Registro Informatizado de la Enfermedad TromboEmbolica venosa (RIETE) registry

The diagnostic strategy for pulmonary embolism (PE) includes a D‐dimer test when PE probability is low or intermediate, but false‐positive D‐dimer results are frequent and can result in an unnecessary computed tomography pulmonary angiogram. The PE rule‐out criteria (PERC) rule excludes PE without D‐dimer testing when pretest probability is <15%. The aim of this study was to assess the safety of the PERC rule strategy in patients included in the Registro Informatizado de la Enfermedad TromboEmbolica venosa (RIETE) registry.


INTRODUC TI ON
Pulmonary embolism (PE) is a common and potentially fatal disease.
The validated diagnostic strategy to identify PE in hemodynamically stable patients combines the assessment of the pretest probability of PE derived from the revised Geneva score, Wells' criteria, [1][2][3][4] or from the unstructured clinician's gestalt. 1,5 When PE probability is low or intermediate, D-dimers are measured. 1 If they are above the predefined cutoff, pulmonary vascular imaging, mainly pulmonary computed tomography angiography (CTPA), rules in or out PE.
However, D-dimer results are false positive in >50% of cases, 6 resulting in unnecessary CTPAs. Fear of missing PE, combined with the ubiquitous availability of D-dimers and CTPA, has led to more frequent testing, thus leading to an increasing number of CTPAs performed over the past decades 7 but a decreasing diagnostic yield. 8 To reduce the proportion of false-positive D-dimer results, the positivity threshold has been increased, based on older age or a lower clinical probability. [9][10][11] In a more radical approach, the PE ruleout criteria (PERC) rule, an eight-item set of clinical criteria (Table 1), proposes to rule out PE without D-dimer testing when no criteria are met (PERC-negative [PERC-N]) in patients with a pretest PE probability of <15%. 12,13 The risk-benefit ratio of further testing is unfavorable as PE probability is lower than the test threshold for PE, estimated at 1.8%. 12,14,15 The safety of a PERC rule diagnostic strategy is critical. In 2012, a meta-analysis based on 12 studies including 14,844 patients from six countries supported PERC rule use when pretest PE probability was low (pooled specificity and negative likelihood ratios of 97% and 0.17, respectively). 13 Nevertheless, its external validity and transportability have been questioned as six of the 12 studies, which represented 74% of all patients, were authored by at least one of the original authors of the rule. 16 A recent meta-analysis with over 35,000 patients showed that the PERC rule was safe to use in emergency care centers when patients presented without referral by a general physician or specialist. In this setting with a PE prevalence of 7%, the PERC rule had a failure rate of 1.12% when combined with Wells' criteria or 0.90% with clinical gestalt. 17 However, with a PE prevalence of 20% in referred secondary care, the failure rate was 6.01%, with the upper bound of the 95% confidence interval (CI) reaching 8.8%, thus indicating an insufficient safety profile when used in a population with a higher PE prevalence. [17][18][19][20] Of note, this point remains controversial as some studies showed that the PERC rule was safe when PE prevalence was <15% or even close to 30% when combined with a low probability assessed by gestalt. 21 The safety of the PERC was also confirmed in a randomized noninferiority study, 22 but the number of nonenrolled eligible patients was not reported and the proportion of low-probability patients in the PERC arm was higher, which was suggestive of an inclusion bias. 23 Finally, a large European prospective observational study found that the PERC failure rate was only 1.2%, with a PE prevalence of 11.4%. 24 Although the UK National Institute for Health and Care Excellence (NICE) guidelines encourage clinicians to consider using the PERC rule if clinical suspicion of PE is <15% based on their gestalt, 23 it is not yet recommended in the latest European guidelines. 1 Evidence regarding the PERC rule is based on publications from clinical trials, but their generalizability to the real-life clinical setting may be limited. 25 The Registro Informatizado de la Enfermedad TromboEmbolica venosa (RIETE) registry provides data on the management of venous thromboembolism (VTE) in such a setting with an unselected patient population. 26 Notably, the registry offers a unique opportunity to assess the performance of the PERC rule by using its large international database of VTE patients enrolled in diverse types of hospitals over the past 20 years. The purpose of this study was to assess the safety of the PERC rule in patients included in the RIETE registry.

Study design
This retrospective cohort study used data from the RIETE registry from March 1, 2001, through December 31, 2021. The methodology of the RIETE registry has been described elsewhere. 27  committees at all participating sites approved the protocol for enrollment and all patients or their health care proxies provided informed consent. As of January 2022, the registry contained data from over 101,000 patients from 210 hospitals in 26 countries, followed up for at least 3 months.

Measures
The primary outcome of our study was the overall percentage of PE patients included in the RIETE registry who were PERC-N and therefore represented the failure rate of the PERC rule. The registry does not provide all the items required to estimate the pretest probability of PE as clinicians' gestalt, "PE as the most likely diagnosis" in the Wells' score, and "unilateral lower leg pain" in the revised Geneva score are not recorded. Therefore, we draw a parallel with the two-tier Wells' score: given the shared criteria between the Wells' score and the PERC rule, PERC-N could have a maximum of 4 points (Appendix S1), i.e., PE unlikely using the two-tier rule with a prevalence of <15%.   in 890 (4.3%) patients despite all PERC criteria being negative, except for a missing O 2 Sat, and they were excluded from our analyses.

Characteristics of study subjects
PERC-P patients were similar in age to PERC-N patients, with similar VTE risk factors, apart from less frequent immobilization (Appendix S1). They had similar underlying diseases but presented less frequently with dyspnea and had a lower respiratory rate, while their heart rate was similar. PERC-P patients also had similar rates of measured and positive troponins, as well as echocardiogram results, with similar pulmonary pressures or signs of right ventricular dysfunction, and the same proportion was classified at low risk of a poor outcome. By contrast, PERC-P patients had a few more severe characteristics than PERC-N patients, such as a higher proportion of active cancer and more frequent syncope, hypotension, and PE located in the lobar and pulmonary arteries (Appendices S1 and S1).

Primary and secondary outcomes
A total of 48,903 patients were included in the final analysis.
Among these, 48,557 (99.3%) had at least one positive PERC item.
The three most frequently positive criteria were age ≥ 50 years, O 2 Sat ≤ 94%, and a heart rate of ≥100 beats/min. Only 346 patients were PERC-N, representing a failure rate of 0.7% (95% CI 0.6%-0.8%). These patients were predominantly male and younger and females were more likely to be pregnant or postpartum, but they were less likely to have an active cancer, leg varicosities, chronic heart failure or chronic pulmonary disease, anemia, or renal failure (Table 2).
In terms of clinical presentation, PERC-N patients had chest pain and a lower respiratory rate more often but dyspnea, syncope, or hypotension less often (Table 2). Regarding diagnostic tests, they received the same percentage of CTPAs, a lower percentage of compression ultrasonography that was more frequently negative, and more pulmonary arteriographies.
Diagnosed PEs were more likely to be subsegmental or segmen- For their initial anticoagulation, PERC-N patients were treated more often with direct oral anticoagulants (Table 3). They less often received mechanical or pharmacological thrombolysis. Long-term anticoagulation relied more frequently on direct oral anticoagulants and less often on vitamin K antagonists (VKA) or low-molecularweight heparin (LMWH). Patients in all groups were anticoagulated for slightly more than 6 months on average. PERC-N patients had a lower incidence of recurrent deep vein thrombosis (DVT), a 10fold lower incidence of major bleeding, and a sevenfold lower incidence of death attributed to PE (  Figure 2A; Appendix S1). In the 2016-2021 period, PERC-N patients had less hypotension than during the precedent periods, less positive D-dimer levels and more were at low risk of 30-day mortality (Table 5). Since 2012, more PEs were located distally in the pulmonary arteries. Finally, PE in PERC-N patients was diagnosed progressively more often in larger hospitals.

DISCUSS ION
In this retrospective cohort study based on the RIETE registry that included the highest number of PERC-N patients to date, we found that the PERC rule had a low failure rate when used in a large group of patients diagnosed with PE in diverse emergency department (ED) settings and mostly located in Europe. In addition, PERC-N patients had low 3month rates of hemorrhagic complications or recurrence of VTE, but similar rates of PE recurrence once anticoagulation was discontinued.  index or recurrent PE during follow-up. They had a low risk of bleeding during anticoagulation, concordant with a low RIETE score, and none had a major bleeding event during follow-up. In addition, the overall mortality rate was much lower than for PERC-P patients. The rate of nonfatal VTE during the first 3 months of anticoagulation was usually around 3.8%, while the rate of fatal PE was 0.5%, both higher than observed in PERC-N patients. 43 During and after being anticoagulated, PERC-N patients had a similarly low rate of PE recurrence but a lower recurrence of DVT. These differences reflect the PERC-N patients' younger age, fewer comorbidities, and lower proportion of anticoagulation with LMWH and VKA. [44][45][46][47] According to published clinical studies, PERC-N patients have a very low risk of recurrent PE or death at 3 months without anticoagulation. 22,24 However, the prevalence of PE was <5% in these studies and they had probably very few PERC-N patients. Our study showed that PERC-N patients had nonnull rates of recurrent PE or death, albeit much lower than in PERC-P patients during anticoagulation, but the rates of recurrent PE were similar after discon-

CON CLUS IONS
A low failure rate of the pulmonary embolism rule-out criteria rule was observed in the Registro Informatizado de la Enfermedad TromboEmbolica venosa registry, thus supporting its use to safely identify patients with an unlikely probability of pulmonary embolism according to a Wells' score of 4 points or less. However, our study was observational and the proportion of missing data is a source of uncertainty around the pulmonary embolism rule-out criteria rule true failure rate. Prospective and interventional studies are still needed before the use of the pulmonary embolism rule-out criteria rule can be universally recommended.

ACK N OWLED G M ENTS
The authors express gratitude to Sanofi Spain and ROVI for supporting the RIETE registry with an unrestricted educational grant. We also thank the RIETE Registry Coordinating Center and S&H Medical Science Service for their quality control data and logistic and administrative support and Professor Salvador Ortiz, Universidad Autónoma, Madrid, and statistical advisor in the S&H Medical Science Service, for the statistical analysis of the data presented in this paper. Open access funding provided by Universite de Lausanne.

CO N FLI C T O F I NTER E S T S TATEM ENT
The authors declare no conflicts of interest.