Differentiating Low-Risk and No-Risk PE Patients: The PERC Score

Christopher R. Carpenter, MD, MSC, FAAEM; Samuel M. Keim, MD, MS; Rawle A. Seupaul, MD; Jesse M. Pines, MD, MBA, MSCE; The Best Evidence in Emergency Medicine Investigator Group

J Emerg Med. 2009;36(3):317-322. 

Abstract

Background: Pulmonary embolism (PE) remains one of the most challenging diagnoses in emergency medicine. The Pulmonary Embolism Rule-out Criteria (PERC) score, a decision aid to reliably distinguish low-risk from very low-risk PE patients, has been derived and validated.
Clinical Question: Can a subset of patients with sufficiently low risk for PE be identified who require no diagnostic testing?
Evidence Review: The PERC score derivation and validation trials were located using PubMed and Web of Science. A critical appraisal of this research is presented.
Results: One single-center and another multi-center validation trial both confirmed that the eight-item PERC score identified a very low-risk subset of patients in whom PE was clinically contemplated with a negative likelihood ratio 0.17 (95% confidence interval 0.11-0.25) in the larger trial. If applied, the rule would have identified 20% of potential PE patients as very low risk.
Conclusion: The PERC score provides clinicians with an easily remembered, validated clinical decision rule that allows physicians to forego diagnostic testing for pulmonary embolus in a very low-risk population.

Case

A 44-year-old woman presents to the emergency department (ED) with a cough, non-pleuritic chest pain, and dyspnea. She is a smoker currently using Depo-Provera, but has no other past medical history. Her physical examination is unremarkable, including clear lung auscultation and a non-tender chest wall. Although you contemplate the diagnosis of pulmonary embolism, her well appearance and normal vital signs argue for bronchitis-related symptoms. You consider whether a D-dimer is indicated in this patient.

Clinical Question

Can a subset of patients with sufficiently low risk for pulmonary embolism be identified who require no diagnostic testing?

Context

Although 600,000 individuals have a pulmonary embolism (PE) identified every year, the diagnosis remains one of the most elusive for emergency physicians.[1,2] In 2003, American College of Emergency Physicians guidelines indicated that a negative D-dimer could be used to exclude PE in low-risk patient subsets. In randomized controlled trials, D-dimer testing has demonstrated favorable diagnostic test characteristics, but multiple D-dimer tests exist of variable diagnostic efficacy. Additionally, the false-positive rate increases with increasing age.[3,4] This is particularly important given recent concerns of radiation exposure and the subsequent need to limit computed tomography (CT) scanning.[5] Despite systematic incorporation of D-dimer testing into PE algorithms, ED diagnostic and therapeutic management often might be inappropriate, resulting in suboptimal care.[5,6] One key differentiation emergency physicians must make is low-risk from no-risk PE to avoid the potential of false-positive D-dimer results with resultant unnecessary pulmonary vascular imaging and systemic anticoagulation.

Evidence Search

Using PubMed clinical queries, category: clinical prediction guides, and scope: broad, the keywords 'pulmonary embolism' were entered on April 11, 2008 and 12,011 items were displayed. These results were combined with the search term 'D-dimer,' with the number of items reduced to 687. One derivation trial was acquired and appraised along with one validation trial.[6,7] Finally, a Web of Science search was conducted for those articles referencing the original derivation article, which identified yet another validation trial.[8] Decision rules were graded according to a published hierarchy of evidence.[9]

Evidence Review

PERC Score Derivation

Clinical criteria to prevent unnecessary diagnostic testing in ED patients with suspected pulmonary embolism. Journal of Thrombosis and Haemostasis, 2004.[6]

Population. ED patients from Carolina Medical Center (CMC) and Brigham & Women's Hospital (BWH) with 3148 subjects enrolled from January 2001 through June 2003. Subjects' mean age was 48 years, with an 11% prevalence of venous thromboembolism. In this derivation set, 69% were female, 72% had dyspnea, 31% cough, 50% pleuritic chest pain, and 12% prior PE or deep venous thrombosis (DVT).

Study Design. From 21 PE-related variables collected prospectively on all subjects, the authors derived an eight-item 'all-negative' clinical decision rule using logistic regression analysis. The authors called the rule the Pulmonary Embolism Rule-out Criteria (PERC) score ( ). They subsequently validated the PERC score on 382 very low-risk subjects at the CMC and 1427 low-risk ED patients from CMC and the BWH. Low risk was defined as 'enough clinical suspicion for pulmonary embolism that a board-certified emergency physician thought that a formal evaluation for pulmonary embolism was necessary.' Very low-risk subjects were identified by a research nurse and enrolled if the patient confirmed that 'shortness of breath was the most important or just as important as the main reason you came to the ED today' and the emergency physician told the research nurse that PE was not the most likely diagnosis. The authors sought to reduce their post-test probability below 1.8% based upon quantitative test-treatment threshold theories of Pauker and Kassirer[10] (Figure 1).

Table 1.  PERC Rule*

Age < 50 years
Pulse < 100 beats/min
Pulse ox > 94%
No unilateral leg swelling
No hemoptysis
No recent surgery
No prior DVT or PE
No oral hormone use

PERC = pulmonary embolism rule-out criteria; min = minute; pulse ox = pulse oximeter reading; DVT = deep venous thrombosis; PE = pulmonary embolism.
*Adapted from [6]: Kline JA, Mitchell AM, Kabrhel C, Richman PB, Courtney DM. Clinical criteria to prevent unnecessary diagnostic testing in emergency department patients with suspected pulmonary embolism. J Thromb Haemost 2004;2:1247-55.

Figure 1.

 

Test and treatment thresholds.

Primary Outcome

The primary outcome was the ED diagnosis of PE using the criterion standard of CT angiography (196 subjects), CT angiography with CT venography (1116 subjects), ventilation-perfusion scintillation lung scanning (1055 subjects including 372 with confirmatory duplex ultrasonography), pulmonary angiogram (110 subjects), or autopsy (21 subjects). Alveolar dead space measurement and D-dimer with 90-day follow-up was used in 650 subjects to exclude PE without pulmonary imaging.[6] All low-risk and very low-risk subjects were followed for 90 days after enrollment using a combination of telephone follow-up, medical record review, and contact with patients' personal physicians. The interval development of a PE at follow-up was defined by initiation of anticoagulation, vena caval interruption for venous thromboembolism (VTE), or death by PE.

Exclusion Criteria. No clear exclusion criteria or cohort are described, though very low-risk validation subjects had to answer 'yes' to the question, 'Was shortness of breath the most important or just as important as the main reason you came to the ED today?,' and the emergency physician informed the research nurse that PE was not the most likely diagnostic possibility.

Main Results. Patients with any positive PERC item were considered non-very-low-risk and further testing would be required to exclude PE. When tested on the low-risk validation cohort (with 90-day VTE prevalence of 8%), the PERC score had a positive likelihood ratio (LR) of 1.31 (95% confidence interval [CI] 1.25-1.38) and a negative LR of 0.16 (95% CI 0.07-0.38). In the very low-risk cohort (with 90-day VTE prevalence of 2.4%), the PERC score had a positive LR of 1.18 (95% CI 0.97-1.30) and a negative LR of 0 (95% CI 0.02-4.90) with no negative PERC score cases identified. The PERC score at the time of this publication met Level 4 criteria ().

Table 2.  Hierarchy of Evidence for Clinical Decision Rules*

Level 1:
   Rules that can be used in a wide variety of settings with confidence that can change clinician behavior and improve patient outcomes. Prospectively validated in at least one different population and also one impact analysis, demonstrating change in clinician behavior with beneficial consequences.
Level 2:
   Rules that can be used in various settings with confidence in their accuracy. Validated by a demonstration of accuracy in either one large prospective study (including a broad spectrum of patients and clinicians) or in several smaller settings that differ from one another.
Level 3:
   Rules that clinicians may consider using with caution if patients in the study are similar to those in the clinician's clinical setting. Validated in only one narrow prospective sample.
Level 4:
   Rules that need further evaluation before they can be applied clinically. Derived but not validated or validated only in split samples, large retrospective databases, or by statistical techniques.

*Reprinted with permission from [9]): McGinn TG, Guyatt GH, Wyer PC, Naylor CD, Stiell IG, Richardson WS. Users' guides to the medical literature: XXII: how to use articles about clinical decision rules. Evidence-Based Medicine Working Group. JAMA 2000;284:79-84.

Colorado PERC Validation

Assessment of the pulmonary embolism rule-out criteria rule for the evaluation of suspected pulmonary embolism in the ED. American Journal of Emergency Medicine, 2008.[7]

Population. Consecutive patients presenting to a single residency-affiliated community-based Denver ED from August 2001 through June 2002. Patients were enrolled over 120 randomly generated 8-h shifts with clinical suspicion for PE after history, physical examination, chest X-ray study, and electrocardiogram were obtained. A total of 134 patients were enrolled, with 65% over age 50 years.

Study Design. Results were derived from a secondary analysis of a prospective database for estimating the pretest probability of PE. All variables were collected before the PERC derivation study had been published. At the time of data collection, all subjects had their probability for PE estimated using Well's criteria and a D-dimer, as well as imaging specific for PE.[11] Data collection was performed by a research assistant blinded to D-dimer and radiographic results. Investigators modified the original PERC score by defining an abnormal oxygen level to be < 90% to reflect Denver's altitude.

Primary Outcome(s). The primary outcomes included the diagnosis of PE after initial ED evaluation, VTE during 3-month follow-up, or VTE-related death. The criterion standards for PE were any of the following: high probability ventilation/perfusion (V/Q) scan using Prospective Investigation of Pulmonary Embolism Diagnosis criteria; diagnostic PE-protocol CT scan; intermediate probability V/Q scan with high pre-test probability as determined by the treating clinician; positive pulmonary angiogram; self-reported VTE by 3-month telephone follow-up; medical record review documentation of VTE performed only after three failed attempts at telephone follow-up.[12]

Exclusion Criteria. Eligible subjects were excluded if they did not speak English, were pregnant within the preceding 6 months, weighed more than 350 pounds, had a pre-established diagnosis of thrombophilia, were younger than 18 years or older than 85 years, were critically ill or unable to consent, or were known to have a recently elevated or normal D-dimer assay.

Main Results. From the total cohort, 45% had a low pretest probability for PE, with 2 patients subsequently having a PE. The PERC score identified 22% (13/60) of low-risk subjects as very low-risk, with a positive LR of 1.28 and a negative LR of 0 for the primary outcome. The overall prevalence of PE in this study population was 12% (16/134). Only 1% of the study population was lost to follow-up. The PERC score at the time of this publication met Level 3 criteria ().

Table 2.  Hierarchy of Evidence for Clinical Decision Rules*

Level 1:
   Rules that can be used in a wide variety of settings with confidence that can change clinician behavior and improve patient outcomes. Prospectively validated in at least one different population and also one impact analysis, demonstrating change in clinician behavior with beneficial consequences.
Level 2:
   Rules that can be used in various settings with confidence in their accuracy. Validated by a demonstration of accuracy in either one large prospective study (including a broad spectrum of patients and clinicians) or in several smaller settings that differ from one another.
Level 3:
   Rules that clinicians may consider using with caution if patients in the study are similar to those in the clinician's clinical setting. Validated in only one narrow prospective sample.
Level 4:
   Rules that need further evaluation before they can be applied clinically. Derived but not validated or validated only in split samples, large retrospective databases, or by statistical techniques.

*Reprinted with permission from [9]): McGinn TG, Guyatt GH, Wyer PC, Naylor CD, Stiell IG, Richardson WS. Users' guides to the medical literature: XXII: how to use articles about clinical decision rules. Evidence-Based Medicine Working Group. JAMA 2000;284:79-84.

Multicenter PERC Validation

Prospective multicenter evaluation of the pulmonary embolism rule-out criteria. Journal of Thrombosis and Haemostasis, 2008.[8]

Population. Eligible patients were recruited from 12 United States EDs and one New Zealand ED between July 2003 and November 2006. Some sites enrolled all patients consecutively, aiming for 85% capture, whereas other sites recruited patients during randomly selected 8-h time blocks. Eligibility was defined by the ordering of an objective test for PE (CT, V/Q, or D-dimer) under the direction of a board-certified emergency physician after the initial history and physical examination were obtained.

Study Design. This was a prospective, observational trial with an intent-to-study preset plan to review medical records for any individual site failing to enroll 85% of eligible subjects. The purpose of the medical record review was to permit comparison of VTE frequency and patient demographics for enrolled vs. missed subjects. Clinical data, including physician gestalt, were collected and recorded on a web-based collection form before knowledge of test results. Patients were followed up to 45 days by telephone or mail contact with the patient, family, or primary care physician. The social security death index was queried if medical records could not be reviewed.

Primary Outcome. The primary outcome was the identification of PE with the criterion standard being a high-probability V/Q scan, positive CT angiogram, a positive pulmonary angiogram demonstrating a pulmonary arterial filling defect, or autopsy evidence of PE. A secondary outcome was the development of VTE within 45 days as determined by agreement between two independent clinicians who reviewed imaging results, medical records, and follow-up reports. A DVT was diagnosed by a positive venous duplex Doppler ultrasound or CT venogram in the popliteal, femoral, or axillary veins with written documentation of either anticoagulation or a vena caval filter.

Exclusion Criteria. Subjects were excluded for any of the following reasons: clinician knowledge of a positive PE imaging study in the preceding 7 days; the enrollment hospital was not the patient's hospital-of-choice for follow-up; or any circumstance that might compromise follow-up (e.g., homelessness or lack of a telephone).

Main Results. From 12,213 eligible subjects, 67% were enrolled with a mean age of 49 years. Two-thirds of enrolled subjects were female. The most common chief complaints were chest pain (53%) and dyspnea (33%). Clinician pre-test probability labeled 67% of patients as low risk and 74% had a D-dimer ordered. During 45 days of follow-up, 0.5% were diagnosed with VTE who were not diagnosed during the index ED evaluation or hospital admission. The incidence of VTE within 45 days was 6.9%, including 3% of those labeled as low risk by clinical gestalt. Among those with an alternative diagnosis more likely than PE and PERC-negative, 16 developed VTE, for a prevalence of 0.9% (95% CI 0.5-1.5%). The PERC score had a negative LR of 0.17 (95% CI 0.11-0.25) and would have identified 20% of these subjects as very low risk. The investigators conclude that in order for the PERC score to be useful, clinicians must first risk-stratify patients as pre-test low risk based upon validated clinical decision rules or clinical gestalt because the PERC score will reduce post-test risk below 1% only if the pre-test probability is 6% or less. The PERC score at the time of this publication met Level 2 criteria ().

Table 2.  Hierarchy of Evidence for Clinical Decision Rules*

Level 1:
   Rules that can be used in a wide variety of settings with confidence that can change clinician behavior and improve patient outcomes. Prospectively validated in at least one different population and also one impact analysis, demonstrating change in clinician behavior with beneficial consequences.
Level 2:
   Rules that can be used in various settings with confidence in their accuracy. Validated by a demonstration of accuracy in either one large prospective study (including a broad spectrum of patients and clinicians) or in several smaller settings that differ from one another.
Level 3:
   Rules that clinicians may consider using with caution if patients in the study are similar to those in the clinician's clinical setting. Validated in only one narrow prospective sample.
Level 4:
   Rules that need further evaluation before they can be applied clinically. Derived but not validated or validated only in split samples, large retrospective databases, or by statistical techniques.

*Reprinted with permission from [9]): McGinn TG, Guyatt GH, Wyer PC, Naylor CD, Stiell IG, Richardson WS. Users' guides to the medical literature: XXII: how to use articles about clinical decision rules. Evidence-Based Medicine Working Group. JAMA 2000;284:79-84.

Conclusion

The PERC score provides clinicians with an easily remembered, validated clinical decision rule that allows physicians to forgo diagnostic testing for pulmonary embolus in a very low-risk population. By appropriately using this tool, emergency physicians can diminish the cost, time, and ancillary test risk related to a false-positive D-dimer result with a miss rate under 1%. However, the PERC score should be applied only to low-risk PE patients, which requires application of another PE clinical decision rule, such as Well's criteria, to the general ED population with suspected PE. Using the PERC score on the general ED population will not reduce the post-test probability below 1%. For example, the Colorado validation trial had a VTE prevalence of 12%. Because the negative LR 95% CI for the multi-center PERC validation trial was 0.11-0.25, a negative PERC score would reduce the post-test probability from 12% to 1.4-3.3%.[13] Recognizing this two-step process for appropriate clinical application, additional studies are needed to define the rate of acceptance of this decision rule by practicing emergency physicians. Whether the rule will impact test ordering or alter patient-important outcomes remains undetermined.

Commentary by Jesse M. Pines, MD, MBA, MSCE

The detection of occult PE is a challenge. I use 'occult' because diagnosing PE in patients with clear signs of PE (e.g., unilateral leg swelling, tachycardia, hypoxia, sudden chest pain) is usually not a mystery. Occult PE is a different animal. It represents the diagnostic 'perfect storm:' 1) it can present atypically, 2) it may be lethal if missed, and 3) the diagnostic pathway involves a highly sensitive but poorly specific test (D-dimer) where more than half with positive D-dimers will be false positives and be subjected to the radiation of a negative chest CT scan. The recently validated PERC criteria shed new light on the challenge of occult PE by providing an intuitive rule that guides us to forgo testing in some low-risk patients. However, for those who intend to integrate the PERC into practice, a number of caveats should be discussed ().

Table 3.  Evidence-based Medicine Teaching Points

Pretest Probability
   The probability of disease for a given patient before the physician knows the results of a specific diagnostic test. Generally thought to equal the prevalence of the disease in the population of patients that most resemble the patient before the test is conducted.
Post-test Probability
   The probability of disease for a given patient after the physician knows the results of a specific diagnostic test. The post-test probability is different from the pretest probability as a function of the magnitude of the likelihood ratio characteristic of the diagnostic test. This post-test probability is quickly ascertained using a Fagan nomogram.[13]
Positive Likelihood Ratio (LR+)
   The amount by which the pretest probability is increased in a patient with a positive result for a specific test. It represents the ratio of a positive test result in patients with disease to the same test result in patients without the disease. Likelihood ratios combine the sensitivity and specificity of a diagnostic test so that it can be used, according to Bayes' theorem, to calculate the post-test probability of disease.
Negative Likelihood Ratio (LR-)
   The amount by which the pretest probability is decreased in a patient with a negative result for a specific test. It represents the ratio of a negative test result in patients with disease to the same test result in patients without disease. Likelihood ratios combine the sensitivity and specificity of a diagnostic test so that it can be used, according to Bayes' theorem, to calculate the post-test probability of disease.
Test and Treatment Thresholds
   Two thresholds involved in clinical decision-making can be assigned from data on the reliability and potential risks of the diagnostic test and the benefits and risks of a specific treatment. Treatment should be withheld if the probability of disease is smaller than the testing threshold, and treatment (if proven useful) should be given without further testing if the probability of disease is greater than the test-treatment threshold. The test should be performed only if the probability of disease is between the two thresholds (Figure 1)[10].
Risk Acceptance
   An attitude held by both physicians and patients that pertains to the amount of risk of harm or bad outcome they are willing to accept given a specific clinical scenario. Physician and patient values influence the amount of risk they are willing to accept. Patients and physicians may not understand probabilities well and frequently prefer risk categories or intervals, for example, very low risk.
Web of Science
   A searchable database product of the Thompson Corporation (Stamford, CT). Web of Science is a subset of the product ISI Web of Knowledge and consists of five databases containing information gathered from thousands of international scholarly journals.

Adapted from [10]): Pauker SG, Kassirer JP. The threshold approach to clinical decision making. N Engl J Med 1980;302:1109-17.
Adapted from [13]): Centre for Evidence-Based Medicine (CEBM). EBM tools: Interactive nomogram. Oxford, UK: CEBM. Available at: http://www.cebm.net/index.aspx?o=1161. Accessed April 24, 2008.

The first issue is the miss rate. Kline and colleagues calculated that a 1.8% (a little less than 1 in 50) miss rate was acceptable in the PERC derivation.[6] This number was based on a formula that incorporated the following risks from CT angiography: 1) cancer from radiation exposure, 2) anaphylaxis or severe pulmonary edema requiring intubation, and 3) requiring hemodialysis from the dye. It also considered the risk of death from missing a PE (defined as a 20% risk reduction for detected vs. missed). In the validation, the false-negative rate of the PERC was 1%. In English, this means that if you think a patient is low risk for PE and the patient is PERC negative, there is still a 1% chance that the patient actually has one. The question you have to ask yourself (and your patient) is: Is it really acceptable to send 1 in 100 patients home with a PE? Or, for that matter, 1.8 in 100? There is really no right answer. However, this perspective provides a more practical explanation of the risk behind the rule: PERC negative is not 100% perfect; there is still a small chance of PE.

An additional point that should be highlighted is that the PERC needs to be applied to the correct population. That is, it should be used only for 'low-risk' patients (< 15% pre-test probability). But differentiating who is truly 'low-risk' vs. 'low-risk but high-risk enough not to apply the PERC' is a difficult and subjective decision. Recent literature has demonstrated that inter-rater reliability for pre-test probability in PE is poor.[14] In addition, there may be patients who are at higher risk for PE due to other factors that were not common enough in the derivation set to be included in the PERC. For example, consider a 42-year-old woman with sudden-onset chest pain on standing after a transatlantic flight whose brother and sister both had a PE in their 40s. This patient meets the PERC criteria. Would you still order a test? What this illustrates is that applying the PERC to higher prevalence (pre-test probability) populations is associated with unacceptably high miss rates. When applying the PERC, consider the clinical scenario and ask yourself, is this patient really 'low risk?'

With these caveats in mind, we need to be practical and not throw out the baby with the bath water with important clinical decision rules like the PERC. The PERC is an important addition to the emergency medicine literature and, if used properly, may safely identify low-risk patients and prevent many from undergoing the all-too-common positive D-dimer, negative CT, 8-h-later pathway. In summary, although the PERC does not eliminate the challenge of diagnosing the occult PE, it does provide us with a compass to help navigate the waters for this complex and challenging diagnosis.

References

  1. Chunilal SD, Eikelboom JW, Attia J, et al. Does this patient have pulmonary embolism?. JAMA. 2003;290:2849-2858.

  2. Fesmire F, Kline J, Wolf S. Critical issues in the evaluation and management of adult patients presenting with suspected pulmonary embolism. Ann Emerg Med. 2003;41:257-270.

  3. Kearon C, Ginsberg JS, Douketis J, et al. An evaluation of D-dimer in the diagnosis of pulmonary embolism: a randomized trial. Ann Intern Med. 2006;144:812-821.

  4. Righini M, Goehring C, Bounameaux H, Perrier A. Effects of age on the performance of common diagnostic tests for pulmonary embolism. Am J Med. 2000;109:357-361.

  5. Brenner DJ, Hall EJ. Computed tomography—an increasing source of radiation exposure. N Engl J Med. 2007;357:2277-2284.

  6. Kline JA, Mitchell AM, Kabrhel C, Richman PB, Courtney DM. Clinical criteria to prevent unnecessary diagnostic testing in emergency department patients with suspected pulmonary embolism. J Thromb Haemost. 2004;2:1247-1255.

  7. Wolf SJ, McCubbin TR, Nordenholz KE, Naviaux NW, Haukoos JS. Assessment of the pulmonary embolism rule-out criteria rule for evaluation of suspected pulmonary embolism in the emergency department. Am J Emerg Med. 2008;26:181-185.

  8. Kline JA, Courtney DM, Kabrhel C, et al. Prospective multicenter evaluation of the pulmonary embolism rule-out criteria. J Thromb Haemost. 2008;6:772-780.

  9. McGinn TG, Guyatt GH, Wyer PC, Naylor CD, Stiell IG, Richardson WS. Users' guides to the medical literature: XXII: how to use articles about clinical decision rules (Evidence-Based Medicine Working Group). JAMA. 2000;284:79-84.

  10. Pauker SG, Kassirer JP. The threshold approach to clinical decision making. N Engl J Med. 1980;302:1109-1117.

  11. Wells PS, Anderson DR, Rodger M, et al. Derivation of a simple clinical model to categorize patients probability of pulmonary embolism: increasing the models utility with the SimpliRED D-dimer. Thromb Haemost. 2000;83:416-420.

  12. Value of the ventilation/perfusion scan in acute pulmonary embolism. Results of the prospective investigation of pulmonary embolism diagnosis (PIOPED) (The PIOPED Investigators). JAMA. 1990;263:2753-2759.

  13. Centre for Evidence-Based Medicine (CEBM). EBM tools. interactive nomogram. Oxford, UK: CEBM. http://www.cebm.net/index.aspx?o=1161 Accessed April 24, 2008.

  14. Rodger MA, Maser E, Stiell I, et al. The interobserver reliability of pretest probability assessment in patients with suspected pulmonary embolism. Thromb Res. 2005;116:101-107.