Article Text
PDF
Abstract
Objective To establish the diagnostic value of prespecified ECG changes in suspected pulmonary embolism (PE).
Methods Retrospective case–control study in a district general hospital setting. We identified 189 consecutive patients with suspected PE whose CT pulmonary angiogram (CTPA) was positive for a first PE and for whom an ECG taken at the time of presentation was available. We matched these for age±3 years with 189 controls with suspected PE whose CTPA was negative. We considered those with large (n=76) and small (n=113) clot load separately. We scored each ECG for the presence or absence of eight features that have been reported to occur more commonly in PE.
Results 20%–25% of patients with PE, including those with large clot load, had normal ECGs. The most common ECG abnormality in patients with PE was sinus tachycardia (28%). S1Q3T3 (3.7%), P pulmonale (0.5%) and right axis deviation (4.2%) were infrequent findings. Right bundle branch block (9.0%), atrial dysrhythmias (10.1%) and clockwise rotation (20.1%) occurred more frequently but were also common in controls. Right ventricular (RV) strain pattern was significantly more commonly in patients than controls, 11.1% vs 2.6% (sensitivity 11.1%, specificity 97.4%; OR 4.58, 95% CI 1.63 to 15.91; p=0.002), particularly in those with large clot load, 17.1% vs 2.6% (sensitivity 17.1%, specificity 97.4%; OR 7.55, 95% CI 1.62 to 71.58; p=0.005).
Conclusion An ECG showing RV strain in a breathless patient is highly suggestive of PE. Many of the other ECG changes that have been described in PE occur too infrequently to be of predictive value.
- electrocardiogram
- right ventricular strain pattern
- pulmonary embolism
Statistics from Altmetric.com
Introduction
Pulmonary embolism (PE) is one of the most common cardiovascular diseases with 47 594 cases reported in the UK in 2013.1 The majority of pulmonary emboli result from the propagation of a deep vein thrombosis (70%–80%).2 PE is the most common cause of vascular death after myocardial infarction (MI) and stroke and the leading cause of preventable death in patients admitted to hospital.3 It is a common mode of death in patients with cancer, stroke and pregnancy.4
PE may be categorised as massive, submassive or low risk based on the presence or absence of hypotension and right ventricular (RV) dysfunction.5 Most but not all patients with PE will have risk factors for venous thromboembolism, the most frequently reported of which is immobilisation.3 The most common presenting symptoms are dyspnoea and pleuritic pain but both may be absent even in patients with large clot load.6 Clinical probability scores and D-dimer are recommended in order to determine which patients should have CT pulmonary angiography (CTPA) but it is well recognised that patients with low probability scores can have PE with large clot load6 while D-dimer has low specificity for PE and is best regarded as a rule-out test.7
The role of the ECG in PE remains controversial. It is generally accepted that the ECG is of limited value in diagnosis but may predict a poor outcome when abnormal. A number of ECG changes have been reported to occur more commonly in PE including sinus tachycardia, right bundle branch block (RBBB), RV strain pattern, right axis deviation (RAD), P pulmonale, S1Q3T3 pattern, clockwise rotation and atrial dysrhythmias.8 9 RV strain pattern is recognised by simultaneous T wave inversion (TWI) in the inferior (II, III, avF) and right precordial leads (V1–V4) (figure 1). These changes are considered to reflect RV dysfunction in patients with large clot load10–12 and have been shown to predict a poorer outcome when present.8–10 13–15 Clockwise rotation is said to be present when the transition zone from dominant S wave to dominant R wave occurs after V4. This may be seen in normal subjects, in heart failure and also in patients with acute or chronic pulmonary disease.
ECG showing right ventricular (RV) strain pattern but not S1Q3T3 as there is a small R wave in lead III (see table 1 for definitions).
ECG changes said to occur more commonly in PE
It is our belief that these ECG changes might also be a clue to diagnosis and that failure to recognise the significance of these changes in a breathless patient might lead to a delayed or incorrect diagnosis. A difficulty here is that the changes described can occur in patients with other causes of RV dysfunction such as chronic lung disease16 and that similar but subtly different changes may be present in acute coronary syndrome (ACS), though in ACS with TWI in leads V1–V4 it would be unusual to find TWI in leads III and aVF as well.17 18
In order to address these issues further we have investigated the diagnostic value of prespecified ECG changes for PE in a cohort of patients all of whom were suspected of having PE and all of whom underwent CTPA.
Methods
This was a retrospective study carried out in Dumfries and Galloway Royal Infirmary, a district general hospital covering a population of 147 000 in southwest Scotland. We selected patients for inclusion if they had presented as an emergency and had both D-dimer and CTPA for suspected PE between September 2012 and March 2016.7 From a cohort of 1397 patients with suspected PE we identified 189 whose CTPA was positive for a first PE and for whom an ECG taken at the time of presentation was available. We considered patients with large (n=76) and small (n=113) clot load separately12 (figure 2). We then matched patients for gender and age ±3 years with 189 controls whose CTPA was negative. Patients were either admitted to hospital or discharged following CTPA either because this was negative or because we felt their PE could be managed safely in an ambulatory setting.
Study flow chart. CTPA, CT pulmonary angiogram; PE, pulmonary embolism.
We screened for the presence of cardiorespiratory disease by retrospective analysis of the electronic case note in order to identify conditions that might contribute to altered signal or ‘noise’ on the ECG. We defined pre-existing cardiorespiratory disease as MI, chronic obstructive pulmonary disease (COPD), pulmonary hypertension, pulmonary fibrosis or obstructive sleep apnoea (OSA). If we were unable to find this information in the admission document we searched general practitioner referrals and clinic letters. To be certain that we did not miss cases of OSA, we cross-checked a database of patients who were known to the local sleep service. It had been our intention to match cases and controls for age and sex and for cardiorespiratory disease though there were too few controls to do this. We were able, however, to match 87 CTPA-positive patients with 87 controls suspected of having PE whose CTPAs were negative, for age, sex and absence of cardiorespiratory disease.
We printed CTPA reports for all cases and controls and assigned each a unique reference number. We defined large clot load as PE in pulmonary trunk/main pulmonary artery and either the lobar arteries or the remote branches; and small clot load as PE confined to lobar arteries or remote branches.12 The radiologist (PH) who reviewed the CTPAs did so without knowledge of the ECG findings. We printed ECGs for all cases and controls from the electronic case note for analysis by a cardiologist (AM). Most patients had more than one ECG during their admission. When more than one ECG was present we selected the highest quality ECG that was most closely related in time to its respective CTPA. AM scored each ECG for the presence or absence of eight features that have been reported to occur more commonly in PE9 (table 1) without knowledge of the CTPA result.
For each ECG finding and each patient group (all patients and subgroups by clot load and absence of cardiorespiratory disease), we calculated sensitivity and specificity for identifying PE, along with binomial CIs, ORs with 95% confidence limits and p values from Fisher’s exact tests comparing cases and controls.19
Results
Our cohort of 189 patients with PE comprised 82 men and 107 women, average age 66 years with age range of 20–93 years. Seventy-six (40%) were judged to have large clot load and 87 (46%) no evidence of pre-existing cardiorespiratory disease. The distribution of pre-existing cardiorespiratory disease in patients and their age and sex-matched controls is shown in figure 3 which illustrates that more control patients had COPD.
Cardiorespiratory disease in cases and controls. COPD, chronic obstructive pulmonary disease; DVT, deep vein thrombosis; MI, myocardial infarction; OSA, obstructive sleep apnoea; PF, pulmonary fibrosis; PHTN, pulmonary hypertension.
Twenty to twenty-five per cent of patients with PE, including those with large clot load, had normal ECGs (table 2). The most common ECG finding in patients with PE was sinus tachycardia though this was not invariable: around 70% patients with large clot load had heart rates <100/min. S1Q3T3 (3.7%), P pulmonale (0.5%) and RAD (4.2%) were infrequent findings. RBBB (9.0%), atrial dysrhythmias (10.1%) and clockwise rotation (20.1%) occurred more frequently but were also common in controls. RV strain pattern occurred more commonly in patients than controls (11.2% vs 2.7%, p=0.002), particularly in those with large clot load (17.1% vs 2.6%, p=0.005) (tables 2 and 3).
Frequency of ECG findings for all patients and controls, and for subgroups by clot load and absence of cardiorespiratory disease
ORs for ECG as predictor of pulmonary embolism for all patients and subgroups by clot load and absence of cardiorespiratory disease
Tables 3 and 4 show sensitivities, specificities and ORs for each of the eight ECG abnormalities for all patients, and for patients with large clot load, small clot load and no pre-existing cardiorespiratory disease separately. Sensitivity for each individual ECG finding was always lower than specificity and never more than 31% which is too low for the ECG to be considered a rule-out test in patients suspected of having PE. Specificity approached 100% for RV strain, RAD, P pulmonale and S1Q3T3 but with the possible exception of RV strain the prevalence of these ECG abnormalities was too low for the ECG to be of value as a rule-in test. RV strain, which was present in 11.1% of all cases, 17.1% cases with large clot load and 10.3% of those with no pre-existing cardiorespiratory disease, was associated with specificities of 97.4% (95% CI 93.9% to 99.1%), 97.4% (95% CI 90.8% to 99.7%) and 98.9% (95% CI 93.8% to 100.0%), respectively (table 4). Corresponding ORs were 4.58, 7.55 and 9.82 though CIs were wide (table 3). Sensitivities were higher (76%–79%) and specificities lower (30%–35%) when we combined all ECG findings and considered any ECG abnormality. We did not calculate positive and negative predictive values as these are prevalence-dependent risk factors and this was a case–control study.
Sensitivities and specificities in per cent for all patients and subgroups by clot load and absence of cardiorespiratory disease, with binomial 95% CIs
Discussion
We believe that ours is the first study of the ECG as a predictor of diagnosis in PE to use age and sex-matched controls and also to consider clot load. The ECG finding that best predicted PE was RV strain pattern. S1Q3T3 was uncommon in our study. More of our control patients had COPD, suggesting that PE is frequently suspected but not confirmed in patients with COPD who present with worsening respiratory symptoms.
S1Q3T3 and RV strain pattern, when present, are the two ECG changes most frequently associated with RV dysfunction in PE. S1Q3T3, probably the most well-known ECG finding, was first described in 1935.20 The frequency with which it occurs varies from 3.7%21 to 50%,10 the higher figure reflecting the high proportion of patients with massive PE (76%) in that study. The prevalence of S1Q3T3 in our patients was 4% which is at the lower end of this range. We recognise that the determination of S1Q3T3 pattern can sometimes be difficult, particularly in relation to the presence or absence of a small Q wave in lead III and wonder if this might contribute at least in part to the variation in prevalence.
The literature on the role of the ECG in the diagnosis of PE includes some who believe the ECG is of limited value,22 23 and others who feel that ECG changes are a possible clue to a diagnosis that must then be confirmed by CTPA.21 24 Rodger and colleagues examined 212 consecutive inpatients and outpatients with suspected PE. All had a ventilation–perfusion scan and anyone with an indeterminate scan had pulmonary angiography. PE was confirmed in 49 patients and excluded in 163. Only sinus tachycardia and incomplete RBBB predicted PE in this data set. Patients and controls were not age or sex matched and no attempt was made to stratify by clot load or pre-existing cardiorespiratory disease. These authors concluded that the ECG was of limited value in the diagnosis of PE.22 Chan et al, reviewing the literature, came to a similar conclusion, namely that the overall utility of the ECG is limited due to the variable presence, frequency and transient nature of most of the ECG abnormalities associated with the disease.23
By contrast, Marchick and colleagues were able to show positive likelihood ratios for S1Q3T3 pattern and precordial TWI among 6049 emergency department patients suspected of having PE, in 354 of whom a diagnosis was subsequently confirmed. No attempt was made to age or sex-match patients and controls or to stratify by clot load, though their results were not influenced one way or another by the presence of pre-existing cardiorespiratory disease.24 More recently, Co et al used patients as their own controls by comparing ECG at diagnosis with a previous ECG in a cohort of 352 patients with PE. Seventy-six per cent had a significant change in their ECG, most notably TWIs in just over one-third of cases. These TWIs occurred in all leads but were most common inferiorly.21
How may these apparently conflicting views be reconciled? The prevalence of RV strain varies from 7.3% (24) to 68% (10), reflecting the fact that most studies include a mixture of patients with large and small clot load. In our study, 76/189 (40%) patients were judged to have large clot load, a proportion similar to that reported by Co et al.21 We might expect the ECG to be less predictive for PE in studies with a higher proportion of small PEs. It is also likely that studies with smaller numbers of patients may not have the power to detect the findings of RV dysfunction. Six of the 11 studies reviewed by Chan et al comprised 50 or fewer patients.23
Strengths and limitations
Our study has strengths and limitations. Strengths are our comparison with age and sex-matched controls, and stratification by clot load. The fact that more control patients had pre-existing cardiorespiratory disease may also be a strength as we were still able to show that RV strain predicted PE despite using controls likely to have RV strain for other reasons. We would have preferred to match cases and controls for presence or absence of cardiorespiratory disease but had too few control patients to enable us to do this. Our main limitation is that we could find no electronic record for 96 of our 313 CTPA-positive patients. It is likely that this group includes more patients with massive and submassive PE who died early during the course of their admission as case sheets for such patients are often sent straight to file and may not always be scanned. We used an estimate of clot load rather than RV dilation as our index of PE severity. RV dilation and biochemical markers such as high sensitivity troponin and NT-Brain natriuretic peptide could further refine the assessment of severity. We did not attempt to assess interobserver variability when analysing ECGs and CTPAs but eliminated interpretation bias by ensuring that our radiologist and cardiologist were unaware of each other’s findings. As this was a case–control study, we made no attempt to calculate positive and negative predictive values as these are prevalence-dependent factors.
Conclusion
In conclusion, an ECG showing RV strain in a breathless patient with no history of cardiorespiratory disease is highly suggestive of PE with specificity of 97.4% for large clot load. S1Q3T3 also has high specificity for PE but was a relatively uncommon finding in our study. Many of the other ECG changes that have been described in PE occur too infrequently to be of predictive value.
Main messages
The ECG finding that best predicted pulmonary embolism (PE) in our study was right ventricular (RV) strain pattern.
S1Q3T3 was uncommon in our study.
An ECG showing RV strain when present in a breathless patient is highly suggestive of PE.
Many of the other ECG changes that have been described in PE occur too infrequently to be of predictive value.
Current research questions
It would be interesting to repeat the study in all patients undergoing CT pulmonary angiography (CTPA) and not simply those who had both D-dimer and CTPA.
We used an estimate of clot load rather than right ventricular (RV) dilation as our index of pulmonary embolism severity. It is possible that RV dilation would have been a better marker of severity.
The use of biochemical markers such as high sensitivity troponin and N-terminal pro b-type natriuretic peptide could be used to further refine the assessment of severity.
What is already known on the subject box
The role of the ECG in identifying patients who may have pulmonary embolism (PE) remainscontroversial.
The S1Q3T3 pattern is perhaps best known.
The literature on the role of the ECG in the diagnosis of PE ranges from those who believe the ECG is of limited value to those who feel that ECG changes are a possible clue to a diagnosis that must then be confirmed by CTPA.
References
Footnotes
Contributors CI, DT, GK and RT designed the study. PH and AM were responsible for scoring the CTPAs and ECGs, respectively. CMM undertook the statistical analyses. CI wrote the first draft and worked with DT and GK on the second draft. All authors contributed to the final draft.
Funding The authors have not declared a specific grant for this research from any funding agency in the public, commercial or not-for-profit sectors.
Competing interests None declared.
Patient consent for publication Not required.
Ethics approval In keeping with the policies of the Scottish health boards,
ethics approval is not required as there were no patient identifiable data.
Provenance and peer review Not commissioned; internally peer reviewed.