Unexpected pericardial effusions are often found by frontline providers who perform computed tomography. To study the hypothesis that electrocardiographic findings and whether cancer is known or suspected importantly change the likelihood of tamponade for such providers, all unique patients with moderate or large pericardial effusions determined by transthoracic echocardiography during a 6-year period were retrospectively identified. Electrocardiograms were evaluated by blinded investigators for electrical alternans (total and QRS), low voltage (limb leads only, precordial leads only, and both), and tachycardia (>100 QRS complexes/min). Medical records were reviewed to determine whether cancer was known or suspected and whether tamponade was diagnosed. Tamponade was present in 66 patients (27% of 241) with moderate or large pericardial effusions. No tachycardia lowered the odds of tamponade the most (likelihood ratio 0.4, 95% confidence interval 0.3 to 0.6) but by a degree less than any single diagnostic element increased it when present. The combined presence of all 3 electrocardiographic findings and cancer increased the odds of tamponade 63-fold (likelihood ratio 63, 95% confidence interval 33 to 150), whereas their combined absence decreased the odds only fivefold (likelihood ratio 0.2, 95% confidence interval 0.2 to 0.3). In conclusion, electrocardiography findings and cancer rule in tamponade better than they rule it out. Combining these diagnostic elements improves their discriminatory power but not sufficiently enough to rule out tamponade in patients with moderate or large pericardial effusions.
Pericardial effusions are often found unexpectedly when computed tomography is performed for nonspecific cardiopulmonary symptoms. Some unexpected pericardial effusions may be benign, whereas others may herald cardiac tamponade. Frontline clinical providers must therefore quickly determine which patients require immediate echocardiography. Previous diagnostic accuracy studies of electrocardiographic (ECG) findings for cardiac tamponade were limited by small sample sizes, absent comparison groups, and summary statistics for single findings that could not be combined. We therefore aimed to determine the diagnostic value of combining multiple ECG findings for cardiac tamponade in a large sample of patients with moderate or large pericardial effusions identified by echocardiography.
Methods
We conducted this retrospective study at a public teaching hospital after approval by our committee on human subjects research. We first identified all patients who had moderate or large pericardial effusions on standard transthoracic echocardiography from March 2005 to October 2010. We excluded small pericardial effusions, because when found unexpectedly, they rarely progress to tamponade. We then retrieved the electrocardiograms that were recorded just before each eligible echocardiogram. Electrocardiograms were eligible if they were obtained <1 day before echocardiographic studies of hospitalized patients or 1 week before echocardiographic studies of unhospitalized patients; we reasoned that unhospitalized patients’ effusions and ECG findings would remain stable. For patients with >1 eligible echocardiogram, we chose the earliest one that had an eligible electrocardiogram.
We reviewed patients’ medical records to determine whether an attending physician diagnosed or suspected active cancer at the time eligible electrocardiograms were obtained. Cancer is the most common cause of moderate or large pericardial effusion and tamponade in the United States, is already suspected in most patients who present with associated moderate or large pericardial effusions, and is most likely to be unambiguously recorded in the medical record. We used the clinical notes of the attending physician cardiologist or, if none was involved, the attending physician of record to classify each patient’s final pericardial effusion diagnosis. Attending physician cardiologists were involved in all tamponade diagnoses, which were based on direct examination, echocardiography, and in almost all cases right-sided cardiac catheterization and percutaneous pericardiocentesis.
All electrocardiograms were recorded with standard 12-lead placement at a paper speed of 25 mm/s and a standardization of 1 mV/cm. Blinded to identifying information, 2 investigators independently reviewed each electrocardiogram; a third investigator resolved any disagreements. Each electrocardiogram was manually evaluated for heart rate, low voltage, and electrical alternans. Heart rate was dichotomized as tachycardia or not on the basis of whether the average rate of QRS complexes was >100/min. Low voltage and electrical alternans were evaluated as ordinal outcomes. The highest ordinal category for low voltage was assigned when both sets of limb and precordial leads met low-voltage criteria for QRS amplitudes: <5 and <10 mm, respectively, in all corresponding leads. QRS amplitude for each lead was measured from the maximum positive to the maximum negative deflection. Intermediate ordinal categories were assigned when only 1 set of corresponding leads met low-voltage criteria. The highest ordinal category for electrical alternans was total electrical alternans: a peak-to-peak amplitude change of ≥1 mm in each successive P wave, QRS complex, and T wave of ≥1 lead. The intermediate ordinal category for electrical alternans was QRS electrical alternans: similar amplitude changes for just the QRS complexes of ≥1 lead. All patients underwent standard 2-dimensional transthoracic echocardiography. Pericardial effusions were classified by their largest echo-free dimension at end-diastole: small if as much as 10 mm posteriorly (with or without accumulation elsewhere) and moderate or large if greater. Probable tamponade was based on exaggeration of chamber collapse or respirophasic flow variation.
We based our sample size on the diagnostic accuracy of low voltage, the ECG finding with the most available data for tamponade. We used the lower bounds of the 95% confidence intervals (CIs) from a pooled estimate of sensitivity (32%) and a single estimate of specificity (95%), and we assumed that patients with and without tamponade would occur in a 1:1 ratio. Thus, to achieve a 95% CI for a positive likelihood ratio (LR) that excluded 2.0, our lower bound for a clinically meaningful result, we needed ≥60 patients with and without tamponade. Because we estimated 1 case of tamponade per month in our hospital, we reviewed 6 years of our echocardiography laboratory database. We measured interobserver variability for ECG findings with an unweighted κ statistic and constructed CIs with a bias-corrected nonparametric bootstrap. We calculated LRs and, for dichotomous findings, sensitivity and specificity using clinical documentation of tamponade as the reference standard. To allow readers to combine LRs from multiple findings, we used Spiegelhalter-Knill-Jones multivariate regression modeling to account for correlations between multiple diagnostic tests. We collapsed categories with similar LRs within each ECG finding and constructed CIs for adjusted LRs using a bias-corrected nonparametric bootstrap. All data were analyzed using Stata version 12.0 (StataCorp LP, College Station, Texas).
Results
Our study group consisted of 241 unique patients with moderate or large pericardial effusions ( Figure 1 ). Neoplastic effusions were diagnosed more often in patients with tamponade ( Table 1 ), and having known or suspected cancer increased patients’ odds of tamponade more than twofold ( Table 2 ). Among the 224 patients with available ECG data, the interobserver agreement for the assessment of ECG findings was excellent: unweighted κ statistics were 0.92 (95% CI 0.85 to 0.96) for electrical alternans and 0.97 (95% CI 0.95 to 0.99) for low voltage. Analysis of electrical alternans and low voltage as ordinal findings revealed 2 characteristics. First, total electrical alternans had perfect diagnostic accuracy but was too infrequent (3 of 224) to remain a stand-alone category. Second, low voltage in limb leads only or precordial leads only changed the odds of tamponade little more than no low voltage. Thus, all 3 categories were collapsed into 1 for multivariate analysis. The adjusted LRs were, because of expected positive correlations between diagnostic tests, closer to 1 than the unadjusted LRs ( Table 3 ).
| Characteristic | Tamponade | |
|---|---|---|
| Present (n = 66) | Absent (n = 175) | |
| Age (yrs) | 52 ± 14 | 56 ± 15 |
| Women | 32 (48%) | 91 (52%) |
| Left ventricular ejection fraction <50% ∗ | 7 (11%) | 28 (16%) |
| Hospitalized during echocardiography | 66 (100%) | 173 (99%) |
| Cause of pericardial effusion | ||
| Neoplasm | 34 (52%) | 38 (22%) |
| Connective tissue disease | 7 (11%) | 22 (13%) |
| Idiopathic | 6 (9%) | 66 (38%) |
| Infectious | 5 (8%) | 6 (3%) |
| Tuberculosis | 5 (8%) | 4 (2%) |
| Chronic kidney disease | 4 (6%) | 25 (14%) |
| Acquired immune deficiency syndrome | 3 (5%) | 3 (2%) |
| Hypothyroidism | 2 (3%) | 9 (5%) |
| Cirrhosis | 0 | 2 (1%) |
∗ Echocardiographic assessment of the left ventricular ejection fraction was indeterminate or missing for 2 patients with and 20 without tamponade.
| Finding | Adjusted LR (95% CI) † | |
|---|---|---|
| Present | Not Present | |
| Electrical alternans ‡ | 3.9 (2.7–5.0) | 0.7 (0.7–0.8) |
| Low voltage § | 3.5 (2.1–5.5) | 0.9 (0.8–0.9) |
| Tachycardia || | 2.2 (1.6–2.5) | 0.5 (0.5–0.7) |
| Cancer ¶ | 2.1 (1.6–2.7) | 0.7 (0.6–0.8) |
| All 3 ECG findings #,∗∗ | 34 (19–68) | 0.3 (0.3–0.4) |
| All 4 findings ∗∗ | 63 (33–150) | 0.2 (0.2–0.3) |
∗ Includes all 241 study patients. Missing ECG data (2 patients with tamponade and 15 patients without tamponade) were not considered diagnostic.
† The reported LRs have been adjusted for other findings in the model and can therefore be multiplied to combine >1 finding for patient-specific LRs. For example, electrical alternans present with tachycardia not present would have an LR of 2.0 (3.9 × 0.5).
‡ Peak-to-peak amplitude change ≥1 mm in each successive QRS complex of ≥1 lead. Thus, this definition includes total and QRS electrical alternans.
§ QRS amplitudes (measured from the maximum positive to the maximum negative deflection) <5 mm in all limb leads and <10 mm in all leads. Thus, this definition includes limb-lead and precordial-lead low voltage.
|| Average rate of QRS complexes >100/min by manual inspection of the electrocardiogram.
¶ Based on attending physicians’ clinical notes that were written before electrocardiography was performed.
# Whether patients had cancer was omitted from the multivariate model.
∗∗ Five patients had all 3 ECG findings, and 3 patients had all 4 findings. All 5 of these patients had tamponade.
Stay updated, free articles. Join our Telegram channel
Full access? Get Clinical Tree
