Background
Echocardiography is important for the diagnosis of infective endocarditis (IE), for which transesophageal echocardiography (TEE) is superior to transthoracic echocardiography (TTE).
Methods
A systematic review and meta-analysis of observational studies was performed with the objective of evaluating diagnostic properties of TTE, with transesophageal findings of IE as the reference standard in patients with suspected IE.
Results
The literature search yielded 377 unique articles, of which 16 met the inclusion criteria. The 16 studies included 2,807 patients, of whom 793 (28%) had vegetations on TEE. For detecting vegetations, harmonic TTE had sensitivity of 61% (95% CI, 45%–75%) and specificity of 94% (95% CI, 85%–98%) with a negative likelihood ratio (NLR) of 0.42 (95% CI, 0.26–0.61). NLR for harmonic TTE can be improved by including only patients without prosthetic valves (NLR = 0.36; 95% CI, 0.22–0.55) or by having strict criteria for conclusively negative results on TTE (NLR = 0.17; 95% CI, 0.10–0.28). In the setting of patients without prosthetic valves, harmonic TTE had likelihood ratios of 0.14 (95% CI, 0.09–0.23) for a conclusively negative result, 0.66 (95% CI, 0.53–0.81) for an indeterminate result, and 14.60 (95% CI, 3.37–70.40) for a positive result.
Conclusions
Modern harmonic TTE still has the potential to miss many vegetations detected on TEE. When limited to patients without prosthetic valves, a conclusively negative TTE under optimal view greatly decreases likelihood of IE. All other transthoracic results are not useful for ruling out IE, and subsequent TEE is almost always required.
Highlights
- •
Modern harmonic TTE still has the potential to miss many vegetations detected on TEE.
- •
Completely normal results on TTE in patients without prosthetic valves can aid in ruling out endocarditis.
- •
Any results on TTE other than completely normal require subsequent TEE to rule out endocarditis.
Attention ASE Members:
The ASE has gone green! Visit www.aseuniversity.org to earn free continuing medical education credit through an online activity related to this article. Certificates are available for immediate access upon successful completion of the activity. Nonmembers will need to join the ASE to access this great member benefit!
Methods
Search Strategy and Study Selection
A literature search was done for English articles in MEDLINE, Embase, and the Cochrane Library from inception to April 2, 2016. A combination of the following Medical Subject Headings or equivalent terms was used in the search strategy: “endocarditis,” “echocardiography,” “transthoracic echocardiography,” “diagnostic accuracy,” “sensitivity and specificity,” and “screening.” The search strategy for each database is listed in the Appendix . The references from appropriate articles were hand searched for additional references.
Inclusion and Exclusion Criteria
Studies were included if they met all of the following criteria:
- 1.
inclusion of adult patients with clinically suspected IE due to any cause,
- 2.
inclusion of patients with and without IE,
- 3.
clear explanation of TTE as the index test and TEE as the reference standard for IE, and
- 4.
sufficient data to allow extraction of a 2 × 2 diagnostic table.
Case reports, commentaries, letters, and reviews were excluded.
These inclusion and exclusion criteria were used for vegetation and valve dehiscence. For descriptive analysis of abscess, we included studies that described only IE cases.
Data Extraction and Quality Assessment
Titles and abstracts of all articles were screened. Full text and reference lists of relevant articles fulfilling the inclusion and exclusion criteria were reviewed in detail. Data for 2 × 2 tables were extracted. Individual authors of studies were contacted for any additional data to generate the 2 × 2 tables, which were then used in the analysis.
For each included full-text article, quality was assessed by the Rational Clinical Examination level of evidence and the Quality Assessment of Diagnostic Accuracy Studies criteria for potential biases.
We also extracted from the study description the type of echocardiography machine and time from TTE to TEE.
Definitions
Patients were classified into suspected native or prosthetic valve IE. Patients with any prosthetic valve undergoing echocardiography evaluation were considered to have suspected prosthetic valve IE.
On the basis of the description in the methodology of the source studies, the imaging modality was classified as either harmonic imaging or fundamental imaging.
Major echocardiographic findings on TTE and TEE were based on the modified Duke criteria for IE, which include oscillating intracardiac mass on valve or supporting structure, abscess, new partial dehiscence of prosthetic valve, and new valvular regurgitation.
Definitions for optimal view, suboptimal view, and indeterminate TTE were defined as the following on the basis of previous studies : optimal view was defined as adequate visualization of anatomic structures, chamber morphology, endocardial borders, and cardiac function from standard acoustic windows that was graded as satisfactory quality by the cardiologist at the time of clinical read. Any echocardiographic study not fulfilling these criteria was classified as suboptimal view. Indeterminate TTE was defined as TTE with no major echocardiographic findings on the basis of modified Duke criteria and any of the following: more than trace valvular regurgitation, any abnormal structural findings (valve thickening, valve prolapse, valve stenosis, valve or annular calcification, or valve sclerosis), or suboptimal views ( Supplemental Table 1 , available at www.onlinejase.com ). In studies that allowed indeterminate TTE, conclusively negative results were defined as TTE with no major echocardiographic findings on the basis of modified Duke criteria that did not fulfill any of the criteria above for indeterminate TTE.
The diagnostic test is the major echocardiographic finding on TTE. The reference standard is the major echocardiographic finding on TEE. As an example, for the major echocardiographic finding of vegetation, a true positive is vegetation on TTE and TEE. A false positive is vegetation on TTE and no vegetation on TEE. A true negative is no vegetation on TTE and TEE. A false negative is no vegetation on TTE and vegetation on TEE.
Likelihood Ratios
Likelihood ratios were used to describe diagnostic properties in addition to sensitivity and specificity. Likelihood ratio is the ratio of the proportion of patients with disease having a specific test result to the proportion of patients without disease having the same test result. For a binary diagnostic test (positive or negative), the likelihood ratio associated with a positive test result is the positive likelihood ratio, and the likelihood ratio associated with a negative test result is the negative likelihood ratio (NLR). Likelihood ratios convert the pretest probability of disease to the posttest probability of disease using the following formula :
Pretest odds = pretest probability / ( 100 % − pretest probability ) Posttest odds = pretest odds × likelihood ratio Posttest probability = posttest odds / ( posttest odds + 1 )
In clinical practice, the likelihood ratio can help in ruling in or ruling out a disease. A likelihood ratio >1 increases the probability of disease and helps in ruling in disease. Likelihood ratios of 2, 5, and 10 will increase the probability of disease by approximately 15%, 30%, and 45%, respectively. A likelihood ratio <1 decreases the probability of disease and helps in ruling out disease. Likelihood ratios of 0.5, 0.2, and 0.1 will decrease the probability of disease by approximately 15%, 30%, and 45% respectively.
We used likelihood ratios in our study for two reasons. First, unlike sensitivity and specificity, likelihood ratios can describe tests with multiple levels such as results on TTE, which can be positive, indeterminate, or negative. Second, likelihood ratios are more clinically useful for ruling a diagnosis in or out compared with using predictive values or sensitivity or specificity.
In the case of TTE, the NLR is more clinically important than the positive likelihood ratio, because TTE with a very low NLR will significantly decrease the probability of IE, thereby ruling out IE without need for subsequent TEE.
Statistical Analysis
For detection of vegetations on TTE, sensitivity, specificity, and likelihood ratios were calculated from the abstracted data for each study using the transesophageal finding as the reference standard. In 2 × 2 tables with a zero in one cell, 0.5 was added to all cells before doing the calculations. Ninety-five percent CIs were calculated for sensitivity, specificity, and likelihood ratios using standard methods.
Summary sensitivity, specificity, and likelihood ratios, along with their 95% CIs, were calculated using a bivariate random-effects logit-normal meta-analysis model. Because sensitivity and specificity are negatively correlated with each other, univariate summary estimates of sensitivity and specificity are not recommended. The bivariate logit-normal random-effects model estimates the average values of sensitivity and specificity while assuming that the sensitivity and specificity across studies follow a bivariate logit-normal distribution with a covariance. This model is recommended for meta-analysis of diagnostic test studies and has been extensively used in prior published meta-analysis studies.
We carried out a priori specified subgroup analyses for the following subgroups:
- 1.
harmonic TTE,
- 2.
harmonic TTE for patients without prosthetic valves,
- 3.
harmonic TTE with a strict criterion of conclusively negative results on TTE, and
- 4.
harmonic TTE with a strict criterion of conclusively negative results on TTE in patients without prosthetic valves, for which summary likelihood ratios were calculated for each level of TTE result (negative, indeterminate, and positive).
After the analysis, we carried out an exploratory post hoc subgroup analysis for studies differentiated on the basis of time interval from TTE to TEE.
Heterogeneity was summarized as I 2 statistics separately for positive likelihood ratio and NLR. Publication bias was assessed using the method described by Deeks et al . for diagnostic accuracy studies.
A minimum of four studies was required for meta-analysis and calculation of a pooled estimate. For transthoracic findings for which fewer than four studies were done, diagnostic properties were presented for each study.
All statistical tests were two sided with a significance level of .05. All analyses were done with R version 3.2.3 (R Foundation for Statistical Computing, Vienna, Austria) using the mada package ( https://Cran.R-project.org/package=mada ).
Results
Search Results
The literature search yielded 377 unique articles that were systematically screened. Of these, the full text of 25 articles was reviewed, and 16 articles the met inclusion criteria ( Figure 1 , Supplemental Table 2 , available at www.onlinejase.com ).
Among the 16 studies included in the analysis, study quality ranged from 1 to 3 by the Rational Clinical Examination quality checklist ( Supplemental Table 3 , available at www.onlinejase.com ). In nine studies (56%), nonconsecutive patients were recruited because of the exclusion of patients who did not undergo both TTE and TEE.
In total, the 16 studies included 2,807 patients with suspected IE with or without prosthetic valves, of whom 793 (28%) had vegetations on TEE. Study size ranged from 24 patients to 790 patients. The proportion of patients with suspected IE who had vegetations on TEE ranged from 20% to 67%.
Of the 16 studies, seven used harmonic TTE, eight used fundamental TTE, and one used both fundamental and harmonic TTE. Of the eight studies that used harmonic TTE, one study examined patients with or without prosthetic valves, five studies examined only patients without prosthetic valves, and two studies examined patients with or without prosthetic valves with additional subgroup analysis of patients without prosthetic valves. Of the eight studies that used harmonic TTE, seven studies considered a strict definition for conclusively negative results on TTE, and four studies reported numbers for indeterminate and negative results on TTE separately for patients without prosthetic valves.
Detection on TTE of Vegetations Found on TEE
Diagnostic properties of TTE for detecting vegetations that were found on TEE for individual studies are listed in Supplemental Table 4 (available at www.onlinejase.com ). A summary receiver operating characteristic curve was used to describe diagnostic properties for harmonic and fundamental TTE ( Figure 2 ). Summary estimates of diagnostic properties of fundamental and harmonic TTE as well as different criteria for a negative result are listed in Table 1 . In patients without prosthetic valves, a summary likelihood ratio was provided for each level of results (positive, indeterminate, and negative) for the four studies that reported numbers for negative and indeterminate TTE separately ( Table 2 ). For patients without prosthetic valves, conclusively negative results on TTE had the lowest NLR of 0.14.
1. TTE 2. Valves 3. Positive result 4. Negative result | Number of studies | Sensitivity (95% CI) | Specificity (95% CI) | PLR (95% CI) | NLR (95% CI) |
---|---|---|---|---|---|
1. Fundamental TTE 2. Native or prosthetic 3. Positive TTE 4. Negative or indeterminate TTE | 9 | 41% (29%–55%) | 96% (90%–99%) | 10.5 (3.7–30.5) I 2 = 0% | 0.61 (0.47–0.76) I 2 = 36% |
1. Harmonic TTE 2. Native or prosthetic 3. Positive TTE 4. Negative or indeterminate TTE | 7 | 61% (45%–75%) | 94% (85%–98%) | 9.5 (3.5–26.6) I 2 = 36% | 0.42 (0.26–0.61) I 2 = 54% |
1. Harmonic TTE 2. Native only 3. Positive TTE 4. Negative or indeterminate TTE | 6 | 66% (51%–79%) | 95% (85%–98%) | 12.3 (3.8–40.9) I 2 = 51% | 0.36 (0.22–0.55) I 2 = 38% |
1. Harmonic TTE 2. Native or prosthetic 3. Positive or indeterminate TTE 4. Negative TTE | 6 | 90% (80%–96%) | 58% (36%–78%) | 2.1 (1.5–3.6) I 2 = 6% | 0.17 (0.10–0.28) I 2 = 2% |
Result on harmonic TTE | Study | Likelihood ratio (95% CI) |
---|---|---|
Negative | Barton et al . (2014) | 0.14 (0.07–0.26) |
Casella et al . (2009) | 0.22 (0.09–0.55) | |
Jassal et al . (2007) | 0.03 (0.00–0.48) | |
Sivak et al . (2016) | 0.14 (0.05–0.40) | |
Pooled summary estimate | 0.14 (0.09–0.23) | |
Indeterminate | Barton et al . (2014) | 0.71 (0.48–1.06) |
Casella et al . (2009) | 0.23 (0.06–0.97) | |
Jassal et al . (2007) | 0.89 (0.21–3.85) | |
Sivak et al . (2016) | 0.74 (0.63–0.85) | |
Pooled summary estimate | 0.66 (0.53–0.81) | |
Positive | Barton et al . (2014) | 106.79 (30.79–370.36) |
Casella et al . (2009) | 5.35 (2.59–11.07) | |
Jassal et al . (2007) | 29.70 (1.92–460.19) | |
Sivak et al (2016) | 4.47 (3.32–6.01) | |
Pooled summary estimate | 14.60 (3.37–70.40) |
For harmonic TTE, Table 3 lists the diagnostic properties for patients with native or prosthetic valves according to the time interval between TTE and TEE.
Time from TTE to TEE | Study | Sensitivity (95% CI) | Specificity (95% CI) | PLR (95% CI) | NLR (95% CI) |
---|---|---|---|---|---|
At the same time | Chirillo et al . (2005) | 82% (66%–93%) | 98% (93%–100%) | 43.4 (10.9–172.7) | 0.19 (0.09–0.38) |
Within 24 h | Jassal et al . (2007) | 83% (61%–93%) | 97% (78%–100%) | 29.7 (1.9–460.2) | 0.18 (0.07–0.47) |
Within 1 wk | Casella et al . (2009) | 82% (66%–91%) | 86% (72%–93%) | 5.7 (2.7–12.2) | 0.21 (0.10–0.44) |
Within 1 wk | Sivak et al . (2016) | 43% (36%–51%) | 90% (88%–92%) | 4.5 (3.3–6.1) | 0.63 (0.55–0.72) |
Within 1 wk mean of 2.4 d from TTE to TEE | Kini et al . (2010) | 45% (38%–53%) | 80% (75%–84%) | 2.2 (1.7–2.9) | 0.69 (0.60–0.80) |
Detection on TTE of Abscesses Found on TEE
For detection of abscess by TTE, only sensitivity was calculated; summary estimate was not calculated, because no study provided the false positive rate.
Diagnostic properties of TTE for detecting abscesses that were found on TEE for individual studies are listed in Table 4 . Sensitivity of TTE ranged from 0% to 66% for fundamental TTE and from 29% to 100% for harmonic TTE.
TTE | Study | Number of patients | Abscess detected by TEE | Abscess detected by TTE | Sensitivity of TTE |
---|---|---|---|---|---|
Fundamental | Taams et al . (1990) | 33 | 1 (for 1 abscess on TEE, there was no vegetation on TEE) | 0 (for 1 abscess missed on TTE, TTE showed no vegetation) | 0% |
Fundamental | Pedersen et al . (1991) | 24 | 2 (for 2 abscesses on TEE, 1 had a vegetation on TEE) | 0 (for 2 abscesses missed on TTE, TTE showed vegetation in 1 case) | 0% |
Fundamental | Daniel et al . (1991) | 118 | 40 | 13 | 33% |
Fundamental | Birmingham et al . (1992) | 61 | 2 | 0 | 0% |
Fundamental | Shapiro et al . (1994) | 64 | 5 | 1 | 20% |
Fundamental | Morguet et al . (1995) | 73 | 1 (for 1 abscess on TEE, 1 had a vegetation on TEE) | 0 (for 1 abscess missed on TTE, TTE showed vegetation in 0 cases) | 0% |
Fundamental | Lindner et al . (1996) | 105 | 3 (for 3 abscesses on TEE, 3 had a vegetation on TEE) | 2 (for 1 abscess missed on TTE, TTE showed vegetation in 1 case) | 66% |
Fundamental | Fowler et al . (1997) | 103 | 2 | 0 | 0% |
Fundamental | Lengyel et al . (1997) | 31 | 14 | 4 | 29% |
Fundamental | Pachirat et al . (2000) | 105 | 7 | 2 | 29% |
Fundamental | Chirillo et al . (2005) | 139 | 7 | 1 | 14% |
Harmonic | Chirillo et al . (2005) | 139 | 7 | 2 | 29% |
Harmonic | Mocchegia et al . (2007) | 14 | 13 | 6 | 46% |
Harmonic | Casella et al . (2009) | 75 | 3 | 3 | 100% |
Harmonic | Barton et al . (2014) | 622 | 15 | 7 | 47% |
Findings on TTE of Prosthetic Valve Abnormalities Found on TEE
For transthoracic detection of prosthetic valve abnormalities, summary estimate was not calculated, because of the small number of studies. Diagnostic properties of TTE for detecting prosthetic valve abnormalities found on TEE for individual studies are listed in Table 5 . Both harmonic and fundamental TTE had low sensitivity for valve vegetation and dehiscence <50%.
TTE finding | Study | Sensitivity (95% CI) | Specificity (95% CI) | PLR (95% CI) | NLR (95% CI) |
---|---|---|---|---|---|
Fundamental vegetation | Morguet et al . (1995) | 20% (9%–38%) | 99% (91%–100%) | 18.5 (1.1–321.5) | 0.81 (0.68–0.98) |
Fundamental vegetation | Taams et al . (1990) | 10% (1%–54%) | 94% (63%–99%) | 1.8 (0.0–77.7) | 0.95 (0.68–1.33) |
Fundamental dehiscence | Morguet et al . (1995) | 41% (21%–64%) | 99% (92%–100%) | 47.9 (2.8–806.6) | 0.60 (0.40–0.90) |
Fundamental dehiscence | Taams et al . (1990) | 7% (1%–44%) | 93% (56%–99%) | 1.0 (0.0–43.7) | 1.0 (0.75–1.34) |
Harmonic vegetation | Kini et al . (2010) | 46% (33%–60%) | 79% (67%–87%) | 2.2 (1.2–3.9) | 0.68 (0.51–0.91) |
Methods
Search Strategy and Study Selection
A literature search was done for English articles in MEDLINE, Embase, and the Cochrane Library from inception to April 2, 2016. A combination of the following Medical Subject Headings or equivalent terms was used in the search strategy: “endocarditis,” “echocardiography,” “transthoracic echocardiography,” “diagnostic accuracy,” “sensitivity and specificity,” and “screening.” The search strategy for each database is listed in the Appendix . The references from appropriate articles were hand searched for additional references.
Inclusion and Exclusion Criteria
Studies were included if they met all of the following criteria:
- 1.
inclusion of adult patients with clinically suspected IE due to any cause,
- 2.
inclusion of patients with and without IE,
- 3.
clear explanation of TTE as the index test and TEE as the reference standard for IE, and
- 4.
sufficient data to allow extraction of a 2 × 2 diagnostic table.
Case reports, commentaries, letters, and reviews were excluded.
These inclusion and exclusion criteria were used for vegetation and valve dehiscence. For descriptive analysis of abscess, we included studies that described only IE cases.
Data Extraction and Quality Assessment
Titles and abstracts of all articles were screened. Full text and reference lists of relevant articles fulfilling the inclusion and exclusion criteria were reviewed in detail. Data for 2 × 2 tables were extracted. Individual authors of studies were contacted for any additional data to generate the 2 × 2 tables, which were then used in the analysis.
For each included full-text article, quality was assessed by the Rational Clinical Examination level of evidence and the Quality Assessment of Diagnostic Accuracy Studies criteria for potential biases.
We also extracted from the study description the type of echocardiography machine and time from TTE to TEE.
Definitions
Patients were classified into suspected native or prosthetic valve IE. Patients with any prosthetic valve undergoing echocardiography evaluation were considered to have suspected prosthetic valve IE.
On the basis of the description in the methodology of the source studies, the imaging modality was classified as either harmonic imaging or fundamental imaging.
Major echocardiographic findings on TTE and TEE were based on the modified Duke criteria for IE, which include oscillating intracardiac mass on valve or supporting structure, abscess, new partial dehiscence of prosthetic valve, and new valvular regurgitation.
Definitions for optimal view, suboptimal view, and indeterminate TTE were defined as the following on the basis of previous studies : optimal view was defined as adequate visualization of anatomic structures, chamber morphology, endocardial borders, and cardiac function from standard acoustic windows that was graded as satisfactory quality by the cardiologist at the time of clinical read. Any echocardiographic study not fulfilling these criteria was classified as suboptimal view. Indeterminate TTE was defined as TTE with no major echocardiographic findings on the basis of modified Duke criteria and any of the following: more than trace valvular regurgitation, any abnormal structural findings (valve thickening, valve prolapse, valve stenosis, valve or annular calcification, or valve sclerosis), or suboptimal views ( Supplemental Table 1 , available at www.onlinejase.com ). In studies that allowed indeterminate TTE, conclusively negative results were defined as TTE with no major echocardiographic findings on the basis of modified Duke criteria that did not fulfill any of the criteria above for indeterminate TTE.
The diagnostic test is the major echocardiographic finding on TTE. The reference standard is the major echocardiographic finding on TEE. As an example, for the major echocardiographic finding of vegetation, a true positive is vegetation on TTE and TEE. A false positive is vegetation on TTE and no vegetation on TEE. A true negative is no vegetation on TTE and TEE. A false negative is no vegetation on TTE and vegetation on TEE.
Likelihood Ratios
Likelihood ratios were used to describe diagnostic properties in addition to sensitivity and specificity. Likelihood ratio is the ratio of the proportion of patients with disease having a specific test result to the proportion of patients without disease having the same test result. For a binary diagnostic test (positive or negative), the likelihood ratio associated with a positive test result is the positive likelihood ratio, and the likelihood ratio associated with a negative test result is the negative likelihood ratio (NLR). Likelihood ratios convert the pretest probability of disease to the posttest probability of disease using the following formula :
Pretest odds = pretest probability / ( 100 % − pretest probability ) Posttest odds = pretest odds × likelihood ratio Posttest probability = posttest odds / ( posttest odds + 1 )