Cardiac Computed Tomography Versus Transesophageal Echocardiography for the Detection of Left Atrial Appendage Thrombus: A Systemic Review and Meta-Analysis.

Background Transesophageal echocardiography (TEE) has been considered the gold standard for left atrial appendage (LAA) thrombus detection. Nevertheless, TEE may sometimes induce discomfort and cause complications. Cardiac computed tomography has been studied extensively for LAA thrombus detection. We performed this systemic review and meta-analysis to assess the diagnostic accuracy of cardiac computed tomography for LAA thrombus detection compared with TEE. Methods and Results A systemic search was conducted in the PubMed, Embase, and Cochrane Library databases from January 1977 to February 2021. Studies performed for assessment diagnostic accuracy of cardiac computed tomography on LAA thrombus compared with TEE were included. Summary sensitivity, specificity, and posterior probability of LAA thrombus was calculated by using bivariate random-effects model. The Quality Assessment of Diagnostic Accuracy Studies-2 tool was used for the quality assessment. A total of 27 studies involving 6960 patients were included in our study. The summary sensitivity of early imaging studies was 0.95 (95% CI, 0.79-0.99), and the specificity was 0.89 (95% CI, 0.85-0.92). The positive posterior probability was 19.11%, and the negative posterior probability was 0.16%. The summary sensitivity of delayed imaging studies was 0.98 (95% CI, 0.92-1.00), and the specificity was 1.00 (95% CI, 0.98-1.00). The positive posterior probability was 95.76%, and the negative posterior probability was 0.12%. The delayed imaging method significantly improved the specificity (1.00 versus 0.89; P<0.05) and positive posterior probability (95.76% versus 19.11%; P<0.05). Conclusions Cardiac computed tomography with a delayed imaging is a reliable alternative to TEE. It may save the patient and health care from an excess TEE. Registration URL: https://www.crd.york.ac.uk/PROSPERO; Unique identifier: CRD42021236352.

cardiac computed tomography

left atrial appendage

pulmonary vein isolation

Clinical Perspective

What Is New?

This updated meta‐analysis demonstrated that compared with transesophageal echocardiography, cardiac computed tomography showed a high diagnostic accuracy for left atrial appendage thrombus detection when delayed imaging was used.

What Are the Clinical Implications?

Cardiac computed tomography with a delayed imaging method is a reliable alternative tool for left atrial appendage thrombus detection.

Doing a delayed computed tomography scan adds nominal radiation exposure (<1 millisievert) and allows a single test to perform both tasks (pulmonary vein assessment and rule out left atrial thrombus), saving the patient and health care from an excess transesophageal echocardiography before pulmonary vein isolation.

Left atrial appendage (LAA) thrombus, which may present in conditions resulting in left atrial flow stasis, especially in atrial fibrillation, is an important source of cardioembolic stroke. Transesophageal echocardiography (TEE) is currently considered the gold standard for the detection of LAA thrombus, based on 2 large prospective studies. , However, TEE is a semi‐invasive and time‐consuming procedure. Although generally safe when performed by experienced operators, TEE carries physical discomfort for some patients and is associated, although rarely, with potentially life‐threatening complications.

In the past 2 decades, cardiac computed tomography (CCT) has been studied extensively for LAA thrombus detection. Almost all of the studies reported that CCT has a high sensitivity for LAA thrombus detection, whereas the specificity has been reported variable. Studies using delayed imaging method reported higher specificity than studies using early imaging method. Moreover, it only takes a few minutes for CCT scan, far less than that of TEE. This will reduce time cost for examiners and patients. Some researchers have assessed the diagnostic accuracy of CCT by conducting meta‐analyses. , , , , The results of these studies have reported that CCT has good diagnostic accuracy for LAA thrombus detection, , , , , especially when the delayed imaging method is used. , However, there are reasons to conduct a new meta‐analysis. First, all studies included in these meta‐analyses were conducted before the year of 2014, and studies using delayed imaging method were relative few. Second, the pooled sensitivity and specificity of 2 meta‐analyses , were relatively low. Third, 2 meta‐analyses , included 1 study that did not meet the criteria because CCT was used for cardiogenic embolus detection, not LAA thrombus detection. In this study, CCT was used for cardiogenic thrombus but not LAA thrombus detection. Moreover, there have been some new studies (at least 10) on LAA thrombus detection using CCT in recent years, some of which reported higher specificity and narrower CIs. , , We therefore conducted this systematic review and meta‐analysis to determine the diagnostic accuracy of CCT versus TEE for LAA thrombus detection.

METHODS

Authors declare that they will make the data, methods used in the analysis, and materials used to conduct the research available to any researcher for purposes of reproducing the results or replicating the procedure. The data that support the findings of this study are available from the first author on reasonable request.

This meta‐analysis was performed on the basis of guidelines from the Preferred Reporting Items for a Systematic Review and Meta‐Analysis of Diagnostic Test Accuracy Studies statement (Tables S1 and S2). The literature search, article screening, study selection, quality assessment, and data extraction were performed by 2 authors (S.Y. and H.Z.) independently. Disagreements were resolved by discussion, and a consensus was reached in the selection of the articles for analysis.

Search Strategy

PubMed, Embase, and Cochrane Library databases were searched from January 1977 to February 2021. The search terms are shown in Table S3. In addition, we searched relevant studies from references of the retrieved articles.

Study Selection

Studies fulfilling the following criteria were included: (1) assessment of left atrial thrombus; (2) patients who underwent both CCT and TTE; and (3) sensitivity, specificity, positive predictive value, and negative predictive value data were provided or could be calculated.

Data Extraction and Quality Assessment

Data extraction was performed by 2 authors (S.Y. and H.Z.) independently. We extracted demographics of patients, indications of left atrial thrombus, and CCT method (eg, electrocardiogram (ECG) gated versus non–ECG gated).

Quality Assessment of Diagnostic Accuracy Studies‐2 was used for the quality assessment of the included studies. Two authors (S.Y. and H.Z.) assessed the risk of bias and applicability concerns independently. The following domains were used to assess bias risk and applicability concerns: patient selection, performance of the index test, performance of the reference standard, and flow and timing (the interval between index test and standard reference, for risk of bias assessment only).

Data Synthesis and Statistical Analysis

Metandi and midas commands in Stata 15.0 (StataCorp, College Station, TX) were used for data synthesis and analysis. , The analysis was implemented mainly by midas, and metandi was used to construct hierarchical summary receiver operating characteristic curve. Sensitivity, specificity, and likelihood ratio (LR), along with 95% CIs, were calculated from the contingency 2×2 tables of true‐positive, false‐positive, false‐negative, and true‐negative results using a bivariate random‐effects model estimation. Random effects model was selected because heterogeneity is expected in meta‐analysis of diagnostic accuracy studies.

Primarily, midas uses an exact binomial rendition of the bivariate mixed‐effects regression model developed by Van Houwelingen for treatment trial meta‐analysis and modified for synthesis of diagnostic test data. It fits a 2‐level model, with independent binomial distributions for the true positives and true negatives conditional on the sensitivity and specificity in each study and a bivariate normal model for the logit transforms of sensitivity and specificity between studies. The standard output of the bivariate model includes the following: mean logit sensitivity and specificity with their SEs and 95% CIs; and estimates of the between‐study variability in logit sensitivity and specificity and the covariance between them. On the basis of these parameters, we can calculate other measures of interest, such as the likelihood ratio for positive and negative test results, the diagnostic odds ratio (OR), and the correlation between logit sensitivity and specificity. Summary sensitivity, specificity, and the corresponding positive likelihood, negative likelihood, and diagnostic ORs are derived as functions of the estimated model parameters. The derived logit estimates of sensitivity, specificity, and respective variances are used to construct a hierarchical summary ROC curve to display the variation in diagnostic accuracy among studies.

Posterior probability of LAA thrombus was also calculated to assess the diagnostic accuracy. The formula is as follows: posterior probability=pretest probability (LAA thrombus incidence)×LR/(pretest probability×LR+1). I2 index was used to assess the heterogeneity. Heterogeneity sources among studies was investigated by using multiple univariable meta‐regression and subgroup analysis. Publication bias was assessed by the Deek method.

RESULTS

Search Results

We identified 588 potentially eligible articles. In total, 555 articles were excluded by reviewing the titles and abstracts. The remaining 33 articles were evaluated in detail. Finally, 27 articles that met the inclusion criteria were identified (Figure 1). Six studies were excluded because not all patients underwent TEE, , , no thrombus was found, the sensitivity and specificity could not be calculated because the reference test was surgical finding, and one study was not limited to LAA thrombus detection.

Figure 1

Flowchart of selection of studies.

CCT indicates cardiac computed tomography; LAA, left atrial appendage; sen, sensitivity; spe, specificity; and TEE, transesophageal echocardiography.

Baseline Characteristics of the Included Studies

The baseline characteristics of the included studies are shown in Table 1. , , , , , , , , , , , , , , , , , , , , , , , Seventeen studies (63%) had a prospective design, and 10 studies (37%) had a retrospective design. Nineteen studies (70%) were performed with patients scheduled for pulmonary vein isolation (PVI), 4 studies (15%) were performed with patients recently experiencing stroke, 1 study was performed with patients scheduled to direct current cardioversion, and the remaining 3 studies had mixed populations. The ECG‐gated method was used in 16 studies (59%). CCT with delayed imaging was performed in 11 studies (41%). The incidence of LAA thrombus was 3.68% (251/6960).

Table 1

Baseline Characteristics of Included Studies

Study	Year	Design	No. of patients	Men, %	Age, y	Indication	CT type	Slice thickness, mm	Diagnostic criteria for TEE
Achenbach 28	2004	Prospective	52	64	66±10	DCCV	ECG‐gated EBCT; early phase	1.5	LAT/LAAT
Kim 29	2007	Retrospective	223	82	57±10	PVI	ECG‐gated 16‐, 40‐, 64‐slice MDCT; early phase	1.2, 0.75, 0.6	LAAT+SEC
Shapiro 30	2007	Retrospective	21	N/A	N/A	No restrict	ECG‐gated 64 slice; MDCT; early phase	0.6	LAAT
Feuchtner 31	2008	Prospective	64	68	58±13	PVI/valve surgery	ECG‐gated 64 slice; MDCT	0.6	LAT/LAAT
Tang 32	2008	Prospective	170	72	56±12	PVI	Non–ECG‐gated 64 slice; MDCT; early phase	N/A	LAT/LAAT
Hur 33	2008	Retrospective	101	62	67	Stroke	ECG‐gated 64‐section CCTA; early phase	0.6	LAAT
Patel 34	2008	Prospective	72	69	56±10	PVI	ECG‐gated 64 slice; MDCT; early phase	0.625	LAAT+SEC
Martinez 35	2009	Prospective	402	76	56±10	PVI	64 Slice; MDCT; early phase	0.6	LAAT
Hur 36	2009	Prospective	55	65	61	Stroke	ECG‐gated 64‐section CCTA; late phase	0.6	LAAT
Kim 37	2010	Prospective	314	59	65±13	Stroke	ECG‐gated 64‐slice MDCT; late phase	0.625	LAAT
Kapa 38	2010	Prospective	255	78	59±11	PVI	ECG‐gated DSCT; early phase	0.6	LAAT
Maltagliati 39	2011	Prospective	171	83	60±11	PVI	64‐Slice MDCT; early phase	N/A	LAA/LAAT
Hur 40	2011	Prospective	83	67	63±10	Stroke	ECG‐gated DSCT; late phase	0.6	LAAT+SEC
Swait 41	2012	Retrospective	70	N/A	N/A	PVI	ECG‐gated (patient in sinus rhythm) and nongated (patients in AF) 256‐, 128‐, and 64‐slice CCT; late phase	N/A	LAT+LAAT
Hur 42	2013	Prospective	101	70	62±10	PVI	ECG‐gated 128‐, 64‐slice CCT; late phase	0.6	LAT+LAAT
Dorenkamp 43	2013	Prospective	329	65	62±10	PVI	ECG‐gated 64‐slice MDCT; early phase	0.625	LAT+LAAT
Budoff 44	2014	Retrospective	86	81	66	PVI	64‐Slice CCTA; late phase	N/A	LAAT
Hong 45	2014	Retrospective	678	78	57±11	PVI	ECG‐gated 64‐slice MDCT; early phase	0.6	LA/LAAT+SEC
Homsi 46	2016	Prospective	124	83	58±12	AF/stroke	64‐Slice MDCT; early phase	0.9	LAAT+SEC
Lazoura 11	2016	Retrospective	122	78	60	PVI	ECG‐gated DSCT; late phase	0.5	LAAT
Wang 47	2016	Retrospective	831	75	61±10	PVI	Non–ECG‐gated 64 slice; MDCT; early phase	0.625	LAAT+SEC
Zhai 48	2017	Retrospective	783	72	55±11	PVI	ECG‐gated 64 slice; MDCT; late phase	0.625
Kottmaier 49	2019	Prospective	622	69	60±10	PVI	ECG‐gated (patient in sinus rhythm) and nongated (patients in AF) 64‐slice DSCT early phase image	0.6	LAT
Kuronuma 12	2019	Prospective	81	75	68±11	PVI	ECG‐gated CCT; late phase	N/A	LAAT
Li 50	2019	Prospective	302	54	64±7	PVI	64‐Slice DSCT; late phase	0.6	LAAT
Guha 51	2020	Retrospective	480	66	63	PVI	64‐Slice MDCT; early phase	0.63	LAAT
Spagnolo 10	2020	Prospective	260	77	59±11	PVI	ECG‐gated 64‐slice CCT; late phase	N/A	LAAT

AF indicates atrial fibrillation; CCT, cardiac CT; CCTA, coronary CT angiography; CT, computed tomography; DCCV, direct current cardioversion; DSCT, dual‐source CT; ECG, electrocardiogram; EBCT, electron‐beam CT; LAAT, left atrial appendage thrombus; LAT, left atrial thrombus; MDCT, multidetector CT; N/A, not available; PVI, pulmonary vein isolation; SEC, spontaneous echo contrast; and TEE, transesophageal echocardiography.

Quality Assessment

The results of the quality assessment are summarized in Table S4. In total, 3.70% (1/27) of the studies showed an unclear risk of bias in the patient selection domain, 7.41% (2/27) of the studies showed an unclear risk of bias in the index test domain, 33.33% (9/27) of the studies showed an unclear risk of bias in the reference standard domain, and 7.41% (2/27) of the studies showed a high or unclear risk of bias in the flow and timing domain.

Main Analysis

Analysis was based on study design (prospective or retrospective), imaging methods (early or delayed imaging; ECG gated or non–ECG gated), indication (PVI or not PVI), and sample size (patient number >100 or ≤100). The results are shown in Table 2. Sensitivity and negative LR (LR−) were not influenced by any factors, but the delayed imaging method had a significant impact on specificity and positive LR (LR+). The pooled sensitivity and specificity of the early and delayed imaging subgroups are also shown in Figures 2A and 2B and 3A and 3B.

Table 2

Sensitivity and Specificity of Each Subgroup

Subgroup	Sensitivity (95% CI)	Specificity (95% CI)	LR+ (95% CI)	LR− (95% CI)	Incidence of thrombus
Prospective	0.97 (0.82–1.00)	0.97 (0.93–0.99)	29.91 (13.36–66.96)	0.03 (0.00–0.20)	169/3467
Retrospective	0.98 (0.85–1.00)	0.92 (0.82–0.97)	12.63 (5.30–30.12)	0.02 (0.00–0.19)	82/3493
Early	0.95 (0.79–0.99)	0.89 (0.85–0.92)*	8.99 (6.61–12.21)*	0.06 (0.01–0.26)	120/4695
Delayed	0.99 (0.92–1.00)	1.00 (0.98–1.00)*	368.27 (41.94–3233.86)*	0.01 (0.00–0.08)	131/2265
ECG gated	0.98 (0.87–1.00)	0.97 (0.93–0.99)	36.30 (12.99–101.46)	0.02 (0.00–0.14)	158/3604
Non–ECG gated	0.97 (0.73–1.00)	0.91 (0.85–0.95)	11.32 (6.61–19.39)	0.03 (0.00–0.36)	93/3356
PVI	0.98 (0.84–1.00)	0.95 (0.91–0.97)	19.69 (10.13–38.3)	0.03 (0.00–0.19)	134/6146
Non‐PVI	0.99 (0.76–1.00)	0.96 (0.86–0.99)	28.11 (6.75–117.02)	0.01 (0.00–0.30)	117/814
PVI delayed	0.99 (0.79–1.00)	1.00 (0.93–1.00)	302.20 (14.3–6386.8)	0.01 (0.00–0.25)	52/1511
Stroke	0.99 (0.87–1.00)	0.99 (0.93–1.00)	172.40 (13.8–2151.4)	0.01 (0.00–0.14)	77/553
Small sample	0.99 (0.70–1.00)	0.94 (0.83–0.98)	17.56 (5.47–56.43)	0.01 (0.00–0.44)	84/592
Large sample	0.98 (0.85–1.00)	0.96 (0.92–0.98)	23.98 (11.61–49.54)	0.03 (0.00–0.17)	167/6368

ECG, electrocardiogram; LR indicates likelihood ratio; and PVI, pulmonary vein isolation.

indicates a statistical difference.

Figure 2

Forest plot of diagnostic accuracy of cardiac computed tomography (CCT) with early imaging method vs transesophageal echocardiography (TEE).

A, Sensitivity of CTT with the early imaging method vs TEE. B, Specificity of CCT with the early imaging method vs TEE. C, Posterior probability of CCT with the early imaging method vs TEE. D, The Deek method for assessment of publication bias. ESS, effective sample size; LR indicates likelihood ratio; Post Prob Neg, negative posterior probability; Post Prob Pos, positive posterior probability; and Prob, probability.

Figure 3

Forest plot of the diagnostic accuracy of cardiac computed tomography (CCT) with the delayed imaging method vs transesophageal echocardiography (TEE).

A, Sensitivity of CTT with the delayed imaging method vs TEE. B, Specificity of CCT with the delayed imaging method vs TEE. C, Posterior probability of CCT with the delayed imaging method vs TEE. D, The Deek method for assessment of publication bias. ESS, effective sample size; LR indicates likelihood ratio; Post Prob Neg, negative posterior probability; Post Prob Pos, positive posterior probability; and Prob, probability.

The incidence of LAA thrombus in the early imaging subgroup and delayed imaging group was 2.56% (120/4695) and 5.78% (131/2265), respectively. The positive posterior probability of the early imaging subgroup was 18.70%, and the negative posterior probability of the early imaging subgroup was 0.15% (Figure 2C). P=0.11 suggests no strong evidence of publication bias has been found (Figure 2D). The positive posterior probability of the delayed imaging subgroup was 95.51%, and the negative posterior probability of the delayed imaging subgroup was 0.12% (Figure 3C). P=0.14 suggests no strong evidence of publication bias has been found (Figure 3D).

Compared with the early imaging group, the delayed imaging method had a significantly higher LR+ and similar LR−, meaning that the delayed imaging method significantly improved the diagnostic accuracy. The positive posterior probability of the delayed imaging group was significantly higher than that of the early imaging group.

The hierarchical summary receiver operating characteristic curves of the early imaging group and delayed imaging group are shown in Figure 4A and 4B. The 95% prediction region and confidence region of the delayed imaging group (Figure 4A) were smaller than those of the early imaging group (Figure 4B), indicating that the diagnostic accuracy of the delayed imaging group was better than that of the early imaging group.

Figure 4

Hierarchical summary receiver operating characteristic (HSROC) curve of studies using the early imaging method (A) and delayed imaging method (B).

 

Analysis Based on Indications

Most patients in these studies were patients scheduled for PVI or patients experiencing stroke. Because these 2 indications have different LAA thrombus incidence (pretest probability), the posterior probability may be different. The incidence of LAA thrombus in the PVI with delayed imaging subgroup was 3.44% (52/1511). And the incidence of LAA thrombus in the stroke subgroup was 13.92% (77/553). Because the early imaging method has a low LR+ value, we mainly analyzed data from the delayed imaging group. The results are shown in Table 2 and Figures 5 and 6. The pooled sensitivity and specificity were similar between the 2 subgroups (Table 2 and Figures 5A and 5B and 6A and 6B). Although the estimated LR+ of the stroke subgroup was lower than that of the PVI subgroup, the positive posterior probability of the stroke subgroup was higher (96.00% versus 91.22%) because of the higher LAA thrombus incidence (Figures 5C and 6C). P=0.71 and P=0.09 suggested no strong evidence of publication bias has been found (Figures 5D and 6D).

Figure 5

Forest plot of the diagnostic accuracy of cardiac computed tomography (CCT) in patients with pulmonary vein isolation (PVI) using delayed imaging method vs transesophageal echocardiography (TEE).

A, Sensitivity of CCT in patients with PVI using the delayed imaging method vs TEE. B, Specificity of CCT in patients with PVI using the delayed imaging method vs TEE. C, Posterior probability of CCT in patients with PVI using the delayed imaging method vs TEE. D, The Deek method for assessment of publication bias. ESS, effective sample size; LR indicates likelihood ratio; Post Prob Neg, negative posterior probability; Post Prob Pos, positive posterior probability; and Prob, probability.

Figure 6

Forest plot of the diagnostic accuracy of cardiac computed tomography (CCT) in patients with stroke using delayed imaging method vs transesophageal echocardiography (TEE).

A, Sensitivity of CCT in patients with stroke using the delayed imaging method vs TEE. B, Specificity of CCT in patients with stroke using the delayed imaging method vs TEE. C, Posterior probability of CCT in patients with stroke using the delayed imaging method vs TEE. D, The Deek method for assessment of publication bias. ESS, effective sample size; LR indicates likelihood ratio; Post Prob Neg, negative posterior probability; Post Prob Pos, positive posterior probability; and Prob, probability.

Meta‐regression was performed to explore the source of heterogeneity. The results showed that the delayed imaging method, ECG‐gated method, and PVI may be the source of heterogeneity (Figure S1). When the delayed imaging method was defined as the interval between contrast injection and image capture of >1 minute, the heterogeneity of the delayed imaging subgroup decreased significantly (Figure S2).

DISCUSSION

In this comprehensive meta‐analysis of 27 studies, we assessed the diagnostic accuracy of CCT compared with TEE. The results demonstrated that CCT showed a high diagnostic accuracy for LAA thrombus detection when delayed imaging was used. In the delayed imaging subgroup, the positive posterior probability was 95.76%, and the negative posterior probability was 0.12%. Accurate identification of LAA thrombi is important for patients with atrial fibrillation and suspected cardiogenic stroke. For patients with atrial fibrillation, it can change the subsequent treatment strategy; for patients with suspected cardiogenic stroke, it can clarify a diagnosis. In the subgroup analysis based on these 2 indications, the positive posterior probabilities of PVI with delayed imaging and stroke subgroups were 91.22% and 96%, respectively. The negative posterior probabilities of these 2 subgroups were 0.34% and 0.14%, respectively. In the stroke subgroup, the LR+ value of the study with the early imaging method was 25. This relatively low LR+ value underestimated the positive posterior probability. Therefore, the actual positive posterior probability would be higher. If the LR+ value of PVI with the delayed imaging subgroup was used, the positive posterior probability would be 98%. This means that CCT with delayed imaging method has a better diagnostic accuracy for LAA thrombus detection in patients with stroke.

Although TEE is currently considered the gold standard for LAA thrombus detection, it is time‐consuming. In the past 2 decades, an increasing number of studies on the diagnostic accuracy of CCT for the detection of LAA thrombi have been performed. Most of these studies reported a high sensitivity and negative predictive value. LAA thrombus detection by CCT relies on filling defects. However, low blood flow velocity may also present as filling defects. It may be difficult to differentiate thrombi from low blood flow for early imaging method because the interval between contrast arrival and LAA image capture is too short. The delayed imaging method helps to differentiate thrombi and low blood flow. Our results showed that the delayed imaging method significantly improved the positive posterior probability compared with the early imaging method. In the subgroup analysis based on indications, our results showed that CCT with delayed imaging method has good diagnostic accuracy for LAA thrombus detection in patients with PVI and stroke. According to our results, we believe that CCT with a delayed imaging method is a reliable alternative tool for LAA thrombus detection. Furthermore, CCT has been recommended to assess left atrial and pulmonary vein anatomical features before PVI. In addition, the cost of CCT is only a few minutes. So, doing a delayed scan at the same time adds nominal radiation exposure (<1 millisievert) and allows a single test to perform both tasks (pulmonary vein assessment and rule out left atrial thrombus), saving the patient and health care from an excess TEE. TEE can be reserved for those with positive CCT to confirm the diagnosis of clot when needed. Given the high diagnostic accuracy and efficiency for LAA thrombus detection, TEE may be prevented in patients before PVI or in patients with stroke.

In previous studies, , , , the diagnostic accuracy of CCT was assessed by sensitivity, specificity, positive predictive value, and negative predictive value. However, the diagnostic accuracy of a test not only depends on sensitivity, specificity, positive predictive value, and negative predictive value but also depends on disease prevalence. The diagnostic accuracy of a test may vary in different populations because of different disease prevalence. The posterior probability calculated on the basis of disease prevalence may be more accurate. In our study, diagnostic accuracy was assessed by calculating the positive posterior probability and negative posterior probability based on the prevalence of LAA thrombi in each group. Our results did show the difference in posterior probability between patients before PVI and patients with stroke. Moreover, all studies included in previous meta‐analyses were conducted before 2014, and studies using delayed imaging method were relatively few. In this study, we included studies published until February 2021, including 11 studies using delayed imaging method.

There are some disadvantages in the use of CCT for the detection of LAA thrombi. First, the contrast agent used during CCT examination may cause contrast‐induced nephropathy and anaphylaxis. The risk of contrast‐induced nephropathy is relatively low in patients with normal renal function. Although the risk may increase in patients with chronic kidney disease, most kidney injuries are reversible. Second, patients are exposed to radiation. Currently, however, as technology has advanced, the level of radiation exposure is relatively low. CCT is most often done in <3 millisieverts, a marked reduction from early reports of ≥15 millisieverts in earlier studies.

There are some limitations to our study. First, the heterogeneity was high, and the results of the meta‐regression showed that the heterogeneity was from the delayed imaging method, ECG‐gated method, and PVI. Second, the reference standard was TEE, not surgical validation.

CONCLUSIONS

CCT with a delayed imaging method is superior to CCT with an early imaging method for LAA thrombus detection. It is a reliable alternative to TEE.

Sources of Funding

This work was supported by the National Natural Science Foundation of China (81800292).

Disclosures

None.

Tables S1–S4

Figures S1–S2

Click here for additional data file.