Diagnostic accuracy of ultrasound, CT and their combination in detecting cervical lymph node metastasis in patients with papillary thyroid cancer: a systematic review and meta-analysis.

OBJECTIVES: To determine the diagnostic accuracy of ultrasound (US), CT and their combination in detecting cervical lymph node metastasis (CLNM) in patients with papillary thyroid cancer (PTC).
METHODS: Medline (via PubMed), Web of Science, Embase were searched to identify studies published till 5 December 2021 that used US and CT to detect CLNM in patients with PTC. The primary outcomes were sensitivity, specificity and diagnostic ORs in neck-level-based (lymph nodes are analysed by neck level) or patient-based (lymph nodes are analysed by patient) analysis. Secondary outcomes were sensitivity, specificity and DORs in the central and lateral compartments.
RESULTS: Fourteen studies (6167 patients with 11 601 neck lymph nodes) met the inclusion criteria. Based on the neck-level-based analysis, the pooled sensitivity, specificity and DORs were 0.35 (95% CI 0.34 to 0.37), 0.95 (95% CI 0.94 to 0.95) and 13.94 (95% CI 9.34 to 20.82) for US, were 0.46 (95% CI 0.44 to 0.47), 0.88 (95% CI 0.87 to 0.89) and 7.24 (95% CI 5.46 to 9.62) for CT, were 0.51 (95% CI 0.49 to 0.52), 0.85 (95% CI 0.84 to 0.86), 6.01 (95% CI 3.84 to 9.40) for the combination of US and CT. In the patient-based analysis, the pooled estimates of sensitivity, specificity and DOR were 0.41 (95% CI 0.36 to 0.46), 0.92 (95% CI 0.89 to 0.94) and 7.56 (95% CI 4.08 to 14.01) for US, were 0.49 (0.44 to 0.54), 0.91 (0.89 to 0.94), 9.40 (5.79 to 15.27) for CT, and were 0.64 (95% CI 0.57 to 0.71), 0.83 (95% CI 0.77 to 0.88), 8.59 (95% CI 5.37 to 13.76) for the combination of US and CT. DISCUSSION: These findings suggest US, with a DOR almost twice that of CT in the neck-level-based analysis, was superior to CT in detecting CLNM in patients with PTC, especially in the lateral compartment. The combination of US and CT increased the sensitivity from 41%-49% for the individual modalities to 64% for combined modalities in the patient-based analysis. © Author(s) (or their employer(s)) 2022. Re-use permitted under CC BY-NC. No commercial re-use. See rights and permissions. Published by BMJ.

Only studies that analysed CT and ultrasound were included.

The analyses were performed based on the neck level and the patient level.

Heterogeneity was observed due to study design and timing of the examinations.

The use of CT for cervical lymph node metastasis screening is not recognised everywhere globally.

Introduction

Papillary thyroid carcinoma (PTC) is an endocrine neoplasia with a high incidence of lymphatic metastasis and is associated with regional recurrence.1–3 The incidence of cervical lymph node metastasis (CLNM) in patients with thyroid cancer has been reported to be 20%–90%.4 The presence of CLNM might increase the risk of locoregional recurrence after surgery,5 6 worsening prognosis and survival.7 Therefore, it is of great clinical importance to accurately evaluate CLNM and determine the extent of neck dissection.8 Although prophylactic central compartment neck (groups VI and VII) dissection (ipsilateral or bilateral) is recommended by the American Thyroid Association guidelines in patients with clinically positive central nodes, especially for those with advanced primary tumours, the information regarding prophylactic lateral compartment (groups I–V) neck dissection has not been clearly stated.8 Thus, the indications for neck dissection, especially the lateral compartment, should be carefully assessed as it might lead to severe postoperative complications.9

Preoperative staging with ultrasound (US) for cervical lymph nodes, including both central and lateral neck compartments, is the most widely accepted first imaging technique for patients with thyroid or suspicious malignancies cytologic or molecular findings. It can observe node enlargement, loss of fatty hilum, a rounded rather than oval shape, hyperechogenicity, cystic change, peripheral vascularity and calcifications, which are all indicators of malignant invasion.10 In addition, US is inexpensive, widely available, can be carried out bedside, and provide real-time imaging. Still, US is operator-dependent, and the images will vary depending on the angle and pressure of the probe on the neck. CT provides three-dimensional neck images that avoid operator-dependency issues. On the other hand, the analysis of each layer takes time and the use of contrast carries a risk of kidney injury. The preoperative use of CT with intravenous contrast is complementary to US in patients with advanced disease.8 11 12 A suspicious node on US can be confirmed by CT, and CT can detect nodes that were not visible because they were behind solid or air-containing structures or were not considered suspicious for various reasons. Although several studies have failed to prove the benefit of CT over US in detecting lateral lymph node metastasis,13–15 some studies suggested a superior diagnostic performance of the combination of preoperative US with CT over US alone.14 16–19

Some meta-analyses examining the diagnostic accuracy of US and CT in detecting CLNM in patients with PTC have been previously conducted.20–24 However, these meta-analyses studies integrated the findings of US and CT from different studies and populations.

To the best of our knowledge, no previous meta-analysis included studies that evaluated CLNM in patients with PTC using both US and CT, which could minimise the confounding effect of an operator in interpreting the diagnostic accuracy of preoperative imaging. This meta-analysis aimed to evaluate the sensitivity, specificity and diagnostic ORs of US, CT and their combination in detecting positive CLNM in patients with PTC based on the central and lateral neck levels and by using neck-level-based (lymph nodes are analysed by neck level) and patient-based (lymph nodes are analysed by patient, irrespective of the level) analyses.

Methods

Systematic literature research

This meta-analysis was reported according to the Preferred Reporting Items for Systematic reviews and Meta-Analysis guidelines.25 A systematic search of Medline (via PubMed), Web of Science and Embase was conducted to identify studies published up to 5 December 2021 that assessed the accuracy of US and CT in detecting CLNM in patients with PTC. The search strategy was developed in collaboration with a hospital librarian and included subject headings and text words: (“thyroid cancer” OR “thyroid carcinoma” OR “thyroid tumor” OR “papillary thyroid cancer” OR “thyroid neoplasm”) AND (“cervical lymph node” OR “neck lymph node”) AND (“metastasis” or “metastatic”) AND (“ultrasonography” or “ultrasound” or “US”) AND (computed tomography” or “CT”) (online supplemental table S1). The studies were initially screened by examining their titles and abstracts, and the full texts of potentially eligible studies were retrieved for further review. No language restriction was applied. A manual search of additional records and reference lists was also performed to include more relevant studies.

Study selection

The inclusion criteria of the studies were (1) prospective or retrospective studies that evaluated the diagnostic accuracy of both US and CT for detecting CLNM in patients with PTC, using neck-level-based or patient-based analysis; (2) studies with >10 patients; (3) studies with a reference standard of histopathology or cytology (the diagnostic gold standard was the pathological diagnosis of the resected lymph nodes); (4) studies that reported the absolute numbers of true-positive (TP), true-negative (TN), false-positive (FP), and false-negative (FN) results directly or derived from the reported data or communicated by the authors in response to our request; (5) studies published in English.

The exclusion criteria were (1) case reports, case series, review articles, pictorial essays, letters to editors, unpublished data, conference abstracts and proceedings on the topic of interest; (2) studies that used only US or only CT; (3) insufficient data regarding TP, FP, FN and TN; (4) duplicate publications using the same databases and studies; (5) if the patient population of one article is overlapping with the patient population of other or multiple articles, then the article with the largest sample size was included; (6) studies with less than 10 cases confirmed by the reference standard. One reader reviewed the full texts of the candidate articles and selected those that met the inclusion criteria. A second reader reviewed the process of the inclusion of articles in the meta-analysis. No inter-reader disagreements were observed.

Primary and secondary outcomes

The primary outcomes were sensitivity, specificity and DORs（Diagnostic Odds Ratios） in a neck-level-based or patient-based analysis. Secondary outcomes were sensitivity, specificity and DORs in central and lateral compartments in neck-level-based or patient-based analysis.

Data extraction and quality assessment

Two reviewers independently performed the data extraction. Data such as study characteristics, clinical and patient characteristics, reference standard or standards, cervical lymph node compartment, technical characteristics of CT and US and contrast enhancement, the definition of CLNM according to CT and US image findings, and the diagnostic performance of CT and US, such as TP, FP, FN, TN were obtained from each study.

Two reviewers who were not blinded to the journal names, author names, and year of publication assessed the methodologic and reporting quality of each study by using the Quality Assessment of Diagnostic Accuracy Studies 2.26 Each study was independently assessed by two reviewers after a tutorial meeting on the guidelines for interpreting the items. Any disagreements were resolved by discussion with an experienced third reviewer.

Statistical analysis

The pooled sensitivity, specificity, diagnostic OR, positive likelihood ratio (LR+), and negative LR (LR−) were calculated for US and CT in a neck-level-based analysis (at neck level, central neck level and lateral neck level) and a patient-based analysis (patient level, central patient level and lateral level). The heterogeneity of pooled sensitivity, specificity, diagnostic OR, LR+ and LR− was measured by the inconsistency (I). Heterogeneity in the included articles was defined as small I2 <25%, moderate I 25%–50%, and obvious I2 >50%. If heterogeneity was detected (p<0.10 or I2 ≥50%), a random-effects model was applied; otherwise, a fixed-effects model was used. A bivariate logistic regression model was used for meta-analysis of diagnostic test accuracy,27 and forest plots were created. The pooling of sensitivity, specificity, diagnostic OR, LR+ and LR− was performed using the Meta-Disc software (V.1.4, Madrid, Spain). Forest plots and summary receiver operating characteristic curves were obtained using RevMan V.5.3. A p value of <0.05 was considered statistically significant.

Patient and public involvement

The patients and the public were not involved in this study.

Results

Characteristics of included studies

The initial literature search yielded 1135 potential studies for this meta-analysis. A total of 449 articles were screened after removing the duplicates. Of these, 372 studies were excluded after reviewing the titles and abstracts, and 63 articles were excluded after reviewing the full texts (figure 1). Fourteen studies were ultimately selected for inclusion13 14 16–19 28–35: 10 studies used a neck-level-based analysis, two studies used a patient-based analysis, and two studies used both. Five studies reported the diagnostic performance by combining both US and CT.14 16–19 A total of 6167 patients with 11 601 neck lymph nodes were included, and all patients were diagnosed with PTC except one who was diagnosed with medullary thyroid cancer. The earliest study was started in 1997, whereas the latest one was started in 2012. The median number of patients per study was 171 (range 20–3668), while the median number of lymph nodes per study was 331 (range 107–6557). Eleven were retrospective studies, and 3 were prospective studies; 13 studies were performed preoperatively, and 1 study was performed postoperatively. Twelve, one and one were conducted in Korea, the USA and Japan, respectively (table 1). The studies included in this meta-analysis were of moderate quality (online supplemental sfigures 1,2).

Figure 1

Flow chart of the literature search process.

Table 1

Characteristics of included studies

Study	Country	Study design	Timing of imaging	Duration of patient recruitment	Sample (n) (males/ females)	Age (range)	No of lymph node	Diagnosis	Analysis methods
Jeong et al 200628	Korea	Retrospective	Preoperative	July 2004-March 2005	26 (7/19)	44 (17–73)	312	All PTC	L
Kim et al 200816	Korea	Retrospective	Preoperative	April 2006–October 2006	165 (25/140)	48 (16–78)	277	All PTC	L+P
Ahn et al 200813	Korea	Retrospective	Preoperative	January 2005–December 2005	37 (7/30)	47(20-68)	181	All PTC	L
Choi et al 200914	Korea	Retrospective	Preoperative	February 2006–April 2007	299 (44/255)	45 (20–74)	352	All PTC	L
Choi et al 201029	Korea	Retrospective	Preoperative	January 2007–December 2008	589 (121/468)	46	589	All PTC	P
Morita et al 201030	Japan	Prospective	Preoperative	January 2007–December 2009	74 (12/62)	66 (16–84)	349	All PTC	L
Yoon et al 201131	Korea	Retrospective	Preoperative	February 2007–December 2007	113 (16/97)	46 (15–83)	122	All PTC	L
Seo et al 201232	Korea	Retrospective	Postoperative	August 2008–August 2011	20 (4:16)	49.8	107	19 PTC, 1 MTC	L
Lee et al 201317	Korea	Retrospective	Preoperative	January 2007–May 2010	252 (45/207)	49 (15–82)	558	All PTC	L+P
Lesnik et al 201433	USA	Prospective	Preoperative	2003–2008	95 (NA)	NA	196	All PTC	L
Na et al 201519	Korea	Retrospective	Preoperative	March 2011–February 2012	176 (44/132)	43 (23–74)	352	All PTC	P
Kim et al 201734	Korea	Retrospective	Preoperative	January 1997–June 2015	3668 (NA)	NA	6557	All PTC	L
Eun et al 201835	Korea	Retrospective	Preoperative	January 2013–December 2015	302 (76:226)	44	308	All PTC	L
Lee et al 201818	Korea	Prospective	Preoperative	November 2011–December 2012	351 (78:273)	47.1	801	All PTC	L

L, neck-level-based analysis; MTC, medullary thyroid carcinoma; NA, not available; P, patient-based analysis; PTC, papillary thyroid carcinoma.

Neck-level-based diagnostic accuracy of US and CT

Eleven studies used both CT and US for detecting CLNM in patients with PTC, and five of them assessed the diagnostic accuracy of the combination of CT and US. The pooled sensitivity, specificity, diagnostic OR, LR+ and LR− were 0.35 (95% CI 0.34 to 0.37), 0.95 (95% CI 0.94 to 0.95), 13.94 (95% CI 9.34 to 20.82), 6.79 (95% CI 4.79 to 9.63) and 0.50 (95% CI 0.41 to 0.60) for US, were 0.46 (95% CI 0.44 to 0.47), 0.88 (95% CI 0.87 to 0.89), 7.24 (95% CI 5.46 to 9.62), 3.77 (95% CI 2.08 to 6.84) and 0.52 (95% CI 0.45 to 0.61) for CT, and were 0.51 (95% CI 0.49 to 0.52), 0.85 (95% CI 0.84 to 0.86), 6.01 (95% CI 3.84 to 9.40), 3.04 (95% CI 1.93 to 4.80), and 0.52 (95% CI 0.45 to 0.60) for the combination of US and CT, with marked heterogeneity (figures 2 and 3, table 2).

Table 2

Pooled estimates of sensitivity, specificity, diagnostic OR, LR+, LR−

	Studies, n	Sensitivity (95% CI)	Specificity (95% CI)	Diagnostic OR (95% CI)	LR+ (95% CI)	LR− (95% CI)
Diagnostic accuracy of CT or US on neck level
All neck level
US	11	0.35 (0.34 to 0.37, 97.5)	0.95 (0.94 to 0.95, 90.8)	13.94 (9.34 to 20.82, 81.0)	6.79 (4.79 to 9.63, 84.1)	0.50 (0.41 to 0.60, 95.2)
CT	11	0.46 (0.44 to 0.47, 97.6)	0.88 (0.87 to 0.89, 97.9)	7.24 (5.46 to 9.62, 72.2)	3.77 (2.08 to 6.84, 98.5)	0.52 (0.45 to 0.61,89.7)
US/CT	5	0.51 (0.49 to 0.52, 92.8)	0.85 (0.84 to 0.86, 97.8)	6.01 (3.84 to 9.40, 89.2)	3.04 (1.93 to 4.80, 96.3)	0.52 (0.45 to 0.60, 78.8)
Central neck level
US	4	0.28 (0.24 to 0.32, 94.3)	0.97 (0.96 to 0.98, 53.0)	14.07 (6.66 to 29.71, 53.1)	14.07 (6.66 to 29.71, 53.1)	14.07 (6.66 to 29.71, 53.1)
CT	4	0.32 (0.28 to 0.36, 88.2)	0.89 (0.86 to 0.91, 84.6)	5.48 (2.15 to 13.98, 84.3)	3.71 (1.79 to 7.66, 83.0)	0.74 (0.62 to 0.89, 86.7)
US/CT	3	0.40 (0.35 to 0.45, 82.3)	0.85 (0.82 to 0.88, 37.0)	4.32 (2.09 to 8.92, 81.0)	2.85 (1.75 to 4.65, 77.5)	0.67 (0.52 to 0.86, 83.9)
Lateral neck level
US	6	0.74 (0.69 to 0.78, 78.1)	0.92 (0.90 to 0.94, 89.6)	24.41 (11.16 to 53.42, 71.9)	6.67 (2.91 to 15.30, 90.5)	0.35 (0.28 to 0.43, 30.1)
CT	6	0.73 (0.68 to 0.77, 90.7)	0.89 (0.87 to 0.91, 91.8)	15.55 (7.98 to 30.32, 64.6)	5.54 (2.95 to 10.39, 82.9)	0.35 (0.21 to 0.59, 91.4)
US/CT	4	0.88 (0.83 to 0.91, 64.1)	0.79 (0.73 to 0.84, 89.6)	22.59 (11.29 to 45.19, 46.6)	3.31 (1.53 to 7.17, 91.1)	0.19 (0.14 to 0.25, 0)
Diagnostic accuracy of CT or US on patient level
All patient level
US	4	0.41 (0.36 to 0.46, 95.5)	0.92 (0.89 to 0.94, 0)	7.56 (4.08 to 14.01, 51.3)	4.48 (3.31 to 6.05, 0)	0.65 (0.53 to 0.80, 75.5)
CT	4	0.49 (0.44 to 0.54, 95.9)	0.91 (0.89 to 0.94, 66.8)	9.40 (5.79 to 15.27, 34.4)	4.84 (3.66 to 6.39, 0)	0.53 (0.37 to 0.75, 85.6)
US/CT	2	0.64 (0.57 to 0.71, 0)	0.83 (0.77 to 0.88, 0)	8.59 (5.37 to 13.76, 0)	3.71 (2.72 to 5.08, 0)	0.43 (0.36 to 0.53, 0)
Central patient level
US	2	0.21 (0.16 to 0.28, 24.8)	0.95 (0.91 to 0.97, 0)	4.53 (2.34 to 8.77, 0)	3.78 (2.08 to 6.86, 0)	0.84 (0.76 to 0.93, 36.4)
CT	2	0.38 (0.32 to 0.46, 92.6)	0.90 (0.85 to 0.93, 87.2)	5.02 (0.46 to 54.54, 94.4)	3.52 (0.52 to 23.84, 93.8)	0.71 (0.41 to 1.22, 95.3)
US/CT	2	0.47 (0.39 to 0.54, 80.8)	0.85 (0.80 to 0.89, 0)	4.88 (2.58 to 9.23, 46.4)	3.14 (2.23 to 4.41, 0)	0.64 (0.47 to 0.86, 78.4)
Lateral patient level
US	2	0.87 (0.74 to 0.95, 88.2)	0.89 (0.83 to 0.93, 91.4)	20.11 (6.77 to 59.70, 0)	3.58 (0.85 to 15.16, 91.0)	0.22 (0.05 to 1.08, 58.0)
CT	2	0.92 (0.81 to 0.98, 88.6)	0.88 (0.83 to 0.93, 96.8)	36.88 (11.40 to 119.35, 0)	3.44 (0.29 to 40.77, 98.3)	0.23 (0.10 to 0.52, 0)
US/CT	2	0.98 (0.89 to 0.99, 58.9)	0.92 (0.87 to 0.96, 97.5)	78.10 (2.82 to 2160.4, 65.2)	5.30 (0.15 to 186.11, 98.6)	0.08 (0.02 to 0.41, 0)

Results are presented as n (95% CI, I2 %).

LR+, positive likelihood ratio; LR-, negative likelihood ratio; US, ultrasound.

Forest plots for the sensitivities and specificities of US, CT, and combination in neck-level-based analysis. FN, false-negative; FP, false-positive; TN, true-negative; TP, true-positive; US, ultrasound.

SROC of US, CT and combination in detecting CLNM in patients with PTC in neck-level-based analysis. CLNM, cervical lymph node metastasis; PTC, papillary thyroid cancer; SROC, summary receiver operating characteristic; US, ultrasound.

Subgroup analyses of central and lateral neck levels were performed to investigate the effects of cervical lymph node compartment based on the diagnostic accuracy of US and CT. The subgroup analysis of the central neck level revealed that the pooled sensitivity, specificity and DOR of US were 0.28 (95% CI 0.24 to 0.32), 0.97 (95% CI 0.96 to 0.98) and 14.07 (95% CI 6.66 to 29.71) from four studies. For CT, the pooled sensitivity, specificity and DOR were 0.32 (95% CI 0.28 to 0.36), 0.89 (95% CI 0.86 to 0.91) and 5.48 (95% CI 2.15 to 13.98) from four studies. The pooled sensitivity, specificity and DOR of the combination of US and CT were 0.40 (95% CI 0.35 to 0.45), 0.85 (95% CI 0.82 to 0.88) and 4.32 (95% CI 2.09 to 8.92) from three studies (table 2, online supplemental sfigures 3,4).

In contrast, the subgroup analysis of the lateral neck level revealed that the pooled sensitivity, specificity and DOR of US were 0.74 (95% CI 0.69 to 0.78), 0.92 (95% CI 0.90 to 0.94) and 24.41 (95% CI 11.16 to 53.42) from six studies; the values for CT were 0.73 (95% CI 0.68 to 0.77), 0.89 (95% CI 0.87 to 0.91) and 15.55 (95% CI 7.98 to 30.32) from six studies; the values for the combination of US and CT were 0.88 (95% CI 0.83 to 0.91), 0.79 (95% CI 0.73 to 0.84) and 22.59 (95% CI 11.29 to 45.19) from four studies (table 2, online supplemental sfigures 5,6).

Patient-based diagnostic accuracy of US and CT

Four studies included both US and CT in detecting CLNM in patients with PTC, and two of them assessed the diagnostic accuracy by combining both CT and US. The pooled estimates of sensitivity, specificity and DOR of US were 0.41 (95% CI 0.36 to 0.46), 0.92 (95% CI 0.89 to 0.94) and 7.56 (95% CI 4.08 to 14.01); the values for CT were 0.49 (95% CI 0.44 to 0.54), 0.91 (95% CI 0.89 to 0.94) and 9.40 (95% CI 5.79 to 15.27); the values for the combination of US and CT were 0.64 (95% CI 0.57 to 0.71), 0.83 (95% CI 0.77 to 0.88) and 8.59 (95% CI 5.37 to 13.76) (figures 4 and 5, table 2).

Forest plots for the sensitivities and specificities of US, CT, and combination in patient-based analysis. FN, false-negative; FP, false-positive; TN, true-negative; TP, true-positive; US, ultrasound.

SROC of US, CT and combination in detecting CLNM in patients with PTC in patient-based analysis. CLNM, cervical lymph node metastasis; PTC, papillary thyroid cancer; SROC, summary receiver operating characteristic; US, ultrasound.

Only two studies assessed the diagnostic accuracy of US, CT and their combination on a patient basis. On the patient level, the pooled estimates of sensitivity, specificity and DOR were 0.21 (95% CI 0.16 to 0.28), 0.95 (95% CI 0.91 to 0.97) and 4.53 (95% CI 2.34 to 8.77) for US, were 0.38 (95% CI 0.32 to 0.46), 0.90 (95% CI 0.85 to 0.93) and 5.02 (95% CI 0.46 to 54.54) for CT, and were 0.47 (95% CI 0.39 to 0.54), 0.85 (95% CI 0.80 to 0.89) and 4.88 (95% CI 2.58 to 9.23) for the combination of CT and US (table 2, online supplemental sfigures 7,8).

In contrast, the pooled estimates of sensitivity, specificity, and DOR of US were 0.87 (95% CI 0.74 to 0.95), 0.89 (95% CI 0.83 to 0.93) and 20.11 (95% CI 6.77 to 59.70); the values for CT were 0.92 (95% CI 0.81 to 0.98), 0.88 (95% CI 0.83 to 0.93) and 36.88 (95% CI 11.40 to 119.35); the values for the combination of US and CT were 0.98 (95% CI 0.89 to 0.99), 0.92 (95% CI 0.87 to 0.96) and 78.10 (95% CI 2.82 to 2160.4) (table 2, online supplemental sfigures 9,10).

Discussion

This meta-analysis revealed that the DORs of US in the neck-level-based analysis was higher than for CT or their combination on the central, lateral and neck levels. Differentiated thyroid carcinoma, particularly PTC, involves CLNMs in 20%–50% of the patients,36–39 which could prevent small and intrathyroidal primary tumours.40 Still, the clinical implications of macrometastases (≥2 mm) are more significant than micrometastases, in which 90% of patients might reach the criteria according to the sensitivity of the imaging methods.41 42 The combination of US features might increase the likelihood of detecting CLNM as several US features are suggestive of metastatic lymph nodes, including enlargement, loss of fatty hilum, a rounded rather than oval shape, hyperechogenicity, cystic change, peripheral vascularity and calcifications.10 The preoperative US identifies lymph node or soft-tissue metastases in up to 39% of patients who had no physical examination43 and changed the operative management in 23% of patients.44

Previous meta-analyses examined CT and US. Suh et al20 and Cho et al21 demonstrated the value of CT for CNLM but did not include US. Raijmakers et al22 only examined the detection of the sentinel lymph node. Wu et al23 and Zhao et al24 examined the value of US for CLNMs but did not include CT. Therefore, these studies did not examine CT and US simultaneously. Our data found that the DORs of CT were higher than US and the combination, and the DORs of the combination remained higher than US and CT by patient-based analysis. This was reasonable because the sensitivity of CT in the patient-based analysis was higher than that of US in the central, lateral and patient analyses. This result might still need further investigation because of the inclusion of a small number of studies in the subgroup analysis. The operator-independent CT could be used as an adjunct in imaging deep anatomic structures, including the mediastinum, infraclavicular, retropharyngeal and parapharyngeal regions and the structures that are acoustically shadowed by bone or air. In addition, preoperative knowledge on the extent of laryngeal, tracheal and oesophageal involvement, as well as bulky nodal disease from neck CT with contrast, significantly influences the surgical plan by indicating the need for sternotomy, tracheal or laryngeal resection and reconstruction.45

The results suggested that the sensitivity on the lateral compartment tended to be higher than for the central compartment regardless of the use of US, CT or their combination in the neck-level-based and patient-based analyses. The location of the lymph nodes helps in decision-making as most of the metastatic nodes are found in the lower third of the neck, and reactive enlarged lymph nodes are found in the upper part of the neck.46 Besides, the lateral compartment should be carefully evaluated for skip metastases located in the upper pole or are ≤1 cm in diameter.47 For patients who had preoperative CT and US and subsequently underwent total thyroidectomy and neck dissection, the sensitivity of CT was much better than US for evaluating CLNM on the neck level, but the sensitivity, specificity, and DORs for the lateral neck level tended to be higher than those of the central neck level for both CT and US.13 Dual-energy CT (DECT) for assessing CLNM in patients with PTC was not included in this meta-analysis as it can generate iodine-based material decomposition images and spectral HU curve.48–50 In accordance with the findings from CT, combined gemstone spectral image parameters from DECT also demonstrated better diagnostic accuracy of CLNM in patients with PTC when compared with those that are obtained by combining the US morphological parameters especially in the lateral compartment.50

Our findings revealed that compared with US or CT alone, the combination of both US and CT demonstrated higher sensitivity, that is, a meta-analytic summary sensitivity of 0.51 (95% CI 0.49 to 0.52) and 0.64 (95% CI 0.57 to 0.71), and a lower specificity, that is, a meta-analytic summary specificity of 0.85 (95% CI 0.84 to 0.86) and 0.83 (95% CI 0.77 to 0.88) for evaluating CLNM in patients with PTC using neck-level-based and patient-based analyses, respectively. In patients undergoing primary and revision surgical treatment for PTC, combined preoperative mapping with US and CT yielded significantly higher sensitivity for detecting macroscopic lymph nodes in both lateral and central neck, especially in the central neck.33

It should be noted that the study has strengths. First, Boolean operatives of “AND” rather than “OR” were used for combined datasets for all studies. Namely, only studies of direct head-to-head comparison by US, CT and combination of both in the same patient population were included in this meta-analysis, avoiding bias due to differences in patient and institutional factors. Second, a meta-analysis of the included studies was performed by using neck-level-based and patient-based analyses and on all, central and lateral neck levels. Lastly, our data suggested that future follow-up studies should be performed to determine the comparative role of US and CT in identifying FN nodes that are not biopsied or excised.

Despite great clinical significance, there are several limitations in the current meta-analysis that are mostly associated with the available data and heterogeneity of design, interpretation of results, and reporting of data in primary studies. First, the sources of heterogeneity among primary studies in meta-analyses have been reported by several previous studies, which included contrast amount, scan phase and reconstruction slice thickness for CT,21 and the criteria of lymph node diameter and vascular flow for US.24 Second, the literature included is limited due to the study design and timing of imaging. Eleven of the 14 studies (78.6%) were retrospective, and 1 of the 14 studies was a postoperative imaging study. A large proportion of retrospective studies might increase the sensitivity of CT and US. Twelve of the 14 studies were conducted in Korea, and so ethnic factors might affect the results of this meta-analysis. Thus, the complementary use of CT might be routine in Korea but not necessarily applicable to other parts of the world, especially in developing countries. Third, modern high-resolution US transducers have a lateral resolution of 2 mm, which is not feasible for CT, allowing for the detection of small nodes and the presence of microcalcification. The included CT studies might not be comparable from one study to another, particularly over the decade, depending on the equipment, slice thickness, amount of contrast injected, etc. Fourth, 4 of the 14 included studies were with patient-based results, and 12 of 14 studies were of suboptimal quality, and no definite recommendation could be drawn from this study. Finally, MRI, US-guided FNA and PET-CT were not included in the meta-analysis to directly compare CT and US, although they also play complementary roles in managing CLNMs in PTC.

Despite these potential drawbacks, this meta-analysis demonstrated the unique complementary value of CT secondary to US in detecting CLNMs in patients with PTC in the patient-based analysis. More importantly, the choice of a diagnostic test should be tailored to have feasible access to these imaging modalities at individual healthcare centres.

Conclusion

These findings suggest that US, with a DOR of almost twice that for CT in the neck-level-based analysis, was superior to CT in detecting CLNM in patients with PTC, especially in the lateral compartment. The combination of US and CT increased the sensitivity from 41% to 49% for the individual modalities to 64% for combined modalities in the patient-based analysis. CT might be valid a candidate imaging technique secondary to US in the management of CLNM in patients with PTC.