PDF
Citation
Print
INTRODUCTION
Core needle biopsy (CNB) has been used for thyroid nodules in Korea since the mid-to-late 2000s, and a paper published in 2011 was the first to assess the role of CNB in thyroid nodules [1]. Thereafter, various hospitals in Korea—especially Asan Medical Center—have published studies suggesting that CNB is advantageous over fine-needle aspiration (FNA) cytology. However, at academic meetings, we have encountered discordant opinions across hospitals regarding the role of CNB, with some practitioners feeling that CNB offers enough advantages to be considered as the first-line modality for biopsy and others believing that CNB confers no advantage over FNA. Previous meta-analyses of CNB evaluated its diagnostic accuracy based on its sensitivity or specificity for the diagnosis of malignancy or focused on specific conditions, such as nodules with non-diagnostic or atypia of undetermined significance/follicular lesion of undetermined significance (AUS/FLUS) results on previous FNA [2-11]. Therefore, there are limitations in the degree to which we understand the difference between CNB and FNA in real clinical settings based on those meta-analyses.
This systematic review and meta-analysis included all articles on CNB and/or FNA in thyroid nodules published by Korean authors to analyze the frequency of various Bethesda System for Reporting Thyroid Cytopathology (TBSRTC) categories. The purpose of this study was, through a comprehensive review of consecutive data from Korea, to determine the circumstances in which hospitals favor CNB or FNA and to compare differences in results between CNB and FNA, with the goal of clarifying expectations related to the choice of a biopsy modality in Korea.
MATERIALS AND METHODS
Literature search
The Embase database was searched on September 18, 2019, using the terms “‘thyroid’/exp OR thyroid” AND (“fine needle aspiration” OR “core needle biopsy”) AND “Korea.”
Inclusion criteria
The inclusion criteria were: (1) full-text original articles in English published by Korean authors, (2) presenting data from a specific hospital, (3) frequency of diagnoses reported using consecutive cases, and (4) results on thyroid nodules, especially differentiated thyroid carcinoma. Conference articles, reviews or meta-analyses, and articles reporting multi-center data were excluded. FNA findings were analyzed based on the TBSRTC, while CNB data were analyzed based on the proposal of the Korean Endocrine Pathology Thyroid Core Needle Biopsy Study Group [12]. However, the reporting system proposed by the Korean Study Group uses the same system as TBSRTC, and the only difference is that AUS/FLUS in TBSRTC is expressed as an indeterminate lesion in the Korean Study Group proposal. In this paper, indeterminate results on CNB are collectively referred to as AUS/FLUS.
Data analysis
A proportional meta-analysis was performed to compare the frequencies of diagnoses, and multiple comparisons were performed between six subgroups. The Bonferroni correction was applied to determine the P-values of inter-group differences. The proportional meta-analysis was performed using a comprehensive meta-analysis (Biostat, Englewood, NJ, USA). Forest plots were constructed using Excel (Microsoft Corp., Redmond, WA, USA).
RESULTS
Articles published by Korean authors on FNA or CNB in thyroid diseases
After removing duplicates, 965 records were retrieved from Embase. Screening of titles excluded 721 irrelevant records, leaving 244 records for abstract or full-text review. Before further exclusion from the qualitative synthesis, the frequency of publications on FNA and CNB was analyzed by hospital (Table 1). In total, 204 articles on FNA and 54 on CNB were published by authors from 32 hospitals. Papers on CNB were published by authors from only 11 of those 32 hospitals (34.4%), suggesting that specific hospitals favored CNB. After excluding 188 additional records, the meta-analysis finally included 56 records. The reasons for exclusion and the number of records included in the meta-analysis are described in Fig. 1.
Comparison of CNB and repeated FNA in thyroid nodules with non-diagnostic or AUS/FLUS results on initial FNA
Eleven studies reported biopsy results after initial non-diagnostic or AUS/FLUS results [1,13-22]. Of those papers, the study period of Yeon et al. [20] overlapped with that of another paper [13] published using data from the same hospital and was excluded.
CNB and repeated FNA in nodules with non-diagnostic initial FNA findings
Three papers [13,16,17] compared CNB and repeated FNA of nodules with non-diagnostic initial FNA results, while one paper [21] reported only the results of repeated FNA. The study details are summarized in Table 2. The proportional meta-analysis of the diagnoses for each criterion showed a significantly lower pooled proportion of non-diagnostic results for CNB than for FNA (1.6% vs. 34.4%, P<0.001) (Fig. 2A). The proportions of other diagnostic categories did not differ significantly between CNB and repeated FNA (Supplementary Fig. 1). Although the proportions of follicular neoplasm/suspicious for follicular neoplasm diagnoses (FN/SFN; pooled proportion: 5.8% in CNB and 1.1% in FNA, P-value before the Bonferroni correction=0.021) and malignancy (pooled proportion: 16.6% in CNB and 5.0% in FNA, P-value before the Bonferroni correction=0.022) were higher in CNB than those in FNA, the differences were not statistically significant after the Bonferroni correction.
CNB and repeated FNA in nodules with AUS/FLUS findings on initial FNA
Seven papers [1,14,15,17-19,22] reported CNB or FNA results in nodules initially diagnosed as AUS/FLUS (Table 3). Among these seven papers, two reported only the results of FNA [15,19]. Others compared CNB and repeated FNA results and assessed the usefulness of CNB. However, Yoon et al. [22] concluded that despite the lower rates of inconclusive results from CNB than from repeated FNA, CNB did not decrease the rate of diagnostic surgery, reflecting a weakness of CNB.
The proportional meta-analysis showed significantly lower proportions of non-diagnostic findings on CNB than on FNA (pooled proportions: 3.0% vs. 11.8%, P=0.002) (Fig. 2B). The proportion of AUS/FLUS findings on CNB was lower than that observed for repeated FNA (pooled proportions: 24.1% vs. 35.2%) (Fig. 2C) but the difference was not statistically significant (P=0.379) due to the wide variability in CNB. Meanwhile, the proportion of FN/SFN diagnoses was significantly higher for CNB than for repeated FNA (8.0% vs. 1.7%, P=0.017) (Fig. 2D). The proportions of other diagnostic categories did not significantly differ between CNB and repeated FNA (Supplementary Fig. 2).
Proportions of diagnoses made using CNB or FNA in consecutive patients
Among the 56 studies included in the qualitative analysis, 36 included data on the proportions of diagnoses following the TBSRTC or equivalent criteria [23-57]. Eight papers [24,36,39,40,42,44,51,55] were excluded because their study periods overlapped with those of other papers reporting data from the same hospital; thus, the analysis included 28 papers. Five papers reported diagnostic findings for both CNB and FNA [30,33,43,54,56], while four papers [23,26,27,31] reported CNB results only. The other papers reported FNA data. Table 4 presents a detailed summary of each paper.
Differences in the proportions of diagnoses on consecutive FNA between hospitals favoring or not favoring CNB
Twenty-four papers reporting the proportions of diagnoses made using FNA published from 14 hospitals were included. Seven of these hospitals also published data about CNB and reported its advantages over FNA; these hospitals were categorized as favoring CNB. One hospital focused on the weaknesses or disadvantages of CNB in four papers [22,43,58,59]; despite publishing papers on CNB, we categorized this hospital as not favoring CNB. Head and neck surgeons working at five other hospitals without a paper on CNB were contacted to obtain information on their use of CNB, and they confirmed that they rarely used CNB in clinical settings. Therefore, data from 14 papers [30,32-35,41,46,48-50,52,54,56,57] published by authors from eight hospitals favoring CNB were compared to data from 10 papers [25, 28,29,37,38,43,45,47,53,60] published by authors from six hospitals that did not favor CNB to identify differences in the proportions of diagnoses according to the TBSRTC. Fig. 3A shows the differences in the pooled estimates obtained by applying the TBSRTC criteria between hospitals that did or did not favor CNB. The most significant between-group difference was the higher frequency of AUS/FLUS findings at hospitals favoring CNB than at hospitals not favoring CNB (12.3% vs. 5.1%, P<0.001). In addition, benign diagnoses were more frequent at hospitals not favoring CNB than at those favoring CNB (60.5% vs. 51.3%, P=0.028). The proportions of other diagnoses did not significantly differ between these two groups of hospitals. Forest plots for each diagnosis are shown in Supplementary Fig. 3.
Differences in the proportions of diagnoses between consecutive FNA and CNB
Consecutive data on FNA (24 papers) and CNB (nine papers) were compared in a proportional meta-analysis. The frequency of the proportions of each TBSRTC category differed significantly between FNA and CNB (Fig. 3B). Non-diagnostic, benign, and suspicious for malignancy findings were significantly more common on FNA than on CNB. However, CNB had significantly higher proportions of AUS/FLUS, FN/SFN, and malignancy diagnoses than FNA. The forest plots for the proportional analysis are presented in Supplementary Fig. 4.
Proportional differences in AUS/FLUS diagnoses between FNA and CNB
AUS/FLUS includes various conditions for which cellular atypia/architectural atypia is a representative subcategory. To understand the reasons for the higher frequency of AUS/FLUS diagnoses on CNB, we analyzed proportional differences in subcategories of diagnoses and compared the malignancy risks of cellular/architectural atypia and FN/SFN diagnosed by FNA or CNB. Eight publications included data on the AUS/FLUS subcategory or the role of CNB in FN/SFN [18,24,30,34,61-65]. Park et al. [63] was excluded from the present meta-analysis due to overlap of the study period with other papers analyzing data from the same hospital; thus, the analysis included eight papers (Table 5).
Cellular atypia versus architectural atypia
This analysis included seven papers. Ahn et al. [24] included both FNA and CNB results, Chung et al. [30] reported CNB results, and the others included FNA data [18,34,61,62,64]. Architectural atypia comprised approximately 71.9% of AUS/FLUS cases diagnosed by CNB; in contrast, cellular atypia comprised a pooled rate of 70.1% of AUS/FLUS cases diagnosed by FNA (Fig. 4A).
Malignancy risk of cellular or architectural atypia and FN/SFN diagnosed by CNB or FNA
The malignancy rate of cellular or architectural atypia was analyzed based on six papers [24,30,34,61,62,64]. The pooled risk of malignancy in nodules with both cellular and architectural atypia did not differ significantly depending on whether they were diagnosed using CNB or FNA. The pooled malignancy risk of cellular atypia in CNB and FNA was 27.0% and 32.8%, respectively (Fig. 4B) and 12.7% and 9.6% for architectural atypia, respectively (Fig. 4C). Three papers reported the pooled risk of malignancy of FN/SFN [24,30,65], also without significant differences (34.2% for CNB and 28.5% for FNA) (Fig. 4D).
Complications of CNB and FNA
Nine papers [23,26,36,54,55,58,66-68] reported complications after CNB or FNA. Three papers [26,55,67] included cases overlapping with other papers from the same institution; thus, the final analysis included six papers. While the proportional meta-analysis showed that the pooled complication rate of CNB (1.5%) was higher than that of FNA (0.7%), the difference was not statistically significant (P=0.351) (Supplementary Fig. 5). Table 6 lists the complications observed after FNA and CNB. Hematoma was the only reported complication of FNA. However, in patients who underwent CNB, although very rare, more serious complications were reported, including pseudoaneurysm due to carotid or tracheal injuries.
DISCUSSION
The 2015 guideline of the American Thyroid Association briefly noted papers on the usefulness of CNB for the description of non-diagnostic or AUS/FLUS results [69]. Many papers have reported the results of CNB in thyroid nodules; a PubMed search with the term ‘core needle biopsy[ti] thyroid’ resulted in 117 records, of which 68 (58.1%) were from Korea. However, even with the high interest in CNB in Korea, perceptions of CNB vary among hospitals. Some hospitals apply CNB as an important testing method, while others think it is unnecessary. This systemic review evaluated how hospitals in Korea viewed CNB for thyroid nodules. As shown in Table 1, authors from many hospitals have published studies on FNA, while researchers from only one-third of hospitals (34.4%) have published studies on CNB, illustrating the differences in perceptions of CNB in Korea.
Advantages of CNB in nodules initially classified as non-diagnostic or AUS/FLUS on FNA
The use of CNB as a secondary test in nodules initially classified as non-diagnostic or AUS/FLUS by FNA is a relatively commonly accepted indication. In addition, meta-analyses have shown that CNB of nodules with initial non-diagnostic or inconclusive results by FNA yields significantly fewer non-diagnostic or inconclusive findings than obtained using FNA [6,7,9]. In the present meta-analysis, CNB showed significantly fewer non-diagnostic findings than repeated FNA in nodules with non-diagnostic findings on initial FNA (1.6% vs. 34.4%, P<0.001), with homogeneous results across papers. A comparison of CNB and repeated FNA in nodules initially diagnosed as AUS/FLUS using FNA showed complex results. In this scenario, significantly fewer non-diagnostic results were observed for CNB than for FNA, and the results of CNB were homogeneous. However, the proportion of AUS/FLUS diagnoses did not significantly differ between CNB and repeated FNA, although the pooled estimate of AUS/FLUS diagnoses was lower in CNB than in repeated FNA (24.1% vs. 35.2%). The forest plot showed extensive heterogeneity between studies, which explains the lack of statistical significance. Simultaneously, the proportion of FN/SFN diagnoses was significantly higher on CNB. While previous meta-analyses assessed CNB and FNA after initial AUS/FLUS findings [6,8,9], they focused on the proportions of inconclusive findings, including non-diagnostic findings and AUS/FLUS combined, and showed a decreased frequency of inconclusive findings for CNB. Therefore, whether we should really expect a lower likelihood of repeated diagnosis of AUS/FLUS by selecting CNB rather than FNA is unclear, and the higher frequency of FN/SFN diagnoses obtained by CNB may result in an increased diagnostic surgery rates, as described by Yoon et al. [22]. Therefore, while CNB resulted in significantly fewer non-diagnostic results in nodules with initially non-diagnostic or AUS/FLUS results, its effects on decreasing the frequency of AUS/FLUS findings are unclear.
Factors related to differences in CNB preferences between hospitals
Favoring CNB means using CNB as often as necessary, but not as a primary test. In contrast, not favoring CNB means that there is little use of CNB in clinical settings, which was confirmed by clinicians at each hospital. The reason for including only Korean papers was to confirm each hospital’s preference for CNB by personal contact with head and neck surgeons working at the hospital. A proportional analysis was performed to investigate differences in diagnoses made using the TBSRTC classification across hospitals. Although there was considerable heterogeneity between studies, significantly higher proportions of AUS/FLUS diagnoses were observed in papers published by authors from hospitals favoring CNB than in papers published by authors from other hospitals (12.3% vs. 5.1%). Therefore, we carefully suggest that differences in the patterns of pathological diagnoses may influence the preference for CNB and that the increased rate of inconclusive results from frequent AUS/FLUS findings may contribute to favoring CNB. A meta-analysis by Suh et al. in 2016 [8] compared the sensitivity of FNA between studies originating within and outside of Asia, and reported significantly higher sensitivity outside of Asia than within Asia (85% vs. 64%). Considering that nine out of the 10 Asian papers cited in their paper were by Korean authors, this finding can be interpreted that the sensitivity of FNA performed in Korea is inferior to that of FNA performed in the West. The reason for this difference in sensitivity might be due to differences in the mindset of patients undergoing surgery. In Korea, patients often do not understand why surgery was performed if they hear that the nodule was not cancer after surgery. Therefore, diagnoses may be more conservative than suggested by TBSRTC. The risk of malignancy suggested by TBSRTC in AUS/FLUS, suspicious for malignancy, and malignant nodules is 5%–15%, 60%–75%, and 97%–99%, respectively [70]. However, the corresponding risks of malignancy are much higher in Korea; for instance, the risk of malignancy of AUS/FLUS and suspicious for malignancy nodules is approximately 30% and more than 90%, respectively, according to our unpublished data. Therefore, many nodules that could be diagnosed as suspicious for malignancy according to TBSRTC may be diagnosed as AUS/FLUS in Korea, which may lower the sensitivity of FNA. Moreover, the frequency of non-diagnostic results on CNB was higher in papers outside of Asia. This finding may also explain the relative lack of interest in CNB in Western countries.
Comparison of CNB and FNA based on results from consecutive cases: expectations from first-line CNB
The results of consecutive cases in which CNB and FNA were performed were compared to estimate how the proportional frequency of diagnoses would be affected by using CNB as the first-line modality. The proportions of diagnoses for all categories differed significantly between CNB and FNA even after Bonferroni correction of the P-value. An important point is that the patient population is not the same for CNB and FNA. FNA may be frequently performed in cystic nodules to remove fluid, while CNB may be preferred in solid nodules. This difference should be considered when interpreting these results. Therefore, benign and malignant diagnoses are more frequent for FNA and CNB, respectively. Apart from these two diagnostic results, the frequencies of inconclusive findings also differed. CNB showed significantly fewer non-diagnostic findings, an observation consistent with those for CNB performed in secondary biopsy procedures. Moreover, suspicion for malignancy was less frequently reported for CNB. However, AUS/FLUS and FN/SFN were diagnosed significantly more frequently on CNB than on FNA. Therefore, the rates of inconclusive findings (non-diagnostic and AUS/FLUS) were nearly the same between CNB and FNA (20.7% for both). However, a diagnosis of FN/SFN also results in diagnostic surgery and is sometimes considered to be an inconclusive finding. Therefore, if FN/SFN is included, the proportion of inconclusive results was higher for CNB than for FNA (28.3% vs. 21.8%). Based on these results, it may be difficult to recommend the generalized use of CNB in thyroid nodules if the purpose is to decrease the frequency of inconclusive findings. Another meta-analysis comparing the efficacy of CNB and FNA for diagnosing malignancy also showed no significant difference between the two modalities [2,4].
Increased frequency of AUS/FLUS on CNB due to increased diagnoses of architectural atypia
There is continuing interest regarding differences in the risk of malignancy between subcategories of AUS/FLUS. Authors at our institution also published a meta-analysis showing different risks of malignancy between cellular and architectural atypia [71]. Many researchers have suggested that cellular and architectural atypia should be classified separately in TBSRTC [72], and the CNB diagnostic criteria proposed by the Korean Endocrine Pathology Thyroid Core Needle Biopsy Study Group are divided into IIIA (indeterminate follicular lesion with nuclear atypia) and IIIB (indeterminate follicular lesion with architectural atypia) accordingly. Although only two papers reported AUS/FLUS subcategories in CNB results, a proportional analysis could be performed for comparison with FNA. In this analysis, 72% of the nodules classified as AUS/FLUS by CNB had architectural atypia, while 70% of the nodules classified as AUS/FLUS by FNA had cellular atypia; in other words, the two tests had opposite patterns of findings. Therefore, the higher frequency of AUS/FLUS diagnoses using CNB is in line with the higher frequency of FN/SFN findings, and using CNB appears to increase the likelihood of diagnosis of FN.
Similar risk of malignancy in cellular/architectural atypia and FN/SFN between CNB and FNA
If CNB results in the overdiagnosis of architectural atypia or FN/SFN, the malignancy rate would be expected to be lower for CNB than for FNA. However, the proportional meta-analysis showed no significant difference between CNB and FNA; thus, the increased diagnosis of architectural atypia or FN/SFN is not due to overdiagnosis. The potential to miss cases of FN/SFN by using FNA requires further study.
Complications of CNB and FNA
Although the proportional analysis showed no significant difference in the complication rate between CNB and FNA, the complication rate of CNB was higher than that of FNA (1.5% vs. 0.7%) and serious problems including injury to the carotid artery were reported. Therefore, to avoid complications, caution is required when performing CNB.
The results of this meta-analysis indicate that CNB has the following advantages and disadvantages. First, CNB can avoid non-diagnostic results in many cases, both in secondary biopsies after initial non-diagnostic or AUS/FLUS results and in first-line biopsies. Second, CNB results in significantly increased frequencies of architectural atypia and FN/SFN diagnoses, especially as a first-line modality, and therefore could increase the need for diagnostic surgery. However, as Yoon et al. [65] suggested, if FNA misses FN/SFN, that could be another advantage of CNB, a possibility that requires additional study. Third, the different AUS/FLUS diagnosis patterns between FNA and CNB may explain why the proportion of AUS/FLUS was not lower on CNB than on repeated FNA in nodules initially diagnosed as AUS/FLUS on FNA. Secondary CNB in nodules diagnosed as cellular atypia by FNA may be effective because CNB may decrease the frequency of cellular atypia or suspicion for malignancy findings. However, in other cases, the increased diagnosis of architectural atypia and FN/SFN may increase the frequency of inconclusive findings. In conclusion, CNB has a definite advantage in decreasing the frequency of non-diagnostic results; however, CNB as a first-line biopsy technique should be selected carefully to decrease the risk of inconclusive results. Furthermore, hospitals with low rates of non-diagnostic or AUS/FLUS findings may have a minimal need for CNB.
HIGHLIGHTS
▪ Core needle biopsy (CNB) can significantly decrease the frequency of non-diagnostic findings.
▪ CNB results in an increased frequency of findings of architectural atypia and follicular neoplasm.
▪ The preference for CNB is related to the frequency of atypia of undetermined significance/follicular lesion of undetermined significance results.
▪ The complication rate did not differ between fine-needle aspiration and CNB, but serious complications are possible in CNB.
XML Download