Fine‐Needle Aspiration Versus Core Needle Biopsy for Malignant Cervical Lymphadenopathy: A Meta‐Analysis and Systematic Review

Abstract

Objective

Cervical lymphadenopathy is a common presentation in Otolaryngology–Head and Neck Surgery, often requiring biopsy to exclude malignancy. Fine‐needle aspiration (FNA) is the standard initial diagnostic tool; however, evidence from other anatomical sites suggests that core needle biopsy (CNB) may offer superior diagnostic performance. This systematic review aims to compare the diagnostic accuracy and safety of FNA versus CNB for adult cervical lymphadenopathy.

Data Sources

PubMed, Embase, and Cochrane databases were systematically searched for studies published between 1995 and 2025 evaluating FNA versus CNB for adult cervical lymphadenopathy. Studies lacking extractable cervical lymphadenopathy data were excluded.

Review Methods

Two investigators independently screened studies and extracted data on study characteristics and diagnostic performance. A meta‐analysis was performed in Stata using a random‐effects model, and a quality assessment was conducted using the QUADAS‐2 criteria.

Results

Of 2523 abstracts identified, 6 studies met inclusion criteria, with 5 providing adequate data for meta‐analysis (total n = 649 patients, 331 FNA, 318 CNB). Pooled sensitivity and specificity were higher for CNB (0.94 and 0.98, respectively) compared to FNA (0.72 and 0.96; P < .001). CNB was 1.32 times more sensitive than FNA in detecting malignancy (P < .001). Both techniques had low complication rates. Study heterogeneity was low to moderate, and risk of bias was similarly moderate.

Conclusion

CNB demonstrates significantly greater diagnostic accuracy than FNA in detecting malignant cervical lymphadenopathy without increased complication risk. While further studies are warranted, these findings support greater consideration of CNB as a first‐line diagnostic test to minimize inconclusive results and diagnostic delays.

Adult neck masses are a common clinical presentation encountered in both Otolaryngology–Head and Neck Surgery and primary care settings. ¹ The differential diagnosis is broad, ranging from benign etiologies to malignancy. ¹ Malignancies account for approximately 5% to 10% of adult neck mass cases, ² , ³ necessitating tissue biopsy as a critical step in the diagnostic workup. ⁴

Multiple biopsy techniques are available for evaluating neck masses including fine needle aspiration (FNA), core needle biopsy (CNB), and excisional biopsy. Large studies and the American Academy of Otolaryngology's (AAO's) Adult Neck Mass Clinical Practice Guidelines endorse FNA as the first‐line diagnostic approach due to its low morbidity, acceptable diagnostic accuracy, and high patient acceptability. ⁵ , ⁶ , ⁷ , ⁸ Despite its widespread use, FNA is frequently nondiagnostic, with reported inadequacy rates as high as 32% per previous meta‐analysis of FNA for head and neck lesions. ⁶ The use of rapid onsite evaluation (ROSE) is associated with a reduction in inadequate biopsy rates, improving from an average of 10.3% without ROSE to 6.0% with it.⁶ However, ROSE is not available in many clinical settings where needle biopsies are performed, and data on its availability remain limited. ⁶ This raises important questions about the continued endorsement of FNA as the initial biopsy technique and suggests a need to further investigate the role of other techniques like CNB.

Systematic reviews of the comparative efficacy of FNA versus CNB have demonstrated CNB's superior diagnostic performance across many anatomical sites, including the salivary glands, ⁹ , ¹⁰ thyroid, ¹¹ breast, and lungs. ¹² Although concerns have been raised regarding CNB's potential for increased adverse events—such as bleeding, hematomas, nerve damage, and tumor seeding13, 14, 15—these complications remain rare, with incidence below 1% reported in a meta‐analysis of CNB for head and neck lesions. ¹⁶ To date, no systematic review has directly compared FNA and CNB for the evaluation of cervical lymphadenopathy, contributing to inconsistent biopsy practices and delays in diagnosis and treatment. While differences in the performance of these two biopsy techniques have been established for other head and neck sites and studies on CNB alone have demonstrated its efficacy for cervical lymphadenopathy, the unique tissue architecture, anatomical inaccessibility due to important adjacent vasculature, pathological spectrum of cervical lymph nodes (especially inclusion of lymphoma, for which diagnosis specifically requires architectural preservation), and continued clinical preference for FNA necessitate a site‐specific systematic review and meta‐analysis. ¹⁷

The primary aim of this study is to conduct a systematic review and meta‐analysis comparing both the sample adequacy and diagnostic accuracy of CNB versus FNA (as the standard comparator) in the diagnosis of malignancy in adults presenting with cervical lymphadenopathy requiring needle biopsy. A secondary aim is to evaluate and summarize reported complication rates associated with each modality.

Methods

This study was conducted in accordance with the Preferred Reporting Items for Systematic Reviews and Meta‐Analyses of Diagnostic Test Accuracy (PRISMA) guidelines. ¹⁸ As a review of previously published studies, this work did not require IRB approval or informed consent. A review protocol was not prospectively registered for this study, but our research question, methodology, and outcomes of interest were decided a priori and remained unaltered throughout the review process.

Study Eligibility Criteria

Studies that were conducted between 1995 and 2025 comparing FNA to CNB in adults with cervical lymphadenopathy requiring biopsy were considered eligible. Earlier studies were excluded due to substantially different biopsy techniques and diagnostic standards, which would limit clinical relevance. ¹⁹ Additional inclusion criteria included required reporting of each modality's outcomes in differentiating malignant from nonmalignant cervical lymphadenopathy and a reference test of either an excisional biopsy or ultimate diagnosis identified through clinical follow‐up for at least 6 months. Interventional studies, including retrospective cohort, prospective cohort, and randomized controlled trials (RCTs) were included. Studies were excluded if they had combined data that included other head and neck lesions (such as salivary gland or thyroid nodules) that did not allow isolated extraction of cervical lymphadenopathy biopsy results, were not in English, were unavailable in full text, lacked extractable data for pooled calculation of diagnostic accuracy, or used non‐standard FNA/CNB techniques.

Data Searching Strategy

The literature search was conducted independently by one investigator (SM) and one institutional librarian using PubMed/MEDLINE, EMBASE, and the Cochrane Library, with the last search occurring in August 2025. Keywords and phrases were tailored for each database to capture studies meeting the eligibility criteria (see Supplemental Appendix A, available online). Reference lists from included studies were also reviewed to ensure comprehensiveness.

Data Collection

Reference lists collected from the literature searches were merged and duplicates removed using Rayyan, ²⁰ a web‐based screening tool. Three researchers (KW, SM, and KB) independently screened titles and abstracts, followed by full‐text review. Inter‐rater reliability was assessed using Fleiss' free‐marginal kappa. Disagreements throughout the screening process were resolved through consensus.

Two independent researchers (KW and SM) extracted data using a standardized Excel sheet, with a third researcher (KB) verifying for accuracy. Extracted data included: study characteristics (location, design, institutions); patient demographics and sample size; biopsy methodology (FNA vs CNB); diagnostic performance metrics (true positive, true negative, false positive, false negative); reference standards used for verification of FNA and CNB accuracy; and reported adverse events.

Data Synthesis and Analysis

Study characteristics and outcomes, including rate of nondiagnostic results—defined as samples that were inadequate for analysis—and indeterminate results—defined as samples that could not be delineated as either benign or malignant—were summarized in tabular form and pooled. Studies that did not report on diagnostic adequacy or indeterminate outcomes were excluded from the respective pooled analyses. Diagnostic accuracy data were synthesized using 2 × 2 contingency tables. In studies where biopsy results were reported as nondiagnostic or indeterminate, these outcomes were included in the 2 × 2 contingency tables to reflect real‐world diagnostic performance. This approach ensures that test sensitivity and specificity estimates incorporate test failure rates and avoid selection bias. ²¹ Meta‐analysis was conducted using the metadta package in Stata, appropriate for diagnostic test accuracy studies. ²² Using a bivariate random‐effects model, which accounts for potential data heterogeneity, we visualized the comparative performance of these two biopsy methods using the area under the curve (AUC) of our data's summary receiver operating characteristics (SROC) curves. Pooled sensitivity and specificity were also calculated and visualized via forest plots, with statistical significance set to P < .05. ChatGPT (openai.com) was utilized to confirm the appropriateness of our statistical methods and to identify bugs in the metadta code used for analysis. A post hoc subgroup analysis of diagnostic test accuracy based on malignancy type (lymphoproliferative versus metastatic) was attempted with the studies that made this disease distinction in their reported data; however, the hierarchical bivariate model did not converge due to the small number of studies and data sparsity. We therefore present a descriptive comparison of CNB and FNA for the diagnosis of lymphoma versus metastatic lymphadenopathy using the data from these 3 studies.

Assessment of Heterogeneity, Study Quality, and Risk of Bias

Study heterogeneity of both sensitivity and specificity was calculated using the I ² statistic, which quantifies the proportion of study variation attributable to true heterogeneity rather than chance. ²³ The Cochrane criteria defines I ² values from 0% to 40% as indicative of low heterogeneity, 30% to 60% as moderate heterogeneity, 50% to 90% as substantial heterogeneity, and 75% to 100% as considerable heterogeneity. ²⁴

Two reviewers (SM and KB) independently assessed study quality using the Quality Assessment of Diagnostic Studies (QUADAS)‐2 checklist and criteria, a tool developed and validated for quality assessment of studies included in systematic reviews of diagnostic test accuracy specifically. ²⁵ The main factors of importance within the QUADAS‐2 checklists are: (1) patient selection, (2) index test, (3) reference standard, and (4) flow and timing. Given that excisional biopsy is invasive and often not indicated, clinical follow‐up of at least 6 months to confirm negative or positive results was deemed an acceptable reference standard, per established diagnostic test accuracy systematic review and meta‐analysis guidelines described by the Cochrane Collaboration. ²⁶ Discrepancies were resolved by discussion until consensus was reached. To identify potential publication bias, the midas Stata package for meta‐analysis was used to create a Deeks' funnel plot with a regression test of diagnostic log odds ratio against effective sample size. This regression test assesses for asymmetry in diagnostic test accuracy data included in meta‐analyses more accurately than a funnel plot alone given its ability to account for each study's sample size. ²⁷ , ²⁸ Of note, however, reliability of Deeks’ funnel plot is predicated on the inclusion of ten or more studies, so any asymmetry in our meta‐analysis was interpreted with caution.

Results

Literature Search and Characteristics of Included Studies

A PRISMA flow diagram (Figure 1) illustrates the literature screening process. From 2523 initial records, 5 met all inclusion criteria after title, abstract, and full‐text screening. ¹⁵ , ²⁹ , ³⁰ , ³¹ , ³² One additional study was included qualitatively but excluded from meta‐analysis due to insufficient extractable data. ³³ A Fleiss' kappa score of 0.78 (95% CI 0.70, 0.86) was achieved for title and abstract screening, indicating good inter‐rater agreement during the screening process.

Figure 1.

Preferred reporting items for systematic reviews and meta‐analyses flow diagram.

Study characteristics are presented in Table 1. Three studies originated from otolaryngology and three from radiology/ultrasound departments at tertiary medical institutions. Three studies were performed in China, and the other three in South Korea. While five studies were retrospective, one was prospective, which still aligned with our a priori eligibility criteria regarding interventional study design, and which we ensured would be comparable to the retrospective studies by reviewing their data collection approaches and by using a random effects model for our analysis. ³⁰ A sensitivity analysis with the prospective study excluded showed no effect on our results. Only 5 studies provided raw data sufficient for meta‐analysis, with one excluded from quantitative synthesis. ³³

Table 1.

Characteristics of Included Studies

Reference and study location	Study design	Participant criteria	N	Biopsy method	Image guidance	Biopsy performer	Complications
Ryu et al, 2015 South Korea	Retrospective consecutive case series	Patients >15 years who presented with cervical lymphadenopathy without signs of acute febrile illness/URI, apparent benign mass, or known lymphoma.	CNB: 75 (17 of which were after FNA) FNA: 150	Techniques not described	US only for CNB	Unknown if Oto‐HNS or Radiology	Adverse effects not discussed
Oh et al, 2016 South Korea	Retrospective observational study	LAD suspects with unresolved symptoms after two weeks of antibiotic therapy and 4 weeks of observation, with history of recurrent lymphadenopathy mass >1 cm in diameter, “without suspicion of metastatic LN.” Goal was for all patients to receive both biopsy types; same patient sample for both CNB and FNA.	CNB: 79 FNA: 62	CNB: 18 or 20‐ US gauge needle with ACECUT FNA: “Usual method”	US	Radiology	“No neurological, vascular, or tumor‐seeding complications.”
Park et al, 2018 South Korea	Retrospective study	Pts 15 or older, with at least 2 years of follow‐up after initial diagnosis, in which cervical lymph nodes were palpable for more than 1 month without signs of acute febrile illness, obvious benign mass, definite malignant head and neck tumor, or previous history of malignancy.	CNB: 122 FNA: 62	Techniques not described	US	Radiology	CNB: 3 patients (1.8%) across both cervical LAD and salivary gland mass: 1 hematoma, 2 paresthesia FNA: 2 patients (1.8% across both cervical LAD and salivary gland tumor groups) with post‐biopsy infections
Teng et al, 2021 China	Prospective trial	Patients with US revealing suspicious cervical lymph nodes ≤1.5 cm indiameter and adjacent to blood vessels and nerves without significant comorbidities.	CNB: 42 FNA: 42	CNB: 21‐gauge US needle with 20‐mL syringe for hydrodissection FNA: 22‐gauge needle	US	Radiology	Slight swelling at the CNB puncture site, self‐resolving in 30 minutes: 2 patients No major complications during or after the process in either group.
Mu et al, 2022 China	Four‐armed retrospective study: Conventional vs CEUS ultrasound technique, with CNB vs FNA in each US group	Patients requiring pathological diagnosis of enlarged lymph nodes (≥0.5 cm in length) and who are willing to undergo puncture.	CNB: 90 (but only 39 included in our analysis because 51 in CEUS group) FNA: 78 (but only 37 included in our analysis because 41 in CEUS group)	Techniques not described Surgical resection used as reference test	US, comparing conventional to CEUS (CEUS results not included in our analysis)	Unknown if Oto‐HNS or Radiology	Adverse effects not discussed
Mu et al, 2023 China	Retrospective analysis	Adult patients with cervical lymphadenopathy.	CNB: 40 FNA: 40	CNB: 18‐gauge needle FNA: 22‐gauge fine needle	US	Unknown if Oto‐HNS or Radiology	CNB: Bleeding (4, 10%), fever (2, 5%), abscess formation (1, 2.5%), aggravated pain (2, 5%) FNA: Bleeding (1, 2.5%), fever (1, 2.5%)

Across the five meta‐analyzed studies, 331 patients underwent FNA and 318 underwent CNB, with studies ultrasound for all biopsies except one which used ultrasound only for CNB. ²⁹ Age and sex data were inconsistently reported postexclusion of nonlymph node neck masses. Three studies classified malignancies as metastases (n = 86) or lymphomas (n = 58). All studies employed ultrasound‐guided biopsy. One study compared traditional versus contrast‐enhanced US; only data from traditional ultrasound methods were included in our analysis. ³¹ Another study had reported findings on both cervical lymph nodes and salivary gland lesions but was eligible for inclusion because the data for these 2 anatomical sites were presented separately, allowing us to extract only the results for cervical lymphadenopathy. ¹⁵

Comparative Diagnostic Test Accuracy

Table 2 presents data on diagnostic yield and accuracy. Across the four studies reporting diagnostic adequacy (which does not include one that is otherwise included in the meta‐analysis), CNB had a pooled inadequacy rate of 2.9% (8/279) compared to 11.2% (33/294) for FNA. ¹⁵ , ²⁹ , ³⁰ , ³² The pooled indeterminate rate was 0% (0/279) for CNB and 5.4% (16/294) for FNA. Therefore, only 2.9% (8/279) of CNB results versus 16.7% (49/294) of FNA results were diagnostically inconclusive.

Table 2.

Diagnostic Performance of Core Needle Biopsy (CNB) Versus Fine‐Needle Aspiration (FNA) for Cervical Lymphadenopathy

	CNB					FNA
Reference	Diagnostic	Nondiagnostic**	Indeterminate	Sensitivity	Specificity	Diagnostic	Nondiagnostic	Indeterminate	Sensitivity	Specificity
Ryu et al	75 (100%)	0 (0%)	0 (0%)	95.7%	96.2%	131 (87.3%)	11 (7.3%)	8 (5.3%)	66.1%	94.8%
Oh et al	73 (92.4%)	6 (7.6%)	ND	91.6%	100%	35 (56.5%)	23 (37.0%)	4 (6.5%)	50%	100%
Park et al	115 (94.3%)	0 (0%)	7 (5.7%)	95.6%	100%	48 (77.4%)	3 (4.8%)	11 (17.7%)	ND	ND
Teng et al	42 (100%)	0 (0%)	0 (0%)	100%	100%	31 (73.8%)	2 (4.8%)	9 (21.4%)	79.2%	100%
Mu et al 2022	ND	ND	ND	89.1%	91.4%	ND	ND	ND	86.1%	85.7%
Mu et al 2023	39 (97.5%)	0 (0%)	1 (2.5%)	100%	93.8%	35 (87.5%)	0 (0%)	5 (12.5%)	86.7%	90%

Abbreviations: CNB, core needle biopsy; FNA, fine‐needle aspiration; ND, not described; nondiagnostic, inadequate sample.

**Indicate that nondiagnostic was defined as “inadequate sample or inconclusive result”.

For the five studies with extractable data, a random effects model was employed given its ability to account for study heterogeneity. We analyzed the SROC curves created using the pooled data, which show that CNB had a greater area under the curve (AUC) than FNA, indicating superior overall diagnostic accuracy (Figure 2). We also conducted a pooled analysis of sensitivity and specificity for FNA versus CNB for detecting malignancy. ¹⁵ , ²⁹ , ³⁰ , ³¹ , ³² Pooled sensitivity was 0.72 (95% CI: 0.59, 0.82) for FNA and 0.94 (95% CI: 0.88, 0.97) for CNB, and pooled specificity was 0.96 (95% CI: 0.74, 1.00) for FNA and 0.98 (95% CI: 0.85, 1.00) for CNB (Table 3 and Figure 3). CNB demonstrated significantly greater sensitivity than FNA, with a relative sensitivity ratio of 1.32 (95% CI: 1.13‐1.54, P < .001), using FNA as the reference standard (Table 4). The relative specificity ratio was 1.02 (95% CI: 0.97‐1.07, P = .17).

Figure 2.

Summary receiver operating characteristic (SROC) curves for FNA versus CNB. CNB, core needle biopsy; FNA, fine‐needle aspiration.

Table 3.

Summary of Diagnostic Test Accuracy of FNA Versus CNB

Effect	Group	Proportion (95% CI)	P‐value
Sensitivity
	FNA	0.72 (0.59, 0.82)	.00
	CNB	0.94 (0.88, 0.97)	.00
Specificity
	FNA	0.96 (0.74, 1.00)	.00
	CNB	0.98 (0.85, 1.00)	.00

Abbreviations: CI, confidence interval; CNB, core needle biopsy; FNA, fine‐needle aspiration.

Figure 3.

Forest plots of pooled sensitivity and specificity for FNA versus CNB. CNB, core needle biopsy; FNA, fine‐needle aspiration.

Table 4.

Relative Measures of Diagnostic Test Accuracy of FNA Versus CNB

Effect	Group	Relative ratio (95% CI)	P‐value
Relative sensitivity
	FNA	1.00
	CNB	1.32 (1.13, 1.54)	.00
Relative specificity
	FNA	1.00
	CNB	1.02 (0.97, 1.07)	.17

Abbreviations: CI, confidence interval; CNB, core needle biopsy; FNA, fine‐needle aspiration.

A post‐hoc subgroup analysis was attempted to compare diagnostic accuracy based on malignancy type using the three studies that reported this distinction. ¹⁵ , ²⁹ , ³⁰ Due to the limited number of studies and small pool of data, the hierarchical bivariate random effects model did not converge. Therefore, we conducted a descriptive analysis of sensitivity and specificity for FNA versus CNB for detecting lymphoma versus metastatic lymphadenopathy (Table 5). Of note, FNA biopsies with an indeterminate result later found to be malignant were grouped as false negatives, per standard guidelines, as a conservative measure to ensure FNA's performance was not overestimated. The pooled performance data for lymphoma demonstrated a sensitivity and specificity of 44.4% and 99.1%, respectively, for FNA and 97.5% and 99.0% for CNB. The pooled data for metastatic lymphadenopathy demonstrated a sensitivity and specificity of 81.6% and 100% for FNA and 100% and 100% for CNB.

Table 5.

Performance of CNB versus FNA based on Malignancy Type

Lymphoma
	CNB						FNA
Reference	TP	FP	TN	FN	Sensitivity	Specificity	TP	FP	TN	FN	Sensitivity	Specificity
Ryu et al	10	2	62	1	90.9%	96.9%	6	2	122	7	46.2%	98.3%
Park et al	26	0	96	0	100%	100%	2	0	57	0	100%	100%
Teng et al	3	0	39	0	100%	100%	0	0	39	3	0%	100%
Pooled data	39	2	197	1	97.5%	99.0%	8	2	218	10	44.4%	99.1%

Carcinoma/Metastatic Lymphadenopathy
Ryu et al	3	0	72	0	100%	100%	4	0	143	3	57.1%	100%
Park et al	19	0	103	0	100%	100%	8	0	49	2	80.0%	100%
Teng et al	26	0	16	0	100%	100%	19	0	21	2	90.5%	100%
Pooled data	48	0	191	0	100%	100%	31	0	213	7	81.6%	100%

Adverse Events

Four studies reported adverse events (Table 1). One study reported no neurological, vascular, or tumor‐seeding complications with either biopsy technique, while another found a 1.8% complication rate for both FNA (2 post‐biopsy infections) and CNB (1 hematoma and 2 paresthesias), though their report included both cervical lymphadenopathy and salivary masses. ¹⁵ , ³³ One study reported transient swelling in 2 CNB patients that resolved within 30 minutes, and the final study reported only minor complications including bleeding (FNA: 2.5%, CNB: 10%), fever (FNA: 2.5%, CNB: 5%), 1 CNB abscess (2.5%), and aggravated pain in 2 CNB cases (5%). ³⁰ , ³²

Only the data from one study provided enough granularity to statistically compare complication rates between FNA and CNB, showing no significant differences in rates of bleeding (X ² = 1.92, P = .17), fever (X ² = 0.35, P = .56); abscess (X ² = 1.01, P = .31), and pain (X ² = 2.05, P = .15). ³²

Study Heterogeneity, Quality, and Risk of Bias Assessment

Across the studies included in our meta‐analysis, pooled heterogeneity, as calculated using I ², was low to moderate: 12.31% and 49.08% for FNA's versus 28.74% and 17.11% for CNB's sensitivity and specificity, respectively.

Figure 4 summarizes the results of the QUADAS‐2 risk of bias assessment for each study included in the meta‐analysis. All five studies demonstrated a low risk of bias in the patient selection domain. Three studies were rated as having unclear risk of bias in the index test domain due to insufficient reporting of biopsy techniques, such as needle gauge, which limited the assessment of comparability of methods across studies. ¹⁵ , ²⁹ , ³¹ One study had a high risk of bias in the reference standard domain because positive tests were confirmed with CNB, which—as one of the index tests—should not be used as the reference test. ³⁰ Two additional studies had an unclear risk of bias in this domain because their gold standard reference test was vaguely described as “pathological results” from excision or lymph node biopsy. ³¹ , ³² The lack of clarity around these reference standards, along with the limited description of clinical follow‐up duration for another study, then contributed to an unclear risk of bias in the flow and timing domain for 3 studies in total. ²⁹ , ³¹ , ³²

Figure 4.

Quality assessment for studies included in meta‐analysis: QUADAS‐2 results.

Based on our Deeks' funnel plot with linear regression test (Figure 5), there was no statistically significant evidence of publication bias for FNA, indicated by the symmetric distribution of study findings (intercept = −0.3, P = .39). The distribution for CNB was more asymmetric (intercept = 15.4, P = .04). The overall certainty in the pooled estimates was qualitatively judged to be moderate to high. This assessment was based on consistent findings across studies, low complication rates, low to moderate heterogeneity, and moderate risk of bias.

Figure 5.

Publication bias assessment: Deeks' funnel plot asymmetry test results.

Discussion

To our knowledge, this is the first systematic review and meta‐analysis of studies directly comparing CNB to FNA for the diagnosis of malignant cervical lymphadenopathy in terms of adequacy, diagnostic accuracy, and adverse events. Previous meta‐analyses of head and neck lesions in general have focused on just FNA or CNB performance alone. A meta‐analysis of 78 FNA studies reported inadequacy rate of 10.3% without ROSE and 6.0% with ROSE, though they did not report on indeterminate findings such as “atypical” or “suspicious.” ⁶ On the other hand, a meta‐analysis of 16 CNB studies with 1267 patients showed a lower inadequacy rate of 5%. ¹⁶ Our meta‐analysis included studies published after the previous two meta‐analyses and directly compared FNA to CNB done by the same research team. Our results demonstrate that in this small number of available studies, CNB has superior sample adequacy in comparison to FNA, with an aggregate inadequate and indeterminate rate of 2.9% for CNB versus 16.7% for FNA. These findings are consistent with studies of other anatomical sites showing higher rate of inadequate FNAs compared to CNB. ⁹ , ¹¹ , ¹³ , ³⁴ , ³⁵ The lower inadequate and indeterminate rate for CNB has been shown to reduce the need for repeat procedures, diagnostic delays, and institutional costs. ³¹ , ³⁶ , ³⁷ , ³⁸ , ³⁹ , ⁴⁰ , ⁴¹ Thus, while concerns for greater costs with CNB biopsies and their pathological analysis in comparison to FNA have been noted, studies of the financial burden of CNB in other anatomical sites has shown that, due to the high rate of inadequate or unsuccessful FNAs, the diagnostic workup may be more cost‐efficient if CNB, with its higher first‐pass yield, is the standard preliminary testing modality. ³⁶ , ³⁹

Our pooled analysis shows CNB has a sensitivity of 0.94 compared to 0.72 for FNA, translating to a 32% greater likelihood of correctly diagnosing malignancy with CNB (relative ratio = 1.32). Both techniques showed high specificity with no significant difference, and the SROC curves corroborate these findings. Comparing the results of the random‐effects model with a fixed‐effects model (to assess for selection bias or unobserved, invariant confounders) yielded similar sensitivity and specificity estimates for CNB, supporting the reliability of the random‐effects model. The superior diagnostic performance of CNB aligns with findings from systematic reviews of other anatomic sites, including breast, ³⁴ , ⁴⁰ lung, ⁴¹ thyroid, and salivary gland. ⁹ , ¹¹ , ¹³ , ¹⁴ , ¹⁶ , ³⁵ , ³⁹ , ⁴² , ⁴³ Given these findings, diagnostic standards have shifted for other anatomical sites: for example, in recent years, CNB has become the preferred needle biopsy technique for breast masses as outlined by the National Comprehensive Cancer Network's diagnostic guidelines. ⁴⁴ Our findings suggest that reconsideration of current standards for cervical lymphadenopathy, for which FNA remains the first‐line diagnostic test, may be similarly warranted, though further evaluation via larger studies with more heterogeneous patient populations is still required.

This reconsideration of current guidelines is particularly important given the performance of FNA in the diagnosis of lymphoma: as seen in our pooled data from the 3 studies that distinguished results based on malignancy type, FNA performed substantially worse in the diagnosis of lymphoma (sensitivity 44.4%) than metastatic lymphadenopathy (sensitivity 81.6%). These findings are consistent with previous studies that have highlighted FNA's low sensitivity for lymphoma in particular, as well as its inferiority in the pathological typing of cervical lymph nodes, due to the absence of tissue architecture for morphological assessment and inadequate material for immunohistochemistry. ¹⁷ , ³⁰ , ⁴⁵ , ⁴⁶ Nevertheless, the limited availability of data for disease‐specific analysis in our study underscores the need for further evaluation of diagnostic performance as distinguished by malignancy type.

Concerns about CNB's safety include primarily risk of pain, bleeding or nerve damage. ¹² Our review, however, found a negligibly higher rate of minor complications with CNB, with no major complications (including nerve damage) reported. This is consistent with other studies showing no significant difference of adverse events. ⁴⁷ , ⁴⁸ , ⁴⁹ Moreover, the universal use of ultrasound guidance for CNB may mitigate these risks. ⁵⁰ Thus, the data from our included studies suggests that safety concerns should not preclude broader adoption of CNB.

Our study's assessment of heterogeneity indicated a low amount of heterogeneity in the data available for the calculation of sensitivity for both biopsy modalities and specificity for CNB. These results demonstrate that the studies had some level of consistency in study population, design, and biopsy technique. Heterogeneity was moderate for the calculation of FNA's specificity, indicating that some variation between studies was not due to chance and may be due to study‐level factors such as differences in technique, cytopathologist experience, sample adequacy rates, and inclusion of lymphoproliferative cases of malignancy (which as aforementioned potentially contributes to FNA's lower rate of accuracy in comparison to metastatic lymph nodes). Since CNB results are more consistent across studies than FNA, this may further validate our finding that CNB provides more reliable diagnostic performance for identifying malignancy. Subgroup or meta‐regression analyses would aid in more definitively identifying potential sources of heterogeneity that influence the interpretability of the comparison of CNB to FNA; however, given the small number of studies available for inclusion, such analyses were not feasible or recommended. ⁵¹ While our utilization of a bivariate random‐effects model accounts for some of this heterogeneity, the lack of feasible subgroup analyses further supports the need for higher‐powered studies with standardized methodologies.

Our quality assessment showcased that the studies in our meta‐analysis had an unclear risk of bias across many of the domains included in the QUADAS‐2 criteria. This is largely due to incomplete reporting of their biopsy and clinical follow‐up protocols, as well as differences in biopsy performers and their established experience or training with the two techniques, which limit our ability to fully assess methodological rigor. On publication bias assessment, some asymmetry was observed in the CNB data, suggesting potential bias; however, the analysis included only 5 studies—fewer than the recommended minimum of ten for reliable evaluation. ²³ Therefore, the results of this analysis should be interpreted with caution due to limited statistical power. The limited interpretability of the quality and publication bias results substantiates the need for standardization of biopsy performance and methodological reporting in future studies.

Limitations

This study had several limitations. First, the available literature directly comparing FNA and CNB for cervical lymphadenopathy is limited, with only six eligible studies, most of which were small, retrospective, and single‐institutional, increasing the risk of selection bias. Technical factors—such as needle gauge, image modality, and operator experience—varied across studies and were not consistently reported, making it difficult to control for these possible confounding variables, and techniques that ensure the adequacy of biopsy samples, like ROSE, were absent. Prospective randomized controlled trials with standardized methodologies would help minimize these limitations. Although study heterogeneity was low for CNB, moderate heterogeneity was observed for FNA specificity, suggesting potential variability in FNA technique across studies and highlighting the need for more standardized implementation in clinical settings and reporting of techniques in research. Additionally, differences in how indeterminate and inadequate results were addressed—with one study reporting these results as separate categories from the 2 × 2 tables used to calculate sensitivity and specificity—may limit assessment of real‐world diagnostic performance. However, by including this study's indeterminate and inadequate results in our meta‐analysis, we ensured that our calculations consistently incorporated test failure rates to minimize selection bias. Finally, all included studies were conducted in South Korea or China, which may limit generalizability to other healthcare settings and populations, and all were in English, potentially excluding eligible studies published in other languages.

Conclusion

This study demonstrates that CNB offers significantly greater diagnostic efficacy than FNA for the evaluation of cervical lymphadenopathy, with markedly higher sensitivity and substantially lower rates of non‐diagnostic and indeterminate results. CNB was associated with a low incidence of minor complications and no major adverse events, suggesting that concerns about safety should not deter from its use. Thus, our findings suggest that CNB may be considered as a first‐line diagnostic approach for cervical lymphadenopathy given its potential to improve diagnostic accuracy and reduce delays in care when compared to FNA. Given a persistent gap in the literature on cervical lymph node biopsy technique efficacy as demonstrated by the limited number of studies available for meta‐analysis, we call for further studies with diverse populations, clinical contexts, and distinction between malignancy classifications to better evaluate the comparative diagnostic performance of CNB to FNA.

Author Contributions

Sacha Moufarrej, led the study design, data collection, data analysis and visualization, and writing of the manuscript; Kavenpreet S. Bal, contributed to data collection, data analysis, and editing of the manuscript; Alexander Rivero, contributed to the study design and editing of the manuscript; Miranda L. Weintraub, contributed to the editing of the manuscript; Thomas L. Haupt, contributed to the editing of the manuscript; Kevin H. Wang, contributed to the study design and data collection, oversaw project administration, and contributed to the editing of the manuscript.

Disclosures

Competing interests

None.

Funding source

None.

Supporting information

Appendix A. Database Search Strategies.