Diagnostic accuracy and complication rate of image-guided percutaneous transthoracic needle lung biopsy for subsolid pulmonary nodules: a systematic review and meta-analysis

Abstract

Objectives:

To determine the diagnostic accuracy and complication rate of percutaneous transthoracic needle biopsy (PTNB) for subsolid pulmonary nodules and sources of heterogeneity among reported results.

Methods:

We searched PubMed, EMBASE, and Cochrane libraries (until November 7, 2020) for studies measuring the diagnostic accuracy of PTNB for subsolid pulmonary nodules. Pooled sensitivity and specificity of PTNB were calculated using a bivariate random-effects model. Bivariate meta-regression analyses were performed to identify sources of heterogeneity. Pooled overall and major complication rates were calculated.

Results:

We included 744 biopsies from 685 patients (12 studies). The pooled sensitivity and specificity of PTNB for subsolid nodules were 90% (95% confidence interval [CI]: 85–94%) and 99% (95% CI: 92–100%), respectively. Mean age above 65 years was the only covariate significantly associated with higher sensitivity (93% vs  85%, p = 0.04). Core needle biopsy showed marginally higher sensitivity than fine-needle aspiration (93% vs  83%, p = 0.07). Pooled overall and major complication rate of PTNB were 43% (95% CI: 25–62%) and 0.1% (95% CI: 0–0.4%), respectively. Major complication rate was not different between fine-needle aspiration and core needle biopsy groups (p = 0.25).

Conclusion:

PTNB had acceptable performance and a low major complication rate in diagnosing subsolid pulmonary nodules. The only significant source of heterogeneity in reported sensitivities was a mean age above 65 years.

Advances in knowledge:

This is the first meta-analysis attempting to systemically determine the cause of heterogeneity in the diagnostic accuracy and complication rate of PTNB for subsolid pulmonary nodules.

Introduction

Subsolid pulmonary nodules, which consist of pure ground-glass nodules and part-solid nodules, are reportedly found in 9% of lung cancer screening CT scans.^1,2 Subsolid nodules are clinically significant due to their high incidence of malignancy when compared with solid nodules.^1,3 However, accurate diagnosis of subsolid nodules remains difficult, as they result from a variety of disorders, including inflammatory disease, pulmonary fibrosis, alveolar haemorrhage, or neoplastic conditions – from preinvasive lesions such as atypical adenomatous hyperplasia or adenocarcinoma in situ to invasive lesions such as minimally invasive or invasive adenocarcinoma.^4,5

Large clinical trials showed that screening with low-dose CT (LDCT) could reduce mortality from lung cancer,^6,7 and currently many organisations recommend lung cancer screening with LDCT for high-risk groups. Given the expected increase in the detection rate of lung cancer using the LDCT screening protocol, accurate diagnosis of subsolid nodules has become more important.

Percutaneous transthoracic needle biopsy (PTNB) is an established technique known to be accurate and relatively safe for diagnosing both benign and malignant peripheral pulmonary lesions.^8–11 However, it has been reported that the diagnostic yield of PTNB for subsolid nodule is significantly lower than that of solid lesions due to their low cellularity,¹² and expert opinions on the role of PTNB in assessing subsolid nodules remain diverse.¹³ To date, a few studies regarding the diagnostic performance of PTNB for subsolid nodules have been published. However, to the best of our knowledge, there have been no large cohort studies or meta-analyses of studies focusing on the heterogeneity of the reported diagnostic yield or the complication rate of PTNB for subsolid nodules. In this meta-analysis, we assessed the diagnostic accuracy and complication rate of PTNB for subsolid pulmonary nodules and identified factors that influence the diagnostic accuracy and complication rate through subgroup analysis.

Methods and materials

This study was conducted in accordance with the Preferred Reporting Items for a Systematic Review and Meta-analysis of Diagnostic Test Accuracy Studies (PRISMA-DTA). Institutional review board approval was not required. The study was registered on the International Prospective Register of Systematic Reviews (PROSPERO) (CRD42020191109).

Search strategy

Our target population was composed of patients who underwent image-guided percutaneous biopsy for subsolid pulmonary nodules suspected of malignancy.

We conducted online searches of PubMed, EMBASE, and Cochrane library from conception until November 7, 2020. The following terms were used to search abstracts or titles: (‘ggo’ OR ‘ground glass opacit*’ OR ‘ground glass nodul*’ OR ‘ground glass attenuatio*’ OR ‘ground glass lesio*’ OR ‘ground glass opacificati*’ OR ‘ground glass pulmonary’ OR ‘ground glass lung’ OR ‘subsolid nodul*’ OR ‘subsolid lesio*’ OR ‘subsolid lung’ OR ‘subsolid pulmonary’ OR ‘part solid nodul*’ OR ‘part solid lesio*’ OR ‘part solid lung’ OR ‘part solid pulmonary’) AND (lung OR pulmonary) AND (biops*). The literature search was not restricted to any publication date or study setting.

Study selection

Two investigators (both with 8 years of experience in radiology research) independently screened the titles and abstracts. They excluded articles not written in English, animal studies, case reports or series, letters, editorials, conference abstracts, review articles, guidelines, or consensus statements. The full-texts of the remaining articles were assessed, and the articles that had appropriate target populations and tabulated the data in a 2-by-2 contingency table were included. In case of disagreement, consensus was reached through discussion between the investigators.

Data extraction

The two investigators independently extracted data from the selected studies using the following standardised form: (a) study characteristics: authors, year of publication, country of origin, study period, study design (retrospective vs prospective), patient inclusion criteria, number of patients, number of biopsies, number of undetermined diagnoses; (b) demographic characteristics: mean age, sex; (c) lesion characteristics: mean size, type of subsolid nodule (ground-glass opacity [GGO] dominant vs solid dominant); (d) biopsy procedure: biopsy technique (fine-needle aspiration [FNA] vs core needle biopsy [CNB]), needle size, image-guidance method (CT fluoroscopy vs CT); (e) 2-by-2 contingency table (number of true-positive, false-positive, false-negative, and true-negative results); and (f) complications: number of complications, type of complications (major complications vs minor complications). Post-procedural pneumothorax and haemoptysis requiring interventional treatment and air embolism were defined as major complications.¹⁴ Minor pulmonary haemorrhage without symptom was not regarded as a complication. Clinical diagnosis was the reference standard used in our study, including pathologic confirmation by surgery or repeated biopsy, imaging follow-up, or intended treatment for the lesion. The 2-by-2 contingency table was recorded per-biopsy.

Quality assessment

Each included study was assessed independently by the two investigators using the Quality Assessment of Diagnostic Accuracy Studies (QUADAS)−2, and discrepancies were resolved by consensus.¹⁵ Pre-specified criteria for applicability were as follows. In the patient-selection domain, the authors assigned a rate of high concern for applicability if the study included only small lesions or pure GGO lesions. The concern regarding applicability was rated high in the index-test domain if the precancerous lesion (i.e. atypical adenomatoid hyperplasia) was considered positive using image-guided percutaneous lung biopsy. A minimum follow-up period shorter than 2 years in patients with negative biopsy result was regarded as a high risk of bias in the domain of reference standard. Due to patients lost to follow-up, most studies included patients with indeterminate clinical diagnoses; therefore, the risk of bias was considered high in the flow-and-timing domain only for studies in which the number of patients with indeterminate clinical diagnoses exceeded 10% of the study sample.¹⁶ Additionally, studies which excluded non-diagnostic results from the analysis were evaluated high risk of bias in the flow-and-timing domain.

Statistical analysis

We used a bivariate random-effect model to calculate the pooled per-biopsy sensitivity and specificity and a univariate random-effect model to calculate the pooled complication rate. Diagnostic accuracies (sensitivity and specificity) and complication rates of the included studies were reviewed using forest plots. Heterogeneity was assessed via visual inspection of the hierarchical summary receiver operating characteristic curve (HSROC) and was judged present when there was a significant difference between the 95% prediction and confidence interval regions. The presence of a threshold effect, which may have affected heterogeneity across the studies, was determined by visual assessment of the forest plot and Spearman’s correlation coefficient between the false-positive rate and sensitivity. A Spearman’s correlation coefficient more than 0.6 was considered a considerable threshold effect. Cochran’s Q test and the I² test were performed to evaluate the heterogeneity between studies for both diagnostic accuracy and complication rate. Heterogeneity was considered significant if p < 0.1 in Cochran’s Q test; the I² test results were interpreted as follows: 0–25%, may not be important; 25–50%, may represent low heterogeneity; 50–75%, may represent moderate heterogeneity; 75–100%, high heterogeneity.¹⁷ Deeks’ funnel plot was used to evaluate the potential publication bias.

Bivariate meta-regression was performed to identify factors associated with heterogeneity, if present. The following variables were evaluated in the subgroup analysis: (a) number of patients (≥60 vs <60), (b) year of publication (before 2013 vs 2013 and beyond), (d) origin of study (Asian vs Non-Asian country), (e) mean age (≥65 years vs <65 years), (f) proportion of males (≥40% vs <40%), (g) mean size (≥2 cm vs <2 cm), (h) type of subsolid nodule (GGO dominant only vs inclusion of solid dominant), (i) imaging modality used for biopsy (CT vs CT fluoroscopy), (j) biopsy method (FNA vs CNB), (k) pathologic type of malignancy (lung cancer only vs lung cancer, metastasis, lymphoma), and (l) minimum follow-up period (≥2 years vs <2 years).

All statistical analyses were performed using Stata v. 15.0 (StataCorp, College Station, TX) with ‘midas’, ‘metandi’, ‘metaprob’ modules and R 3.6.1 (R Foundation for Statistical Computing, Vienna, Austria)) with ‘mada’ and ‘meta’ modules. Two-tailed p < 0.05 were considered statistically significant.

Results

Study search

Our search of the online literature yielded 3378 studies: 882 in PubMed, 2350 in EMBASE, and 46 in the Cochrane library. After removing 797 duplicate studies, 2569 were excluded after review. Finally, 12 studies were included in the meta-analysis.^18–29 The literature search and selection that we conducted is summarised in Figure 1.

Figure 1.

Study search and selection.

Study characteristics

A total of 744 biopsies from 685 patients in the 12 studies were included in the meta-analysis. The study characteristics and demographics of the 12 studies are summarised in Table 1. Of the 12 studies included, 11 were retrospective^{18–22,24–29} and 1 was prospective.²³ 11 studies examined subsolid nodules and one study examined pure GGO nodules only. Eight studies used CNB and four used FNA. Only 5 out of 12 studies fulfilled the minimum follow-up period of at least 2 years.

Table 1.

Study characteristics and demographics

Study	Year	Country	Study design	No. of patients	No. of biopsy	No. of indetermined biopsy	Mean age (years)	Male patients (%)	Mean size (range) (mm)	Nodule typea	No. of pure GGOb	CNB vsFNA	Needle size (gauge)	Image guidance	Histology of malignancy
Kim et al23	2013	Korea	prospective	42	46	4	60	50%	20 (N/A)	GGO dominant	12 (26%)	FNA	20 or 22	CT	Lung cancer, metastasis, lymphoma
Hur et al19	2009	Korea	retrospective	28	28	0	62	57%	18 (N/A)	GGO dominant	7 (25%)	FNA	22	CT fluoro	Lung cancer only
Inoue et al21	2012	Japan	retrospective	79	83	16	65	33%	12 (5–35)	pure GGO only	83 (100%)	CNB	20	CT fluoro	Lung cancer only
Yamagami et al24	2013	Japan	retrospective	73	85	0	69	48%	14 (4–30)	GGO dominant, solid dominant	23 (27%)	CNB	18	CT fluoro	Lung cancer only
Maxwell et al25	2014	US	retrospective	32	40	8	67	56%	21 (8–62)	GGO dominant, solid dominant	5 (13%)	FNA	22	CT fluoro	Lung cancer, metastasis
Yun et al27	2018	Korea	retrospective	54	54	0	65	54%	36 (8–115)	Part solid, solidc	N/A	CNB	18	CT	Lung cancer only
Kiranantawat et al28	2019	US	retrospective	86	89	3	71	31%	25 (9–95)	GGO dominant, solid dominant	N/A	FNA	22	CT	Lung cancer, lymphoma
Munir et al26	2017	England	retrospective	51	52	4	71	24%	33 (N/A)	GGO dominant, solid dominant	7 (13%)	CNB	20	CT	Lung cancer, metastasis
Kim et al18	2008	Korea	retrospective	50	53	7	61	48%	19 (7–45)	GGO dominant	21 (40%)	CNB	18 or 20	CT	Lung cancer, metastasis
Lu et al22	2012	Taiwan	retrospective	55	55	5	62	20%	17 (5–30)	GGO dominant	30 (55%)	CNB	20	CT	Lung cancer only
Yamauchi et al20	2011	Japan	retrospective	67	90	23	65	46%	17 (6–47)	GGO dominant	36 (40%)	CNB	18	CT fluoro	Lung cancer only
Halpenny et al29	2020	US	retrospective	68	69	1	71	40%	19 (13–29)	GGO dominant, solid dominant	27 (39%)	CNB	18 or 20	CT	Lung cancer only

CNB, core needle biopsy; FNA, fine-needle aspiration; GGO, ground glass opacity.

^aGGO dominant means solid component within the GGO lesion is below 50%, and solid dominant means solid component within the GGO lesion is above 50%.

^bThe definitions of pure GGO are as follows: GGO >90%,^18–20,23 GGO >95%,²¹ with no solid part.^{22,24–26,29}

^cOnly part solid lesions, i.e. GGO dominant or solid dominant, were extracted for analysis. Solid lesions were excluded.

Quality assessment of selected studies

The reporting quality of the studies was assessed using the QUADAS-2 scoring system and is summarised in Table 2. Of the 12 studies, 7 satisfied 5 or more of the 7 QUADAS criteria. The reference standard was the domain where a majority of the studies had high risk of bias and applicability concerns.

Table 2.

Assessment of reporting quality by QUADAS-2 scoring system

Study	Risk of bias				Applicability concerns
	Patient selection	Index test	Reference standard	Flow and timing	Patient selection	Index test	Reference standard
Kim et al., 201323	Low risk	Low risk	Low risk	High risk	Low risk	Low risk	Low risk
Hur et al., 200919	Low risk	Low risk	High risk	Low risk	Low risk	Low risk	High risk
Inoue et al., 201221	Low risk	Low risk	High risk	High risk	High risk	High risk	High risk
Yamagami et al., 201324	Low risk	Low risk	High risk	Low risk	Low risk	Low risk	High risk
Maxwell et al., 201425	Low risk	Low risk	Low risk	High risk	Low risk	Low risk	Low risk
Yun et al., 201827	Low risk	Low risk	High risk	Low risk	Low risk	Low risk	High risk
Kiranantawat et al., 201928	Low risk	Low risk	Low risk	Low risk	Low risk	High risk	Low risk
Munir et al., 201726	High risk	Low risk	High risk	Low risk	Low risk	Low risk	High risk
Kim et al., 200818	Low risk	Low risk	High risk	High risk	Low risk	High risk	High risk
Lu et al., 201222	Low risk	Unclear risk	High risk	Low risk	High risk	Unclear risk	High risk
Yamauchi et al., 201120	Low risk	Unclear risk	Low risk	High risk	Low risk	Unclear risk	Low risk
Halpenny et al., 202029	Low risk	Unclear risk	Low risk	Low risk	Low risk	Unclear risk	Low risk

QUADAS, Quality Assessment of Diagnostic Accuracy Studies.

Overall diagnostic performance of image-guided percutaneous biopsy

The sensitivity and specificity of the 12 included studies ranged from 70–100% and 90–100%, respectively. The pooled sensitivity and specificity were 90% (95% confidence interval [CI]: 85–94%) and 99% (95% CI: 92–100%), respectively. The forest plot and HSROC of sensitivity and specificity are shown in Figures 2 and 3, respectively. Cochran’s Q test (Q = 36.39, p < 0.001), and the I² test (I² = 69.77%, 95% CI: 51.82–87.72%) revealed moderate and significant heterogeneity in sensitivity, respectively. However, the heterogeneity in specificity within studies was not significant (Q = 11.69, p = 0.39; I² = 5.85).³⁰ The overall arrangement of the coupled forest plot did not follow a V or inverted-V shape, and the Spearman’s correlation coefficient between the false-positive rate and sensitivity was 0.22 (95% CI: −0.11–0.55). Since the contribution of threshold effect to heterogeneity was not remarkable, subgroup analysis was warranted to determine the sources of heterogeneity in sensitivity. Results of the subgroup analysis and meta-regression for each covariate are presented in Table 3. A mean age above 65 years was the only covariate significantly associated with higher sensitivity (93% vs  85%) (p = 0.04). CNB showed higher sensitivity than FNA (93% vs  83%), but with marginal statistical significance (p = 0.07). The sensitivity was not significantly different comparing 18G CNB vs 20G CNB (89% vs  92%, p = 0.51).^{20–22,24,26,27} The Deeks’ funnel plot and asymmetry test (p = 0.17 for the slope coefficient) indicated no influence of publication bias on our meta-analysis (Figure 4).

Figure 2.

Forest plot for sensitivity and specificity of image-guided percutaneous transthoracic needle lung biopsy for subsolid pulmonary nodules.

Figure 3.

Hierarchical summary receiver operating characteristic (HSROC) curve of image-guided percutaneous transthoracic needle lung biopsy for subsolid pulmonary nodules.

Table 3.

Subgroup analyses and meta-regression for sensitivity

	No. of studies	Sensitivity (%) [95% confidence interval (%)]	p-value
Included number of patients			0.31
≥60	5	93 [84; 97]
<60	7	87 [80; 92]
Year of publication			0.99
Before 2013	5	90 [80; 95]
2013 and beyond	7	91 [83; 95]
Country			0.22
Asian country	8	88 [81; 92]
Non-Asian country	4	94 [87; 97]
Mean age			0.04a
≥65 years	6	93 [89; 96]
<65 years	6	85 [77; 91]
Proportion of male			0.33
≥40%	7	88 [80; 93]
<40%	5	93 [85; 97]
Mean size			0.64
≥2 cm	5	88 [81; 93]
<2 cm	7	92 [84; 96]
Type of GGN			0.22
GGO dominant only	6	87 [77; 94]
Including solid dominant	6	92 [88; 95]
Imaging modality used for biopsy			0.50
CT	7	92 [85; 96]
CT fluoroscopy	5	88 [79; 93]
Biopsy method			0.07
FNA	4	83 [71; 91]
CNB	8	93 [87; 96]
Pathology of malignancy			0.96
Lung cancer only	5	89 [85; 93]
Lung cancer, metastasis, lymphoma	7	91 [82; 96]
Minimum follow-up period			0.78
≥2 years	4	92 [74; 98]
<2 years	8	90 [85; 93]

CNB, core needle biopsy; FNA, fine-needle aspiration.

^ap < 0.05

Figure 4.

The Deeks’ funnel plot and asymmetry test for publication bias of the 12 studies included in the meta-analysis.

Overall complication rate of image-guided percutaneous biopsy

The pooled rate of overall complications during image-guided percutaneous biopsy was 43% (95% CI: 25–62%). A subgroup analysis of biopsy method showed that CNB had a higher overall complication rate than FNA (52% vs  20%, p = 0.02) (Figure 5A). The overall pooled major complication rate was 0.1% (95% CI: 0–0.4%) and did not differ significantly according to biopsy method (p = 0.25) (Figure 5B). Other variables had no influence on the overall or major complication rates. The study by Kim et al²³ was excluded from the meta-analysis for complication rate due to a lack of data.

Figure 5.

The pooled rate and subgroup analysis of overall (A) and major (B) complications during image-guided percutaneous transthoracic needle lung biopsy for subsolid pulmonary nodules.

Discussion

We performed a meta-analysis of studies on the diagnostic performance of PTNB for detecting malignant subsolid pulmonary nodules when the reference standard was set to clinical diagnosis, including surgery, repeated biopsy, imaging follow-up, or clinical treatment for the lesion. The pooled sensitivity and specificity of PTNB for the diagnosis of malignancy of subsolid nodules were 90 and 99%, respectively. In a recent multicentre large-cohort study, which was not included in our analysis due to unextractable data form for subgroup analysis on subsolid nodules, Lee et al³¹ reported the sensitivity and specificity of PTNB for subsolid pulmonary nodules to be 86 and 86%, respectively. The sensitivity is comparable; however, the specificity is much higher in our meta-analysis. This may be because the specificity of this meta-analysis was calculated according to the results of each study, excluding unevaluable results^18,20,21,23 or considering unevaluable results to be negative,^19,22,26 while Lee et al³¹ calculated the specificity including unevaluable results due to insufficient specimens and considered them to be false-positives. Excluding the differences in statistical calculations, the high specificity of PTNB in our meta-analysis is consistent with previously reported values of 98–100%.^32–37 Given the known high and uniform specificity of PTNB, it is understandable that the heterogeneity in specificity within the studies is not significant in this meta-analysis.

In the subgroup analysis, a mean age above 65 years was the only covariate significantly associated with higher sensitivity. This is contrary to our expectation that, as patient age increases, breathing control worsens, resulting in increased technical failures and lower sensitivity. Larger lesion size in older patients was hypothesised as the reason for the increased sensitivity of PTNB, but there was no significant relationship between mean age (≥65 vs <65) and mean size (≥2 cm vs <2 cm, p = 0.69). Therefore, the higher likelihood of a larger invasive component in the subsolid nodules of older patients was speculated to be the reason for the higher diagnostic yield of PTNB.³⁸ The sensitivity of CNB was higher than FNA (93% vs  83%), with marginal statistical significance (p = 0.07) in a meta-regression analysis using a bivariate random-effect model. The results of our study were not able to prove the difference between FNA and CNB; this may be due to the low statistical power of the included studies, or it may be in keeping with the results of a study by Kiranantawat et al²⁸ that showed CNB is not always necessary and that accurate results can be obtained with carefully performed FNA. Further studies are required to clarify this issue.

Current management guidelines do not advise biopsy for pure GGO nodules until they have formed a considerable solid component due to potential problems with inadequate sampling and false-negative results.^39,40 In that respect, we tried to investigate the diagnostic accuracy of image-guided percutaneous transthoracic needle lung biopsy for pure GGO nodules. However, the definition of pure GGO varied from study to study, and some studies did not analyse pure GGO separately (Table 1). As a result, quantitative meta-analysis on the diagnostic accuracy of biopsy for pure GGO lesions could not be performed and this could be a topic for future research.

Our meta-analysis showed the pooled rate of overall complications during PTNB to be 47%, and CNB had a higher overall complication rate than FNA (52% vs  20%). The pooled major complication rate was low (0.1%) and did not differ significantly according to biopsy method (p = 0.25). These results are consistent with a recent meta-analysis regarding complication rates of CT-guided transthoracic lung biopsy.⁴¹ Radiologists should choose carefully between CNB and FNA, taking into account the diagnostic accuracy and complication rate of each method.

There were some potential limitations to this review. First was the relatively small number of included publications evaluating the diagnostic performance of PTNB for subsolid pulmonary nodules. Some studies suggested the diagnostic yield of PTNB for subsolid pulmonary nodules but were excluded because a 2-by-2 contingency table could not be tabulated.^31,42–45 Second, there were biases in extracting data due to variable standards in the included studies. Risk of bias was the highest in the reference standard domain, due to various minimum follow-up periods across studies for patients with negative PTNB results. However, subgroup analysis yielded that sensitivity was not affected whether a minimum follow period was above 2 years or below 2 years. Additionally, non-diagnostic results with PTNB were excluded^18,20,21,23 or included^19,22,26 in the analysis as negative results; these differences would notably affect sensitivity and such risks of bias are demonstrated in the flow and timing domain. Third, some of the included studies showed that GGO component and lesion size could affect the diagnostic yield of PTNB for subsolid pulmonary nodules.^{18–20,22,24–26} Since the raw data of the lesion size and GGO component were unavailable for each paper, our meta-analysis was only able to utilise the provided data and criteria, which varied across studies. Further studies on the characteristics of subsolid lesions that could affect the diagnostic yield of PTNB are warranted.

In conclusion, our meta-analysis showed that PTNB has acceptable performance and a low major complication rate in diagnosing subsolid pulmonary nodules. The only significant source of heterogeneity in reported sensitivities was a mean age above 65 years. Of the two PTNB methods, CNB had marginally higher sensitivity and a similar major complication rate when compared to FNA.