Institutional Review Board approval (Groupe de Sélection des Projets Cliniques, EMBOPSIE-IPC 2021–088) was obtained to study this retrospective multicenter cohort from three tertiary referral centers. All consecutive PCTLB performed between June 2012 and December 2021 in the first center and between January 2018 and January 2021 in the second and third centers were included. The TE technique was introduced during 2018; prior to this period, no TE was performed, and after this date, the use of this technique remained at the discretion of the operating physicians. The clinical outcome was the occurrence of an in-hospital pneumothorax and the indication for insertion of a chest tube. The outcome was determined at the end of the procedure using computed tomography (CT) scan of the entire lung parenchyma. A retrospective analysis of a multicenter database was performed.

The exclusion criteria were mediastinal biopsies, pleural tumors, and tumors adjacent to the pleura without pleural crossing. Contrary to previous single-center studies, the occurrence of pneumothorax before needle removal was not an exclusion criterion. Biopsies of tumors adjacent to the pleura are known to be less at risk for pneumothorax and chest tube insertion. Making a difference between biopsies with or without the transpleural tract in this location is sometimes difficult. The evaluation of the outcomes of these procedures could be biased, which is why they were excluded.

All procedures were performed under CT guidance (SOMATOM Definition 128 scanner, SOMATOM Confidence, Siemens Healthineers, Erlangen, Germany; Revolution EVO, GE Healthcare, WI, USA; and SOMATOM Definition, Siemens Medical Solutions, Erlangen, Germany). Prior to the procedure, all patients were informed of the risk of and hemoptysis and pneumothorax. All procedures were performed by 10 interventional radiologists with > 5 years of experience. The patients were given oral information on the use of gelatin sponge slurry to reduce the risk of pneumothorax.

First, helical acquisition of the entire lung was performed before the start of the biopsy trajectory planning. Local anesthesia was performed by injecting approximatively 15 mL of 1% lidocaine on the skin, along the path and in contact with the pleura using a 22-G needle. A coaxial needle (an inner stylet and an outer cannula) [27] was then inserted through the skin into the lesion, a semi-automatic needle with a 1 to 2 cm cutting area was used (Temno Evolution Merit Medical, South Jordan, Utah; Quick Core, Cook Medical, Bloomington, IN; Supercore™, Argon Medical Devices, Athens, USA), and the specimen was placed in a formaldehyde tube for analysis. Sequential acquisitions were performed during needle positioning. Breath holding was not required from the patient during the biopsy or when the needle was removed. Immediately after the removal of the coaxial needle, a final acquisition was performed on the entire lung parenchyma to detect pneumothorax or intra-alveolar pulmonary hemorrhage. Following the procedure, patients were monitored at least for a minimum of 4 h in the medical department. In case of dyspnea during monitoring, a new CT scan was performed to ensure that no pneumothorax had appeared or increased.

Gelatin sponge slurry preparation was standardized at the three centers (supplemental data). Two thirds of the single gelatin sponge plate (Gelitaspon, Gelita Medical, Amsterdam, The Netherlands) was cut into small pieces (5 × 5 mm), loaded into a 10-mL syringe, and mixed with no more than 2 mL of iodinated contrast agent (Iopamiron, Guerbet) with a three-way stopcock to obtain a compact and dense texture. Then, the mixture was loaded into a 3-mL syringe. To test the quality of the preparation before each use, the mixture was pushed out of the syringe. If the mixture was sufficiently dense and cohesive to adhere to the syringe, the product was considered satisfactory and could be used (Fig. 1). If the mixture did not stick to the syringe, it was not validated for use and discarded, and a fresh mixture was prepared. Once the biopsy was completed, the gelatin sponge slurry was injected into the needle channel until resistance to the injection was felt (less than 1 mL). The needle trocar was then inserted at the base of the coaxial needle, and the coaxial needle was withdrawn while holding the trocar so that the gelatin sponge slurry replaced the needle tract. No other maneuvers were performed to reduce the risk of pneumothorax. The final helical CT scan also verified the absence of non-target embolization or systemic air embolism.

Pictures showing the preparation steps of the gelatin sponge slurry. a 3 × 10 mL syringes, 1 × 3 mL syringe, 1 × gelatine sponge plate, 1 × 3-way stopcock, 1 × iodine cup. b One gelatine sponge plate is cut with a scalpel. c 2/3 of the gelatine sponge plate is cut into 5 × 5 mm pieces. d The pieces of gelatine sponge are put into a 10-mL syringe and the iodine into a 2-mL syringe. e The two syringes are mixed using the 3-way stopcock. f The mixture is loaded into a 3-mL syringe. g The syringe is tested for cohesion

Variables related to the tumors were retrospectively collected from the initial procedure images (fissural tract, lesion size, patient position, needle-pleural angle, distance to the pleura, emphysema) in each center. Variables related to the patients were retrospectively collected from the computerized patient file (sex, age, history of thoracic surgery). Variables related to the procedures (time of procedure, pneumothorax, chest tube insertion, size of coaxial/needle) were retrospectively collected from the biopsy report in each center.

Puncture time was defined as the time between the first CT a with the co-axial needle in the lung and the final acquisition. The distance to the pleura was defined as the distance from the pleural puncture to the target lesion along the biopsy-track axis. Emphysema along the needle tract was evaluated qualitatively according to the Fleischner Society classification [26] and was assigned as no significant emphysema (group a) or significant emphysema (groups b to e). The needle-pleural angle was defined as the angle (°) between the line defined by the needle and the line perpendicular to the pleural tangent. A diagnosis of immediate pneumothorax was established if any air density was visible in the pleural space on the final chest CT scan. Systemic gas embolism was defined as the presence of gas in the arterial or venous circulatory system on the final acquisition. Non-target embolization was defined as the presence of an iodinated opacity within the vascular system and adjacent bronchi peripheral to the sampling area. The indication to insert a chest drain was standardized in the three centers and based on the recommendations of the Pleural Disease Guidelines 2010 established by the British Thoracic Society, according to the volume of pneumothorax and/or clinical symptoms [28]. No intervention was performed for patients without pneumothorax or for patients with a small pneumothorax (distance < 2 cm between the parietal and visceral pleura at the level of the hilum). Exsufflation was performed in asymptomatic patients with large pneumothorax. A chest tube was positioned for patients with exsufflation failure, large pneumothorax (distance > 2 cm between the parietal and visceral pleura at the level of the hilum), respiratory symptoms, or increased pneumothorax on the CT scan performed during monitoring. A chest radiograph 4 h after the procedure was not routinely performed in the three centers.

Results are presented as mean ± standard deviation (SD) or median [Q1-Q3], according to the distribution for continuous variables and as number and frequency for categorical variables. Categorical variables were compared using the chi-square test, and continuous variables were compared using Student’s t-test. If a variable had less than 5% missing data, we performed a median imputation for quantitative variables and a mode imputation for qualitative variables. If a variable had between 5 and 20% missing data, a multiple chain equation multiple imputation (MICE) was performed. The predictive factors for chest tube insertion were assessed using univariable and multivariable analyses. Odds ratios (OR) and 95% confidence intervals are reported as appropriate. For the multivariable analysis, logistic regression was performed, and we selected the candidate variables from the set of collected variables in such a way that less than 20% of patients had missing data or variables with less than 5% missing values. The predictor variables TE, emphysema, tract length, and prone position were introduced in the multivariable model based on p < 0.2 in the univariable analysis and according to a priori data from the literature. Statistical significance was set at p < 0.05. All statistical analyses were conducted using the R +  + software.

Propensity score-based matching was developed using a logistic regression model that included 8 variables (age, lesion size, tract length, sex, prone position, fissural tract, emphysema, needle size) and/or to the outcome (chest tube insertion), regardless of their statistical significance using a non-parsimonious approach. Using the propensity score, non-TE patients were matched to TE patients. A nearest-neighbor 1:1 matching algorithm was applied, with a caliper width of 0.2 SD of the logit of the propensity score. Standardized difference before and after matching were estimated (with their 95% confidence intervals) to assess the quality of the propensity score matching procedure. A mirrored histogram of distribution of propensity scores for non-TE (bars above the zero line) versus TE (bars below the zero line) was plotted. The R package MatchIt was used for the propensity score-based matching.