Patients
This was a retrospective comparative analysis of MIAB and EUS-FNAB using data for patients with gastric SELs with intraluminal growth patterns who underwent either MIAB or EUS-FNAB at five hospitals in Japan (Kyushu University Hospital, Kyushu Medical Center, Harasanshin Hospital, Kyushu Rosai Hospital, and Kitakyushu Municipal Medical Center) between January 2010 and January 2018. No patients who met the above criteria were excluded from the study. The decision to use MIAB or EUS-FNAB was left to the primary physician for each patient.
The intraluminal growth pattern of the SELs was confirmed by preoperative gastroscopy and EUS, and SEL size was measured on EUS images. All patients who underwent biopsy (either MIAB or EUS-FNAB) were suspected before biopsy of having tumors of mesenchymal origin, such as GISTs, leiomyomas, schwannomas, and glomus tumor, on the basis of EUS findings. Written informed consent for biopsy procedures was obtained from all patients. This study was approved by the ethics committees of all five hospitals.
MIAB and EUS-FNAB procedures
For MIAB, to lift the mucosa covering the SEL and to create a safer incision, normal saline or glycerol supplemented with diluted epinephrine was injected into the submucosal layer above the lesion. The target mucosal and submucosal tissues were incised with an endoscopic submucosal dissection knife (Needle knife, Olympus, Tokyo, Japan; Flush knife, Fuji Film, Tokyo, Japan) using electrosurgical current generated by a high-frequency power supply (ICC or VIO300D; ERBE, Tubingen, Germany). After exposing the lesion, tissue samples were obtained by biopsy forceps (Radial jaw, Boston Scientific, Natick, MA, USA). Approximately 3–7 biopsy samples were obtained from each lesion.
For EUS-FNAB, 19- to 25-gauge FNAB needles (described later) were used to obtain SEL samples through the mucosa under the guidance of oblique-view EUS imaging (GF-UCT 260; Olympus). Rapid on-site evaluation (ROSE) by cytologists or pathologists was performed for all EUS-FNAB procedures. Half of each FNAB sample was used for ROSE and the other half reserved for later histological examination. The procedures were repeated a maximum of six times until either the biopsy team considered they had obtained enough samples for histology, for example, because the cytologists/pathologists had seen numerous spindle cells, indicating that the needle had reached a GIST, or the team decided to end the procedure due to difficulties in sample collection.
All samples in both groups were later evaluated histologically by pathologists. All patients were monitored daily for symptoms and signs of hematomesis and hematochezia.
Comparison between MIAB and EUS-FNAB
Three aspects of MIAB and EUS-FNAB were compared: diagnostic yield, procedural time, and adverse event rate. The procedural time was defined as the time from start to finish of the biopsy procedures. The data were collected from the operational records of patients. Major bleeding was defined as a ≥ 2 g/dL drop in blood hemoglobin.
Comparison between EUS-FNA and EUS-FNB
The following needles were categorized as FNA and FNB needles: FNA; SonoTip, Mediglobe GmbH, Rosenheim, Germany; Expect, Boston Scientific; Ez-shot, Olympus, Tokyo, Japan; and FNB: Echo-Tip Procore, Cook Medical, Bloomington, USA; Acquire, Boston Scientific. The diagnostic yields achieved by FNA and FNB needles were compared.
Statistical analysis
All statistical analyses were performed using the JMP software program version 13.0 (SAS Institute, Cary, NC, USA). Comparisons between MIAB and EUS-FNAB were performed before and after propensity-score matching of the lesion sizes in the two groups (Fig. 1). The chi-squared test or Fisher’s exact test were used to compare the categorical data (patients' characteristics, lesion locations, and histology types). Student’s t-test was used to compare continuous data (age, lesion size, and procedural time) before propensity-score matching. We used a multivariate logistic regression analysis to evaluate the relationship between the diagnostic yield and lesion size. After propensity-score matching, Student's paired t-test was used to compare continuous data in the two groups. P < 0.05 indicated statistical significance, for all tests.
To estimate propensity scores, lesion sizes (mm) were entered as independent variables into a multivariate logistic regression model. This model yielded an area under the receiver operating characteristic curve score of 0.67. Once the propensity scores were estimated, we matched patients in the two groups by setting calipers, using the stringency scores in the JMP software program, at a width equal to a distance of 0.2 from the standard deviation of the logit of the propensity score, without replacement. The effect of matching was evaluated in terms of the absolute standardized difference.