Splash M-knife versus Flush Knife BT in the technical outcomes of endoscopic submucosal dissection for early gastric cancer: a propensity score matching analysis

This study was a single-center retrospective cohort study without randomization. We collected and retrospectively reviewed data from medical records, endoscopic reports, and the database of ESD at Kitakyushu Municipal Medical Center, Fukuoka, Japan.

Between January 2012 and August 2016, ESD for early gastric cancer was performed on 445 patients at the Kitakyushu Municipal Medical Center. We excluded 15 patients with lesions in the remnant stomach or the gastric tube. We also excluded 27 patients in whom two lesions were simultaneously treated by ESD. We further excluded 249 patients who were treated by ESD using the insulated-tip or scissor-type knife. Finally, 149 patients who underwent ESD either by Flush Knife BT or Splash M-knife were enrolled in the study. Flush Knife BT was mainly used from January 2012 to June 2015 while Splash M-knife was used from July 2015 to August 2016. The flowchart of patient selection is shown in Fig. 2.

The primary endpoint in this study was the usage rate of the hemostatic forceps. The usage rate of the hemostatic forceps was defined as the percentage of lesions that required the use of hemostatic forceps at least once during the ESD procedure. According to the previous report, less than 20% of ESD using Flush Knife BT could be completed without the use of hemostatic forceps [8]. However, there were no data on the efficacy of M-knife in achieving hemostasis. We therefore hypothesized that 50% of ESD using M-knife could be completed without the use of hemostatic forceps, resulting in a 30% reduction in the usage of hemostatic forceps in this study. By assuming a 5% significance level and a statistical power of 80%, power analysis estimated that a minimum number of 39 patients would be required in each group. The secondary endpoints were procedure time for ESD, en bloc and complete resection rates, and adverse events rates (perforation and post-procedural bleeding). Procedure time for ESD was measured from the first injection after the mucosal marking to the completion of submucosal dissection. En bloc resection was defined as the whole tumor being removed in one-piece fashion. Complete resection was defined as the whole tumor being removed in one-piece fashion with tumor-free lateral and vertical margins. Perforation was diagnosed if extramural organ or fat protruding through the gastric wall was visualized during endoscopy, or if free air was seen on abdominal radiography or CT imaging. Post-procedural bleeding was diagnosed on the basis of clinical symptoms such as melena or hematemesis, or a decrease of > 2 g/dL in hemoglobin level after ESD, or bleeding proven with routine or emergency second-look endoscopy.

There were potential confounding biases between the two groups due to the retrospective nature of the study. In previous reports, including meta-analyses, large lesions, ulcerative (scar) lesions, lesions in the lesser curvature, lesions in the upper or middle position, flat or depressed morphology, the presence of gastric malignancy, male gender, antithrombotic drugs use, patients with underlying diseases (cardiomyopathy/cirrhosis/chronic kidney disease), and skill of operators were identified as prognostic factors for perioperative bleeding or technical difficulty of ESD [11‐24]. Differences in these factors between the two groups might affect clinical outcomes. To compensate for these differences, we calculated propensity scores [25] using a logistic regression model (Fig. 2) for the following covariates: age (≧75 years vs < 75 years), gender (male vs female), underlying disease (none vs with cardiomyopathy or cirrhosis), antithrombotic drugs use (not receiving or discontinuation vs continuation), tumor size (≧21 mm vs < 21 mm), lesion location (in the upper or middle third of the stomach vs in the lower stomach), lesion position (in the lessor curvature of the stomach vs other positions), macroscopic type (flat or depressed vs protruding), ulceration (presence vs absence), skill of operators (expert vs trainee), and electrosurgical unit (VIO300D vs ICC200). Patients with similar propensity scores were matched in a 1:1 ratio from the ESD-M and ESD–F groups. Calipers (0.21) with a width equal to 0.25 of the standard deviation of the logit of the propensity score were used. We estimated the area under the receiver operating characteristics curve to validate the model in this study. The area under the curve was 0.744 and had predictive power. The two groups were evaluated in terms of the absolute standardized differences before and after matching to ensure balance of propensity scoring. Characteristics were considered to be in balance if the absolute standardized difference (ASD) was within 1.96√2/n after matching [26]. In this study, all ASDs were within 1.96√2/n after matching. To analyze the baseline characteristics of the patients and outcomes of this study, The Mann-Whitney U test or t test was used for continuous data, whereas the χ² test and Fisher’s exact test were used for categorical data. Values of p < 0.05 were considered statistically significant for all tests. All analyses were performed using the JMP Pro 11.0 software.

This study was conducted in accordance with the Declaration of Helsinki. This study protocol was approved by the institutional review board of Kitakyushu Municipal Medical Center on November 18, 2016. Written informed consent was obtained from all patients before ESD.

The baseline characteristics of the 149 patients and lesions are shown in Table 1. ESD-M was conducted in 73 patients and ESD-F in 76. The percentage of expert operators was significantly higher in the ESD-M group than in the ESD-F group. There were no significant differences between the two groups with respect to other factors. Treatment outcomes before propensity matching are shown in Table 2.

En bloc resection was successfully performed in all patients from both groups. The overall rate of complete resection was 97.3%, with no significant difference between the two groups. Perforation occurred in only one patient in ESD-F. In contrast, post-procedural bleeding occurred in nine patients. Four of these were in ESD-M and the remaining five in ESD-F. Although three patients required blood transfusion, all bleeding was successfully controlled using endoscopic procedures with hemostatic forceps, and no patients required open surgical treatment. There were no treatment-related deaths. Of the 73 patients in the ESD-M group, hemostatic forceps were used during the ESD procedure in 10 patients (13.7%), compared with 65 of the 76 patients (85.5%) in the ESD-F group. Furthermore, hemostatic forceps were only required in one (3.8%) of the 26 ESD procedures performed by trainees.

The matching factors between ESD-M and ESD-F after propensity score matching are shown in Table 3. Propensity score matching analysis created 45 matched pairs, which was an adequate number for statistical requirements. The treatment outcomes are shown in Table 4. Adjusted comparisons between the two groups showed a significantly lower usage rate of hemostatic forceps in the ESD-M group compared to the ESD-F group (6.7% vs 84.4%, p < 0.001). Treatment outcomes were similar in the two groups, (complete resection rate: 95.6% vs 100%, p = 0.49; median procedure time: 74.0 min vs 71.0 min, p = 0.90 in the ESD-M and ESD-F groups, respectively). With respect to adverse events, there were no perforations in either group; post-procedure bleeding was 2.2% vs 2.2%, p = 1 in the ESD-M and ESD-F groups, respectively; and there were no treatment-related deaths. The en bloc resection rate was 100% in both groups as before matching.