Is Totally Laparoscopic Total Gastrectomy Better Than Laparoscopy- Assisted Total Gastrectomy For Clinical Stage I Gastric Cancer? A Propensity-Score Matched Analysis

Shin-Hoo Park Seoul National University College of Medicine Yun-Suhk Suh (  ysksuh@gmail.com ) Seoul National University College of Medicine Tae-Han Kim Gyeongsang National University Changwon Hospital Yoon-Hee Choi Seoul National University Hospital Jong-Ho Choi Seoul National University Hospital Seong-Ho Kong Seoul National University College of Medicine Do Joong Park Seoul National University College of Medicine Hyuk-Joon Lee Seoul National University College of Medicine Han-Kwang Yang Seoul National University College of Medicine


Introduction
The Korean national cancer screening program contributed to an increase in the diagnosis of early gastric cancer (EGC), reaching 61% in 2014. In particular, the incidence of upper one-third EGC gradually increased from 11.2% in 1995 to 16.0% in 2014, according to the Information Committee of Korean Gastric Cancer [1], [2] The global incidence of cardia cancer has also grown seven-fold over the past decades [3]. In the era of minimally invasive surgery, laparoscopy-assisted total gastrectomy (LATG) or totally laparoscopic total gastrectomy (TLTG) have been highlighted with the expectation of minimal invasiveness. TLTG has not been fully standardized yet due to the technical di culty of intracorporeal esophagojejunostomy. Even a recent large prospective multicenter phase II trial (KLASS-03) reported acceptable postoperative morbidity and mortality for patients with clinical stage I gastric cancer, the procedures for esophagojejunostomies were not standardized yet [4].
Previous studies using those various surgical procedures reported the potential advantages of TLTG, such as less pain, less blood loss, and shorter operation time, than LATG [5-10]. Besides, upper abdominal pain interferes with diaphragmatic movement and subsequently worsens pulmonary complications, which can be the typical morbidity after TLTG or LATG [11-13]. However, level I evidence for these morbidities has not been established by a randomized critical trial yet, and a reasonable case matching study with a su cient sample size also has not even been reported. On the other hand, evaluating the postoperative quality of life (QoL) may provide meaningful implications for minimally invasive surgery.
Less adhesion after laparoscopic gastrointestinal surgery was reported to improve QoL such as global health status, re ux symptom, and appetite loss by enabling a comfortable diet with less pain and better peristalsis [14][15][16][17]. However, QoL after TLTG was rarely compared with that after LATG, especially at multiple time points after surgery [18].
This study aimed to evaluate and compare surgical outcomes and QoL of TLTG with those of LATG in patients with clinical stage I gastric cancer using propensity-score matching (PSM).

Study design
We reviewed the prospectively collected morbidity database of consecutive patients who underwent TLTG for clinical stage I gastric cancer between 2012 and 2018 at Seoul National University Hospital (SNUH).
Clinical staging was evaluated by preoperative esophagogastroduodenoscopy, endoscopic ultrasonography and computed tomography. In this study, TLTG was de ned as the case in which esophagojejunostomy was reconstructed intracorporeally, irrespective of intra-or extracorporeal jejunostomy. In SNUH, TLTG was performed since 2013, and four surgeons have gradually adopted TLTG according to the trend of minimal invasiveness demand, rather than separate indications for TLTG or LATG (Supplementary Figure S1). All TLTG cases, including the rst starting case with intracorporeal esophagojejunostomy by using the hemi-double stapling technique (hDST) were enrolled in this study to elucidate the safe adoption of new laparoscopic surgical skill and minimize the selection bias. TLTGs with intracorporeal esophagojejunostomy other than hDST or reduced port laparoscopic total gastrectomies were excluded (Supplementary Figure S2).
Clinicopathologic data and other operative parameters were retrospectively reviewed. Each case from the TLTG group was 2:1 propensity score matched to control cases of the LATG group. The matching variables included age, sex, body mass index (BMI), combined organ resection, and pathological T and N stages. A propensity score of each patient was estimated by logistic regression (SPSS version 25; IBM Inc., Chicago, IL, USA) and matched nearest-neighbor value within a caliper 0.02 times the standard deviation of the estimated score. After PSM, the balance of covariates between TLTG and LATG group were evaluated by calculating the standardized mean difference. Detailed method for statistical analysis was described in supplementary methods.

Surgical procedures
Laparoscopic total gastrectomy was conducted with D1+ lymph node dissection according to the Korean practice guideline for gastric cancer and Japanese gastric cancer guidelines [19,20].
For TLTG, a 3-4 cm laparotomy was made in the umbilicus or through the left lower port site after transecting the duodenum. The anvil head of the circular stapler (EEA, 25-4.8 mm, Covidien, Mans eld, MA, USA), with its rod knotted several times using a 2-0 Prolene, were brought into the peritoneal cavity.
The distal esophagus was fastened tightly with umbilical tape (32 mm width, 15 cm length, Ethicon, USA) and stretched in the direction of left lower quadrant. Then, the anterior wall of the distal esophagus was opened along the circumferential direction. The prepared anvil was inserted through the esophagotomy site and advanced into the esophagus higher than the expected proximal resection margin. By piercing the needle through the medial side of the esophageal wall, a spike of the anvil rod could be retrieved outside. The esophagus was transected by the linear stapler with 60-mm AMT, a purple cartridge (Endo GIA TM , Covidien, Mans eld, MA, USA) above the esophagotomy site. As a result, the anvil rod is located at the medial end of the staple line ( Figure 1).
After the resected stomach was brought out through the mini-laparotomy, side-to side jejuno-jejunostomy was performed at approximately 40 cm distal to the expected esophagojejunostomy site using the linear stapler with 45-or 60-mm AVM, tan cartridge (Endo GIATM, Covidien). The circular stapler was inserted into the jejunal Roux limb, fastened with a rubber band to prevent slippage. Then, the Roux limb with the circular stapler was brought into the abdominal cavity, and pneumoperitoneum was reestablished. Under a secure laparoscopic view, the jejunal Roux limb was connected to the anvil, and intracorporeal anastomosis was nally performed. The jejunal stump was closed by the linear stapler with 60-mm AVM, a tan cartridge.
For LATG, about 8.5 cm sized upper midline incision was made at the epigastrium [21]. Under the direct vision through the mini-laparotomy, a purse-string suture and device were applied to the distal esophagus, and the stomach was transected distal to the purse-string device. The anvil head of the circular stapler (EEA, 25-4.8 mm, Covidien, Mans eld, MA, USA) was inserted into the esophagus and the purse-string suture was secured to fasten the anvil rod. Then, the extracorporeal esophagojejunsotomy was performed with a 25 mm circular stapler through a mini-laparotomy incision. Extracorporeal side to side jejuno-jejunostomy was performed through the mini-laparotomy incision with a similar manner to jejunojejunostomy in TLTG.
Surgical outcome and quality of life Complication data have been prospectively collected and recorded with the consensus of the entire gastrointestinal surgical team of SNUH through the weekly conference. General postoperative management including oral care, usage of prophylactic antibiotics, and pulmonary rehabilitation was the same over the study period. Morbidity and mortality were evaluated according to the Clavien-Dindo classi cation, and comprehensive complication index (CCI) calculated by the CCI formula (https://www.assessurgery.com/) [22]. Detailed methods for determining the learning curve based on cumulative sum score and evaluating the quality of life were described in supplementary methods.

Surgical outcome
Before matching, the TLTG group (n=223) and LATG group (n=114) had no signi cant differences in baseline clinicopathologic variables (Table 1). After 2:1 PSM, the 213 patients in the TLTG group were matched to the 111 patients in the LATG group. The propensity scores, matching variables, and other remaining variables became highly balanced between TLTG and LATG groups (Supplementary Figure  S3). In terms of oncological safety, the number of retrieved lymph nodes and the distribution of TNM stage was not signi cantly different between TLTG and LATG (Table 1). In addition, the number of retrieved lymph nodes per each station was not signi cantly different across all stations between TLTG and LATG (Supplementary Figure S4).
Regarding surgical complications within postoperative 1 month, grade I pulmonary complications of the TLTG group were signi cantly lower than those of the LATG group before (0.9% vs. 5.3%, P=0.020) and after matching (0.5% vs. 5.4%, P=0.007). However, the overall rate of pulmonary complications was not different between two groups before (10.8% vs. 12.3%, p = 0.717) and after matching (9.4% vs. 12.6%, p = 0.445). The other complications, including anastomosis-related complications, were not signi cantly different between the two groups before or after matching. Regarding complications detected between the postoperative 1 month and 1 year, the incidence and detection date of delayed stenosis of esophagojejunostomy were not different between the two groups before and after matching ( Table 2).

Learning curve for TLTG
The CUSUM graph using the CCI showed two negatively sloping curve during the observation period with the trend line (y=-0.3118x+55.602) (Supplementary Figure S5a). In the rst phase, the CUSUM score gradually increased and reached the rst highest peak at case 26 (score, 118.00), then two more peak values at case 50 (score, 113.77) and case 73 (score, 80.77), and decreased until case 103 (score, -112.357). The TLTG group did not show clear decreasing pattern in operation time over chronological cases (Supplementary Figure S5b). We de ned the 26 th case as a point of overcoming the learning curve, and rationales for this was described in Supplementary Table 1. Table 3 presents the postoperative morbidity between the late TLTG group after overcoming the learning curve and the re-matched LATG group since 2012 when LATG was actively performed. The overall rate of grade I complication (2.1% vs 8.5%, P=0.016), especially pulmonary complication (0.5% vs. 4.7%, P=0.024), was still signi cantly lower in the late TLTG group than in the LATG group after matching. The overall rate of pulmonary complications was not different between two late groups (9.0% vs. 11.3%, P=0.546). Other complications, including anastomosis-related complications, were not different between the late TLTG and LATG groups before and after matching.

Quality of Life
The TLTG (n=63) and LATG (n=21) groups were matched to prospectively collected QoL data. The clinicopathologic characteristics and complications were not different between the two groups (Supplementary Table 2). During postoperative 1 year, the rates of STO22 dysphagia (P=0.028), STO22 pain (P=0.028), STO22 eating restriction (P=0.006), OG25 eating (P=0.004), and OG25 odynophagia (P=0.023) were signi cantly lower in the TLTG group (n=63) than in the LATG group (n=21) (Figure 2a (Table 4). For more robust validation of the role of TLTG, we used the anastomosis related complication and motility disorder as covariates for multivariate analysis, instead of CCI as overall complications. Still, TLTG remained as the only common independent risk factor for better QoL (Supplementary Table 3).

Discussion
This study successfully demonstrated the advantage of TLTG compared with matched LATG in terms of lower grade I pulmonary complication rate and better QoL of dysphagia, pain, or eating during postoperative 1 year. Retrospective studies cannot usually be sensitive enough to analyze parameters such as minor complications or changes in QoL and may provide false-negative or biased results. This study utilized prospectively collected complication data and QoL cohort, both of which had been recruited independently of the original purpose of this study. We believe that our study can provide less biased and more sensitive results than other unmatched retrospective studies.
The postoperative pulmonary complication was reported as one of the greatest risk factors for postoperative mortality in gastric cancer patients [12, 23,24]. In addition, total gastrectomy was an independent risk factor for pulmonary complications following laparoscopic gastrectomy [13]. Previous meta-analysis comparing LATG with open TG reported that LATG was associated with a signi cant reduction in medical complications, but a contribution from respiratory complications was not signi cant [25]. Other retrospective study limitedly demonstrated the lower incidence of pulmonary complications in the LATG group than in the OTG group, only in patients aged over 65 [26]. On the other hand, previous studies comparing TLTG and LATG mainly focused on anastomotic complications, and rarely addressed issues with pulmonary complications [1][2][3][4][5]. Upper abdominal incision causes decreased pulmonary function more frequently than lower abdominal surgery [11, 27, 28]. The mini-laparotomy wounds of the LATG are inevitably larger and located closer to epigastrium than those of TLTG. In addition, the LATG group had a higher score of STO22 pain than the TLTG group (Figure 2b). The larger incisions in the epigastrium and worse pain score may explain the limited movement of the diaphragm and deep breathing, followed by a decreased pulmonary function in the LATG group. This is the rst study comparing QoL over consecutive multiple time points during the year after operation between TLTG and LATG groups. Previous studies reported better QoL scores of C30 pain and STO22 dysphagia in the TLTG group than in the LATG group, but only investigated the QoL at a single time point and did not include OG25, more sensitive in evaluating QoL after total gastrectomy [18,29]. In this study, TLTG only determined a better QoL for dysphagia, eating, or odynophagia. DST without pursestring suture was rst introduced in 1994 as an easier alternative technique to single stapling technique (SST), but has a risk of high postoperative anastomotic stenosis rate [30,31]. To overcome this limitation, hDST was proposed, but previous studies still reported high rates of stenosis (7.3%-21%) and leakage In this study, we analyzed the CUSUM and learning curve based on the CCI, rather than operation time. In the past, standardizing multiple complications into a single variable seemed impossible due to the absence of adequate methods. However, through introducing CCI, one representative complication index per patient can be estimated. To our knowledge, this is the rst study to evaluate the learning curve using CCI. Because CCI is directly related to the patient's outcome, this approach to the learning curve is more reasonable and intuitively understandable, than previous ones based on operation time. Our study can imply that simple effort to shorten the operation time may be less meaningful during the adoption and stabilization of a novel and complex surgical technique.
This study has some limitations. Firstly, since LATG and TLTG were performed in different time periods, there might be a discrepancy in laparoscopic surgical skills or chronologic changes in clinicopathologic factors between the TLTG and LATG groups. This time trend was inevitable when comparing old and new surgical techniques in retrospective analysis. To minimize this bias, we included all patients in the TLTG group from the rst case, and all patients in the LATG group during the same period of TLTG for analysis. Besides, all surgeons at SNUH started performing TLTG in a similar period. Secondly, the sample size for QoL evaluation between TLTG and LATG was limited. In SNUH, an independent prospective cohort study was conducted to analyze only QoL, regardless of the purpose of the present study. Of these prospectively collected cohorts, we could separately selected 84 patients who met the inclusion criteria of the current study, not based on speci c criteria or intentions. Therefore, the independence of QoL data can be the unbiased evidence for current study. Despite the small sample size of QoL data, this is the rst study comparing QoL over consecutive multiple time points during the year after operation between TLTG and LATG groups. Considering that a small sample size usually has a risk of yielding false-negative or low sensitivity results, the signi cant difference in QoL between TLTG and LATG, even in the multivariate analysis, still can be valuable. However, large-scale prospective RCTs are necessary to validate more robust evidence for QoL differences.
In conclusion, TLTG with hDST were associated with reduced pulmonary complications and better QoL in terms of dysphagia, pain, eating, and odynophagia than LATG for patients with clinical stage I gastric cancer.

Consent for publication
Not applicable.

Availability of data and materials
The datasets generated and/or analyzed during the current study are not publicly available due to the governmental policy regarding the individual information, but are available from the corresponding author upon reasonable request. All authors con rmed that the content has not been published elsewhere and does not overlap with or duplicate their published work.   Stage IV 0 (0) 0 (0) 0 (0) 0 (0) *TNM stage according to AJCC, the 7 th edition.