The DRains After Gastrectomy (DRAG) trial (Clinical Trials.gov NCT 04288661) was a prospective, non-randomized, controlled clinical trial involving patients diagnosed with gastric or esophagogastric junctional neoplasm. All surgeries were performed by experienced Foregut Surgeons in The University Surgical Department at Hippocration General Hospital in Athens, Greece, a leading referral centre for upper gastrointestinal tract diseases. Patients underwent open total gastrectomy with D2 lymph node dissection, followed by Roux-en-Y gastrointestinal tract reconstruction. This was performed in accordance with a pre-defined, ERAS-compliant perioperative departmental protocol. The study design was conducted according to the following sequence: i) A literature review was carried out to identify commonly reported risk factors for major intra-abdominal complications following total gastrectomy; ii) The most consistently cited risk factors were selected to form the basis for the drain placement criteria; iii) Eligible patients were screened according to predefined preoperative and intraoperative criteria; iv) The final decision regarding drain placement was made intraoperatively-after completion of the resection and reconstruction, and immediately prior to abdominal wall closure-based on the selected criteria; v) All patients were followed for 30 days postoperatively, in line with the institutional monitoring protocol.

As aforementioned, the first step involved a comprehensive review of the literature, focused on the use of gastrectomy for the treatment of gastric cancer, and on the identification of risk factors associated with post-operative complications, such as esophagojejunal anastomotic leak or bleeding. The present review included 11 retrospective studies and two randomized controlled trials, and these highlighted numerous risk factors associated with the aforementioned complications. Of these, the most consistently reported factors were selected for subsequent analyses. The following criteria were suggested for the placement of a drain, as the presence of these risk factors signified an increased risk of post-operative complications (15,35-47): i) Chronic cardiopulmonary disease; ii) chronic oral steroid use (≥5 mg/day prednisone equivalent for >1 month); iii) intraoperative hemodynamic instability requiring vasopressors; iv) intraoperative blood loss exceeding 250 ml (mean amount of estimated blood loss in all included studies), referring to intra-operative non-traumatic, non-vascular blood loss; v) vessel injury involving the celiac axis or the associated branches; vi) injury to adjacent structures, such as the pancreas, spleen or duodenum; vii) tension at the anastomosis, or the requirement for anastomosis reconstruction; and viii) concerns regarding the integrity of the duodenal stump, particularly due to potential staple misfiring or compromised tissue quality.

All patients that met the eligibility criteria underwent pre-operative evaluations, including clinical examinations, blood tests, imaging studies and consultations with specialist clinicians. The eligibility criteria included the following: Histologically confirmed gastric adenocarcinoma; planned open total gastrectomy with D2 lymphadenectomy; no evidence of distant metastasis (M1); absence of massive ascites or severe cachexia; no unplanned organ resections; no severe cardiovascular, respiratory, renal, hepatic, or psychiatric conditions; not currently enrolled in another clinical trial; and no contraindications such as pregnancy or poor compliance with clinical protocols. Intraoperative criteria that ruled out eligibility included findings of unresectable or metastatic disease and cases requiring conversion to subtotal gastrectomy.

Intra-operatively, all dissections were carried out with an ultrasonic vibration scalpel to minimise blood loss. Saline was used for irrigating the surgical field, and volume was monitored and excluded from the total calculated blood loss. Notably, the protocol followed by the University Surgical Department at Hippocration General Hospital does not routinely incorporate exploratory laparoscopy or peritoneal cytology. In some cases, high procedural complexity may necessitate lymphadenectomy beyond a standard D2 dissection (48). However, to maintain uniformity, only patients undergoing D2 gastrectomy were included in the present study.

Final allocation to the drain or non-drain group was made at the end of the procedure and prior to abdominal wall closure. The decision for drain was based on the aforementioned criteria. The patients for whom drain placement was decided, received a standard 10-mm (~30 Fr) silicone tube drain near the esophagojejunal anastomosis (Shanghai International Holding Corp GmbH). By contrast, no drain was placed in the patients who did not meet the criteria. A CONSORT flowchart illustrating patient enrolment, exclusion, and group allocation is provided in Fig. 1.

According to the departmental protocol, patients were gradually mobilized directly following surgery, when feasible. At 48 and 72 h post-surgery, an oral gastrografin fluoroscopy study was performed for each patient, for the detection of early anastomotic leaks. Following a routine radiological study, patients were initiated on a liquid diet, which was subsequently advanced to pureed food on the fourth post-operative day, and a soft diet on the fifth post-operative day. In the drain group, drains were monitored daily for 24-h output volume and fluid characteristics, including amylase and bilirubin concentrations. They were removed from patients in the drain group on the fifth post-operative day, when drainage volume was <50 ml within the preceding 48 h. Drain output was considered excessive if the volume exceeded 100 ml on postoperative day five (18,26) or if the drain fluid amylase level was greater than three times the upper limit of normal serum amylase on or after postoperative day three, consistent with the definition of a postoperative pancreatic fistula (49). Patients were discharged on the fifth post-operative day when no complications were observed.

Drain and non-drain groups were compared within the following cohorts: i) The overall study population; ii) patients with complications; and iii) patients without complications.

Primary outcomes were as follows: i) Pain level assessment during the first 5 post-operative days, using the Visual Analog Scale (VAS). Notably, the VAS is a 10-point scale, with 0 representing no pain and 10 indicating the highest level of pain. A score of >7 is considered high (50); ii) post-operative nausea and vomiting (PONV) within the first 5 post-operative days; iii) initiation of feeding, carried out between 48 and 72 h post-surgery. According to the ERAS-compliant institutional protocol, initiation of feeding beyond the third post-operative day is considered delayed; iv) post-operative bowel mobilization, primarily measured by the time to the first passage of flatus; v) patient mobilization, following a specific post-operative mobilization schedule for each day. Failure to meet the specific criteria was indicative of a delay in mobilization; and vi) length of hospital stay, referring to the time from the day of the operation to the day of discharge.

Secondary outcomes included: i) mortality, referring to death occurring within 30 days post-surgery; ii) readmissions due to surgical complications, occurring within 30 days post-surgery, and iii) reoperations due to surgical complications, occurring within 30 days post-surgery.

The present study was conducted in accordance with the principles outlined in the Declaration of Helsinki (51) and the Guidelines of Good Clinical Practice (52). The final study protocol and the informed consent form for participant inclusion were approved by the Institutional Review Board of Hippocration General Hospital (approval no. 1678/31-01-2020; Athens, Greece). The IRB also conducted regular assessments as required, to ensure ongoing compliance with lawful medical practice throughout the trial.

Statistical analysis was performed using R software (version, 4.3.0; R foundation for Statistical Computing). Descriptive statistics for quantitative data included descriptive characteristics, and these were expressed as the median and minimum/maximum values and quartile 1 (Q1) to quartile 3 (Q3) range, and/or for completeness reasons the mean ± standard deviation (SD). Descriptive statistics for qualitative data were expressed as the frequency of occurrence and the relevant percentage. Comparisons were performed between patients with drainage and those without drainage, and additional analyses were performed between patients with and without complications. Qualitative data were analysed using Chi-squared and Fisher's exact tests when required and quantitative data were analysed using the Mann Whitney U test, as data normality was not always confirmed (via the Shapiro Wilk test). P<0.05 was considered to indicate a statistically significant difference. For time to event analysis, that is to visualize and report time to gut mobilization and days of hospitalization, the Kaplan-Meier estimator was applied.

To address potential confounding and strengthen causal inference between study groups, a propensity score matching (PSM) approach was applied using R version 4.4.0(53). The binary clinical covariates were blood loss, vessel damage, anastomosis, stump and respiratory issues, the criterion of adjacent structures was not involved since none of the patients had such issues in addition drainage volume was also not possible to account as only two patients had abnormal drainage volume. Eventually, the application of a drain was modelled using logistic regression to estimate the propensity scores and matching was performed using 1:1 nearest neighbour matching without replacement through the MatchIt package (54) especially developed for the R environment. Covariate balance was evaluated using standardized mean differences. Moreover, to explore potential impact of possible confounders, a Rosenbaum sensitivity analysis was conducted using the rbounds package (55). Finally, subgroup analysis was performed between patients with and without a drain, separately for the patients that had complications and those who did not experience such complications.

The present study was conducted from February 2020 to March 2023. Of the 60 eligible participants, 40 were assigned to the drain group, while 20 were included in the non-drain group, as per the aforementioned criteria. The mean age of patients was 71±9.6 years. Pre-operative histological diagnosis primarily identified intestinal-type gastric cancer in 46.6% of cases, with the body of the stomach being the most common location of observed lesions (43.3%). In addition, 25% (n=15) of patients had undergone pre-operative chemotherapy due to nodal positivity (N≥1) at diagnosis, or due to unresectability observed during exploratory laparotomy (Table I). As there were no patients on chronic steroids or patients that developed intra-operative hemodynamic instability, these factors were not included in the results. Results of the present study revealed no statistically significant differences in pre-operative characteristics between the drain and non-drain groups (Table II).

The mean duration of gastrectomy was 197±37 min. Blood transfusion was required in 13.3% of patients, and 8.3% received fresh frozen plasma (FFP) only. The most common intra-operative indications for drain placement were significant blood loss (33.3%; n=20) and tension at the anastomosis site (18.3%; n=11). Further details on intra-operative characteristics are provided in Table I. Regarding age distribution, Table II shows no statistically significant difference between the two groups: the median age was 73.5 years (IQR: 63-78) in the drain group and 73 years (IQR: 66.5-77.5) in the non-drain group (P=0.8138). No other statistically significant differences were observed when comparing intra-operative characteristics between the two groups (Table II).

Post-operatively, ~56% (n=34) of patients reported severe pain during the first 5 days post-surgery. Results of the present study also revealed that high VAS scores (>7) were significantly less frequent in the non-drain group (15%; n=3) when compared with the drain group (77.5%; n=31; P<0.0001; Fig. 2).

PONV was common in the overall study population, affecting 48.3% (n=29) of patients. Significant differences were observed between groups, with a lower incidence of PONV in the non-drain group (25%; n=5) compared with the drain group (60%; n=24; P=0.0142). Thus, there was a ~5X higher risk of PONV in the drain group [odds ratio (OR), 0.22].

Surgical site infections (SSI) were observed in 12 patients (20%), with no statistically significant difference between the drain and non-drain groups.

Timely mobilization was impeded in 37% (n=22) of patients in the overall study population. Delays in mobilization were notably less frequent in the non-drain group, occurring in only 10% (n=2) of patients, compared with 50% (n=20) in the drain group (P=0.0038). Notably, these results were statistically significant.

In 28.3% of patients (n=17), feeding initiation was delayed, despite the recommendation that feeding should be initiated 48-72 h post-surgery. Notably, the delay in feeding initiation was less common in the non-drain group, occurring in 10% (n=2) of patients, compared with 37.5% (n=15) in the drain group (P=0.0340). These results were statistically significant.

Gut motility resumed at a mean of 3.58±1 days, as illustrated by the Kaplan-Meier plot (Fig. 3). The initiation of bowel movement occurred significantly earlier in the non-drain group, with a median time of 3 days, compared with 4 days in the drain group (P=0.0142).

Intra-abdominal complications, Clavien-Dindo ≥II (56,57) occurred in 15% (n=9) of patients, accounting for 10% (n=2) of the non-drain group and 17.5% (n=7) of the drain group. Notably, there was no statistically significant difference between the two groups.

Among the intra-abdominal complications observed, the most common was esophagojejunal anastomotic leak, affecting four patients in the drain group. In total, two patients required reoperation with hybrid intra-operative stent suturing (58), one patient received endoscopic stenting, and one patient succumbed during the study period. The second most common complication was post-operative bleeding, occurring in two patients. Notably, one patient experienced bleeding on Day 0, and bleeding was evident in the drain. This patient required reoperation. In addition, one patient experienced delayed bleeding on Day 9, and this was managed non-surgically. A duodenal stump leak was identified in one patient via computed tomography on Day 7, and this patient was managed conservatively. Necrotizing pancreatitis developed in one patient from the non-drain group due to a pancreatic leak detected on Day 3, and this led to reoperation. Notably, this patient died during the study period. Post-operative ileus requiring nasogastric tube placement occurred in one patient. A detailed summary of the aforementioned findings is provided in Table I.

Extra-abdominal complications occurred in 8.3% (n=5) of patients, comprising 5% (n=2) of the non-drain group and 10% (n=3) of the drain group. All extra-abdominal complications involved the respiratory system, including three cases of atelectasis, one pleural effusion and one case of respiratory failure in a patient who also experienced the pancreatic leak. However, there was no statistically significant difference between the drain and non-drain groups.

The mean length of hospital stay was 9.5±9 days (Fig. 3). Patients in the non-drain group experienced a significantly shorter length of stay with a median of 5 days, compared with 7 days in the drain group (P=0.0001).

The majority of cases were classified as stages 1B (16.7%), 2A (26.7%) and 2B (20.0%) according to the AJCC. Comparisons between the drain and non-drain groups revealed no statistically significant differences in the distribution of AJCC stages.

In the overall study population, the 30-day post-operative outcomes included a 5% mortality rate, an 8.33% readmission rate and a 10% reoperation rate we reviewed the ages of the patients who died compared with those who survived. The mean age of deceased patients was 69.7 years (76, 66 and 67 years, one female and two males), while the mean age of survivors was 71.1 years. This indicates that there was no statistically significant association (P>0.05) between age and mortality within our cohort. No other statistically significant differences were observed between the drain and non-drain groups.

Of the 40 patients who received a drain, 4 (10%) exhibited excessive drain output by postoperative day five, and 5 (12.5%) met the criteria for postoperative pancreatic fistula. Among those with excessive output, 2 patients-who also had a pancreatic fistula-developed intra-abdominal complications. All comparisons made between the two groups, including demographic, histological and intra-operative characteristics are summarized in Fig. 2.

Propensity score matching was feasible for 20 cases from each group, the covariate balance as measured via standardized mean differences had values as follows: blood loss: 1.4, vessel damage: 0.38, anastomosis: 0.12, stump: 0.57, respiratory issues: 0.57 indicative for not perfect balance between the two categories, however we applied additional sensitivity analysis.

The Rosenbaum sensitivity test evaluated how sensitive the outcomes of the matched data were to a potential presence of unmeasured confounding. Under the assumption of no hidden bias, Gamma=1 the test yielded a P-value of 0.0899 (marginally non-significant) suggesting that, in the absence of unmeasured confounding, there is weak evidence that the treatment effect differs from zero. At increased Gamma=2, the bounds of the P-value were in the range 0.0134 to 0.2635, meaning that if there is an unmeasured factor that doubles the odds of being assigned to non-drain group, the true P-value could vary from being statistically significant (0.0134) to clearly non-significant (0.2635). In summary, the sensitivity analysis indicated that there results moderately sensitive to hidden bias.

Of the 60 eligible patients, nine (15%) experienced intra-abdominal complications, classified as Clavien-Dindo ≥II. In the drain group, complications included one grade IIIa, three grade IIIb, two grade II and one grade IV. In the non-drain group, complications included one grade V and one grade II.

Patients with complications were compared with those without complications to identify perioperative factors that may influence or be influenced by complications. Fisher's exact test was used for the analysis of categorical data. Results of the present study revealed no significant differences in pre-operative characteristics between the two groups (Table III and Fig. 4).

Intra-operatively, although FFP transfusion was not initially considered a potential risk factor for developing intra-abdominal complications, we later observed that patients who developed such complications had received FFP transfusion more frequently than those without complications (P=0.0208; OR, 12.3). In addition, vessel injury was the most common indication for drain placement (P=0.05; OR, 14.3; Table III).

The analysis of post-operative characteristics revealed differences between patients with and without intra-abdominal complications. Notably, differences were observed in oral contrast, SSIs, patient mobilization delays, extra-abdominal complications, length of hospital stay (LOS), readmissions and reoperations (Fig. 4).

As aforementioned, nine patients experienced intra-abdominal complications, with seven patients included in the drain group and two patients included in the non-drain group. The most common complications included anastomotic leak and bleeding, and these significantly affected the LOS and rates of reoperation. The median day of complication detection was 5.8 (range, 1-11; Table SI). Although the number of cases in this subgroup is relatively small, the statistical analysis was performed to ensure completeness and the outcomes revealed no statistically significant differences in the detection of anastomotic leaks, post-operative day of complication detection, diagnosis based on drain content, hospitalization duration, reoperation rates or mortality between the two groups.

Drains were closely monitored for signs suggestive of intra-abdominal complications. Notably, four patients with complications were identified through drain content analysis, while three were detected based on clinical presentation and blood tests. A comparison between patients diagnosed via drain analysis and those diagnosed through other methods revealed no significant differences in terms of detection method, post-operative day of diagnosis, LOS or rates of mortality (Table SII).

A total of 51 patients did not exhibit any intra-abdominal complications. A comparative analysis was performed within this subgroup to determine whether non-abdominal adverse effects were more prevalent in the drain or non-drain group. When comparing the pre- and intra-operative characteristics of each group, no statistical significance was observed (Table SIII). Results of the present study revealed that a blood loss of >250 ml was the most common indication for drain placement in this group (P<0.001).

Moreover, statistically significant differences in pain score, delay of patient mobilization, oral feeding delay, initiation of gut motility and LOS were observed between groups post-surgery (Table IV).