Propensity score-matching analysis comparing laparoscopic and open pancreaticoduodenectomy in elderly patients

There is little evidence on the safety and benefits of laparoscopic pancreaticoduodenectomy (LPD) in elderly patients; therefore, we evaluated the feasibility and efficacy of this procedure by comparing perioperative and oncological outcomes between LPD and open pancreaticoduodenectomy (OPD) in elderly patients. We retrospectively reviewed the data of 1,693 patients who underwent PD to manage periampullary tumours at a single institution between January 2014 and June 2017. Of these patients, 326 were elderly patients aged ≥70 years, with 56 patients allocated to the LPD group and 270 to the OPD group. One-to-one propensity score matching (56:56) was used to match the baseline characteristics of patients who underwent LPD and OPD. LPD was associated with significantly fewer clinically significant postoperative pancreatic fistulas (7.1% vs. 21.4%), fewer analgesic injections (10 vs. 15.6 times; p = 0.022), and longer operative time (321.8 vs. 268.5 minutes; p = 0.001) than OPD in elderly patients. There were no significant differences in 3-year overall and disease-free survival rates between the LPD and OPD groups. LPD had acceptable perioperative and oncological outcomes compared with OPD in elderly patients. LPD is a reliable treatment option for elderly patients with periampullary tumours.


Results
We performed 1,693 consecutive PDs (262 LPDs, 39 robotic PDs, 1,392 OPDs) in patients with periampullary tumours at our institution between January 2014 and June 2017. Of these patients, 326 (146 women, 180 men) were elderly patients who were undergoing PD for resectable periampullary tumours, with a mean age of 74.6 years (standard deviation [SD]: 3.5 years), a mean body mass index (BMI) of 23.5 kg/m 2 (SD: 8.4 kg/m 2 ), and mean American Society of Anesthesiologists (ASA) score of 2.1 (SD: 0.4). Histological diagnosis was confirmed in all patients, and the most common indication for PD was distal common bile duct cancer (n = 111; 34%), followed by pancreatic cancer (n = 94; 28.8%) and ampulla of Vater cancer (n = 67; 20.6%). The mean operative time was 294.3 minutes (SD: 67.9 minutes), and the mean length of postoperative hospital stay (POHS) was 16 days (SD: 13.1). The overall complication rate was 57.7%. Most patients had either no postoperative complications (n = 138; 42.3%) or minor Clavien-Dindo grade I/II events (n = 152; 46.6%). Postoperative pancreatic fistula (POPF) was the most frequent complication, with 20.2% of patients (n = 66) having clinically significant (grade B/C) pancreatic fistula. Major morbidity (Clavien-Dindo grades III/IV/V) occurred in 36 patients (11%). No 90-day mortality occurred. overall comparison between LpD and opD in terms of perioperative outcomes in elderly patients. Of the 326 elderly patients who underwent PD, 56 (17.2%) underwent LPD and 270 (82.8%) underwent OPD. The demographic data and perioperative outcomes of patients in both groups are presented in Table 1. Patients in the LPD group had lower BMI (22.8 vs. 23.6 kg/m 2 ; p = 0.045), lower rate of preoperative biliary drainage (55.4% vs. 73.3%; p = 0.07), and a lower incidence of pancreatic cancer (25% vs. 42.6%; p = 0.0174). The LPD group had longer operative time than the OPD group (321.8 vs. 288.6 minutes; p < 0.001). There were no significant differences in age, sex, ASA score, or POHS between the two groups. Table 2 presents the in-hospital complications in both groups, classified and graded according to the Clavien-Dindo system. The LPD group showed a significant lower overall complication rate (35.8% vs. 62.2%; p < 0.001). However, the rate of major complications (Clavien-Dindo grade III-V) did not differ between the two groups (5.4% in the LPD group vs. 12.2% in the OPD group; p = 0.136). The most common postoperative complication was POPF in both groups, with a higher incidence in the OPD group (p < 0.001). Finally, the rate of clinically significant pancreatic fistula (grade B/C) was significantly lower in the LPD group (7.1%) than in the OPD group (23%; p = 0.007).
propensity score-matched comparison of perioperative outcomes in elderly patients who underwent PD (LPD vs. OPD). To control for selection bias, one-to-one propensity score matching was used (56 patients in each group), adjusted for sex, age, BMI, ASA score, preoperative biliary drainage, and pathologies, so that all baseline parameters were balanced after matching, as shown in Table 3.
The perioperative outcomes are summarised in Table 4. The LPD group had a significantly longer mean operative time than the OPD group (321.8 vs. 268.5 minutes; p < 0.001). The mean number of analgesic injections was lower in the LPD group than in the OPD group (9.6 vs. 14.3; p < 0.026). There were no significant differences in estimated blood loss, soft diet starting time, and POHS. Table 4 also presents the postoperative complications classified and graded according to the Clavien-Dindo classification system in the two groups. The overall morbidity rate was significantly lower in the LPD group than in the OPD group (35.7% vs. 57.1%, p = 0.023). There was no significant difference in major complication (Clavien-Dindo grade III-V) rate between the two groups (5.4% vs. 10.7%; p = 0.297). www.nature.com/scientificreports www.nature.com/scientificreports/ The most common postoperative complication was POPF in both groups, with a higher incidence in the OPD group (p < 0.001). Four of 56 patients (7.1%) in the LPD group and 12 of 56 patients (21.4%) in the OPD group developed a clinically significant pancreatic fistula (p = 0.031). Table 5 shows the oncological outcomes in elderly patients with periampullary tumours. There was no difference in microscopic positive margin (R1) rate between the two groups (LPD: 10.2% vs. OPD: 6%; p = 0.443). In the LPD group, five patients had positive resection margin (RM) findings. Three had an RM-positive hepatic duct (two with high-grade dysplasia, one with adenocarcinoma). Another two patients had periductal radial RM-positive findings. In the OPD group, three patients had RM-positive findings. Two had an RM-positive hepatic duct (one with high-grade dysplasia, one with adenocarcinoma). Another one patient had pancreatic radial RM-positive findings. The numbers of lymph nodes in the resected specimens did not differ between the two groups (LPD: 15.3 vs. OPD: 17.7; p = 0.152). The 3-year overall survival rates of patients in the LPD and OPD groups were 68.8% and 83.2%, respectively (p = 0.383). The 3-year disease-free survival rates of patients in the LPD and OPD groups were 53.3% and 65.6%, respectively (p = 0.71). There were no differences in Kaplan-Meier curves of overall and disease-free survival between the two groups ( Fig. 1).

Discussion
Populations are aging worldwide, with individuals living longer and the surgical needs of elderly patients increasing. In particular, elderly patients show higher prevalence of multiple morbidities and geriatric syndromes and lower physiological reserve and preoperative nutritional conditioning than those of young patients 17,18 . Several reports have found that age itself is an important risk factor for postoperative morbidity and mortality, and that it decreases the tolerability for surgical stress [19][20][21] . Minimally invasive surgery has proven to be safe and effective and has largely replaced open surgery in many procedures 8,[22][23][24] . The typical benefits of laparoscopic surgery include reduced postoperative pain and POHS length, improved mobilization, faster return to normal activity, and fewer abdominal wall complications 22,[25][26][27] . Because of these benefits of laparoscopic surgery, LPD can be considered as alternative surgical option for elderly patients. However, it is questionable whether the advantages of laparoscopic surgery are applicable to LPD performed in elderly patients.
In many large volume hospitals, improved outcomes after LPD have broadened the selection criteria for surgery, leading to the inclusion of more elderly patients, and the procedure is becoming increasingly popular among experienced pancreatic surgeons 8 . Thus, researchers must establish whether LPD is a suitable surgery in elderly patients by researching this matter in large-volume hospitals.
To evaluate the feasibility and efficacy of LPD in elderly patients with periampullary tumours, we compared perioperative and oncological outcomes between LPD and OPD in elderly patients aged ≥70 years. To our knowledge, this was the first large-scale cohort study including elderly patients that investigated the clinical feasibility of LPD to manage periampullary tumours. Although a randomised controlled trial (RCT) would be ideal to address this matter, the rarity of periampullary tumours as a surgical indication in elderly patients may prompt investigators to avoid including frail patients in prospective randomised trials. To circumvent this impracticability, we conducted a propensity score-matched study.
PD is one of the most complex general surgical procedures involving various anastomoses, and POPF caused by failure of pancreatic enteric anastomosis is the most common complication of PD 28,29 . In the present study, the overall and clinically significant rates of POPF in the LPD group were lower than those in the OPD group (p < 0.001 and p < 0.031). Previous studies have reported that laparoscopic surgery leads to similar or superior perioperative outcomes with regards to POPF 8,10,30-32 . Most surgeons recognise from experience that the development of bowel wall oedema plays a key role in anastomotic failure. One comparative study evaluated the development of bowel wall oedema during laparoscopic and open visceral surgeries, showing that laparoscopic surgery is associated with lower rates of this finding 33 . The authors concluded that prevention of bowel wall oedema may be one advantage of minimally invasive surgery, as it leads to faster anastomotic healing. Nonetheless, since there  www.nature.com/scientificreports www.nature.com/scientificreports/ was no evidence-based data, we need further research of the effects of laparoscopic surgery on anastomotic healing in the future.
Postoperative ileus and delayed gastric emptying (DGE) are the most common complications of PD. Laparoscopic surgery is associated with a lower rate of postoperative ileus than open surgery 34 , and several studies have reported that clinically significant DGE was less frequent in LPD than in OPD 12 . These decreased complication rates may occur because LPD involves minimal manipulation of the bowel and is associated with less postoperative adhesion.
In any case, it is still unclear whether laparoscopic surgery affects the incidence of postoperative pneumonia. In several studies, fewer postoperative pulmonary complications developed after laparoscopic surgery than after open surgery [35][36][37] . In this regard, diaphragmatic splinting due to postoperative pain may lead to the increased incidence of pneumonia after open surgery 38 . In a previous study, we reported the obvious benefits of LPD, including significantly less postoperative pain and faster recovery 8 . Indeed, in the present study, the mean number of analgesic injections was lower in the LPD group than in the OPD group (9.6 vs. 14.3; p < 0.026). Ambulation with an erect posture tends to exacerbate abdominal pain, especially after OPD involving a larger incision. In elderly patients, early ambulation after surgery should be encouraged, and delays in this regard have been related to the development of pneumonia and to increased POHS length 39,40 . In patients who have undergone OPD, severe abdominal pain likely causes such a delay, leading to the high rate of complications in such patients, including pneumonia. However, our data were inadequate to show any statistically significant benefits of LPD to in the prevention of postoperative pulmonary complications.
Moreover, we evaluated our results to determine whether the survival data justify the implementation of LPD in elderly patients. In particular, the important operative oncological measures of PD include the retroperitoneal margin, superior mesenteric artery margin, number of resected lymph nodes, and survival 41,42 . The number of harvested lymph nodes and the microscopic negative margin (R0) resection rate were not different between the LPD and OPD groups. The mean number of lymph nodes in the resected specimens was 15.3 in the LPD group and 17.7 in the OPD group, which is consistent with the consensus adequate lymph node retrieval number 43 . In addition, median overall survival and relapse-free survival were not significantly different between the LPD and OPD groups. The laparoscopic approach was similar to the open procedure with respect to oncological surrogates and survival in elderly patients with periampullary tumours.
In conclusion, the results of the present study indicated that LPD in elderly patients leads to acceptable postoperative and oncological outcomes, justifying the use of LPD in elderly patients with periampullary tumours.

Methods
The present study was a single-centre, retrospective, propensity score-matched case-control study that included patients who underwent PD for periampullary tumours between January 2014 and June 2017 at Asan Medical Center, Seoul, Korea. The study was approved by the institutional review board of Asan Medical Center (Approval number: 2018-0486). It was exempted from the need for informed consent for the following reasons: (1) the risk expected was not greater than Level I; (2) Gaining consent from the research participants was deemed practically www.nature.com/scientificreports www.nature.com/scientificreports/ impossible during the course of the research, and it may have seriously affected the validity of the research; (3) There was no reason to assume the subjects would refuse consent, and the risks to the subjects were extremely low, even after the need for consent was waived; (4) The exemption from consent did not infringe upon the rights or    www.nature.com/scientificreports www.nature.com/scientificreports/ welfare of the subject; (5) The study was not carried out to gain approval of medicines or medical devices, and it was not regulated by regulatory agencies. All experiments were performed in accordance with relevant guidelines and regulations.
Data were collected on patient demographics, preoperative management, operative variables, postoperative outcomes, pathological findings, and postoperative follow-up details, including survival status. The collected demographic data included age, sex, BMI in kg/m 2 , and ASA score. Operative details obtained from the anaesthesia record included surgical approach (laparoscopic/open), operative time (from incision to wound closure), estimated blood loss, and packed red blood cell transfusion. Estimated blood loss was calculated using the equation from a previous paper 44 .
The collected pathological specimen data included final pathological diagnosis, tumour size at its largest diameter, and margin status. The outcomes used to assess the oncological adequacy of LPD included oncological surrogates, such as pathological measures (i.e., number of harvested lymph nodes) and R0 vs. R1. RMs were classified by a pathologist, according to previous reports, as follows: R0, no cancer cells or high-grade dysplasia seen microscopically at the resection margin; R1, cancer cells or high-grade dysplasia present microscopically at the resection margin or within 1 mm of the margin 45 . For staging, we referred to the TNM Classification of Malignant Tumours, 7 th edition, published by the Union for International Cancer Control.
Our hospital is a high-volume centre, and 1,693 consecutive PDs were performed in patients with periampullary tumours by experienced surgeons between January 2014 and June 2017. We defined the patient over the age of 70 as elderly patient. Ultimately, the study included 326 patients over 70 years of age who had resectable periampullary tumours who underwent curative PD. We excluded 1,257 patients for the following reasons: age <70 years, major vessel invasion, adjacent organ invasion, robotic PD, double primary cancer, and previous major abdominal surgery (Fig. 2). The remaining elderly patients who underwent PD to manage periampullary tumours at our institution were divided into the following two operative groups for analysis: 56 patients in the LPD group and 270 patients in the OPD group. To control for selection bias, a comparative study was performed in 112 patients using one-to-one propensity score matching (56 patients in the LPD group, 56 in the OPD group). The primary end point of the study was perioperative outcomes.
Postoperative complications were graded according to the Clavien-Dindo classification system (minor complications: grades I and II, major complications: grades III-V) 46 . The highest complication grade in each patient was considered the final overall complication grade. POPFs were defined according to the modified 2016 International Study Group of Pancreatic Surgery consensus definitions 47 . DGE was defined as the inability to Statistical analysis. Propensity scores were estimated using logistic regression. The following covariates were included in the regression model 49,50 : age, sex, BMI, ASA score, preoperative biliary drainage, and pathologies (tumours of the pancreas, ampulla of Vater, and distal common bile duct). These covariates were selected because they can affect the choice of surgical approach and perioperative outcomes. The operative approach (laparoscopic vs. open) was entered into the regression model as a dependent variable. A 1:1 "nearest neighbour", casecontrol match without replacement was used 51,52 . Each patient treated using laparoscopic surgery was matched with the patient treated using open surgery who had the closest estimated propensity scores. Post-match baseline characteristics and operative and postoperative variables were compared between the groups using bivariate analysis. Categorical data were reported as either number or percentage and significance tested using two-tailed Fisher's exact test or the chi-square test. Continuous variables were expressed as mean and standard deviation (SD). A p-value of <0.05 was used to determine significance.
Overall survival was defined as the time interval between the date of surgery and the date of death and was censored at the last follow-up date for patients who were alive. All patients were accounted for in the follow-up. Disease-free survival was defined as the time interval between the date of surgery and the date of recurrence or death, whichever came first, and was censored at the last follow-up date for patients who were alive without recurrence. Statistical analyses were performed using SPSS version 21.0 (IBM Corp., Armonk, NY, USA).