Introduction

Minimally invasive distal pancreatectomy (DP) was described for the first time in 1994 [1] and it has been increasingly performed over the last 2 decades for treating lesions located in the pancreatic body–tail. A recent analysis from the American College of Surgeons National Surgical Quality Improvement Program (ACS-NSQIP) Hepato-Pancreato-Biliary (HPB) Collaborative has shown that almost half of DPs are currently performed using a minimally invasive approach [2]. Laparoscopy, thanks to the high definition of optical cameras, has improved the precision of the surgical procedure and increased the accessibility to the deep posterior part of the gland [3]. Nowadays, patients are selected for open DP (ODP) only when laparoscopic DP (LDP) is not technically feasible [4] or when absolute contraindications for laparoscopy are present. Performing LDP requires the completion of a substantial learning curve that can be achieved through a stepwise training program aimed at increasing the proportion of operations performed without need of conversion [5, 6]. Several studies have reported that a number of pre- and intraoperative features, including characteristics of the disease, increasing parenchymal thickness at the resection line, tumor location, proximity to major vessels, and patient's body habitus can affect the difficulty of LDP [7,8,9,10]. The safety and feasibility of laparoscopy have been widely reported, with an advantage of LDP as compared to ODP in terms of surgical outcomes, including reduced intraoperative blood loss, lower postoperative complications rate, decreased time to functional recovery (TFR), and shorter length of hospital stay (LOS) [11,12,13,14,15,16]. Nevertheless, it is still unclear who are the patients who really benefit from a laparoscopic approach. Specifically, those factors affecting the loss of benefit in patients submitted to LDP have never been investigated. Aim of the present study was to identify those factors influencing the loss of benefit after LDP compared to ODP.

Materials and methods

Study design

This study was conducted in accordance with the STrengthening the Reporting of OBservational studies in Epidemiology statement (STROBE) guidelines [17]. All the patients who underwent DP, with both open and laparoscopic approach, between January 2015 and December 2019 at San Raffaele Hospital (Milan, Italy), for benign or malignant lesions of the pancreatic body–tail, were retrospectively analyzed from a prospectively maintained database. An informed consent for surgery was obtained from each patient. Institutional review board approval was waived due to the retrospective nature of the study.

Definition of outcomes

Short-term surgical outcomes (intra- and postoperative) were analyzed. Intraoperative blood loss was defined as the amount of blood suctioned during the operation. TFR was defined as the time between the day of the operation and the achievement of the following items: adequate control of postoperative pain with oral medications; adequate selfmobilization (to an independent level or to the preoperative level if already impaired); adequate bowel function (passage of flatus and stool) and ability to tolerate at least two solid meals without nausea and vomiting (in the absence of intravenous fluid administration); absence of sepsis signs [16, 18]. A patient was considered as “fit for discharge” only when all these items were achieved. The severity of postoperative complications was assessed according to the Clavien–Dindo classification of surgical complications [19]. Postoperative pancreatic fistula (POPF) was defined and graded according to the latest International Study Group on Pancreatic Fistula (ISGPF) classification [20]. Postpancreatectomy hemorrhage (PPH) and delayed gastric emptying (DGE) were defined according to the International Study Group on Pancreatic Surgery (ISGPS) classifications [21, 22]. Surgical site infections (SSI) included infections of the surgical wound, intraabdominal collections, and sepsis. LOS was calculated from postoperative day 1 (POD 1) to the day of discharge. The same standardized enhanced recovery perioperative pathway [23] was applied to all the patients, regardless of the type of surgical approach. Postoperative mortality was defined as death occurred within 90 days of surgery or any in-hospital death. Readmission was defined as a new hospital admission after discharge within 30 days of surgery. A benefit was defined as an outcome that resulted significantly better in the LDP group compared to the ODP group (after PSM), thus providing a clinical advantage. Possible predictors of loss of benefit were then evaluated in the LDP group.

Data collection

Preoperative, intraoperative, and postoperative details were retrospectively retrieved from a prospectively maintained electronic database. Preoperative data included: gender, age, body mass index (BMI), type of diagnosis (other vs. pancreatic ductal adenocarcinoma), neoadjuvant therapy, and American Society of Anesthesiologists physical status classification (ASA) [24]. Elderly patients were defined as those older than 70 years. Intraoperative data included duration of surgery, calculated from skin incision to closure of the last surgical wound, intraoperative blood loss, need for blood transfusions, resection extended to other organs, and spleen preservation (for nonmalignant lesions). C-reactive protein (CRP) and serum pancreatic amylase (AMS) were routinely measured on POD 1, POD 2, and POD 3.

All cases were completed directly or under the supervision of experienced surgeons who had performed at least 30 ODPs and 30 LDPs. The choice of the surgical approach was based on tumor and patient’s characteristics, with the laparoscopic approach being adopted whenever technically feasible. Patients who required conversion from the laparoscopic to the open approach were included in the LDP group based on the intention-to-treat principle.

Statistical analysis

Categorical data were expressed as numbers and percentages. For continuous variables, median and interquartile range (IQR) were reported. Comparisons between the two groups (LDP and ODP) were performed using the Chi-square test or Fisher’s exact test, as appropriate, for categorical data. Continuous variables were compared using the Student’s t test or the Mann–Whitney U test, as appropriate. Normality of data distribution was checked by histogram visual inspection. Bonferroni correction was applied to account for multiple testing when necessary. To reduce the effect of treatment selection bias, a propensity score matching (PSM) was performed. The PSM (nearest-neighbor matching algorithm) was computed using a logistic regression model without variable selection. The following variables, related to treatment selection or to the outcomes, were included in the PSM model: gender, age, BMI, ASA score (1–2 vs. 3–4), neoadjuvant treatment, multivisceral resection, and final diagnosis (other vs. pancreatic ductal adenocarcinoma). Mean standardized differences were calculated for each variable included in the model to assess the balance of the covariates before and after PSM. Those outcomes that, after PSM, displayed a statistically significant clinical advantage from the laparoscopic approach were defined as benefit of LDP compared to ODP. The loss of benefit was defined categorizing the outcomes according to their median value in the ODP group (> 5 days for TTF and > 250 mL for intraoperative blood loss). Univariate and multivariable logistic regression analyses were performed to identify predictors of loss of benefit in the LDP group. The final multivariable models were obtained with a backward selection of variables. The following factors were evaluated as possible predictors of benefit loss: gender, age, BMI, ASA score, neoadjuvant treatment, final diagnosis, year of treatment, multivisceral resection, and duration of surgery. Duration of surgery was categorized according to the median value in the LDP group (after PSM). The year of treatment was categorized as 2017–2019 vs. 2015–2016, as by 2017, all the operating surgeons had completed their learning curve for the laparoscopic approach. To test the effect of the same predictors in the ODP group, linear regression (for intraoperative blood loss) and negative binomial generalized linear regression (for TFR) were performed. The intraoperative blood loss variable was used in the natural logarithmic scale to meet the assumptions of linear regression. Two-tailed p values were considered significant when less than 0.05. Statistical analyses were performed using SPSS for Mac, Version 25.0.

Results

Characteristics of LDP and ODP groups

Overall, 387 patients submitted to DP for benign or malignant lesions of the pancreatic body–tail between 2015 and 2019 were included. LDP was performed in 250 patients (65%), whereas the remaining 137 patients underwent ODP (35%). After PSM, 274 patients (n = 137 LDP, n = 137 ODP) were selected. Comparison between LDP and ODP groups in terms of preoperative and intraoperative variables, before and after PSM, is reported in Table 1. After PSM, all the potential confounders considered in the PS model were balanced between LDP and ODP groups (Fig. 1).

Table 1 Balance of preoperative and intraoperative features between patients who underwent open distal pancreatectomy (ODP) and those who underwent laparoscopic distal pancreatectomy (LDP), before and after propensity score matching
Full size table
Fig. 1
figure 1

Assessment of the balance of patients’ characteristics before and after propensity score (PS) matching

Full size image

Intraoperative and short-term postoperative outcomes

Intraoperative blood loss was significantly lower in the LDP group as compared to the ODP group (200 mL vs. 250 mL, p < 0.001, adj. p < 0.001), both before and after PSM (Table 2). Lower median POD 2 CRP values (116 mg/L vs. 164 mg/L, p < 0.001, adj. p < 0.001) as well as a decreased wound infection rate (1% vs. 9%, p = 0.002, adj. p = 0.044) were observed in patients submitted to LDP after PSM. No statistically significant differences in terms of clinically relevant POPF (p = 0.115, adj. p = 1.000), 30-day reoperation (p = 0.776, adj. p = 1.000) and readmission (p = 0.366. adj. p = 1.000) rates were found between the two groups after balancing for all the potential confounders. After PSM, the median LOS was the same in patients who underwent LDP and ODP (7 days for both, p = 0.493, adj. p = 1.000), whereas median TFR was significantly shorter in the LDP group as compared to the ODP group (4 days vs. 5 days, p < 0.001, adj. p = 0.002).

Table 2 Comparison of intraoperative and postoperative outcomes between patients who underwent open distal pancreatectomy (ODP) and those who underwent laparoscopic distal pancreatectomy (LDP), before and after propensity score matching
Full size table

Risk factor analysis for prolonged TFR and high intraoperative blood loss in the LDP group

Based on median values in the ODP group, TFR > 5 days and intraoperative blood loss > 250 mL were considered as loss of benefit provided by LDP compared to ODP. Logistic regression analysis was then performed to identify risk factors predicting loss of benefit in patients submitted to LDP and selected after PSM (n = 137). A similar analysis was not performed for the outcome wound infection rate, since only one event occurred in the LDP group. At univariate analysis, age > 70 years at the time of surgery was the only factor significantly associated with a TFR > 5 days (p = 0.022). This result was confirmed at multivariable regression analysis [Odds Ratio (OR) 2.774, 95% Confidence Interval (CI) 1.159; p = 0.022] (Table 3). Elderly patients showed a risk of prolonged TFR (> 5 days) after LDP almost three times higher compared to younger patients (Fig. 2).

Table 3 Univariate logistic regression analysis and final multiple regression model for the outcome increased time to functional recovery (> 5 days) in patients who underwent laparoscopic distal pancreatectomy selected after propensity score matching (n = 137)
Full size table
Fig. 2
figure 2

Percentage of patients with time to functional recovery > 5 days vs. ≤ 5 days according to age at the time of surgery

Full size image

Regarding intraoperative blood loss, at univariate analysis, a duration of surgery > 208 min (median value in the LDP group) was the only factor significantly associated with an intraoperative blood loss > 250 mL (p = 0.019). This result was confirmed at multiple regression analysis (OR 2.957, 95% CI 1.193; 7.328; p = 0.019) (Table 4). A laparoscopic operative time > 208 min showed a risk of high intraoperative blood loss (> 250 mL) almost three times higher compared to a laparoscopic operative time ≤ 208 min (Fig. 3).

Table 4 Univariate logistic regression analysis and final multiple regression model for the outcome high intraoperative blood loss (> 250 mL) in patients who underwent laparoscopic distal pancreatectomy (n = 137) selected after propensity score matching
Full size table
Fig. 3
figure 3

Percentage of patients with blood loss > 250 mL vs. ≤ 250 mL according to duration of surgery

Full size image

Comparison between LDP and ODP outcomes in patients at risk of benefit loss

Overall, 84 patients aged > 70 years were submitted to DP. Of these, 39 patients (46%) underwent ODP and 45 (54%) underwent LDP. When only patients > 70 years (n = 84) were considered, no statistically significant difference in terms of rate of prolonged TFR was observed among ODP and LDP groups (ODP n 17, 44% vs. LDP n 14, 31%, p = 0.237). Median TFR was 5 days in both groups (p = 0.102). When only surgical procedures > 208 min were considered, the rate of increased intraoperative blood loss (60% vs. 35%, p = 0.003) as well as the median intraoperative blood loss [250 mL (IQR 150; 400 mL) vs. 200 mL (IQR 100; 300 mL), p = 0.003] were confirmed to be significantly higher in the ODP group as compared to the LDP one.

Predictors of TFR and intraoperative blood loss in the ODP group

Patients submitted to ODP and selected after PSM (n = 137) were considered. No factors influencing TFR after ODP were retained in the final multivariable model after backward variable selection (Online Resource 1). Multivariable linear regression analysis identified gender (female, B = − 0.327, p = 0.011) and BMI (> 25 kg/m2, B = 0.323, p = 0.014) as factors affecting intraoperative blood loss after ODP (Online Resource 1).

Discussion

LDP is associated with an improved postoperative course as compared to ODP [18, 25,26,27,28]. Consistently, the present experience reported a benefit in terms of TFR, intraoperative blood loss and wound infection rate in patients submitted to LDP. The rationale of the present study was based on the hypothesis that specific subgroups of patients may not benefit from the advantages provided by a minimally invasive approach. Thus, identifying preoperative and intraoperative factors associated with this loss of benefit could be helpful in selecting the optimal surgical approach for each patient.

The present study identified age > 70 years as an independent predictor of prolonged TFR after LDP and that an operative time > 3.5 h negatively influenced intraoperative blood loss in the LDP group.

A faster recovery after surgery as well as a decrease intraoperative blood loss have been widely reported as advantages of laparoscopy over the traditional open approach [29,30,31,32]. A multicenter randomized-controlled trial (LEOPARD) assessed the short-term outcomes of minimally invasive DP (MIDP) compared to ODP, showing a significant decrease in intraoperative blood loss and TFR in the MIDP group, with a comparable rate of overall complications [33]. These findings were supported also by the results of a recent single-center randomized-controlled trial reporting shorter LOS and TTF as well as reduced intraoperative bleeding after LDP compared to ODP [34]. The present experience found that LOS was not affected by the type of approach, although TFR was 1 day shorter in the LDP group. This result can be explained by the fact that LOS is less accurate than TFR in determining short-term recovery after surgery. LOS can be affected by different factors related to surgeons’ preferences or patients’ personal needs. In the present series, these elements could have been particularly relevant as more than half of patients lived far from our center. Interestingly, the present experience identified older age (> 70 years) as the only independent predictor of prolonged TFR after LDP. An advanced age is a well-recognized risk factor for increased morbidity and mortality after pancreatic resection, due to its strong correlation with frailty syndrome and potentially severe comorbidities [35,36,37]. However, most of the available studies included different types of pancreatic resections and did not perform stratified analyses according to the type of procedure and surgical approach. Therefore, whether LDP could provide an advantage over ODP in elderly patients remains unclear and available data are inconsistent. In this regard, a recent meta-analysis including 249 subjects showed that elderly patients submitted to LDP had lower blood loss and shorter LOS compared to elderly patients undergoing ODP [38]. However, these data are based on the results of only three retrospective studies [39,40,41] whose methodological quality should be carefully weighted. Among these, Chen et al. [39] reported shorter operative time, decreased blood loss, and reduced LOS for elderly patients undergoing LDP compared to ODP. However, these two groups were not properly comparable, since the significantly higher proportion of malignant tumors and splenectomy rate among elderly patients submitted to ODP. Consistently, another retrospective series [40] reported improved postoperative outcomes for elderly patients undergoing LDP. However, in this study, the comparison between LDP and ODP was also limited by a selection bias, since a significantly higher fraction of malignant lesions, larger tumors, splenectomy, and associated surgical procedures were reported among elderly patients in the ODP group. Finally, Sahakyan et al. [42] observed that elderly patients had postoperative and survival outcomes similar to those of a younger group, but no comparison with the open approach was performed. On the contrary, the present study showed that the advantage in terms of TFR related to the laparoscopic approach is lost for patients aged > 70 years. The selection bias highlighted in the previous series was overcome by the use of a propensity score matching accounting for variables related to treatment selection or to the main outcomes. This methodological approach represents one of the major strengths of the present experience. Of note, it should be pointed out that age > 70 years does not represent a contraindication for LDP, but the present data suggest that the choice of the surgical approach should be carefully weighted in elderly patients, especially when predicted surgical difficulty is high [7].

Intraoperative blood loss represented the other outcome of the present study and it was found to be negatively influenced by prolonged operative time in the LDP group. However, the advantage related to the laparoscopic approach in terms of intraoperative blood loss was retained when a subanalysis considering only operations > 3.5 h was performed.

The present study has several limitations. First, despite the application of a PSM, a possible treatment allocation bias cannot be excluded, due to the retrospective design of the study. Second, detailed data on patient comorbidity (e.g., Charlson comorbidity index) and frailty (e.g., use of Fried frailty index or Risk Analysis index), which may have aided in elderly patient risk stratification, are lacking. Additionally, data regarding other postoperative outcomes (e.g., health-related quality of life) that could represent possible advantages of LDP over ODP were not collected. Finally, the study population belongs to a high-volume center and it can be questioned whether the results obtained on these patients can be generalized also to other institutions.

In conclusion, in this large retrospective series, the advantage in terms of faster postoperative functional recovery provided by LDP over ODP appeared reduced for patients aged > 70 years. However, an advanced age should not be regarded as an absolute contraindication for LDP, as well-selected patients > 70 years might still experience an advantage from minimally invasive procedures. Therefore, the present findings should be taken particularly into account and provide guidance toward the selection of the best surgical approach in elderly patients, especially when undergoing potentially challenging operations (e.g., expecting longer operative time or high blood loss).