Modified Frailty Index to Assess Risk in Elderly Patients Undergoing Distal Pancreatectomy: A Retrospective Single-Center Study

The index population for this study was obtained from the electronic Institutional prospectively maintained database of patients submitted to pancreatic resection at the General and Pancreatic Surgery Unit of the Pancreas Institute of the University of Verona, Verona, Italy. The database was queried for patients ≥70 years of age submitted to MIDP (index population), either laparoscopic and robot-assisted, with or without splenectomy, for any disease from March 2011 to December 2019. The database was then queried for patients ≥70 years who underwent ODP (control group) during the same period.

Demographic, clinical, surgical, pathological, and postoperative data were extracted. Demographic and clinical variables included age, gender, smoking habit, body mass index (BMI), albumin, hemoglobin, and leukocyte values, presence of diabetes or other relevant comorbidities, American Society of Anesthesiologist Physical Status Classification System (ASA) score, Charlson Age Comorbidity index (CACI) [23]. When the final diagnosis was of malignancy, Ca 19–9 values and neoadjuvant therapy were added. According to the mFI, patients were stratified into non-frail (mFI = 0), mildly frail (mFI = 1–2), or severely-frail (mFI > 2), as proposed by Konstantinidis et al. [22]. This stratification was applied to the whole cohort of patients ≥70 years who underwent distal pancreatectomy, and matching was not used due to the small sample size of the MIDP group.

Intraoperative variables included intraoperative blood loss (mL), duration of surgery (min), additional resections (yes/no), conversion to ODP, level of pancreatic transection (pancreatic neck/ gastroduodenal artery level/left aortic border), transection technique (stapler/ultrasonic device/ handsewn), splenic preservation.

Ninety-day postoperative surgical data included pancreatic fistula (POPF) [24], delayed gastric emptying (DGE) [25], post-pancreatectomy hemorrhage (PPH) [26], infected abdominal collections treated with percutaneous drainage or antibiotic therapy, chyle leak [27], surgical site infections (SSI), pulmonary or cardiologic complications, and postoperative confusion. Postoperative pathologic data included final diagnosis, tumor diameter, lymph nodes harvested, lymph node status and the number of metastatic nodes (if any), and R-status.

For the description of the Institutional laparoscopic and robot-assisted techniques adopted, previously published material can be considered [28‐30]. The decision to submit patients to MIDP and the choice of the laparoscopic or the robot-assisted approach was mostly made on a collegial basis at a dedicated preoperative surgical meeting [31]. The level of transection was made based on the presumed diagnosis (at the pancreatic neck for malignant or intermediate-malignant lesion vs. at the projection of the left border of the aorta for benign lesions, when tumor dimensions allowed it). Regarding the technique of transection, two devices were adopted in the vast majority of cases; the triple row stapler reinforced with a PGA felt (NEOVEIL Endo GIA Reinforced Reload with Tri-Staple® Technology 60 mm, Covidien, North Haven, CT, USA), or an ultrasonic scalpel [32, 33]. Typically, for ODP, two Penrose-type capillary drains were placed at the end of the surgery, one close to the pancreatic stump, the other in the splenic fossa. After MIDP, one Penrose-type capillary drain was placed, close to the pancreatic stump and passing through the splenic fossa. When the patient was enrolled in the DIPLOMA trial (MIDP vs. ODP in patients with pancreatic cancer, #ISRCTN44897265), two drains were placed, one to the left into the splenic fossa and one to the right close to the pancreatic stump. Early drains removal was frequently performed based on the quality and quantity of the output, its richness in amylase, and the occurrence of postoperative acute pancreatitis [34]. Generally, in the presence of a non-amylase-rich fluid (e.g., ≤ 2000 U/L) and the absence of a sinister appearance or postoperative acute pancreatitis, the removal was performed on postoperative day three.

According to the normality tests, values were expressed as median and interquartile range, mean and standard deviation (SD), or percentage. Student’s t-test, Mann- Whitney U-test, Kruskal–Wallis test, and χ² test were used as appropriate.

A PSM analysis was performed using the preliminary univariate analysis on the entire cohort. The not balanced variables identified were chosen to compare the MIDP (index) group and the ODP (control) group. The MIDP elderly patients were randomly matched with ODP elderly patients according to a 1:2 matching, with a 0.1 caliper and an average treatment effect (ATE), by the approach (MIDP vs. ODP) as the independent variable. Variables considered for propensity score estimation included: age, CACI, splenectomy, additional resections, pancreatic transection level, management of the pancreatic stump, and presence of malignancy at the final pathology. Balance on covariates between MIDP and ODP groups was assessed and reported using absolute standardized differences (ASD) [35]. An ASD value <0.2 indicates a small difference between the two groups, identifying an excellent balance. An ASD value between 0.2 and 0.5 shows a difference, implying a good balance. An ASD value between 0.5 and 0.8 indicates a high difference, meaning sub-optimal balance. An ASD value > 0.8 resulted in a remarkable difference, suggesting a poor balance between the two groups (Supplementary Fig. 1) [36]. Statistical analyses were performed using SPSS software ver. 22 (IBM, Chicago, IL, United States) and R software (R Foundation for Statistical Computing, Vienna, Austria).

No differences were found in the patients’ baseline characteristics between MIDP and ODP groups after propensity score weighting was found (Table 3). The pancreatic transection was more frequently on the left aortic border in the MIDP group (p = 0.005).

None of the postoperative variables considered stratified differently from a statistical standpoint between the two groups (Table 4). Remarkably, the ODP group still had a higher mean length of stay than the MIDP (16 vs. 10 days, respectively, p = 0.046).

Among the baseline features, BMI and ASA score were higher in severely frail patients (p = 0.001 and p <  0.001, respectively), as shown in Table 5. Surgery’s duration increased as the severity of frailty did (p = 0.013). No differences were found in pathology data among the three groups.

Increasing and statistically significant rates of many complications were found among the three groups, where the severely frail patients worsen, compared to mildly frail and no frail ones (Table 6). In particular, this refers to any complication (76% vs. 60% vs. 44%, p = 0.022), abdominal collection (52% vs. 37% vs. 20%, p = 0.014), pulmonary complication (54% vs. 24% vs. 17%, p = 0.001), postoperative confusion (12% vs. 3% vs. 2%, p = 0.047), and Clavien-Dindo severity ≥3 events (24% vs. 14% vs. 6%, p = 0.036). The length of stay was significantly longer in severely frail patients (16 vs. 13 vs. 12 days, p = 0.018).