This 5-year evaluation provides strong evidence that the classification is valid and applicable worldwide in many fields of surgery. No modification in the general principle of classification is warranted in view of the use in ongoing publications and trials. Subjective, inaccurate, or confusing terms such as "minor or major" should be removed from the surgical literature.

OBJECTIVE To use the concept of benchmarking to establish robust and standardized outcome references after pancreatico-duodenectomy (PD). BACKGROUND Best achievable results after PD are unknown. Consequently, outcome comparisons among different cohorts, centers or with novel surgical techniques remain speculative. METHODS This multicenter study analyzes consecutive patients (2012-2015) undergoing PD in 23 international expert centers in pancreas surgery. Outcomes in patients without significant comorbidities and major vascular resection (benchmark cases) were analyzed to establish 20 outcome benchmarks for PD. These benchmarks were tested in a cohort with a poorer preoperative physical status (ASA class 3) and a cohort treated by minimally invasive approaches. RESULTS Two thousand three hundred seventy-five (38%) low-risk cases out of a total of 6186 PDs were analyzed, disclosing low in-hospital mortality (1.6%) but high morbidity, with a 73% benchmark morbidity rate cumulated within 6 months following surgery. Benchmark cutoffs for pancreatic fistulas (B-C), severe complications (grade 3), and failure-to-rescue rate were 19%, 30%, and 9%, respectively. The ASA 3 cohort showed comparable morbidity but a higher in hospital-mortality (3% vs 1.6%) and failure-to-rescue rate (16% vs 9%) than the benchmarks. The proportion of benchmark cases performed varied greatly across centers and continents for both open (9%-93%) and minimally invasive (11%-62%) PD. Centers operating mostly on complex PD cases disclosed better results than those with a majority of low-risk cases. CONCLUSION The proposed outcome benchmarks for PD, established in a large-scale international patient cohort and tested in 2 different cohorts, may allow for meaningful comparisons between different patient cohorts, centers, countries, and surgical techniques.

Risk adjustment of patient selection and technique in ALPPS resulted in a continuous drop of early mortality and major postoperative morbidity, which has meanwhile reached standard outcome measures accepted for major liver surgery.

Allocation of liver grafts triggers emotional debates, as those patients, not receiving an organ, are prone to death. We analyzed a high-Model of End-stage Liver Disease (MELD) cohort (laboratory MELD score ≥30, n = 100, median laboratory MELD score of 35; interquartile range 31-37) of liver transplant recipients at our center during the past 10 years and compared results with a low-MELD group, matched by propensity scoring for donor age, recipient age, and cold ischemia time. End points of our study were cumulative posttransplantation morbidity, cost, and survival. Six different prediction models, including donor age x recipient MELD (D-MELD), Difference between listing MELD and MELD at transplant (Delta MELD), donor-risk index (DRI), Survival Outcomes Following Liver Transplant (SOFT), balance-of-risk (BAR), and University of California Los Angeles-Futility Risk Score (UCLA-FRS), were applied in both cohorts to identify risk for poor outcome and high cost. All score models were compared with a clinical-oriented decision, based on the combination of hemofiltration plus ventilation. Median intensive care unit and hospital stays were 8 and 26 days, respectively, after liver transplantation of high-MELD patients, with a significantly increased morbidity compared with low-MELD patients (median comprehensive complication index 56 vs. 36 points [maximum points 100] and double cost [median US$179 631 vs. US$80 229]). Five-year survival, however, was only 8% less than that of low-MELD patients (70% vs. 78%). Most prediction scores showed disappointing low positive predictive values for posttransplantation mortality, such as mortality above thresholds, despite good specificity. The clinical observation of hemofiltration plus ventilation in high-MELD patients was even superior in this respect compared with D-MELD, DRI, Delta MELD, and UCLA-FRS but inferior to SOFT and BAR models. Of all models tested, only the BAR score was linearly associated with complications. In conclusion, the BAR score was most useful for risk classification in liver transplantation, based on expected posttransplantation mortality and morbidity. Difficult decisions to accept liver grafts in high-risk recipients may thus be guided by additional BAR score calculation, to increase the safe use of scarce organs.

No abstract