Table 1 Studies deriving risk indices for solid organ pancreas transplantation
Study, year (transplant type) | Risk index (intended use) | Cohort size, source (study dates) | Outcome | Variables entered | Final model variables | Model type | Discrimination | Calibration* |
|---|---|---|---|---|---|---|---|---|
Axelrod D 2010 (All) [3] | Pancreas donor risk index (pre-transplant to inform donor organ acceptance) | 9401 pancreas recipients from SRTR registry, USA (2000–2006) | 1-year pancreas graft survival (All) | Donor: age, BMI, gender, race, height, cause of death, DCD and creatinine Recipient: age, BMI, gender, race, PRA, previous PVD, primary payment type, albumin, previous transplant Other: transplant centre, duct management, degree of HLA matching, preservation time and transplant year | Donor age, BMI, gender, black or Asian race, cause of death, creatinine, height, DCD status, CIT, PAK transplant with a CVA donor | Cox regression model (n = 1) | 0.67 | Observed: 1351 Predicted: NI |
Dorsey SG 1997 (All) [13] | Logistic regression model and a neural network model (pre-transplant) | 160 pancreas recipients from single centre, USA (1991–1996) | 3-month pancreas survival (All) | Donor: age Recipient: age, blood transfusions, smoking and alcohol history, diabetes duration, RRT pre-transplant Other: sex or weight mismatch, having a PAK/PTA transplant, use of nonlocal organ procurement centre, HLA-DR mismatch | All variables used | Logistic regression model (n = 1) | 0.78 Model sensitivity 35.7% | Observed: 23 Predicted: NI H–L p = 0.74 R2 = 0.24 |
|  |  |  |  |  |  | Neural network model (n = 1) | Correctly predicted 92.5% of cases Model sensitivity 68% Model specificity 96% | Observed: 23 Predicted: NI R2 = 0.71 |
Finger EB 2013 (All) [12] | Composite risk model (pre-transplant) | 1115 pancreas recipients from single centre, USA (1998–2011) | 3-month death-censored pancreas failure (All) | Donor: age, gender, race, cause of death, drug/alcohol abuse, pancreatitis history, BMI, DCD status, terminal creatinine, amylase, time to death from admission, CIT, PDRI Recipient: age, gender, BMI, re-transplantation, previous vascular disease, pre-transplant dialysis, smoking status Other: PRA, number of HLA mismatches, type of exocrine drainage | Donor age, BMI, CIT, terminal creatinine, presence of bladder drainage | Cox regression model (n = 1) | 0.6 (≥ 1 risk factors) 0.59 (≥ 2 risk factors) 0.52 (≥ 3 risk factors) | Observed: 10.2% graft failure Predicted: 12.8% (≥ 1 risk factor) 26.7% (≥ 2 risk factors) 42.9% (≥ 3 risk factors) |
Grochowiecki T 2014 (SPK) [32] | Logistic regression model | 112 pancreas recipients from single centre, Poland (1988–2010) | Patient survival (timepoint unclear) (SPK) | Donor: age, gender, cause of death Recipient: age, gender, diabetes duration, RRT type and duration Other: type and time of pancreas/kidney anastomosis, type of enteric anastomosis, total ischaemia time, immunosuppression type | Donor age, duration of pancreas anastomosis (vascular), dialysis duration | Logistic regression model (n = 1) | 0.8 | Observed: 14 Predicted: NI |
Kasiske BL 2013 (All) [33] | 12 Cox regression models (to model survival outcomes by transplant type for viability of prospective pooling of transplant types for future survival analysis) | 6078 pancreas recipients from SRTR registry, USA (2003–2010) | 1-year SPK graft survival | Donor: age, BMI, deceased donor, PVD at listing, cause of death Recipient: age, BMI, gender, race, duration of dialysis, age of diabetes diagnosis, whether working or hospitalized at time of transplant, previous PVD, PRA, HLA mismatch Other: CIT, retransplantation, if pre-emptive kidney transplant | Donor: age, BMI, gender, cause of death, PDRI Recipient: age, BMI, gender, race, preservation method, whether working or hospitalized at time of transplant, pre-emptive kidney transplant, diabetes duration, diabetes type, peripheral vascular disease history, terminal eGFR and eGFR on discharge, PRA, previous pancreas transplant Other: CIT | Cox regression models (n = 12) | 0.63 (0.60–0.67) | H–L p = 0.44 |
|  |  |  | 1-year PAK graft survival |  |  |  | 0.63 (0.58–0.69) | H–L p = 0.32 |
|  |  |  | 1-year PTA graft survival |  |  |  | 0.68 (0.61–0.75) | H–L p = 0.6 |
|  |  |  | 1-year SPK patient survival |  |  |  | 0.62 (0.57–0.69) | H–L p = 0.77 |
|  |  |  | 1-year PAK patient survival |  |  |  | 0.75 (0.62–0.88) | H–L p = 0.83 |
|  |  |  | 1-year PTA patient survival |  |  |  | 0.78 (0.58–0.98) | H–L p = 0.74 |
|  |  |  | 3-year SPK graft survival |  |  |  | 0.59 (0.56–0.62) | H–L p = 0.38 |
|  |  |  | 3-year PAK graft survival |  |  |  | 0.6 (0.56–0.64) | H–L p = 0.81 |
|  |  |  | 3-year PTA graft survival |  |  |  | 0.66 (0.61–0.71) | H–L p = 0.68 |
|  |  |  | 3-year SPK patient survival |  |  |  | 0.64 (0.6–0.68) | H–L p = 0.92 |
|  |  |  | 3-year PAK patient survival |  |  |  | 0.68 (0.59–0.77) | H–L p = 0.71 |
|  |  |  | 3-year PTA patient survival |  |  |  | 0.76 (0.66–0.86) | H–L p = 0.24 |
Smigielska K 2018 (All) [11] | Logistic regression model (pre-transplant) | 408 pancreas recipients, multicentre, Poland (1998–2015) | 1-year pancreas graft survival (All) | No clear specification of all baseline variables (only donor variables considered) | Donor age, BMI | Logistic regression model (n = 1) | 0.61 (0.56–0.66) | Observed: 268 Predicted: NI |
Sousa M 2014 (SPK) [35] | Two Cox regression models (pre-transplant) | 292 pancreas recipients, single centre, Brazil (2000–2010) | 3-month pancreas survival | Donor: age, BMI, gender, creatinine, sodium, amylase, norepinephrine, cause of death Recipient: age, BMI, duration of dialysis, duration of diabetes, need for dialysis, gender, cyclosporine, use of induction therapy, type of preservation fluid Other: CIT of pancreas and kidney, sequence of transplantation (pancreas or kidney first), type of duodenal anastomosis and venous drainage | Donor age, recipient BMI, use of induction therapy, iliac venous drainage, pancreas implantation first | Cox regression model (n = 2) | 0.72 (0.65–0.78) | Observed: 56 Predicted: NI |
|  |  |  | 1-year patient survival | Donor: age, BMI, gender, creatinine, sodium, amylase, norepinephrine, cause of death Recipient: age, BMI, duration of dialysis, duration of diabetes, need for dialysis, gender, cyclosporine, use of induction therapy, type of preservation fluid Other: CIT of pancreas and kidney, sequence of transplantation (pancreas or kidney first), type of duodenal anastomosis and venous drainage | Recipient BMI, use of induction therapy | Cox regression model (n = 2) | 0.67 (0.59–0.75) | Observed: 48 Predicted: NI |
Vinkers MT 2008 (pancreas donors) [4] | Pre-procurement pancreas suitability score (pre-transplant to inform donor organ acceptance) | 2175 pancreas donors from Eurotransplant database (2002–2005) | Donor pancreas acceptance for transplant | Donor: age, BMI, gender, cause of death, cardiac arrest duration, duration of ICU stay, sodium, amylase, lipase, use of vasopressors | Donor age, BMI, duration of ICU stay, duration of cardiac arrest, sodium, amylase/lipase, use of inotropes | Logistic regression model (n = 1) | NI | Observed: 45.3% of grafts declined Predicted: 42.8% risk of P-PASS ≥ 17 and graft declined O/E ratio: 1.06 |