Table 4 Characteristics and results of studies externally validating PDRI and P-PASS
Study, year (transplant type) | Risk index (including any risk groups) | Cohort size, source (study dates) | Outcome | Handling of missing data | Other sources of bias or applicability concerns | Discrimination (or other analysis of association with outcome performed) | Calibration* | Study conclusion |
|---|---|---|---|---|---|---|---|---|
Amaral PHF 2015 (All) [36] | PDRI  ≤ 1.0, > 1.0—< 1.5  ≥ 1.5 | 570 pancreas recipients four hospitals, Brazil (1996–2011) | 1-year pancreas survival | Complete case analysis (Only 27% of cohort with PDRI) | No treatment details No comparison of baseline data to derivation study Graft failure undefined Donor race coded differently to derivation Different PDRI cutoffs to derivation (no reason given) Unclear if graft survival was censored | NI | Observed: 23 pancreas failed Predicted: NI | PDRI not associated with 1-year pancreas survival |
Ayami MS 2018 (All) [37] | PDRI  < 1, 1–1.5, > 1.5 | 327 pancreas recipients at a single centre, Gernany (2002–2015) | 1-, 5-,10-year pancreas, kidney and patient survival | Complete case analysis (Only 98.5% of cohort with PDRI | No treatment details No baseline demographics of predictor variables in cohort No comparison of baseline data to derivation study Different PDRI cutoffs (and not analysed as a continuous model) Donor race set to Caucasian | NI | Observed: 72 pancreas failed Predicted: NI | PDR I > 1.5 associated with worse pancreas survival, but no differences between other risk groups PDRI not associated with kidney or patient survival |
Blok JJ 2016 (All) [8] | PDRI  < 1.24, ≥ 1.24 (median) and as CN | 349 pancreas recipients at a single centre, Germany (1984–2013) | 1-, 5-, 10-year pancreas survival (DC) | Complete case analysis (Only 98.6% of cohort with PDRI) | No treatment details Donor race set to Caucasian as not collected by Eurotransplant, No comparison of baseline data to derivation study Different cutoffs for PDRI (median) | PDRI by categories: C-statistic 0.69 (SE 0.045) PDRI as a CN model: NI | Observed: 54 pancreas failed at 1 year Predicted: NI | PDRI ≥ 1.24 associated with worse 1-, 5-, 10-years pancreas survival but not when analyzed as a CN model |
Franz C 2019 (SPK, PAK) [38] | PDRI  ≥ 1.198 vs < 1.198 (median) | 108 pancreas recipients at a single centre, Germany (2000–2017) | 1-, 5-year pancreas survival (DC) | NMD | No comparison of baseline data to derivation study PDRI with different cut-offs from original (and not analysed as a continuous model) | NI | Observed: 27 pancreas failed at 1 year Predicted: NI | PDRI ≥ 1.198 vs < 1.198 not associated with all outcomes |
Horvath S 2015 (All) [39] | PDRI  > 1.57, ≤ 1.57 | 119 pancreas recipients from single centre, UK (period not clear) | 1-year pancreas survival | Complete case analysis (Only 97% of cohort with PDRI) | No treatment details (abstract) Graft failure undefined PDRI with different cutoffs Unclear if graft survival was censored No baseline demographics of predictor variables in cohort | NI | NI | PDRI not associated with 1-year pancreas survival |
Kopp W 2016 (pancreas donors) [40] | P-PASS < 17, ≥ 17 and as CN | 10,444 pancreas donors from Eurotransplant registry (2004–2014) | Donor pancreas accepted, procured, transplanted | Complete case analysis (Only 93.7% of cohort with P-PASS) | No comparison of baseline data to derivation study Donor race set to Caucasian as not collected routinely | C-statistic for P-PASS in donors: Reported vs non-reported (0.63, 95% CI 0.62–0.63) Accepted vs non-accepted (0.68, 95% CI 0.67–0.69) Procured vs non-procured (0.68, 95% CI 0.67–0.69) Transplanted vs non-transplanted (0.73, 95% CI 0.72–0.74) | Observed: 4354 (41.7%) donor pancreas declined Predicted: 6298 (60.3%) donors with P-PASS ≥ 17 O/E ratio: 0.69 | P-PASS < 17 associated with reported, accepted, procured and transplanted pancreas donors vs P-PASS ≥ 17 |
Lan L 2010 (pancreas donors) [45] | Modified P-PASS (minus serum sodium) (CN) | 220 pancreas donors from a single centre, Australia (2005–2008) | Donor pancreas acceptance | NI | No treatment details (abstract) Serum sodium not included as not collected by registry (different to derivation) No comparison of baseline data to derivation study | NI | Observed: 93 donor pancreas declined Predicted: NI | Modified P-PASS associated with pancreas donor acceptance |
Mittal S 2013 (SPK and PTA) [41] | PDRI 0.64–0.85 0.86–1.15 1.16–1.56 1.57–2.11 2.11–2.56  > 2.57 | 90 pancreas recipients from a single centre, UK (2011) | 1-year pancreas survival | NMD | Graft failure undefined No treatment details | NI | O/E ratio: 0.84–1.1 (SPK) 0.76–1.12 (PTA) | Higher PDRI quintiles associated with 1-year pancreas survival for SPK but not PTA |
Mittal S 2015 (All) [42] | PDRI (CN) | 1265 pancreas recipients from a national registry, UK (2004–2011) | 1-year pancreas survival (DC) | Complete case analysis (Only 76% of cohort with PDRI) | None | NI | Observed: NI Predicted: NI | PDRI associated with 1-year pancreas survival for SPK |
|  | PDRI 0.64–0.85 0.86–1.15 1.16–1.56 1.57–2.11 |  |  |  |  | NI | NI | PDRI quartile 0.64–0.85 associated with better 1-year pancreas survival for SPK vs quartiles 1.16–1.56, 1.57–2.11 |
Rodriguez-Villar C 2018 (Pancreas donors) [34] | P-PASS  < 17, ≥ 17 | 78 pancreas donors at a single centre, Spain (2016–2017) | Donor pancreas acceptance | NMD | No comparison of baseline data to derivation study | NI | Observed: 41 donor pancreas declined Predicted: 57 donor pancreas declined O/E ratio: 0.72 | P-PASS associated with donor pancreas acceptance |
Salamanca-Bustos JJ 2016 (SPK) [9] | PDRI 0.64–0.85 0.86–1.15 1.16–1.56 1.57–2.11 2.12–2.86 | 126 pancreas recipients at a single-centre, Spain (2000–2015) | 1-year pancreas survival | Complete case analysis (Only 94% of cohort with PDRI) | All donors set to Caucasian No treatment details No comparison of baseline data to derivation study | NI | Observed: 16 pancreas failure Predicted: NI | PDRI in quintiles not associated with 1-year pancreas survival |
Smigielska K 2018 (All) [11] | PDRI (CN) | 408 pancreas recipients at multiple centres, Poland (1998–2015) | 1-year pancreas survival | Complete case analysis (Only 73% of cohort with PDRI) | No treatment details No comparison of baseline data to derivation study | AUROC for PDRI: 0.524 | Observed: 139 pancreas failure Predicted: NI | PDRI not associated with 1-year pancreas |