DGF is an important and intricate complication after kidney transplant surgery. Its mechanism is not completely understood. Previous studies have shown that DGF influences the transplant surgery outcome from DCD [25-27]. DGF is universally defined as the need for hemodialysis at 7 days post-transplant. This dialysis-based definition is subjective and has many other causes unrelated to renal function, such as hyperkalemia, volume overload, heart failure and so on. This fact may be a possible reason why different studies about the association of DGF and graft survival have yielded opposite results. Thus, some patients with considerable graft function can be misdiagnosed with DGF. Physicians have put forward some other DGF definitions to fill that gap. This study analyzed 6 DGF definitions based on urine output, creatinine, and necessity of dialysis in the early post-transplant stage and compared their predictive power for the transplant outcome. Furthermore, the present investigation is the first to compare DGF definitions with respect to Chinese DCD transplantation.
Using retrospective cohort data for deceased-donor kidney transplant recipients, we have shown that all DGF definitions were significantly associated with three-year GL and had considerable predictive power for this outcome in the Chinese DCD cohort despite the incidence of DGF fluctuating greatly according to the definition used. DGF was associated with a more than 3-fold three-year GL. This phenomenon could be ascribed to the reduced confidence in the recovery of recipients if they suffered from DGF. In China, kidney transplant surgeries are expensive for most families, and patients usually have great expectations of transplant outcomes. DGF does not occur in China as often as it does in Western countries (20% vs. 70%). Patients suffer from tremendous psychological pressure if DGF occurs, and transplant outcomes are negatively influenced. Our results were consistent with those of some previous studies  and contrasted with the results of others . Decruyenaere et al.  found that dialysis-based DGF was significantly associated with graft failure, with hazard ratios (HRs) ranging from 2.87 to 13.73. However, Mallon et al. found that DGF in DCD kidneys was not associated with inferior graft survival but that DGF was an independent risk factor in DBD cohorts. The authors ascribed this difference to the much more severe warm ischemic damage in DCD organs in the UK transplant population. It is possible that warm ischemia promotes the development of acute tubular necrosis, which was thought to be a characteristic of the delayed graft function [31, 32]. The longer the warm ischemia time (WIT) is, the more severe the damage will be. In our population, there were significant differences in WIT. This finding may explain why DGF has a considerable predictive power for transplant outcome in DCD transplantation, in contrast to the results in other countries.
In this study, GL was defined as resumption of maintenance dialysis, eGFR less than 10 ml/min/1.73 m2, graft excision or retransplantation and patient death. This definition embraced bad endings for both patient and graft. The ROC curve of classical DGF predicting GL overlapped with other curves except for those of Giral and Boom DGF, both of which were based on sCr variation in the early post-transplant stage. This result illustrated that definitions based on objective indicators such as creatinine were not better than the classical definition. Compared with GL, death-censored GL was a more prominent indicator and could reflect graft outcome post-transplant. Nick DGF based on the 48-h creatinine variation had the best performance in prediction because of its largest AUC value, but its ROC curve overlapped with others. Thus, we combined these top two DGF definitions and proved that this new predictive model had the largest AUC for predicting GL and the best predictive accuracy for death-censored GL. The combination of creatinine- and hemodialysis-based DGF definitions can avoid the problems mentioned above and has superior operability.
Definitions based on urine output in our study were combined with creatinine values (Shoskes DGF). In this study, it was a poor indicator for transplant outcome. Using urine output-based definitions may perplex clinicians because it is not possible to differentiate urine outputs from the native kidney and the graft. Consequently, patients with considerable residual renal function may be misdiagnosed as not having DGF because their original kidneys react well to diuretic treatment in the early post-transplant period. Simultaneously, kidneys with severe acute tubular injury might manifest as nonoliguric renal failure, indicating poor renal function companied with polyuria. In sum, urine-based DGF definitions may exhibit deviations.
The associations of creatinine-based definitions of DGF or early-stage creatinine levels and changes in these levels after transplantation have been discussed many times in the previous literature [33-36]. Using creatinine-based definitions leads to bias as well. Physicians may optimize the status of recipients in whom sCr will be reduced after hemodialysis, and existing DGF may be ignored . In addition, the predictive power of the three-year outcome is controversial. Boom DGF showed relatively poor predictive power, with AUCs of 0.72 for GL and 0.75 for death-censored GL, whereas Giral DGF showed a significantly poorer predictive performance with respect to the three-year transplant outcome, while Nick DGF showed a good performance in predicting death-censored GL. Schnuelle et al. compared the creatinine-based DGF definition with the hemodialysis-based definition and found that the creatinine-based definition had a significant association with graft failure, not in accordance with our results.
Previous studies have shown that post-transplant renal function in the first year predicts long-term kidney transplant survival. The one-year post-transplant eGFR, as the best measurement of renal function, was compared between the DGF and non-DGF groups via a Mann-Whitney U test. In 5 of the 6 definitions, a significant decrease in one-year eGFR was observed if DGF occurred, along with a decrease in the three-year eGFR. The only definition not associated with one-year eGFR was Giral DGF, which is based on renal function recovery time, and it was also the only definition not associated with three-year eGFR post-transplant.
The poor predictive value of Giral DGF, based on whether the period required for the kidney to reach creatinine clearance >10 ml/min was greater than one week, illustrated that renal function recovery time was not a good indicator of defined DGF. This definition had obvious limitations. Many other types of abnormal status may exist or coexist with DGF in the same manner as Giral DGF, such as antibody-mediated rejection, drug-toxic graft dysfunction or primary nonfunction. Unlike Nick DGF, which is based on sCr changes during the first three days, Giral DGF is focused much more on the steady result.
In summary, DGF based on the requirement for hemodialysis within the first week had the best predictive value for three-year GL, and DGF based on sCr variation during the first three days post-transplant had the best predictive value for three-year death-censored GL. A combination of the 48-h sCr reduction ratio and classical DGF can improve the AUC for GL and the predictive accuracy for death-censored GL.