Scoring donor lungs for graft failure risk: The Lung Donor Risk Index (LDRI)

Abstract

Lung transplantation lags behind other solid organ transplants in donor lung utilization due, in part, to uncertainty regarding donor quality. We sought to develop an easy-to-use donor risk metric that, unlike existing metrics, accounts for a rich set of donor factors. Our study population consisted of n = 26 549 adult lung transplant recipients abstracted from the United Network for Organ Sharing Standard Transplant Analysis and Research file. We used Cox regression to model graft failure (GF; earliest of death or retransplant) risk based on donor and transplant factors, adjusting for recipient factors. We then derived and validated a Lung Donor Risk Index (LDRI) and developed a pertinent online application (https://shiny.pmacs.upenn.edu/LDRI_Calculator/). We found 12 donor/transplant factors that were independently predictive of GF: age, race, insulin-dependent diabetes, the difference between donor and recipient height, smoking, cocaine use, cytomegalovirus seropositivity, creatinine, human leukocyte antigen (HLA) mismatch, ischemia time, and donation after circulatory death. Validation showed the LDRI to have GF risk discrimination that was reasonable (C = 0.61) and higher than any of its predecessors. The LDRI is intended for use by transplant centers, organ procurement organizations, and regulatory agencies and to benefit patients in decision-making. Unlike its predecessors, the proposed LDRI could gain wide acceptance because of its granularity and similarity to the Kidney Donor Risk Index.

Keywords: deceased-donor; graft failure; lung transplantation; risk prediction.

Figure 1.

Functional forms for continuous donor and transplant factors used in the LDRI. The dots are based on a separate model that, instead of using the continuous predictor, broke the predictor up into categories (8–10 categories) and then refitted the LDRI model. The log HR for each category is plotted against the respective category mid-point. The pattern in the dots represents the implied functional form since (unlike the continuous version of the predictor) no pattern is prespecified. The pattern in the dots was used to select spline knots for the LDRI formula. The line represents the final selected functional form, as applied in the LDRI formula. HR, hazard ratio; LDRI, lung donor risk index.

Figure 2.

LDRI%: Percentiles of the Lung Donor Risk Index (LDRI) based on lung transplants from 2019. The LDRI% is intended to be analogous to the Kidney Donor Profile Index, which is the percentile version of the Kidney Donor Risk Index. As an example, a donor lung with LDRI = 1.1 has LDRI% = 61, implying that this donor lung is at or below 61 % of donor lungs transplanted in 2019. Although the LDRI reflects graft failure (GF) risk relative to a median-risk donor, the LDRI% represents GF risk relative to a pool of transplanted donor lungs.

Figure 3.

Screenshot of Lung Donor Risk Index (LDRI) online application. The user inputs information that is either available or can be estimated using existing knowledge about the donor being considered. Values that are not input by the user are set to their respective reference values, ie, modal values for categorical predictors and approximate medians for continuous factors.

Figure 4.

Five-year graft survival (1 minus GF) by Lung Donor Risk Index percentile (LDRI%) groupings. GF probability is estimated nonparametrically (Nelson-Aalen method) within each of the 6 groupings: LDRI% ≤5, 6 to 25, 26 to 50, 51 to 75, 76 to 95, >95 %. GF, graft failure.

Figure 5.

Paired histograms for utilized (red bars) and non-utilized (blue bars) potential lung donors. For the purposes of this figure, the underlying data set contained 1 record for each donor (from whom at least 1 organ was recovered for transplant). “Utilized” means that at least 1 lung was recovered for transplant, while “non-utilized” refers to donors from whom 0 lungs were recovered for transplant.