Allogeneic hematopoietic cell transplantation risk stratification and eligibility using race-neutral versus race-specific pulmonary equations

Abstract

Abstract Background: Pulmonary function tests, forced expiratory volume in 1 s (FEV₁) and diffusing capacity for carbon monoxide (DLCO), are integral to pre-transplant evaluation of allogeneic hematopoietic cell transplantation (HCT) candidates because they estimate the risk of early post-HCT death and respiratory failure. Interpretation has historically relied on race-specific Global Lung Function Initiative equations from 2012 (GLI-2012). In 2021 the Initiative released race-neutral equations (GLI-Global). The influence of adopting GLI-Global on risk stratification, outcome prediction, and transplant eligibility is unknown. Methods: We retrospectively analyzed 8808 adults who underwent first allogeneic HCT at four United States centers between 1999 and 2021. FEV₁ values were converted to z-scores with both GLI-2012 and GLI-Global, whereas DLCO z-scores were calculated with GLI-2017 which is a race neutral equation. Pulmonary impairment was graded with the Lung Function Score (LFS); each component [FEV₁ and DLCO] receives 1 to 6 points based on percent predicted (> 80 % = 1, 70-79 % = 2, 60-69 % = 3, 50-59 % = 4, 40-49 % = 5, < 40 % = 6) and the sum of 2 to 12 maps to grades I (2), II (3-5), III (6-9), and IV (10-12). Primary endpoints were non-relapse mortality (NRM) at 120 days and 2 years, intensive care unit admission, and invasive mechanical ventilation. Prediction was assessed with Cox proportional hazards models and Harrell c-statistics. Results: Switching to GLI-Global reclassified 4.2 % of patients (368 of 8808) across LFS strata. Reclassification was racially asymmetric: 12.2 % of Black patients moved to higher-risk grades whereas 3.9 % of White patients moved to lower-risk grades. C-statistics for FEV₁ z-scores were nearly identical with GLI-Global versus GLI-2012: 2-year NRM 0.544 vs 0.542, 120-day NRM 0.566 vs 0.568, intensive care admission 0.545 vs 0.538, and mechanical ventilation 0.605 vs 0.605; confidence intervals overlapped for every comparison. Continuous z-score models consistently outperformed categorical cut-offs for each outcome: 2-year NRM 0.544 vs 0.518, 120-day NRM 0.566 vs 0.523, intensive care admission 0.545 vs 0.515, and mechanical ventilation 0.605 vs 0.529; confidence intervals did not overlap for every comparison. The composite LFS outperformed models that used FEV₁ or DLCO alone. Extrapolating cohort reclassification rates to 38,606 adults aged 40 to 84 who received allogeneic HCTs reported to the Center for International Blood and Marrow Transplant Research between 2009 and 2018 predicts that 1521 patients, including 509 Black adults, would shift LFS grades and could face altered transplant eligibility. Conclusions: Race-neutral GLI-Global equations modestly change pulmonary risk categorization, disproportionately raising scores for Black HCT candidates, yet maintain predictive accuracy for early mortality, respiratory failure, and intensive care use. Continuous dual-variable models and the composite LFS provide superior prediction. Although race-neutral interpretation aligns with current society guidelines, its tendency to classify Black individuals as higher risk highlights the need for refined continuous models that limit unnecessary reclassification and increased disparities in transplant eligibility.