A Comparison of Different Comorbidity Indices in Allogeneic Hematopoietic Cell Transplantation (Allo-HCT): Comprehensibility Vs Simplicity Vs Objectivity

Background: The HCT- comorbidity index (HCT-CI) was developed almost two decades ago to capture comorbidity burden before and predict non-relapse mortality (NRM) and overall survival (OS) after allogeneic hematopoietic stem cell transplant (allo-HCT). Subsequently, the HCT-CI was augmented (aHCT-CI) by the addition of age and the values of other biomarkers, becoming even more comprehensive (Vaughn, 2015). Yet, there have been concerns about the burden of collecting data on 20 different comorbidities and the seemingly subjective definition of some comorbidities. Hence, a simplified comorbidity index (SCI) was recently proposed (Shouval, 2022) and validated (Elias, 2023) as an alternative to HCT-CI and aHCT-CI: it focuses only on four comorbidities and is mostly based on lab measures (of liver, kidney, heart, and lung function), in addition to age. This new index raises the question whether sacrificing comprehensibility for simplicity pays off when it comes to overall prognostic evaluation for NRM. Herein, we compare the performance of these three different approaches. To enucleate the role of objective measures, we generated a fourth “objective” score (oHCT-CI) including only the measurable variables of the aHCT-CI and included it in the comparisons.

Methods: we evaluated 3477 consecutive allo-HCT recipients at our center between 2008 to 2022. We calculated HCT-CI, aHCT-CI, oHCT-CI, and SCI scores, and compared the distribution of these four scoring systems in this population. Cumulative incidence curves of NRM and Kaplan-Meier curves of OS were computed for each of the collapsed risk groups: 0, 1 to 2, 3 or more. Cause-specific Cox regression and Cox regression were used to evaluate the association of each of the scores with NRM and OS, respectively. Concordance index (C-index) and time-dependent AUC (from 3 – 132 months) were calculated to evaluate the model performance.

Results: 2,006 (58%) of patients (pts) were male; 1,141 (33%) received a non-myeloablative conditioning regimen; 326 (9.4%) were treated for a non-malignant condition; 1,207 (34.7%) were younger than 40 yo, 1,286 (37%) were 40-60 yo, and 984 (28.3%) were >60 yo. The C-index (95% CI) for NRM was: 0.592 (0.573, 0.612) for the HCT-CI; 0.626 (0.606, 0.645) for the aHCT-CI; 0.615 (0.595, 0.635) for the SCI; and 0.605 (0.585, 0.624) for the oHCT-CI. The C-index (95% CI) for OS was: 0.584 (0.570, 0.599) for the HCT-CI; 0.612 (0.598, 0.627) for the aHCT-CI; 0.588 (0.573, 0.602) for the SCI; and 0.591 (0.577, 0.606) for the oHCT-CI. The averaged time-dependent area under the curve (AUC) of the four systems, calculated for NRM over 3-132 months, was: 0.623 for the HCT-CI; 0.658 for the aHCT-CI; 0.631 for the SCI; and 0.614 for the oHCT-CI. The averaged AUC for OS was: 0.614 for HCT-CI, 0.643 for aHCT-CI, 0.603 for SCI, and 0.598 for oHCT-CI.

For 952 pts with a calculated SCI score of ‘0’, 281 have aHCT-CI score ‘0’, 485 have aHCT-CT score ‘1-2’, and 186 have aHCT-CI score ‘3+’.

These 3 groups are significantly different in OS (log-rank test p-value < 0.0001). The 3-year survival probabilities are 79.7% (74.8%, 84.8%), 71.1% (66.9%, 75.6%), and 57.1% (50.0%, 65.1%), respectively. [unable to compute median survival time because it was not reached]. Compared to the ‘0’ group, the hazard of death in ‘1-2’ group is 1.41 (1.04, 1.91) times higher and ‘3+’ group is 2.49 (1.78, 3.48) times higher.

They are also significantly different in NRM (Gray’s test p-value < 0.001). The cumulative incidences of NRM at 3-year are 7.2% (4.6%, 11%), 12% (8.9%, 15%), and 20% (14%, 26%), respectively. Compared to the ‘0’ group, the hazard of NRM in ‘1-2’ group is 1.56 (1.00, 2.46) times higher and ‘3+’ group is 2.75 (1.68, 4.49) times higher.

NRM and OS were significantly affected by the impact of the other comorbidities included in the HCT-CI and aHCI-CI.

Conclusions: Taken altogether, with all the limits due to a retrospective analysis, our findings show the more comprehensive aHCT-CI to perform the best. Pts labeled as having no comorbidity burden per the SCI (score 0) were further stratified into 3 groups with increasing hazards of NRM and overall mortality per the aHCT-CI, highlighting concerns over simplifying comorbidity evaluation. Moving forward, the aHCT-CI might be the more preferred comorbidity index for pt counseling and trial analyses. The oHCT-CI is not superior to models also including descriptive variables. Still, there is a need for novel biomarkers and even more accurate CIs.