Abstract
There is long-standing interest in estimating non-relapse mortality (NRM) after allogeneic hematopoietic cell transplantation (HCT) for AML, but existing tools have limited discriminative capacity. Using single institution data from 861 adults with AML, we retrospectively examined the treatment-related mortality (TRM) score, originally developed to predict early mortality following induction chemotherapy, as a predictor of post-HCT outcome. NRM risks increased stepwise across the 4 TRM score quartiles (at 3 years: 9% [95% confidence interval: 5–13%] in Q1 vs. 28% [22–34%] in Q4). The 3-year risk of relapse was lower in patients with lower TRM score (26% [20–32%] in Q1 vs. 37% [30–43%] in Q4). Consequently, relapse-free survival (RFS) and overall survival (OS) estimates progressively decreased (RFS at 3 years: 66% [59–72%] in Q1 vs. 36% [29–42%] in Q4; OS at 3 years: 72% [66–78%] in Q1 vs. 39% [33–46%] in Q4). With a C-statistic of 0.661 (continuous variable) or 0.642 (categorized by quartile), the TRM score predicted NRM better than the Pretransplantation Assessment of Mortality (PAM) score (0.603) or the HCT-CI/age composite score (0.576). While post-HCT outcome prediction remains challenging, these findings suggest that the TRM score may be useful for risk stratification for adults with AML undergoing allogeneic HCT.
INTRODUCTION
Allogeneic hematopoietic cell transplantation (HCT) is often considered for adults with acute myeloid leukemia (AML) in morphologic remission.1–3 Still, post-HCT relapse remains common, and a higher risk of non-relapse mortality (NRM) relative to chemotherapy-only post-remission therapy can curtail or negate any benefits associated with allografting. Since the likelihood of post-HCT morbidity and mortality varies greatly across patients, there has long been interest in tools to estimate adverse event risks and survival for individual patients. Many patient- and disease-specific characteristics have individually been associated with increased mortality after allogeneic HCT, and varying combinations thereof have been integrated into validated prognostic scores. These include, for example, the HCT-specific Comorbidity Index (HCT-CI),4–7 which was designed to capture patient co-morbidities and predict NRM and overall survival (OS) post-HCT, the Pretransplantation Assessment of Mortality (PAM) score,8,9 the (modified) European Society for Blood and Marrow Transplantation (EBMT) risk score,10,11 the Comorbidity-EBMT index,12 the revised Disease Risk Index (rDRI),13 the NRM-J index,14 and the Simplified Comorbidity Index (SCI),15 among others. While frequently used, existing tools are useful for risk stratification but, at the individual patient level, have relatively limited discriminative capacity.14,16,17 Hence, there is an ongoing need for new or refined tools to predict NRM and other outcomes following allogeneic HCT.
Here, we used data from a large cohort of adults with AML undergoing allogeneic HCT in first or second morphologic remission to examine the simplified treatment-related mortality (TRM) score as a predictor of post-HCT outcome. This score is composed of weighted information from 8 covariates (age, performance status, white blood cell [WBC] count, peripheral blood blast percentage, type of AML [de novo vs. secondary], platelet count, albumin, and creatinine).18 It was initially developed in a non-HCT cohort of 2,238 adults with previously untreated AML who received intensive induction chemotherapy and predicted the probability of death within 28 days with good accuracy, as indicated by an area under the receiver operator characteristic curve (AUC) of 0.83.18 So far, the TRM score’s ability to predict outcomes in the setting of allogeneic HCT after myeloablative or non-myeloablative conditioning (MAC and non-MAC) has not been studied.
PATIENTS AND METHODS
Study cohort
All patients ≥18 years of age with AML (2016 WHO criteria19) who underwent a first allogeneic HCT while in first or second remission (i.e. <5% blasts in bone marrow) at our institution between 4/2006 and 12/2019 were included in this analysis. Partial data from all but 7 patients in the study cohort were previously reported.20–30 High-resolution HLA-typing was used for donor selection. Post-transplant outcomes were maintained via the Long-Term Follow-Up Program and chart review of those on research studies. All patients were treated with standard of care protocols or on Institutional Review Board (IRB)-approved research protocols (all registered with ClinicalTrials.gov) and gave consent in accordance with the Declaration of Helsinki. Outcomes are updated as of March 30, 2021. This retrospective analysis was approved by the Fred Hutchinson Cancer Research Center IRB (#2562).
AML risk classification and treatment response
Cytogenetic risk at diagnosis was assigned using MRC/NCRI31 criteria. Patients were considered cytogenetically normal when a normal karyotype was present, irrespective of the number of metaphases.29,32 Secondary AML was defined as disease following an antecedent hematologic disorder or treatment with systemic chemotherapy and/or radiotherapy for a different disorder.22,24,29 Treatment responses were defined according to the European LeukemiaNet criteria3 with exception of post-HCT relapse, which was defined as >5% blasts by morphology or multiparameter flow cytometry in blood or bone marrow, re-emergence of previously seen cytogenetic abnormalities, or any level of disease leading to a therapeutic intervention.
Classification of pre-transplant comorbidity scores
The TRM score (online calculator: https://trmcalculator.fredhutch.org) was computed using clinical and laboratory data from the first day of conditioning therapy as described before, and corresponds to the predicted probability of death within 28 days of beginning intensive AML chemotherapy.18 Performance status was designated according to documented physician assessment (more commonly used for later part of study period) or estimated by chart review (more commonly used for earlier part of study period) by one of the authors (L.C.Z.) according to criteria proposed by the Eastern Cooperative Oncology Group (ECOG).33 A composite score between the HCT-specific comorbidity index (HCT-CI) and age (HCT-CI/age composite score) was calculated as described.34 The PAM score, which integrates information on age, donor type, HLA matching, disease risk, forced expiratory volume in the first second (FEV1), and patient/donor CMV status, was calculated, and patients were categorized into 4 groups (<17 vs. 17–23 vs. 24–30 vs. >30), as described.8,9 The modified EBMT risk score, which includes data on age, disease stage, donor type, and gender match, was calculated for all patients except those receiving cord blood or HLA-haploidentical grafts as described.11,35
Types and intensity of conditioning regimens
Regimens including high-dose fractionated total body irradiation (TBI; ≥12 Gy) with or without cyclophosphamide (CY) or fludarabine (FLU), high-dose TBI/thiotepa/FLU, busulfan (4 days) with CY or FLU, treosulfan/FLU with or without low-dose TBI, or any regimen containing a radiolabeled antibody, all of which targeting CD45, were considered MAC. Regimens with 2–3 Gy TBI with or without fludarabine and those with cyclophosphamide as part of conditioning in HLA-haploidentical and cord blood HCT were considered nonmyeloablative (NMA). All others were considered reduced intensity conditioning (RIC) regimens.36,37
Detection of MRD by multiparameter flow cytometry
All patients underwent bone marrow aspirate analysis with ten-color flow cytometry as part of the pre-HCT work-up as described.20–22,24,25,29,30,38 As done before, any detectable MRD was considered positive.20–30 Next-generation sequencing-based MRD testing was not routinely done at our institution, and such data could therefore not be considered in our analyses.
Statistical analysis
Unadjusted probabilities of relapse-free survival (RFS; events = relapse and death) and OS (event = death) were estimated using the Kaplan-Meier method, and probabilities of relapse and NRM were summarized using cumulative incidence estimates. NRM was defined as death without prior relapse and was considered a competing risk for relapse, while relapse was a competing risk for NRM. Categorical patient characteristics were compared using Fisher’s exact test and quantitative characteristics were compared with the Wilcoxon rank sum test. Associations with RFS and OS were assessed using Cox regression; cause-specific regression models were used for relapse and NRM. Besides the TRM score (both continuous and categorized), covariates evaluated were: age at HCT, gender, HCT-CI/age composite score (0 vs. 1–2 vs. 3–4 vs. ≥5), PAM score (<17 vs. 17–23 vs. 24–30 vs. >30), modified EBMT score (0–2 vs. 3 vs. ≥4), conditioning intensity (MAC vs. non-MAC), first vs. second remission at time of HCT, karyotype/cytogenetic risk group at AML diagnosis (favorable/intermediate vs. adverse), type of AML at diagnosis (secondary vs. de novo), cytogenetics at time of HCT (normalized vs. not normalized for patients presenting with abnormal karyotypes), peripheral blood counts at the time of HCT (recovered [i.e. absolute neutrophil count >1,000/μL and platelet count >100,000/μL] vs. not recovered), donor type (related vs. unrelated) and stem cell source. Missing cytogenetic risk at diagnosis, karyotype at HCT, and PAM score were accounted for as separate categories. C-statistics were calculated for regression models to quantify their ability to predict outcomes,39 with values of 0.6–0.7, 0.7–0.8, and 0.8–0.9 commonly considered as poor, fair, and good, respectively. The relative importance of predictors in the regression models was evaluated by the value of the partial Wald Chi-squared statistic minus the predictor’s degrees of freedom. Complete-case analyses were performed for each analysis reported. Two-sided P-values are reported. Statistical analyses were performed using STATA 16.1 (StataCorp LP, College Station, TX) and R (R Foundation for Statistical Computing, Vienna, Austria; http://www.r-project.org).
RESULTS
Characteristics of study cohort
Eight hundred and ninety-two adults met the inclusion criteria for our retrospective study. Of these, 19 did not agree to their data being used for research purposes, 1 did not have sufficient data available to calculate a pre-HCT TRM score, and 11 did not undergo pre-HCT MRD testing at our institution during the pre-HCT work-up, leaving 861 patients for analysis: 544 underwent MAC HCT and 317 underwent non-MAC HCT (RIC HCT: n=98; NMA HCT: n=219), respectively. Table 1 provides a summary of the characteristics of the study cohort, donors, and HCTs, stratified by conditioning intensity (MAC vs. non-MAC). Since we routinely prioritize MAC unless significant comorbidities are present at our institution, there were expected differences between MAC and non-MAC HCT, including age at time of HCT, proportion of secondary AML, and various risk scores. These 2 patient cohorts indeed differed significantly regarding TRM scores, HCT-CI/Age composite scores, PAM scores, and modified EBMT scores (all P<0.001).
TABLE 1.
Pre-transplantation demographic and clinical characteristics of study cohort, stratified by conditioning intensity
| MAC (n=544) | Non-MAC (n=317) | All patients (n=861) | P-value | |
|---|---|---|---|---|
| Median age at HCT (range), years | 48.5 (18.1–72.6) | 65.1 (20.0–79.5) | 54.4 (18.1–79.5) | <0.001 |
| Male gender, n (%) | 282 (52) | 182 (57) | 464 (54) | 0.12 |
| Cytogenetic risk, n (%) | 0.64 | |||
| Favorable/intermediate | 392 (72) | 221 (70) | 613 (71) | |
| Adverse | 129 (24) | 84 (27) | 213 (25) | |
| Missing | 23 (4) | 12 (4) | 35 (4) | |
| Secondary AML, n (%) | 116 (21) | 117 (37) | 233 (27) | <0.001 |
| Remission status at HCT, n (%) | 0.37 | |||
| First remission | 407 (75) | 246 (78) | 653 (76) | |
| Second remission | 137 (25) | 71 (22) | 208 (24) | |
| Pre-HCT MRD status by flow cytometry, n (%) | 0.93 | |||
| MRDneg | 435 (80) | 255 (80) | 690 (80) | |
| MRDpos | 109 (20) | 62 (20) | 171 (20) | |
| Cytogenetics before HCT, n (%) | 0.39 | |||
| Normalized karyotype | 228 (42) | 115 (36) | 343 (40) | |
| Abnormal karyotype | 82 (15) | 53 (17) | 135 (16) | |
| Non-informative karyotype* | 214 (39) | 134 (42) | 348 (40) | |
| Missing | 20 (4) | 15 (5) | 35 (4) | |
| Recovered ANC before HCT**, n (%) | 512 (94) | 289 (91) | 801 (93) | 0.13 |
| Recovered platelet count before HCT**, n (%) | 415 (76) | 209 (66) | 624 (72) | 0.0012 |
| Recovered peripheral blood counts before HCT**, n (%) | 412 (76) | 206 (65) | 618 (72) | <0.001 |
| ECOG performance status, n (%) | <0.001 | |||
| 0 | 113 (21) | 39 (12) | 152 (18) | |
| 1 | 405 (74) | 241 (76) | 646 (75) | |
| 2–3 | 26 (5) | 37 (12) | 63 (7) | |
| Median TRM score (range) | 1.16 (0.03–25.88) | 2.86 (0.09–50.36) | 1.73 (0.03–50.36) | <0.001 |
| 1st quartile (≤0.790), n (%) | 183 (34) | 32 (10) | 215 (25) | <0.001 |
| 2nd quartile (>0.790–1.723), n (%) | 158 (29) | 57 (18) | 215 (25) | |
| 3rd quartile (>1.723–3.311), n (%) | 123 (23) | 93 (29) | 216 (25) | |
| 4th quartile (>3.311), n (%) | 80 (15) | 135 (43) | 215 (25) | |
| HCT-CI/Age Composite score, n (%) | <0.001 | |||
| 0 | 25 (5) | 2 (1) | 27 (3) | |
| 1–2 | 201 (37) | 62 (20) | 263 (31) | |
| 3–4 | 200 (37) | 113 (36) | 313 (36) | |
| ≥5 | 118 (22) | 140 (44) | 258 (30) | |
| Median PAM score (range) | 14.6 (0.9–96.0) | 19.3 (5.4–37.3) | 16.4 (0.9–96.0) | <0.001 |
| <17, n (%) | 342 (63) | 107 (34) | 449 (52) | <0.001 |
| 17–23, n (%) | 141 (26) | 133 (42) | 274 (32) | |
| 24–30, n (%) | 24 (4) | 48 (15) | 72 (8) | |
| >30, n (%) | 3 (1) | 14 (4) | 17 (2) | |
| Missing | 34 (6) | 15 (5) | 49 (6) | |
| Modified EBMT score, n (%) | <0.001 | |||
| 0–1 | 27 (5) | 3 (1) | 30 (3) | |
| 2 | 132 (24) | 34 (11) | 166 (19) | |
| 3 | 208 (38) | 133 (42) | 341 (40) | |
| >4 | 78 (14) | 95 (30) | 173 (20) | |
| Not applicable | 99 (18) | 52 (16) | 151 (18) | |
| Pre-HCT laboratory studies | ||||
| WBC (range), x103/μL | 3.8 (0.0–24.8) | 3.9 (0.1–11.5) | 3.8 (0.0–24.8) | 0.74 |
| Platelet count (range), x103/μL | 126 (10–670) | 110 (5–721) | 121 (5–721) | 0.045 |
| Creatinine (mg/dL) | 0.82 (0.40–1.80) | 0.96 (0.45–2.08) | 0.87 (0.40–2.08) | <0.001 |
| Albumin (g/L) | 3.9 (2.3–5.1) | 3.8 (1.9–4.7) | 3.9 (1.9–5.1) | 0.019 |
| Unrelated donor, n (%) | 371 (68) | 247 (78) | 618 (72) | 0.0022 |
| HLA matching, n (%) | <0.001 | |||
| 10/10 HLA-identical related donor | 160 (29) | 51 (16) | 211 (25) | |
| 10/10 HLA-matched unrelated donor | 233 (43) | 176 (56) | 409 (48) | |
| 1–2 allele/antigen HLA-mismatched unrelated donor | 52 (10) | 38 (12) | 90 (10) | |
| HLA-haploidentical donor | 9 (2) | 17 (5) | 26 (3) | |
| UCB | 90 (17) | 35 (11) | 125 (15) | |
| Source of stem cells, n (%) | <0.001 | |||
| PB | 389 (72) | 273 (86) | 662 (77) | |
| BM | 65 (12) | 9 (3) | 74 (9) | |
| UCB | 90 (17) | 35 (11) | 125 (15) | |
| GVHD prophylaxis, n (%) | <0.001 | |||
| CNI + MMF ± sirolimus | 166 (31) | 262 (83) | 428 (50) | |
| CNI + MTX ± other | 308 (57) | 23 (7) | 331 (38) | |
| PTCy | 58 (11) | 31 (10) | 89 (10) | |
| Other | 12 (2) | 1 (0) | 13 (2) | |
| Transplant year, n (%) | 0.016 | |||
| 2006–2012 | 268 (49) | 129 (41) | 397 (46) | |
| 2013–2019 | 276 (51) | 188 (59) | 464 (54) |
Normal cytogenetics in patient with cytogenetically normal AML or missing cytogenetics at diagnosis.
ANC ≥1,000/μL and platelets ≥100,000/μL.
Abbreviations: BM, bone marrow; CNI, calcineurin inhibitor; EBMT, European Bone and Marrow Transplantation; ECOG, Eastern Cooperative Oncology Group; HCT, hematopoietic cell transplantation; MAC, myeloablative conditioning; MMF, mycophenolate mofetil; MTX, methotrexate; PB, peripheral blood; PTCy, post transplantation cyclophosphamide; TBI, total body irradiation; UCB, umbilical cord blood; WBC, total white blood cell count.
Relationship between TRM score and post-HCT outcome
Four hundred and twenty-one deaths, 282 relapses, and 179 NRM events contributed to the probability estimates for NRM, relapse, RFS, and OS in our dataset. At the time of data cut, the median follow-up time after HCT among survivors was 64 (12–175) months: 72 (12–171) months for MAC and 53 (13–175) months for non-MAC HCT patients, respectively. TRM scores were relatively low across the study cohort overall (median: 1.73, 95% confidence interval [CI]: 0.29–7.85). Unsurprisingly, they were lower in the MAC subset (1.16 [95% CI: 0.21–5.54) than the non-MAC subset (2.85 [95% CI: 0.59–11.15; P<0.001). In a first analysis, we assessed the relationship between TRM score and post-HCT outcome, stratifying patients by quartile of TRM score: 1st quartile, TRM score ≤0.790; 2nd quartile, TRM score >0.790 but ≤1.723; 3rd quartile, TRM score >1.723 but ≤3.311; and 4th quartile, TRM score >3.311. Consistent with the TRM score being affected, among other factors, by age and proportion of secondary AML, many of patient characteristics and HCTs differed across these quartiles (Supplementary Table 1). In the entire cohort, the cumulative incidence of NRM increased in a stepwise fashion across the 4 TRM score quartiles (Figure 1 and Table 2). Relative to the increase in NRM, the relapse risk was only modestly increased in patients with higher TRM scores. Consistent with these increases in NRM and, to a lesser degree, relapse risks, RFS and OS estimates stepwise decreased with increasing quartile of TRM score. Qualitatively similar results to these overall findings were obtained in the MAC and non-MAC patient subsets (Supplementary Figure 1 and Table 2). Like the TRM score, a higher PAM score identified patients with higher risk of NRM and relapse and lower OS and RFS (Supplementary Figure 2). However, unlike the TRM score, only a small proportion of patients (72/812 [9%] and 17/812 [2%]) fell into the higher 2 categories of PAM scores (24–30 and >30, respectively). Compared to TRM and PAM scores, outcomes of patients were much more similar across the HCT-CI/age composite score categories (Supplementary Figure 3).
Figure 1. Post-HCT outcomes for 861 adults with AML undergoing allogeneic HCT while in first or second morphologic remission, stratified by quartile of TRM score.
(A) Risk of non-relapse mortality, (B) time to relapse, (C) relapse-free survival, and (D) overall survival, shown for the entire study cohort.
TABLE 2.
Outcome probabilities (with 95% confidence interval) stratified by conditioning intensity and TRM score
| CI of NRM at 100 days | CI of NRM at 3 years | CI of relapse at 3 years | RFS at 3 years | OS at 3 years | |
|---|---|---|---|---|---|
| All patients (n=861) | 5% (4–7%) | 17% (15–20%) | 31% (28–34%) | 51% (48–55%) | 57% (54–61%) |
| TRM score | |||||
| 1st quartile (n=215) | 4% (2–7%) | 9% (5–13%) | 26% (20–32%) | 66% (59–72%) | 72% (66–78%) |
| 2nd quartile (n=215) | 4% (2–7%) | 15% (10–20%) | 25% (19–31%) | 60% (53–66%) | 65% (58–71%) |
| 3rd quartile (n=216) | 3% (1–6%) | 18% (13–24%) | 37% (30–43%) | 45% (38–51%) | 53% (46–60%) |
| 4th quartile (n=215) | 9% (6–13%) | 28% (22–34%) | 37% (30–43%) | 36% (29–42%) | 39% (33–46%) |
| MAC HCT (n=544) | 5% (3–7%) | 13% (10–16%) | 28% (24–32%) | 59% (55–63%) | 64% (59–68%) |
| TRM score | |||||
| 1st quartile (n=183) | 3% (2–7%) | 8% (4–12%) | 26% (20–33%) | 66% (58–72%) | 71% (64–77%) |
| 2nd quartile (n=158) | 5% (2–9%) | 14% (9–20%) | 22% (16–29%) | 65% (56–72%) | 68% (60–75%) |
| 3rd quartile (n=123) | 2% (0–5%) | 13% (8–20%) | 32% (24–40%) | 55% (45–63%) | 61% (52–69%) |
| 4th quartile (n=80) | 11% (6–19%) | 25% (16–35%) | 37% (26–47%) | 39% (28–50%) | 43% (32–54%) |
| Non-MAC HCT (n=317) | 5% (3–8%) | 25% (20–30%) | 37% (31–42%) | 39% (33–44%) | 47% (41–52%) |
| TRM score | |||||
| 1st quartile (n=32) | 6% (1–18%) | 13% (4–27%) | 23% (10–39%) | 64% (45–79%) | 81% (62–91%) |
| 2nd quartile (n=57) | 2% (0–8%) | 19% (10–31%) | 34% (22–47%) | 47% (33–60%) | 56% (42–69%) |
| 3rd quartile (n=93) | 4% (1–10%) | 25% (17–35%) | 43% (33–53%) | 31% (22–41%) | 43% (32–53%) |
| 4th quartile (n=135) | 7% (4–13%) | 29% (22–37%) | 36% (28–45%) | 34% (26–42%) | 37% (29–45%) |
Abbreviations: CI, cumulative incidence; HCT, hematopoietic cell transplantation; MAC, myeloablative conditioning; MRD, measurable residual disease; NRM, non-relapse mortality; OS, overall survival; RFS, relapse-free survival.
TRM score as independent prognostic factor for post-HCT outcomes
To study the relationship between pre-HCT TRM score and post-HCT outcomes in more detail, we evaluated univariate and multivariable regression models for the endpoints of NRM, relapse, RFS, and OS, accounting for the covariates noted in Patients and Methods. In these models, we used the TRM score both as a continuous as well as a categorical variable. As summarized in Table 3, as a continuous variable, the TRM score was associated with higher risk of NRM and relapse as well as lower RFS and OS (all P<0.001), with a C-statistic value of 0.661 (standard error=0.0227) for the prediction of NRM (univariate associations between individual components of the TRM score and NRM, relapse, RFS, and OS are summarized in Table 4). When categorizing patients based on quartiles of TRM score, consistently higher risks for NRM and relapse and lower RFS and OS were found in patients with TRM scores above the median (i.e. 3rd and 4th quartile) relative to patients within the lowest quartile of TRM scores (all P<0.005). Compared to the use of TRM as a continuous variable, the prediction accuracy was slightly lower for the categorical TRM score (e.g. C-statistic of 0.642 (0.0216) for NRM). As shown in Figure 2, performance status and age were the 2 most important components of the TRM score for prediction of NRM, relapse, RFS, and OS, with the relative importance of other components varying for the different endpoints. In our cohort, outcomes also worsened with increasing PAM score for all 4 endpoints of interest, although the C-statistic was slightly lower compared to the TRM score (e.g. 0.603 [0.0206] for NRM). Likewise, estimates for NRM, RFS, and OS worsened with higher HCT-CI/age composite index although differences only reached statistical significance for the subset of 280 (33%) of patients with HCT-CI/age composite score ≥5 (vs. 0–1) and the magnitude of effect and C-statistic were lower (e.g. 0.576 [0.0215] for NRM).
TABLE 3.
Univariate regression models of entire study cohort
| Non-relapse mortality | Relapse | Relapse-free survival | Overall survival | |
|---|---|---|---|---|
| TRM score (continuous) | 1.12 (1.10–1.15), P<0.001 | 1.07 (1.04–1.10), P<0.001 | 1.09 (1.08–1.11), P<0.001 | 1.11 (1.09–1.14), P<0.001 |
| TRM score (by quartiles) | ||||
| 1st quartile (n=215) | 1 (Reference) | 1 (Reference) | 1 (Reference) | 1 (Reference) |
| 2nd quartile (n=215) | 1.71 (1.03–2.84), P=0.040 | 1.12 (0.78–1.60), P=0.54 | 1.29 (0.96–1.72), P=0.087 | 1.29 (0.95–1.75), P=0.10 |
| 3rd quartile (n=216) | 2.59 (1.58–4.25), P<0.001 | 1.64 (1.17–2.30), P=0.004 | 1.92 (1.46–2.54), P<0.001 | 1.85 (1.38–2.48), P<0.001 |
| 4th quartile (n=215) | 4.33 (2.71–6.90), P<0.001 | 1.87 (1.34–2.62), P<0.001 | 2.56 (1.96–3.36), P<0.001 | 2.66 (2.00–3.53), P<0.001 |
| PAM score | ||||
| <17 (n=449) | 1 (Reference) | 1 (Reference) | 1 (Reference) | 1 (Reference) |
| 17–23 (n=274) | 2.00 (1.45–2.76), P<0.001 | 1.62 (1.24–2.12), P<0.001 | 1.77 (1.44–2.17), P<0.001 | 1.82 (1.47–2.26), P<0.001 |
| 24–30 (n=72) | 2.34 (1.38–3.96), P=0.0016 | 2.87 (1.99–4.13), P<0.001 | 2.68 (1.99–3.62), P<0.001 | 2.55 (1.85–3.50), P<0.001 |
| >30 (n=17) | 3.38 (1.36–8.37), P=0.0086 | 2.84 (1.39–5.82), P=0.0043 | 3.03 (1.73–5.32), P<0.001 | 3.42 (1.94–6.02), P<0.001 |
| HCT-CI/Age Composite score | ||||
| 0 (n=27) | 1 (Reference) | 1 (Reference) | 1 (Reference) | 1 (Reference) |
| 1–2 (n=263) | 1.76 (0.55–5.66), P=0.34 | 0.96 (0.50–1.86), P=0.91 | 1.15 (0.65–2.03), P=0.63 | 1.09 (0.60–1.97), P=0.78 |
| 3–4 (n=313) | 2.19 (0.69–6.99), P=0.16 | 1.02 (0.53–1.95), P=0.96 | 1.29 (0.73–2.27), P=0.38 | 1.23 (0.68–2.21), P=0.49 |
| ≥5 (n=258) | 3.26 (1.02–10.36), P=0.045 | 1.02 (0.57–2.10), P=0.79 | 1.58 (0.90–2.78), P=0.11 | 1.55 (0.86–2.79), P=0.15 |
| Age at HCT | 1.03 (1.02–1.04), P<0.001 | 1.01 (1.00–1.02), P=0.038 | 1.02 (1.01–1.02), P<0.001 | 1.02 (1.01–1.02), P<0.001 |
| Gender | ||||
| Female (n=387) | 1 (Reference) | 1 (Reference) | 1 (Reference) | 1 (Reference) |
| Male (n=464) | 1.47 (1.09–1.98), P=0.012 | 1.14 (0.90–1.44), P=0.29 | 1.25 (1.04–1.51), P=0.017 | 1.23 (1.01–1.49), P=0.038 |
| Cytogenetic risk | ||||
| Favorable/intermediate (n=613) | 1 (Reference) | 1 (Reference) | 1 (Reference) | 1 (Reference) |
| Adverse (n=213) | 0.70 (0.47–1.04), P=0.075 | 1.99 (1.56–2.56), P<0.001 | 1.40 (1.14–1.72), P=0.0012 | 1.27 (1.02–1.57), P=0.032 |
| Type of AML | ||||
| De novo (n=628) | 1 (Reference) | 1 (Reference) | 1 (Reference) | 1 (Reference) |
| Secondary (n=233) | 1.40 (1.03–1.91), P=0.033 | 1.13 (0.88–1.47), P=0.34 | 1.23 (1.01–1.50), P=0.037 | 1.22 (0.99–1.51), P=0.056 |
| Disease status | ||||
| First remission (n=653) | 1 (Reference) | 1 (Reference) | 1 (Reference) | 1 (Reference) |
| Second remission (n=208) | 1.35 (0.97–1.88), P=0.074 | 1.44 (1.11–1.87), P=0.0055 | 1.41 (1.15–1.72), P=0.001 | 1.48 (1.20–1.83), P<0.001 |
| Pre-HCT MRD status | ||||
| MRDneg (n=690) | 1 (Reference) | 1 (Reference) | 1 (Reference) | 1 (Reference) |
| MRDpos (n=171) | 1.48 (1.00–2.20), P=0.051 | 4.18 (3.28–5.32), P<0.001 | 2.99 (2.44–3.65), P<0.001 | 2.44 (1.98–3.01), P<0.001 |
| Pre-HCT karyotype | ||||
| Normalized (n=343) | 1 (Reference) | 1 (Reference) | 1 (Reference) | 1 (Reference) |
| Not normalized (n=135 | 1.62 (1.03–2.55), P=0.035 | 2.29 (1.69–3.11), P<0.001 | 2.06 (1.60–2.65), P<0.001 | 1.92 (1.48–2.51), P<0.001 |
| Pre-HCT ANC * | ||||
| Recovered (n=801) | 1 (Reference) | 1 (Reference) | 1 (Reference) | 1 (Reference) |
| Not recovered (n=60) | 1.76 (1.04–2.99), P=0.037 | 1.90 (1.29–2.81), P=0.0013 | 1.85 (1.35–2.54), P<0.001 | 1.90 (1.37–2.64), P<0.001 |
| Pre-HCT platelet count * | ||||
| Recovered (n=624) | 1 (Reference) | 1 (Reference) | 1 (Reference) | 1 (Reference) |
| Not recovered (n=237) | 1.74 (1.28–2.36), P<0.001 | 1.08 (0.83–1.41), P=0.54 | 1.31 (1.07–1.60), P=0.0077 | 1.37 (1.12–1.69), P=0.0026 |
| Pre-HCT blood counts * | ||||
| Recovered (n=618) | 1 (Reference) | 1 (Reference) | 1 (Reference) | 1 (Reference) |
| Not recovered (n=243) | 1.72 (1.27–2.34), P<0.001 | 1.10 (0.85–1.43), P=0.46 | 1.32 (1.08–1.60), P=0.0061 | 1.39 (1.13–1.71), P=0.0015 |
| Donor | ||||
| Related (n=243) | 1 (Reference) | 1 (Reference) | 1 (Reference) | 1 (Reference) |
| Unrelated (n=618) | 1.36 (0.97–1.88), P=0.074 | 0.92 (0.72–1.19), P=0.53 | 1.07 (0.87–1.31), P=0.53 | 1.14 (0.92–1.41), P=0.24 |
| Stem cell source | ||||
| PB (n=662) | 1 (Reference) | 1 (Reference) | 1 (Reference) | 1 (Reference) |
| BM (n=74) | 0.54 (0.28–1.06), P=0.074 | 1.31 (0.90–1.89), P=0.16 | 1.00 (0.72–1.38), P=0.99 | 0.95 (0.68–1.34), P=0.79 |
| UCB (n=125) | 1.09 (0.73–1.64), P=0.67 | 0.83 (0.58–1.20), P=0.32 | 0.93 (0.71–1.22), P=0.61 | 1.02 (0.77–1.35), P=0.90 |
| HLA matching | ||||
| 10/10 HLA-identical related donor (n=211) | 1 (Reference) | 1 (Reference) | 1 (Reference) | 1 (Reference) |
| 10/10 HLA-matched unrelated donor (n=409) | 1.15 (0.77–1.70), P=0.49 | 1.03 (0.77–1.37), P=0.85 | 1.07 (0.85–1.35), P=0.55 | 1.08 (0.85–1.38), P=0.53 |
| 9/10 HLA-matched unrelated donor (n=90) | 2.63 (1.66–4.19), P<0.001 | 1.22 (0.81–1.86), P=0.34 | 1.70 (1.25–2.30), P<0.001 | 1.88 (1.37–2.59), P<0.001 |
| HLA-haploidentical donor (n=26) | 2.05 (0.86–4.86), P=0.10 | 2.08 (1.17–3.70), P=0.012 | 2.10 (1.30–3.38), P=0.0024 | 2.00 (1.19–3.34), P=0.0086 |
| UCB (n=125) | 1.47 (0.90–2.40), P=0.10 | 0.86 (0.57–1.31), P=0.49 | 1.07 (0.78–1.47), P=0.67 | 1.19 (0.86–1.65), P=0.29 |
| Year of HCT | ||||
| 2006–2012 (n=397) | 1 (Reference) | 1 (Reference) | 1 (Reference) | 1 (Reference) |
| 2013–2019 (n=464) | 1.06 (0.78–1.45), P=0.70 | 0.75 (0.59–0.95), P=0.017 | 0.88 (0.72–1.07), P=0.20 | 0.85 (0.71–1.03), P=0.095 |
Results are presented as hazard radio (95% confidence interval), and P-value;
Recovered: ANC ≥1,000/μL and platelets ≥100,000/μL; not recovered: ANC <1,000/μL and/or platelets <100,000/μL.
Abbreviations: ANC, absolute neutrophil count; BM, bone marrow; HCT, hematopoietic cell transplantation; MAC, myeloablative conditioning; MRD, measurable residual disease; PB, peripheral blood; UBC, umbilical cord blood; WBC, total white blood cell count.
TABLE 4.
Univariate regression models for individual components of TRM score (entire study cohort)
| Non-relapse mortality | Relapse | Relapse-free survival | Overall survival | |
|---|---|---|---|---|
| Age at HCT | 1.03 (1.02–1.04), P<0.001 | 1.01 (1.00–1.02), P=0.038 | 1.02 (1.01–1.02), P<0.001 | 1.02 (1.01–1.02), P<0.001 |
| ECOG performance status | ||||
| 0 (n=152) | 1 (Reference) | 1 (Reference) | 1 (Reference) | 1 (Reference) |
| 1 (n=646) | 2.36 (1.44–3.87), P<0.001 | 1.48 (1.06–2.08), P=0.022 | 1.76 (1.33–2.32), P<0.001 | 1.84 (1.37–2.47), P<0.001 |
| 2–3 (n=63) | 8.79 (4.78–16.18), P<0.001 | 3.80 (2.36–6.12), P<0.001 | 5.24 (3.62–7.60), P<0.001 | 5.85 (3.98–8.60), P<0.001 |
| Type of AML | ||||
| De novo (n=628) | 1 (Reference) | 1 (Reference) | 1 (Reference) | 1 (Reference) |
| Secondary (n=233) | 1.40 (1.03–1.91), P=0.033 | 1.13 (0.88–1.47), P=0.34 | 1.23 (1.01–1.50), P=0.037 | 1.22 (0.99–1.51), P=0.056 |
| WBC at HCT | 0.94 (0.88–1.01), P=0.10 | 0.92 (0.87–0.97), P=0.0026 | 0.93 (0.89–0.97), P<0.001 | 0.93 (0.89–0.98), P=0.0025 |
| Platelet count at HCT | 1.00 (0.99–1.00), P=0.01 | 1.00 (1.00–1.00), P=0.08 | 1.00 (1.00–1.00), P=0.0031 | 1.00 (1.00–1.00), P=0.0037 |
| Serum creatinine at HCT | 2.71 (1.63–4.51), P<0.001 | 0.81 (0.50–1.32), P=0.40 | 1.37 (0.96–1.95), P=0.08 | 1.57 (1.09–2.26), P=0.016 |
| Serum albumin at HCT | 0.39 (0.26–0.58), P<0.001 | 0.71 (0.51–0.98), P=0.035 | 0.56 (0.43–0.72), P<0.001 | 0.55 (0.42–0.71), P<0.001 |
Results are presented as hazard radio (95% confidence interval), and P-value; Abbreviations: ECOG, Eastern Cooperative Oncology Group; HCT, hematopoietic cell transplantation; WBC, total white blood cell count.
Figure 2. Prediction of post-HCT outcomes with TRM score.
Importance of individual covariates to predict (A) non-relapse mortality, (B) relapse, (C) relapse-free survival, and (D) overall survival with TRM score using Chi-squared (χ2) values. “Importance” is evaluated with the Wald χ2 statistic minus the predictor’s degrees of freedom (df). Covariates with larger χ2 values are considered more “important” in predicting the outcome of interest. Covariates are listed on the y-axis in order of their χ2 values, with highest values at the top and lowest values at the bottom.
We then built multivariable models to examine the contribution of risk scores to NRM prediction further. For a basic model that included gender, cytogenetic risk at diagnosis, remission number, pre-HCT MRD status, pre-HCT karyotype, pre-HCT ANC, pre-HCT platelet count, conditioning intensity, donor type/source, and year of HCT, a C-statistic of 0.670 (0.0210) was obtained. With addition of the TRM score as a categorical variable, the C-statistic increased to 0.695 (0.0213). Basic models with inclusion of either PAM score or HCT-CI/age composite score yielded C-statistics of 0.673 (0.0213) and 0.675 (0.0211), respectively. Models that included basic variables and 2 of the 3 risk scores resulted in C-statistics of 0.680 (0.0210) for PAM score together with HCT-CI/age composite score, 0.697 (0.0205) for HCT-CI/age composite score together with TRM score, and 0.693 (0.0211) for PAM score together with TRM score. A full model that included basic covariates and all 3 risk scores yielded a C-statistic of 0.699 (0.0205). For the other endpoints of relapse, RFS, and OS, C-statistics of the full model were 0.715 (0.0155), 0.692 (0.0125), and 0.685 (0.0135; Table 5). After this multivariable adjustment, increasing quartiles of the TRM scores remained statistically significantly associated with higher risks for NRM and relapse and shorter RFS and OS as did the PAM score but not the HCT-CI/age composite score (Table 5). Likewise, after adjustment for the same covariates, the TRM score as continuous variable remained statistically significantly associated with higher risks for NRM (hazard ratio [HR]: 1.10, 95% confidence interval [95% CI]: 1.06–1.13, P<0.001; C-statistic: 0.704 [0.0209]), lower RFS (HR=1.05 [95% CI: 1.02–1.08], P<0.001; C-statistic: 0.692 [0.0125]), and lower OS (HR=1.07 [95% CI: 1.04–1.10], P<0.001, C-statistic: 0.685 [0.0135]) but not higher risk for relapse (HR=1.01 [95% CI: 0.97–1.06], P=0.55; C-statistic: 0.715 [0.0155]).
TABLE 5.
Multivariable regression models of entire study cohort for quartiles of TRM score
| Non-relapse mortality c=0.699 (0.0205) | Relapse c=0.719 (0.0154) | Relapse-free survival c=0.694 (0.0124) | Overall survival c=0.682 (0.0133) | |
|---|---|---|---|---|
| TRM score (ref: 1st quartile) | ||||
| 2nd quartile | 1.50 (0.89–2.54), P=0.13 | 1.00 (0.69–1.47), P=0.99 | 1.16 (0.85–1.57), P=0.29 | 1.16 (0.84–1.61), P=0.36 |
| 3rd quartile | 2.17 (1.28–3.68), P=0.0039 | 1.45 (1.00–2.11), P=0.05 | 1.67 (1.23–2.26), P<0.001 | 1.61 (1.17–2.23), P=0.0035 |
| 4th quartile | 2.96 (1.69–5.21), P<0.001 | 1.21 (0.79–1.86), P=0.38 | 1.72 (1.23–2.40), P=0.0016 | 1.89 (1.33–2.68), P<0.001 |
| PAM score (ref: <17) | ||||
| 17–23 | 1.30 (0.89–1.89), P=0.17 | 1.21 (0.89–1.66), P=0.22 | 1.28 (1.01–1.62), P=0.042 | 1.35 (1.05–1.73), P=0.019 |
| 24–30 | 1.31 (0.70–2.44), P=0.39 | 2.20 (1.43–3.41), P<0.001 | 1.83 (1.28–2.60), P<0.001 | 1.67 (1.15–2.43), P=0.0074 |
| >30 | 1.90 (0.69–5.24), P=0.22 | 1.19 (0.54–2.63), P=0.67 | 1.36 (0.73–2.53), P=0.33 | 1.70 (0.90–3.19), P=0.10 |
| HCT-CI/Age Composite score (ref: 0) | ||||
| 1–2 | 1.17 (0.35–3.89), P=0.80 | 0.83 (0.41–1.66), P=0.60 | 0.91 (0.50–1.65), P=0.75 | 0.77 (0.42–1.44), P=0.42 |
| 3–4 | 1.17 (0.36–3.87), P=0.79 | 0.68 (0.34–1.37), P=0.29 | 0.81 (0.45–1.47), P=0.49 | 0.72 (0.39–1.34), P=0.30 |
| ≥5 | 1.47 (0.44–4.90), P=0.53 | 0.74 (0.36–1.50), P=0.40 | 0.92 (0.51–1.69), P=0.80 | 0.86 (0.46–1.61), P=0.63 |
| Male gender (ref: female) | 1.37 (1.00–1.88), P=0.053 | 0.97 (0.76–1.25), P=0.83 | 1.10 (0.91–1.34), P=0.33 | 1.12 (0.92–1.37), P=0.26 |
| Cytogenetic risk (ref: favorable/intermediate) | ||||
| Adverse | 0.59 (0.37–0.95), P=0.03 | 1.59 (1.14–2.21), P=0.0059 | 1.14 (0.88–1.48), P=0.34 | 1.08 (0.82–1.43), P=0.58 |
| Second remission (ref: first remission) | 1.27 (0.89–1.82), P=0.18 | 1.60 (1.20–2.15), P=0.0015 | 1.42 (1.13–1.78), P=0.0022 | 1.41 (1.12–1.78), P=0.0036 |
| MRD status (ref: MRD neg ) | ||||
| MRDpos | 1.20 (0.79–1.84), P=0.39 | 3.53 (2.69–4.64), P<0.001 | 2.45 (1.96–3.06), P<0.001 | 1.94 (1.54–2.44), P<0.001 |
| Pre-HCT karyotype (ref: normalized) | ||||
| Not normalized | 1.07 (0.66–1.73), P=0.79 | 1.54 (1.10–2.15), P=0.012 | 1.37 (1.05–1.81), P=0.022 | 1.23 (0.92–1.65), P=0.15 |
| Pre-HCT ANC* (ref: recovered) | ||||
| Not recovered | 1.17 (0.65–2.09), P=0.60 | 1.89 (1.20–2.97), P=0.0061 | 1.49 (1.05–2.12), P=0.027 | 1.37 (0.94–1.98), P=0.10 |
| Pre-HCT platelet count* (ref: recovered) | ||||
| Not recovered | 0.96 (0.66–1.40), P=0.82 | 0.67 (0.48–0.93), P=0.017 | 0.78 (0.61–0.99), P=0.045 | 0.81 (0.62–1.04), P=0.099 |
| MAC HCT (ref: non-MAC HCT) | 1.56 (1.10–2.20), P=0.012 | 1.42 (1.09–1.85), P=0.01 | 1.47 (1.19–1.81), P<0.001 | 1.32 (1.06–1.65), P=0.012 |
| Donor type/source (ref: HLA-identical related donor) | ||||
| HLA-identical unrelated donor | 0.97 (0.64–1.46), P=0.88 | 0.97 (0.72–1.32), P=0.86 | 0.97 (0.76–1.24), P=0.83 | 0.98 (0.76–1.26), P=0.86 |
| 1–2 allele/antigen mismatched donor | 1.74 (1.02–2.96), P=0.041 | 0.76 (0.48–1.21), P=0.25 | 1.11 (0.79–1.56), P=0.56 | 1.35 (0.95–1.93), P=0.095 |
| HLA-haploidentical donor | 1.37 (0.54–3.46), P=0.51 | 1.17 (0.63–2.17), P=0.62 | 1.28 (0.77–2.14), P=0.34 | 1.46 (0.84–2.53), P=0.17 |
| Umbilical cord blood | 1.40 (0.83–2.38), P=0.21 | 0.76 (0.49–1.19), P=0.23 | 0.98 (0.70–1.37), P=0.89 | 1.10 (0.77–1.57), P=0.59 |
| Years 2013–2019 of HCT (ref: 2006–2012) | 0.94 (0.68–1.30), P=0.71 | 0.80 (0.62–1.03), P=0.08 | 0.85 (0.70–1.04), P=0.11 | 0.85 (0.69–1.05), P=0.14 |
C-statistic values (standard error) are provided.
Recovered: ANC ≥1,000/μL and platelets ≥100,000/μL; not recovered: ANC <1,000/μL and/or platelets <100,000/μL.
Abbreviations: ANC, absolute neutrophil count; BM, bone marrow; HCT, hematopoietic cell transplantation; MAC, myeloablative conditioning; MRD, measurable residual disease; PB, peripheral blood; UBC, umbilical cord blood; WBC, total white blood cell count.
Pre-HCT TRM score as independent prognostic factor for post-HCT outcomes in specific patient subsets
Finally, we examined the relationship between TRM score and post-HCT outcomes in several discrete patient subsets, specifically the 544 patients who underwent allogeneic HCT after MAC, the 317 patients who underwent allogeneic HCT after non-MAC, and the subset of 620 patients who underwent allogeneic HCT with 10/10 HLA-matched unrelated donor or 10/10 HLA-identical sibling donor allografts; in the latter, we included the modified EBMT score as additional covariate of interest. In multivariable analyses, overall findings in these patient subsets were like those obtained in the entire study cohort for the outcome of NRM. For RFS and OS, the magnitude of effect of the TRM score on outcome appeared greater for non-MAC HCT as compared to MAC HCT (Table 6).
TABLE 6.
Results from multivariable regression models for TRM score in discrete patient subsets
| Non-relapse mortality | Relapse | Relapse-free survival | Overall survival | |
|---|---|---|---|---|
| Subset of MAC HCT patients * | ||||
| TRM score (cont. variable) | 1.15 (1.08–1.22), P<0.001 | 1.05 (0.99–1.11), P=0.11 | 1.07 (1.02–1.12), P=0.002 | 1.07 (1.02–1.12), P=0.0022 |
| TRM score (quartiles) | ||||
| 1st quartile | 1 (Reference) | 1 (Reference) | 1 (Reference) | 1 (Reference) |
| 2nd quartile | 1.65 (0.91–2.98), P=0.099 | 0.81 (0.54–1.23), P=0.32 | 1.03 (0.74–1.45), P=0.84 | 1.03 (0.72–1.46), P=0.89 |
| 3rd quartile | 2.08 (1.12–3.85), P=0.02 | 1.01 (0.66–1.54), P=0.96 | 1.28 (0.90–1.81), P=0.16 | 1.26 (0.88–1.82), P=0.21 |
| 4th quartile | 3.58 (1.90–6.75), P<0.001 | 1.06 (0.66–1.68), P=0.82 | 1.50 (1.02–2.18), P=0.037 | 1.50 (1.01–2.24), P=0.046 |
| Subset of non-MAC HCT patients * | ||||
| TRM score (cont. variable) | 1.10 (1.07–1.14), P<0.001 | 1.10 (0.96–1.05), P=0.87 | 1.05 (1.03–1.08), P<0.001 | 1.09 (1.06–1.12), P<0.001 |
| TRM score (quartiles) | ||||
| 1st quartile | 1 (Reference) | 1 (Reference) | 1 (Reference) | 1 (Reference) |
| 2nd quartile | 1.71 (0.60–4.86), P=0.32 | 1.45 (0.64–3.31), P=0.38 | 1.59 (0.83–3.04), P=0.16 | 1.77 (0.88–3.54), P=0.11 |
| 3rd quartile | 2.88 (1.10–7.56), P=0.031 | 2.55 (1.19–5.45), P=0.016 | 2.76 (1.52–5.02), P<0.001 | 2.57 (1.35–4.90), P=0.0043 |
| 4th quartile | 3.69 (1.47–9.28), P=0.0056 | 1.71 (0.81–3.64), P=0.16 | 2.44 (1.36–4.36), P=0.0027 | 2.86 (1.52–5.36), P=0.0011 |
| Subset of 10/10 HLA-matched HCT patients ** | ||||
| TRM score (cont. variable) | 1.15 (1.09–1.22), P<0.001 | 1.02 (0.97–1.08), P=0.41 | 1.07 (1.03–1.11), P=0.0012 | 1.10 (1.05.−1.14), P<0.001 |
| TRM score (quartiles) | ||||
| 1st quartile | 1 (Reference) | 1 (Reference) | 1 (Reference) | 1 (Reference) |
| 2nd quartile | 1.40 (0.72–2.75), P=0.32 | 1.06 (0.66–1.70), P=0.81 | 1.17 (0.79–1.72), P=0.43 | 1.33 (0.88–2.01), P=0.17 |
| 3rd quartile | 1.78 (0.90–3.51), P=0.096 | 1.44 (0.90–2.29), P=0.13 | 1.54 (1.05–2.26), P=0.026 | 1.53 (1.01–2.32), P=0.042 |
| 4th quartile | 2.75 (1.33–5.70), P=0.0064 | 1.40 (0.82–2.39), P=0.21 | 1.74 (1.13–2.68), P=0.012 | 2.07 (1.31–3.27), P=0.0018 |
Besides TRM score, models included karyotype/cytogenetic risk group at AML diagnosis, first vs. second remission at time of HCT, and pre-HCT MRD status.
Besides TRM score, models included gender, karyotype/cytogenetic risk group at AML diagnosis, first vs. second remission at time of HCT, pre-HCT MRD status, cytogenetics at time of HCT, HCT-CI/age composite score, PAM score, revised EBMT score, ANC recovery before HCT, platelet count recovery before HCT, donor type (related vs. unrelated), and conditioning intensity.
Abbreviations: ANC, absolute neutrophil count; HCT, hematopoietic cell transplantation; MAC, myeloablative conditioning; TRM, treatment-related mortality.
DISCUSSION
For AML in remission, allogeneic HCT provides greater anti-AML efficacy than chemotherapy, making it an attractive post-remission strategy for many patients. However, even though HCT outcomes have gradually improved over the years,40–42 the benefit of reduced relapse rates seen after HCT must still be balanced against HCT-related non-relapse morbidity and mortality. Here, we evaluated the value of the TRM score as a predictor of NRM and other outcomes after allogeneic HCT for adults with AML in morphologic remission. Our findings support three main conclusions. First, this score – used without recalibration as originally developed for adults undergoing intensive AML chemotherapy – was indeed associated with NRM as well as RFS and OS in our cohort, with qualitatively similar results seen in patients receiving myeloablative and non-myeloablative conditioning. Second, while the prediction accuracy of individual tools was overall limited, the TRM score predicted NRM better than the PAM score and the HCT-CI. And third, the accuracy of multicomponent models to predict NRM, RFS, and OS was better than for models using the TRM score alone, indicating the potential to develop refined prediction tools in future research.
The TRM score was originally developed to overcome, at least partially, the limitations of single factors such as PS or age to predict early death after intensive AML chemotherapy in the non-HCT setting.18 With the exception of PS, this score was purposefully built to include objective covariates, obtained mostly from routine laboratory testing of a peripheral blood sample. At our institution, the TRM score is now routinely used to assess medical fitness of adults with AML or other high-grade myeloid neoplasm to inform the treatment decision-making process and to serve as a central eligibility criterion in investigator-initiated AML chemotherapy trials.43–47
Our current data indicate that the TRM score could similarly serve to inform on outcomes of AML patients undergoing allogeneic HCT, with higher scores predicting higher NRM, lower RFS, and lower OS compared to lower scores. Among the factors included in the TRM score, PS (not contained in PAM score or HCT-CI/age composite score) and age (contained in the HCT-CI/age composite score but not the original HCT-CI score) were the most important ones for the early mortality prediction accuracy in the non-HCT setting.18 The studies described herein identify PS and age again as the most important individual predictors in the setting of allogeneic HCT when assessing NRM, RFS, and OS outcomes with the TRM score. While the predictive accuracy of the TRM score for NRM compared favorably relative to that of the PAM score and of the HCT-CI/age composite score in our cohort, the prediction accuracy of all examined instruments was limited, with C-statistics for NRM not exceeding 0.66 (for the TRM score) when used in isolation. With this, albeit helpful to some degree, our data caution against over-reliance on the TRM score (or any other of the assessed prediction tools) to guide decision-making and prognostication in individual patients. This limitation also highlights a need for better predictive tools for outcomes after HCT.
It is plausible that the performance of the TRM score could be improved by recalibrating individual covariates using data from contemporary patients undergoing allogeneic HCT rather than using the TRM score as built with data from patients treated with AML chemotherapy between 1995 and 2008.18 Moreover, the observation that multivariable models that included the TRM score plus additional covariates yielded higher C-statistics than models encompassing the TRM score alone suggests the possibility that a more accurate prediction tool could be developed by combining some or all of the TRM score factors with additional covariates.
Several limitations must be acknowledged. First, while based on many patients, our analyses were retrospective in nature. Validation (ideally, prospectively) of our findings in an independent cohort of patients will be important. Such future analyses should include comparisons with other scoring system not tested in the current study, e.g. the rDRI,13 the NRM-J index,14 and/or the SCI.15 Second, our work was restricted to adults undergoing allogeneic HCT for AML in remission. We purposefully limited our studies to people with AML because the TRM score was developed using data exclusively from adults with AML receiving induction chemotherapy.18 To what degree our findings are generalizable and could be extrapolated to other patient populations (e.g. different disease or disease stage) is unknown. Third, no uniform HCT strategy was pursued in our patients. Rather, patients were assigned to various conditioning regimens of differing intensities in a (largely) non-randomized fashion based primarily on age, comorbidities, and protocol eligibility. Fourth, as an intrinsic limitation of all retrospective analyses, emerging changes in medical practice (e.g. changes in immune suppressive therapies such as the increasing use of post-HCT cyclophosphamide) may not be captured. And fifth, our multivariable models include several scores measuring similar things, thus are colinear, which could partly explain differences in results (hazard ratios, etc.) between univariate and multivariable analyses. Acknowledging these limitations, our data suggest that the TRM score may be useful for risk stratification for adults with AML undergoing allogeneic HCT and, with this, possibly aid in the decision to pursue or not pursue allografting as post-remission therapy. To unify the approach in our retrospective analysis, we calculated the TRM score using data from the first day of conditioning therapy. An earlier time point would need to be chosen for TRM score calculation to provide guidance in decision making. Considering that many patients have relatively stable laboratory findings during the pre-HCT work-up, the TRM score is expected to be relatively stable during this time. However, our findings also highlight the limited accuracy of current tools, including the original TRM score, to predict NRM and other outcomes after allografting. Thus, other factors should be considered in the decision-making process. Efforts to recalibrate the TRM score for use in the setting of allogeneic HCT, and its potential augmentation with additional factors, are ongoing.
Supplementary Material
ACKNOWLEDGEMENTS
Research reported in this publication was supported by grants P01-CA078902, P01-CA018029, and P30-CA015704 from the National Cancer Institute/National Institutes of Health (NCI/NIH), Bethesda, MD, USA. The authors acknowledge the excellent care provided by the physicians and nurses of the HCT teams, the staff in the Long-Term Follow-up office at the Fred Hutchinson Cancer Research Center, the Hematopathology Laboratory at the University of Washington, and the patients for participating in our research protocols.
Research Support:
This work was supported by grants P01-CA078902, P01-CA018029, and P30-CA015704 from the National Cancer Institute/National Institutes of Health (NCI/NIH).
Footnotes
CONFLICT OF INTEREST
The authors declare no competing financial interests.
REFERENCES
- 1.Cornelissen JJ, Gratwohl A, Schlenk RF, Sierra J, Bornhäuser M, Juliusson G, et al. The European LeukemiaNet AML Working Party consensus statement on allogeneic HSCT for patients with AML in remission: an integrated-risk adapted approach. Nat Rev Clin Oncol 2012; 9(10): 579–590. [DOI] [PubMed] [Google Scholar]
- 2.Cornelissen JJ, Blaise D. Hematopoietic stem cell transplantation for patients with AML in first complete remission. Blood 2016; 127(1): 62–70. [DOI] [PubMed] [Google Scholar]
- 3.Döhner H, Estey E, Grimwade D, Amadori S, Appelbaum FR, Buchner T, et al. Diagnosis and management of AML in adults: 2017 ELN recommendations from an international expert panel. Blood 2017; 129(4): 424–447. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 4.Sorror ML, Maris MB, Storb R, Baron F, Sandmaier BM, Maloney DG, et al. Hematopoietic cell transplantation (HCT)-specific comorbidity index: a new tool for risk assessment before allogeneic HCT. Blood 2005; 106(8): 2912–2919. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 5.Sorror ML, Giralt S, Sandmaier BM, De Lima M, Shahjahan M, Maloney DG, et al. Hematopoietic cell transplantation specific comorbidity index as an outcome predictor for patients with acute myeloid leukemia in first remission: combined FHCRC and MDACC experiences. Blood 2007; 110(13): 4606–4613. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 6.Raimondi R, Tosetto A, Oneto R, Cavazzina R, Rodeghiero F, Bacigalupo A, et al. Validation of the Hematopoietic Cell Transplantation-Specific Comorbidity Index: a prospective, multicenter GITMO study. Blood 2012; 120(6): 1327–1333. [DOI] [PubMed] [Google Scholar]
- 7.Sorror ML, Logan BR, Zhu X, Rizzo JD, Cooke KR, McCarthy PL, et al. Prospective validation of the predictive power of the Hematopoietic Cell Transplantation Comorbidity Index: a Center for International Blood and Marrow Transplant Research study. Biol Blood Marrow Transplant 2015; 21(8): 1479–1487. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 8.Parimon T, Au DH, Martin PJ, Chien JW. A risk score for mortality after allogeneic hematopoietic cell transplantation. Ann Intern Med 2006; 144(6): 407–414. [DOI] [PubMed] [Google Scholar]
- 9.Au BKC, Gooley TA, Armand P, Fang M, Madtes DK, Sorror ML, et al. Reevaluation of the pretransplant assessment of mortality score after allogeneic hematopoietic transplantation. Biol Blood Marrow Transplant 2015; 21(5): 848–854. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 10.Gratwohl A The EBMT risk score. Bone Marrow Transplant 2012; 47(6): 749–756. [DOI] [PubMed] [Google Scholar]
- 11.Hemmati PG, Terwey TH, le Coutre P, Vuong LG, Massenkeil G, Dörken B, et al. A modified EBMT risk score predicts the outcome of patients with acute myeloid leukemia receiving allogeneic stem cell transplants. Eur J Haematol 2011; 86(4): 305–316. [DOI] [PubMed] [Google Scholar]
- 12.Barba P, Martino R, Pérez-Simón JA, Fernández-Avilés F, Castillo N, Piñana JL, et al. Combination of the Hematopoietic Cell Transplantation Comorbidity Index and the European Group for Blood and Marrow Transplantation score allows a better stratification of high-risk patients undergoing reduced-toxicity allogeneic hematopoietic cell transplantation. Biol Blood Marrow Transplant 2014; 20(1): 66–72. [DOI] [PubMed] [Google Scholar]
- 13.Armand P, Kim HT, Logan BR, Wang Z, Alyea EP, Kalaycio ME, et al. Validation and refinement of the Disease Risk Index for allogeneic stem cell transplantation. Blood 2014; 123(23): 3664–3671. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 14.Yanada M, Konuma T, Mizuno S, Saburi M, Shinohara A, Tanaka M, et al. Predicting non-relapse mortality following allogeneic hematopoietic cell transplantation during first remission of acute myeloid leukemia. Bone Marrow Transplant 2021; 56(2): 387–394. [DOI] [PubMed] [Google Scholar]
- 15.Shouval R, Fein JA, Cho C, Avecilla ST, Ruiz J, Tomas AA, et al. The Simplified Comorbidity Index: a new tool for prediction of nonrelapse mortality in allo-HCT. Blood Adv 2022; 6(5): 1525–1535. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 16.Shouval R, Fein JA, Shouval A, Danylesko I, Shem-Tov N, Zlotnik M, et al. External validation and comparison of multiple prognostic scores in allogeneic hematopoietic stem cell transplantation. Blood Adv 2019; 3(12): 1881–1890. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 17.Al-Shaibani E, Cyriac S, Chen S, Lipton JH, Kim DD, Viswabandya A, et al. Comparison of the prognostic ability of the HCT-CI, the modified EBMT, and the EBMT-ADT pre-transplant risk scores for acute leukemia. Clin Lymphoma Myeloma Leuk 2021; 21(6): e559–e568. [DOI] [PubMed] [Google Scholar]
- 18.Walter RB, Othus M, Borthakur G, Ravandi F, Cortes JE, Pierce SA, et al. Prediction of early death after induction therapy for newly diagnosed acute myeloid leukemia with pretreatment risk scores: a novel paradigm for treatment assignment. J Clin Oncol 2011; 29(33): 4417–4423. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 19.Arber DA, Orazi A, Hasserjian R, Thiele J, Borowitz MJ, Le Beau MM, et al. The 2016 revision to the World Health Organization classification of myeloid neoplasms and acute leukemia. Blood 2016; 127(20): 2391–2405. [DOI] [PubMed] [Google Scholar]
- 20.Walter RB, Gooley TA, Wood BL, Milano F, Fang M, Sorror ML, et al. Impact of pretransplantation minimal residual disease, as detected by multiparametric flow cytometry, on outcome of myeloablative hematopoietic cell transplantation for acute myeloid leukemia. J Clin Oncol 2011; 29(9): 1190–1197. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 21.Walter RB, Buckley SA, Pagel JM, Wood BL, Storer BE, Sandmaier BM, et al. Significance of minimal residual disease before myeloablative allogeneic hematopoietic cell transplantation for AML in first and second complete remission. Blood 2013; 122(10): 1813–1821. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 22.Walter RB, Gyurkocza B, Storer BE, Godwin CD, Pagel JM, Buckley SA, et al. Comparison of minimal residual disease as outcome predictor for AML patients in first complete remission undergoing myeloablative or nonmyeloablative allogeneic hematopoietic cell transplantation. Leukemia 2015; 29(1): 137–144. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 23.Walter RB, Sandmaier BM, Storer BE, Godwin CD, Buckley SA, Pagel JM, et al. Number of courses of induction therapy independently predicts outcome after allogeneic transplantation for acute myeloid leukemia in first morphological remission. Biol Blood Marrow Transplant 2015; 21(2): 373–378. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 24.Araki D, Wood BL, Othus M, Radich JP, Halpern AB, Zhou Y, et al. Allogeneic hematopoietic cell transplantation for acute myeloid leukemia: is it time to move toward a minimal residual disease-based definition of complete remission. J Clin Oncol 2016; 34(4): 329–336. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 25.Zhou Y, Othus M, Araki D, Wood BL, Radich JP, Halpern AB, et al. Pre- and post-transplant quantification of measurable (‘minimal’) residual disease via multiparameter flow cytometry in adult acute myeloid leukemia. Leukemia 2016; 30(7): 1456–1464. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 26.Hoffmann AP, Besch AL, Othus M, Morsink LM, Wood BL, Mielcarek M, et al. Early achievement of measurable residual disease (MRD)-negative complete remission as predictor of outcome after myeloablative allogeneic hematopoietic cell transplantation in acute myeloid leukemia. Bone Marrow Transplant 2020; 55(3): 669–672. [DOI] [PubMed] [Google Scholar]
- 27.Morsink LM, Bezerra ED, Othus M, Wood BL, Fang M, Sandmaier BM, et al. Comparative analysis of total body irradiation (TBI)-based and non-TBI-based myeloablative conditioning for acute myeloid leukemia in remission with or without measurable residual disease. Leukemia 2020; 34(6): 1701–1705. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 28.Morsink LM, Othus M, Bezerra ED, Wood BL, Fang M, Sandmaier BM, et al. Impact of pre-transplant measurable residual disease on outcome of allogeneic hematopoietic cell transplantation in adult monosomal karyotype AML. Leukemia 2020; 34(6): 1577–1587. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 29.Morsink LM, Sandmaier BM, Othus M, Palmieri R, Granot N, Bezerra ED, et al. Conditioning intensity, pre-transplant flow cytometric measurable residual disease, and outcome in adults with acute myeloid leukemia undergoing allogeneic hematopoietic cell transplantation. Cancers 2020; 12(9). [DOI] [PMC free article] [PubMed] [Google Scholar]
- 30.Paras G, Morsink LM, Othus M, Milano F, Sandmaier BM, Zarling LC, et al. Conditioning intensity and peritransplant flow cytometric MRD dynamics in adult AML. Blood 2022; 139(11): 1694–1706. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 31.Grimwade D, Hills RK, Moorman AV, Walker H, Chatters S, Goldstone AH, et al. Refinement of cytogenetic classification in acute myeloid leukemia: determination of prognostic significance of rare recurring chromosomal abnormalities among 5876 younger adult patients treated in the United Kingdom Medical Research Council trials. Blood 2010; 116(3): 354–365. [DOI] [PubMed] [Google Scholar]
- 32.Breems DA, Van Putten WLJ, De Greef GE, Van Zelderen-Bhola SL, Gerssen-Schoorl KBJ, Mellink CHM, et al. Monosomal karyotype in acute myeloid leukemia: a better indicator of poor prognosis than a complex karyotype. J Clin Oncol 2008; 26(29): 4791–4797. [DOI] [PubMed] [Google Scholar]
- 33.Oken MM, Creech RH, Tormey DC, Horton J, Davis TE, McFadden ET, et al. Toxicity and response criteria of the Eastern Cooperative Oncology Group. Am J Clin Oncol 1982; 5(6): 649–655. [PubMed] [Google Scholar]
- 34.Sorror ML, Storb RF, Sandmaier BM, Maziarz RT, Pulsipher MA, Maris MB, et al. Comorbidity-age index: a clinical measure of biologic age before allogeneic hematopoietic cell transplantation. J Clin Oncol 2014; 32(29): 3249–3256. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 35.Michelis FV, Messner HA, Uhm J, Alam N, Lambie A, McGillis L, et al. Modified EBMT pretransplant risk score can identify favorable-risk patients undergoing allogeneic hematopoietic cell transplantation for AML, not identified by the HCT-CI score. Clin Lymphoma Myeloma Leuk 2015; 15(5): e73–81. [DOI] [PubMed] [Google Scholar]
- 36.Giralt S, Ballen K, Rizzo D, Bacigalupo A, Horowitz M, Pasquini M, et al. Reduced-intensity conditioning regimen workshop: defining the dose spectrum. Report of a workshop convened by the Center for International Blood and Marrow Transplant Research. Biol Blood Marrow Transplant 2009; 15(3): 367–369. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 37.Bacigalupo A, Ballen K, Rizzo D, Giralt S, Lazarus H, Ho V, et al. Defining the intensity of conditioning regimens: working definitions. Biol Blood Marrow Transplant 2009; 15(12): 1628–1633. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 38.Wood BL. Acute myeloid leukemia minimal residual disease detection: the difference from normal approach. Curr Protoc Cytom 2020; 93(1): e73. [DOI] [PubMed] [Google Scholar]
- 39.Gönen M, Heller G. Concordance probability and discriminatory power in proportional hazards regression. Biometrika 2005; 92(4): 965–970. [Google Scholar]
- 40.Gooley TA, Chien JW, Pergam SA, Hingorani S, Sorror ML, Boeckh M, et al. Reduced mortality after allogeneic hematopoietic-cell transplantation. N Engl J Med 2010; 363(22): 2091–2101. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 41.McDonald GB, Sandmaier BM, Mielcarek M, Sorror M, Pergam SA, Cheng GS, et al. Survival, nonrelapse mortality, and relapse-related mortality After allogeneic hematopoietic cell transplantation: comparing 2003–2007 versus 2013–2017 cohorts. Ann Intern Med 2020; 172(4): 229–239. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 42.Cooper JP, Storer BE, Granot N, Gyurkocza B, Sorror ML, Chauncey TR, et al. Allogeneic hematopoietic cell transplantation with non-myeloablative conditioning for patients with hematologic malignancies: Improved outcomes over two decades. Haematologica 2021; 106(6): 1599–1607. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 43.Halpern AB, Othus M, Huebner EM, Buckley SA, Pogosova-Agadjanyan EL, Orlowski KF, et al. Mitoxantrone, etoposide and cytarabine following epigenetic priming with decitabine in adults with relapsed/refractory acute myeloid leukemia or other high-grade myeloid neoplasms: a phase 1/2 study. Leukemia 2017; 31(12): 2560–2567. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 44.Walter RB, Othus M, Orlowski KF, McDaniel EN, Scott BL, Becker PS, et al. Unsatisfactory efficacy in randomized study of reduced-dose CPX-351 for medically less fit adults with newly diagnosed acute myeloid leukemia or other high-grade myeloid neoplasm. Haematologica 2018; 103(3): e106–e109. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 45.Halpern AB, Othus M, Huebner EM, Scott BL, Becker PS, Percival MEM, et al. Phase 1/2 trial of GCLAM with dose-escalated mitoxantrone for newly diagnosed AML or other high-grade myeloid neoplasms. Leukemia 2018; 32(11): 2352–2362. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 46.Halpern AB, Othus M, Huebner EM, Scott BL, Hendrie PC, Percival MEM, et al. Phase I/II trial of cladribine, high-dose cytarabine, mitoxantrone, and G-CSF with dose-escalated mitoxantrone for relapsed/refractory acute myeloid leukemia and other high-grade myeloid neoplasms. Haematologica 2019; 104(4): e143–e146. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 47.Mabrey FL, Gardner KM, Shannon Dorcy K, Perdue A, Smith HA, Davis AM, et al. Outpatient intensive induction chemotherapy for acute myeloid leukemia and high-risk myelodysplastic syndrome. Blood Adv 2020; 4(4): 611–616. [DOI] [PMC free article] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.