CHARM is prognostic of geriatric morbidity and toxicity after allogeneic transplant for older adults: BMT CTN 1704 study

Key Points

• •CHARM is an independent prognostic scoring system for geriatric morbidities after alloHCT in patients aged ≥60 years.
• •CHARM scores are independently associated and informative of functional limitation–free survival through 1 year after alloHCT.

Visual Abstract

Abstract

Despite concerns about the toxicity of allogeneic hematopoietic cell transplantation (alloHCT) in older patients, prospective data characterizing prevalence or risk stratification for geriatric morbidity such as disability or frailty are limited. We prospectively assessed the prognostic impact of the novel composite health assessment risk model (CHARM), a score established to predict 1-year nonrelapse mortality (NRM), among 1105 patients aged ≥60 years enrolled on the Bone Marrow Transplant Clinical Trials Network Study 1704. Secondary end points were assessed post-alloHCT at day 100 (D100), D180, and D365 in multivariable models adjusted with predetermined clinical variables. Among alloHCT survivors, the prevalence of disability by instrumental activities of daily living (IADL), frailty by the Physical Frailty Phenotype, and physical function impairment by Patient Reported Measurement Information System (PROMIS) was highest at D100 and lower on D180 and D365. Higher CHARM scores were independently associated with greater disability (coefficient, −0.64; 95% confidence interval [CI], −0.85 to −0.43; P < .001), increased frailty (coefficient, 0.19; CI, 0.081-0.31; P < .001), worse PROMIS physical function, greater PROMIS depression, increased serious organ toxicity by D100, more cognitive decline at D100, and higher mortality after acute graft-versus-host disease (GVHD) but not significantly associated with PROMIS anxiety or acute GVHD. Higher CHARM scores predicted worse disability-free survival (odds ratio [OR], 2.03; CI, 1.66-2.48; P < .001) and lower frailty-free survival (OR, 2.00; CI, 1.61-2.49). In summary, CHARM is an independent prognostic scoring system not only for NRM but also for geriatric morbidity and functional limitation–free survival through 1 year after alloHCT. Pre-alloHCT CHARM is a novel tool to aid shared decision-making for older patients. This trial was registered at www.clinicaltrials.gov as #NCT03992352.

Introduction

Allogeneic hematopoietic cell transplantation (alloHCT), although potentially curative for high-risk hematologic malignancies, causes well-known complications such as infections, graft-versus-host disease (GVHD), and organ toxicity, all of which may predispose to increased morbidity, impaired quality of life, and nonrelapse mortality (NRM). To risk stratify for the higher rates of NRM among older adults,1, 2, 3, 4 we recently reported the independent value of the composite health assessment risk model (CHARM) in the prospective Bone Marrow Transplant Clinical Trials Network (BMT CTN) 1704 to estimate 1-year NRM in alloHCT recipients aged ≥60 years; a higher CHARM score was also prognostic of lower 1-year overall survival (OS).⁵

However, traditional alloHCT end points of NRM, OS, and even GVHD do not describe the full landscape of older adult HCT complications, such as functional impairments, cognitive changes, and/or functional independence among survivors. Single-institutional data suggest adult alloHCT poses dangers in classic geriatric morbidities such as falls, disability, frailty, physical decline, and delirium, all of which undermine functional independence.6, 7, 8, 9, 10 To date, no large multi-institutional study has characterized geriatric morbidity and toxicity after alloHCT or comprehensively explored how to risk stratify for such morbidity. In addition, the 1-year OS rate of 72% reported under the BMT CTN 1704 study⁵ underscores the importance of quantifying morbidity beyond NRM among most of the older patients who survive alloHCT in the modern era.

Here, we report on the planned secondary end points from BMT CTN 1704, describing the associations between CHARM and posttransplant geriatric morbidity and toxicity.

Methods

This analysis describes the secondary end points and exploratory analysis of novel composite end points from a multicenter prospective observational cohort study (ClinicalTrials.gov identifier: NCT03992352) for older alloHCT recipients. The protocol has been previously published.⁵ Patients were followed for 1 year after alloHCT.

Patients and study details

Key eligibility included age ≥60 years, hematologic malignancy, eligible for alloHCT based on local institutional standards, and ability to speak and read English, Spanish, or Mandarin.

The protocol-specified secondary objectives were to determine the association of CHARM with the following end points: (1) frailty by the Physical Frailty Phenotype,¹¹ (2) disability by any limitation in the instrumental activities of daily living (IADL),¹² (3) skilled nursing facility admission, (4) health-related quality of life (HRQoL) using the Patient Reported Measurement Information System (PROMIS) for physical function, anxiety, and depression, (5) serious organ toxicity by day 100 (D100), (6) incidence of GVHD (acute grades 2-4 at D100, and chronic at D365), (7) survival after acute GVHD (aGVHD), and (8) cognitive decline by D100 using the Montreal Cognitive Assessment (MoCA).¹³ Supplemental Table 1 details study tools and end points for the secondary objectives. Secondary end points were assessed at milestone time points of D100, D180, and D365 (ie, 1 year), except for MoCA that was assessed only at baseline and D100 to minimize testing burden. Serious organ toxicity by D100 was prospectively defined using predefined organ-specific toxicities from registry data forms (supplemental Table 2).

CHARM

The CHARM score comprises 7 health assessment factors evaluated before conditioning: comorbidity by the hematopoietic cell transplantation–comorbidity index,¹⁴ patient weight loss over the prior year, patient reported Karnofsky performance score, c-reactive protein, albumin, cognition by MoCA, and patient age.⁵ An online calculator can be found at https://cibmtr.org/CIBMTR/OffNav/DevSandbox/CHARM-Risk-NRM-Calculator. CHARM score tertiles of <-1.3 for lowest and > 0.8 for the highest can also be determined from online CHARM calculator from point estimates of 1-year NRM reflecting lowest (<9%), intermediate (>10% to 16%), and highest tertiles (>16%).

Statistical methods

Descriptive statistics summarized each group at baseline and at follow-up time points. Logistic regression was used to model serious organ toxicity and cognitive decline by D100. Fine-Gray regression modeled acute and chronic GVHD, with death as a competing risk, whereas Cox regression was used to model survival after the development of aGVHD, starting at the time of aGVHD.

A sequential multiple imputation strategy was implemented to impute end points for survivors at each time point with missing data.¹⁵ Analyses on multiply imputed data sets were conducted, and the results were combined using Rubin’s rule.¹⁶ Multivariable analysis of the impact of CHARM as the main effect on each outcome adjusted for covariates selected using stepwise regression, considering conditioning intensity, GVHD prophylaxis regimen, disease-risk index,¹⁷ donor-recipient sex match, race, ethnicity, prior autologous HCT, graft type, donor type, donor-recipient CMV match, interval from diagnosis to transplant, baseline value of the dependent variable, and study visit time points (ie, D100, D180, and 1 year). Generalized estimating equations were used to model the mean disability, frailty, and PROMIS HRQoL measures for physical function, depression, and anxiety among surviving patients at each time point, with the results summarized as a regression coefficient for the mean response.

Functional limitation–free survival composite end points (ie, frailty-free survival, disability-free survival, and physical function impairment–free survival) at D100, D180, and D365 were analyzed using generalized estimating equations for a longitudinal binary outcome. Multiple imputation was needed for failure-free survival analyses to handle missing data on frailty, disability, or physical function. Failure was defined as either death or impaired by established thresholds (supplemental Table 1) at each time point. Functional limitation–free survival at each time point by CHARM tertiles were summarized by computing the proportions of death, alive with impairment, and alive without impairment for each multiple imputed data set and then averaging these proportions over the multiple imputed data sets.

The trial was approved by the National Marrow Donor Program institutional review board. All participants provided written informed consent.

Results

Patients

Among the 1105 alloHCT patients at 49 centers, median age was 67 years with 349 patients (22%) aged ≥70 years. Table 1 describes the baseline characteristics of the patients.

Table 1.

Patient and transplant characteristics

Characteristic	Total (N = 1105)
Age, y
Median (range)	67 (60-82)
60-69	756 (68%)
≥70	349 (32%)
Recipient sex
Male	704 (64%)
Female	401 (36%)
Race
Asian	46 (4%)
Black or African American	44 (4%)
White	984 (89%)
Other/not reported∗	25 (2%)
Ethnicity
Hispanic or Latino	59 (5%)
Not Hispanic or Latino	1030 (93%)
Not reported	16 (1%)
Time from diagnosis to HCT, median (IQR), mo	7 (5-13)
Primary disease
Acute myeloid leukemia	499 (45%)
Acute lymphoblastic leukemia	58 (5%)
Myelodysplastic syndromes	325 (29%)
Myeloproliferative disorders	108 (10%)
Non-Hodgkin lymphoma	62 (6%)
Other†	53 (5%)
Refined disease risk index
Low	31 (3%)
Intermediate	767 (69%)
High	225 (20%)
Very high	17 (2%)
Missing or n/a	65 (6%)
Number of centers	49
Graft type
Bone marrow	54 (5%)
Peripheral blood	1041 (94%)
Umbilical cord blood	10 (1%)
Donor type
HLA-identical sibling	169 (15%)
Haploidentical	227 (21%)
Matched unrelated	628 (57%)
Mismatched unrelated	81 (7%)
Donor/recipient sex match
Male/male	446 (40%)
Male/female	215 (19%)
Female/male	247 (22%)
Female/male	170 (15%)
Umbilical cord	10 (1%)
Not reported	17 (2%)
Donor/recipient CMV serostatus
+/+	324 (29%)
+/−	125 (11%)
−/+	353 (32%)
−/−	295 (27%)
Not reported	8 (1%)
GVHD prophylaxis
Tacrolimus ± others	588 (53%)
Cyclosporine ± others	52 (5%)
Post-HCT cyclophosphamide ± others	433 (39%)
Ex vivo T-cell depletion/CD34 selection ± others	17 (2%)
Others	15 (1%)
Conditioning regimen
Myeloablative	146 (13%)
Reduced intensity	751 (68%)
Nonmyeloablative	208 (19%)
Survival follow-up, median (range), d	736 (141-1316)

Data are presented as n (%) unless otherwise specified.

CMV, cytomegalovirus; IQR, interquartile range; N/A, not applicable.

^∗Other: American/Indian/Alaskan native = 6, Hawaiian/Pacific Islander = 3, more than 1 race = 8, and not reported = 14.

^†Other acute leukemia = 12, myelodysplastic syndrome/myeloproliferative neoplasm unclassifiable = 18, mantle cell lymphoma = 15, chronic lymphocytic leukemia/prolymphocytic leukemia = 6, Hodgkin lymphoma = 1, and multiple myeloma = 1.

Descriptive outcomes of geriatric morbidity

Evaluable patients

Table 2 presents the number of evaluable patients and corresponding summary scores for geriatric measures and PROMIS HRQoL at each time point. D100 frailty had limited complete data due to the initial survey version mistakenly omitting several patient-reported frailty questions only for this time point, which was later corrected. Supplemental Table 3 further describes the availability of the 5 components for D180 and 1-year frailty.

Table 2.

Evaluable patients and point prevalence by time point

	Baseline	Day 100	Day 180	1 year
Alive	1105 (100%)	1003 (91%)	932 (84%)	790 (71%)
Lost to follow-up	0 (0%)	0 (0%)	1 (0%)	2 (0%)
Died	0 (0%)	102 (9%)	172 (16%)	313 (28%)
Cognition by MoCA
Evaluable	1058 (96%)	830 (83%)
Median (IQR)	26 (24-28)	27 (25-28)
Evaluable for decline		814 (81%)
≥2-point decline		182 (22)
Frailty
Evaluable	721 (65%)	259 (26%)∗	474 (51%)	433 (55%)
0	22 (3%)	6 (2%)	33 (7%)	37 (9%)
1-2 or prefrail	361 (50%)	99 (38%)	285 (60%)	274 (63%)
3-5 or frail	338 (47%)	154 (59%)	156 (33%)	122 (28%)
Disability by IADL
Evaluable	982 (89%)	704 (70%)	761 (82%)	661 (84%)
Median (IQR)	14 (13-14)	13 (11-14)	13 (11-14)	14 (13-14)
Impaired (<14)	359 (37%)	441 (63%)	418 (55%)	269 (41%)
PROMIS tools
Evaluable	982 (89%)	707 (70%)	768 (82%)	664 (84%)
PROMIS physical function
Median (IQR)	44 (39-50)	40 (35-45)	43 (36-50)	45 (39-52)
Impaired (<45)	491 (50%)	505 (72%)	457 (60%)	314 (48%)
PROMIS depression
Median (IQR)	44 (38-50)	46 (38-51)	46 (38-51)	46 (38-50)
Impaired (>55)	80 (8%)	94 (13%)	102 (13%)	83 (13%)
PROMIS anxiety
Median (IQR)	48 (37-54)	46 (37-51)	46 (37-50)	44 (37-50)
Impaired (>55)	187 (19%)	95 (14%)	99 (13%)	73 (11%)

Evaluable among those alive at corresponding time point. Data are presented as n (%) unless otherwise specified.

IQR, interquartile range.

^∗Patient survey mistakenly omitted several frailty questions at D100.

Univariate summary scores

Median MoCA scores were 26 and 27 at baseline and D100, respectively; however, 22% of the patients had cognitive decline from baseline to day 100 based on the prespecified threshold of MoCA decline of at least 2 points (Table 2). The prevalence of functional impairments after alloHCT peaked at the first time point of D100 (59% frail, 63% with disability, and 72% with impaired PROMIS physical function), with lower prevalence for D180 and 1-year survivors (1 year: 40% frail, 41% with disability, and 48% with impaired physical function) (Table 2). For depression, the prevalence of impairment was 13% for each time point after alloHCT. Impairment by anxiety was present for 14%, 13%, and 11% at D100, D180, and 1 year, respectively.

Unadjusted trajectories of geriatric morbidity and HRQoL by CHARM

Figure 1 displays the unadjusted trajectories by mean score for each outcome by CHARM tertile, referencing standard impairment thresholds for the clinical ease of interpretation. Baseline impairments were the highest for patients in the highest CHARM tertile (ie, worst 1-year NRM risk) relative to the lower tertiles, although none of these domains are part of the CHARM score (except MoCA where the outcome was a change score). The patients in the highest CHARM tertile also had a higher prevalence of impairments post-alloHCT for frailty, disability, cognition, and PROMIS physical function. In contrast, emotional health domains of PROMIS depression and anxiety had overlapping confidence intervals (CIs) for CHARM tertiles.

Figure 1.

Longitudinal trajectories after transplant by CHARM tertiles of morbidity outcomes among survivors, mean and 95% CIs, and reference threshold for mild-to-moderate impairment. (A) Physical frailty phenotype omitting D100; (B) disability by IADL; (C) cognition by MoCA; and PROMIS for (D) physical function, (E) depression, and (F) anxiety. The CHARM tertiles are lowest 1-year NRM (green), intermediate (purple), and highest (blue).

Multivariable analysis of secondary end points

Outcomes at each time point were analyzed in multivariable models among survivors to independently assess CHARM and test other prognostic factors.

Frailty and disability

A higher CHARM score was associated with higher frailty score (estimate, 0.193; CI, 0.081-0.31; P = .0008). Similarly, a higher CHARM score was associated with more disability after alloHCT (estimate, −0.64; CI, −0.85 to 0.43; P < .0001) (Table 3). This is interpreted as a 1-point increase in CHARM score being associated with a 0.193-point increase in mean frailty score and a 0.64-point lower mean IADL score. Additional independent prognostic factors included baseline values of the dependent variable (ie, baseline frailty for post-HCT frailty and baseline disability for subsequent disability) and time point of post-alloHCT, indicating earlier time points, independent of other factors, have higher risks of frailty and disability. GVHD prophylaxis using posttransplant cyclophosphamide (PTCy) correlated with lower post-HCT disability vs tacrolimus (estimate, 0.34; CI, 0.051-0.64, P = .021) but PTCy was not associated with frailty.

Table 3.

Multivariate analysis for frailty and disability among survivors

Parameter by outcome	Coefficient	95% CI	P value
Frailty
CHARM	0.193	0.081-0.305	.0008
Intercept	2.475	2.161-2.790	.0000
Baseline frailty∗	0.177	0.107-0.247	.0000
Time point
Day 180	Reference
Day 365	−0.208	−0.299 to −0.117	.0000
Disease-risk index
Low/intermediate	Reference
High/very high	0.160	0.003-0.318	.0461
Donor-recipient sex match
F-F	Reference		<.0001 (4 df)
F-M	−0.790	−1.012 to −0.567	.0000
M-F	−0.053	−0.258 to 0.153	.6117
M-M	−0.797	−0.998 to −0.595	.0000
NA-cord	−0.285	−1.013 to 0.442	.4413
Ethnicity
Non-Hispanic	Reference
Hispanic	0.329	0.030-0.628	.0310
Time from diagnosis to transplant	0.002	0.000-0.003	.0202
Disability by IADL
CHARM	−0.640	−0.846 to −0.433	.0000
Intercept	5.267	3.969-6.565	.0000
Disability at baseline∗	0.428	0.328-0.528	.0000
Time point
Day 100	Reference		<.0001 (2 df)
Day 180	0.325	0.139-0.512	.0010
Day 365	0.901	0.704-1.098	.0000
GVHD prophylaxis
PTCy ± others	Reference		.0180 (4 df)
Tacrolimus ± others	0.344	0.051-0.637	.0219
Cyclosporine ± others	0.508	−0.044 to 1.059	.0710
Ex vivo T-cell depletion	0.084	−0.977 to 1.144	.8764
Other	−0.802	−2.278 to 0.674	.2840

Generalized estimating equations were used to model the mean outcome measures among surviving patients at each time point, with the results summarized as a regression coefficient for the mean response. Results are aggregated across multiply imputed data sets to account for missing data.

CHARM, composite health assessment risk model; df, degrees of freedom; F; female, M; male; NA-cord, not applicable for cord; GVHD, graft-versus-host-disease.

CHARM boldface as the main effect.

^∗Higher baseline frailty score associated with more post-HCT frailty; lower baseline disability score associated with lower post-HCT disability.

PROMIS HRQoL

Higher CHARM scores were associated with the risk of worse physical function (estimate, −0.98; CI, −1.90 to −0.057; P = .038) and depression (estimate, 0.76; CI, 0.042-1.48; P = .038); the association with more anxiety was not significant (estimate, 0.66; CI, −0.070 to 1.39; P = .076) (supplemental Table 4). Consistent with frailty and disability, worse baseline values for each HRQoL domain (ie, physical function, anxiety, and depression) were prognostic of worse HRQoL in the same domain after alloHCT. Physical function HRQoL improved over time (P <.0001 for D180 vs D100 and 1 year vs D100). Anxiety (P = .14) or depression (P = .83) did not differ at D180 vs D100, although patients reported less anxiety (P = .0005) and borderline reduction in depression at 1 year compared to D100 (P = .056).

Serious organ toxicity

Thirteen percent of patients experienced serious organ toxicities by D100, with the most common organ toxicities being endotracheal intubation (8%) and septic shock (5%) (supplemental Table 5). The median time from HCT to organ toxicity was 23 days. Each 1-point increase in the CHARM score was associated with a twofold increase in the risk of serious organ toxicity (odds ration [OR], 2.06; CI, 1.52-2.78; P < .0001) (supplemental Table 6).

Cognitive decline

In multivariable logistic regression, only higher CHARM scores were significantly associated with cognitive decline at D100 (OR, 1.56; CI, 1.16-2.09; P = .0037) (data not shown).

GVHD

CHARM score did not pose higher risks for grade 2 to 4 aGVHD (hazard ratio [HR], 0.97; P = .67) or grade 3 to 4 aGVHD (HR, 1.12; P = .41) (supplemental Table 7). However, a higher CHARM score was linked to increased mortality after the diagnosis of grade 2 to 4 aGVHD (HR, 1.61; CI, 1.2-2.08; P = .0002) (supplemental Figure 1). A higher CHARM score was associated with lower risks of chronic GVHD (HR, 0.83; CI, 0.71-0.98; P = .0264) (supplemental Table 7), potentially related to this higher mortality after aGVHD leaving fewer patients with high CHARM scores at risk for chronic GVHD. GVHD prophylaxis using PTCy reduced the risk of aGVHD grade 2 to 4 and chronic GVHD relative to tacrolimus or cyclosporine, although PTCy did not influence survival after aGVHD.

Skilled nursing facility admission

Only 2% of the patients at D100 and 3% of patients at 1 year self-reported a recent admission to a skilled nursing facility (supplemental Table 9). Due to few events, we did not test CHARM for this outcome.

Multivariable CHARM summary of secondary end points

The Figure 2 forest plot depicts the independent influence of CHARM on all the secondary end points, showing that higher CHARM was linked to higher morbidity (although anxiety was not significant) for all outcomes except for acute and chronic GVHD.

Figure 2.

Association of CHARM score to post-alloHCT morbidity by (A) regression slope and 95% CIs on mean disability by IADL, Physical Frailty Phenotype, and PROMIS anxiety, depression, and physical function across time points of D100, D180, and 1 year, adjusted for baseline characteristics. Frailty is only at D180 and 1 year. (B) Adjusted subdistribution HRs for aGVHD and chronic GVHD, HRs for death after aGVHD, and adjusted ORs for serious organ toxicity by D100 or a decline in cognitive score of ≥2 points by MoCA. A higher CHARM score is associated with less activities of daily living (more disability), greater frailty, greater depression, worse physical function, lower rates of chronic GVHD, higher rates of death after aGVHD, and higher odds of serious organ toxicities and cognitive decline but not with incidence of aGVHD.

Multivariable model of functional limitation–free survival

To quantify the most desirable outcome, we explored the composite end point of functional limitation–free survival across physical function measures. Table 4 lists the factors independently influencing frailty-free survival, disability-free survival, and physical function impairment–free survival. Higher CHARM scores were strongly associated with inferior frailty-free survival (OR, 2.0; CI, 1.61-2.50; P < .0001), disability-free survival (OR, 2.03; CI, 1.66-2.48; P < .0001), and physical function impairment–free survival (OR, 1.74; CI, 1.44-2.1; P < .0001) (Table 4). The multivariable analyses again adjusted for baseline deficits in the functional domain (eg, adjusted for baseline disability for post-alloHCT disability). Higher disease risk adversely influenced all 3 composite functional limitation–free outcomes. Figure 3 presents the proportion of patients meeting frailty- and disability-free survival by CHARM tertiles at each time point to visualize patient-relevant outcomes. For example, at 1-year post-alloHCT, low, intermediate, and high CHARM tertile patients experienced frailty-free survival of 57.1%, 51.6%, and 36.4% and disability-free survival of 51.7%, 44.2%, and 27.3%, respectively. Restricted to surviving patients only at each time point, the proportion of 1-year survivors reporting disability in the low, intermediate, and high CHARM tertiles was 36.3%, 39.6%, and 54.5%, respectively.

Table 4.

Multivariate analysis of functional limitation–free survival by frailty-free survival, disability-free survival, and physical function impairment–free survival

Parameter by outcome	OR	95% CI	P value
Frailty-free survival∗
CHARM	1.998	1.606-2.487	.0000
Intercept	1.679	0.828-3.402	.1489
Frailty at baseline	1.238	1.082-1.416	.0020
Time point∗
Day 180	Reference
Day 365	1.158	0.992-1.352	.0627
Disease-risk index
Low/intermediate	Reference
High/very high	1.803	1.385-2.347	.0000
D-R sex matching			<.001
F-F	Reference
F-M	0.438	0.294-0.652	.0001
M-F	1.082	0.710-1.649	.7118
M-M	0.396	0.267-0.587	.0000
NA-cord	0.742	0.215-2.557	.6351
GVHD prophylaxis			.041
PTCy	Reference
Tacrolimus ± others	0.874	0.676-1.130	.3010
Cyclosporine ± others	1.554	0.876-2.756	.1313
Ex vivo TCD	1.501	0.636-3.542	.3537
Other	1.963	0.808-4.768	.1357
Interval from dx to HCT	1.004	1.001-1.007	.0150
Disability-free survival
CHARM	2.028	1.657-2.482	.0000
Intercept	824.11	191.41-3548.2	.0000
Baseline disability by IADL†	0.707	0.636-0.785	.0000
Time point			<.001
Day 100	Reference
Day 180	0.816	0.688-0.966	.0188
Day 365	0.646	0.540-0.772	.0000
Conditioning intensity			.019
MAC	Reference
RIC	0.649	0.470-0.897	.0089
NMA	0.647	0.425-0.984	.0421
D-R CMV matching			.015
+/+	Reference
+/−	0.874	0.609-1.256	.4669
−/+	0.701	0.511-0.963	.0288
−/−	0.680	0.503-0.919	.0122
Disease-risk index
Low/intermediate	Reference
High/very high	1.474	1.107-1.963	.0084
D-R sex matching			.013
F-F	Reference
F-M	0.758	0.534-1.077	.1214
M-F	1.298	0.890-1.893	.1745
M-M	0.915	0.637-1.315	.6251
N/A	0.828	0.269-2.550	.7408
Race			.034
White	Reference
Asian/Pacific Islander	0.714	0.434-1.175	.1837
Black or African American	1.981	1.001-3.919	.0497
Other/unknown	0.768	0.219-2.696	.6793
Physical function impairment–free survival
CHARM	1.741	1.439-2.105	.0000
Intercept	32.040	16.054-63.946	.0000
Baseline physical function†	0.944	0.930-0.957	.0000
Time point			<.001
Day 100	Reference
Day 180	0.758	0.642-0.895	.0014
Day 365	0.676	0.557-0.821	.0002
Disease-risk index
Low/intermediate	Reference
High/very high	1.462	1.125-1.898	.0047
D-R sex matching			.011
F-F	Reference
F-M	0.828	0.586-1.171	.2853
M-F	1.299	0.869-1.943	.1987
M-M	0.905	0.623-1.316	.5941
N/A	1.671	0.661-4.224	.2768

CHARM, composite health assessment risk model; CMV, seropositive for cytomegalovirus; D-R, donor-recipient; dx, diagnosis; F; female, M; male MAC, myeloablative conditioning; NA-cord, not applicable for cord; NMA, nonmyeloablative conditioning; RIC, reduced-intensity conditioning; TCD; T-cell depletion.

CHARM boldface as the main effect.

^∗Day 100 not modeled due to missing data.

^†Higher baseline frailty associated with worse frailty-free survival; Less baseline disability associated with better disability-free survival; higher baseline physical function associated with better physical function impairment-free survival.

Figure 3.

Proportions of patients experiencing functional limitation–free survival. (A) Frailty-free survival at D180 and 1 year and (B) disability-free survival at D100, D180, and 1 year by CHARM tertiles.

Discussion

This is the first large prospective national longitudinal multicenter study characterizing geriatric morbidity and prognostic factors among patients aged at least 60 years undergoing alloHCT. Here, we found that the composite model (CHARM), recently developed to predict 1-year NRM,⁵ provides independent estimates of the risks of patient-oriented outcomes including frailty, disability, physical function, cognition, and toxicities among survivors. The results fill a major gap in the field to address not only quantity but also quality of survival after alloHCT. Stated differently, we must consider not only life span but appreciate health span, the period of one’s life spent in good health and free of chronic diseases and disabilities that commonly accompany aging.^18,19

We found a high prevalence of geriatric impairments at baseline, including 37% with a disability by IADL (ie, complex tasks required to live independently)¹² and 47% meeting frailty criteria, confirming substantial heterogeneity in health among medically cleared older alloHCT patients. Physical function limitations for disability, frailty, and patient-reported physical function peaked at the first milestone, with 59% to 70% of the patients demonstrating impairments at D100 vs 28% to 48% for those impaired at 1 year. These data are consistent with single-institutional studies of functional decline early after alloHCT and later improvements among survivors.8, 9, 10^,20 In contrast, impairments in HRQoL for emotional health for depression and anxiety were lower and stable at ∼10% to 15% across post-alloHCT time points. We also found that a significant fraction (22%) of patients had cognitive decline using the minimally important difference of ≥2 points in MoCA.²¹

We hypothesized and confirmed an independent association of higher CHARM scores with geriatric morbidities among survivors, including frailty, disability, cognitive decline, patient-reported physical function and depression, serious organ toxicity and a nonsignificant association with anxiety after alloHCT. The strong association of earlier time points (eg, D100 relative to D180 and 1 year) with greater impairments supports progressive functional recovery over the first year. PROMIS emotional health impairments were less pronounced and were more stable over time; D180 values did not differ relative to D100. However, by 1 year, anxiety scores were comparatively lower and there was a borderline reduction in patient-reported depression scores.

To characterize health span, we advanced the novel end point of functional limitation–free survival quantified by disability-free survival, frailty-free survival, and patient-reported physical function impairment–free survival. A higher CHARM score strongly and independently stratified for worse outcomes in all 3 composite end points. We believe that introducing and risk stratifying for functional limitation–free survival end points paves the way for more informed shared decision-making before alloHCT. For example, in the lowest CHARM tertile, >50% of patients were alive, frailty- or disability-free at 1 year, which may reassure older patients and clinicians to disrupt the chronologic age barrier for alloHCT and potentially cure more patients.^22,23 Conversely, among the highest CHARM tertile patients, 1-year disability-free survival was only 27%, requiring a review of the options to dampen morbidity and engage in a broader discussion of patient goals from alloHCT and alternatives.

We assessed physical limitation using patient-reported limitations in IADL, frailty by the Physical Frailty Index, and moderate impairment in PROMIS physical function. No consensus exists on the optimal method to determine acceptable function for older patients after alloHCT. Our results follow the World Health Organization International Classification of Functioning model for a person with a health condition that considers components of functioning and disability as well as components of contextual factors.²⁴ The Behavioral Risk Factor Surveillance System studied functional disability in the context of cancer survivors, defining it as mobility disability by difficulty walking or climbing stairs and self-care disability requiring difficulty bathing or dressing.²⁵ Mobility disability among cancer survivors aged ≥65 years was 31.2% vs 6.6% for self-care disability. Prospective cardiovascular studies have adopted disability-free survival end points for older adult studies using more substantial physical disability defined as dementia, self-care limitations in Katz activities of daily living (eg, bathing, dressing, or toileting) for 6 months, or admission to a nursing care facility for disability.^26,27 Although the optimal approach varies and may differ by treatment context, the inclusion of functional limitation–free end points in future trials should motivate research prolonging health span. Additional research will be needed to evaluate these composite end points in comparison or combined with GVHD-free, relapse-free survival.²⁸

One advantage of standardized functional limitation measures in our study is that they enable insights into alloHCT survivors relative to normative data. For example, the prevalence of frailty was 28% for 1-year survivors in our cohort, which was double of the 13.5% reported for community-dwelling older adults aged 60 to 69 years.²⁹ Similarly, the prevalence of IADL disability of 41% in the CHARM cohort aged >60 years appears markedly higher than the 16% prevalence among community-dwelling older adults aged ≥65 years (mean age of 74.5 years).³⁰ The findings indicate older adult 1-year alloHCT survivors experience a high burden of physical function limitations.

This prospective study allowed us to collect detailed information about patients’ health, transplant characteristics, and hematologic malignancy risk to adjust the models. This ensures accounting for the most important potential confounders and further highlights the robust and reliable predictability of CHARM for these outcomes. These data support CHARM as a robust clinical tool to risk stratify not only older adult NRM but also morbidity and recovery after alloHCT.

How CHARM may mediate post-alloHCT morbidity emerges as a central question to devise interventions. Rather than a set of factors that directly predispose to morbidity (eg, high c-reactive protein and infection risks and baseline cardiovascular comorbidity resulting in cardiovascular events), we propose pre-alloHCT CHARM quantifies physical resilience to post-HCT stressors such as GVHD, infection, organ damage, relapse, or their treatments.³¹ Consistent with this concept, CHARM did not influence the rates of aGVHD but was a strong prognostic factor for mortality after a diagnosis of aGVHD. Similarly, studies in older general medical patients have shown that pre–intensive care unit disability exerts an adverse effect on 1-year survival after intensive care unit discharge.³² Hence, the magnitude of post-alloHCT stressors and degree of physical resilience likely interact to govern morbidity and functional recovery. Such a model guides potential interventions to accelerate post-alloHCT functional recovery by considering physical resilience both before and after alloHCT when physiologic stressors frequently emerge.

We believe interventional strategies based on baseline CHARM scores may ensure timely and safe access to alloHCT. For patients with lower CHARM scores, research into applying CHARM earlier in the disease course to expedite the transplant process or to test higher-intensity therapies for the highest-risk diseases (eg, more intensive conditioning and maintenance) could be pursued. Alternatively for patients with higher CHARM scores, interventions may include bolstering baseline resilience through geriatric-targeted optimization^33,34 and/or abrogating HCT stressors (eg, GVHD prevention, infection prevention, and lower toxicity conditioning). We do not recommend CHARM as a tool to simply select against alloHCT in higher-risk patients. Non-alloHCT approaches in older adults have dismal outcomes and cannot be directly compared to these alloHCT data.

The study has important limitations. Even as the largest prospective study on this topic in alloHCT, generalizability will benefit from external validation, especially in select subsets (eg, age of ≥75 years, specific donor types, uncommon diseases, cord blood, or underrepresented race or ethnicities) with sufficient patient numbers. Our study could not profile all geriatric morbidity or all time points. Early functional decline or delirium before D100, especially when one considers 9% of patients died by D100, requires additional study. Missing data may introduce bias; additional data were missed due to the COVID-19 pandemic. We may have underestimated deficits at specific time points if patients missed evaluations due to being unwell, as evidenced by slightly lower response rates at D100. We designed the study to test a validated score, CHARM. Although additional research may refine the best predictors to risk stratify for specific end points (eg, cognition and frailty-free survival), we believe CHARM offers a unified tool for geriatric morbidity and NRM.

In conclusion, our data emphasize that older patients have a significant burden of geriatric morbidity and toxicity after alloHCT by D100, with gradual improvement among first year survivors. Pre-alloHCT CHARM stratifies for geriatric morbidity and toxicity, empowering patients and clinicians to account for NRM, survival, geriatric morbidity, and functional limitation–free survival. The favorable recovery among those with low CHARM scores further supports alloHCT in appropriate older candidates. Nevertheless, the consequential morbidity among those with high CHARM scores underscores a pressing need to devise novel strategies to mitigate post-HCT morbidity and enhance recovery.

Conflict-of-interest disclosure: A.S.A. reports advisory roles for AstraZeneca and Magenta Therapeutics; and consulting with AbbVie and Daiichi Sankyo. W.A.W. reports research support from Pfizer and Genentech; equity in and consulting for Koneksa Health; and consulting for Teladoc Health, Quantum Health, and the American Society of Hematology Research Collaborative. A.M. reports grant support from Novartis. P.I. reports research support from Janssen. V.R.B. reports participating in the safety monitoring committee for Protagonist; serving as an associate editor for the journal Current Problems in Cancer and a contributor for BMJ Best Practice; consultancy for Imugene, Sanofi, and Taiho; research support from MEI Pharma, Actinium Pharmaceutical, Sanofi US Services, AbbVie, Pfizer, Incyte, Jazz, and National Marrow Donor Program; and drug support (institutional) from Chimerix for a trial. R.O. reports research support from Cellectis; and consulting for Servier and Rigel. J.M. reports research support from Gilead, Atara, CRISPR, Precision Biosciences, Scripps Research Institute, Vor Bio, and Affimed. S.W. reports speaker’s bureau fees from Sobi. M.L.S. reports consultancy and receiving honoraria from Jazz Pharmaceuticals for giving educational talks; and reports receiving research funding from Bluenote. The remaining authors declare no competing financial interests.

Jeremy Walston died on 10 June 2025.