Abstract
Background
Long-term mortality after hematopoietic cell transplantation (HCT) is conventionally calculated from the time of HCT, ignoring temporal changes in survivors’ mortality risks. Conditional survival rates, accounting for time already survived, are relevant for optimal delivery of survivorship care but have not been widely quantified. We estimated conditional survival by elapsed survival time in allogeneic HCT patients and examined cause-specific mortality.
Methods
We calculated conditional survival rates and standardized mortality ratio for overall and cause-specific mortality in 4485 patients who underwent HCT for malignant hematologic diseases at a large transplant center during 1976–2014. Statistical tests were two-sided.
Results
The 5-year survival rate from HCT was 48.6%. After surviving 1, 2, 5, 10, and 15 years, the subsequent 5-year survival rates were 71.2%, 78.7%, 87.4%, 93.5%, and 86.2%, respectively. The standardized mortality ratio was 30.3 (95% confidence interval [CI] = 29.2 to 35.5). Although the standardized mortality ratio declined in longer surviving patients, it was still elevated by 3.6-fold in survivors of 15 years or more (95% CI = 3.0 to 4.1). Primary disease accounted for 50% of deaths in the overall cohort and only 10% in 15-year survivors; the leading causes of nondisease-related mortality were subsequent malignancy (26.1%) and cardiopulmonary diseases (20.2%). We also identified the risk factors for nondisease-related mortality in 1- and 5-year survivors.
Conclusion
Survival probability improves the longer patients survive after HCT. However, HCT recipients surviving 15 years or more remain at elevated mortality risk, largely because of health conditions other than their primary disease. Our study findings help inform preventive and interventional strategies to improve long-term outcomes after allogeneic HCT.
Hematopoietic cell transplantation (HCT) is now an established curative treatment for many malignant and nonmalignant hematologic diseases. The number of survivors continues to grow because of improved efficacy, transplantation strategies, and supportive care (1). There are currently 200 000 HCT survivors in the United States, a number that will exceed 500 000 by 2030 (1).
Despite improvements in long-term outcomes, HCT survivors continue to have substantially higher mortality rates compared with the general population (2–4). Long-term mortality after HCT has conventionally been calculated from the time of HCT, which does not account for temporal variations in survivors’ mortality risks. Conditional survival is the survival probability given that the cohort has already survived a given length of time (5). We previously reported on conditional survival following autologous HCT, demonstrating clinically important changes in causes of late mortality after autologous HCT (6). Allogeneic HCT (alloHCT) survivors are at high risk of morbidity and mortality because of the cumulative effects of pre-HCT therapy and conditioning and post-HCT complications (eg, graft-versus-host disease [GvHD]) (7). Previous studies examining cause-specific conditional survival after alloHCT have been limited to the first 2–5 years after HCT (2,4). A better understanding of cause-specific survival trends can be obtained by stratifying on longer time survived since alloHCT. This would allow for implementation of risk-based screenings and interventions tailored to time-varying mortality risk in long-term survivors. It would also facilitate timely and appropriate discussion between health-care providers and patients as part of comprehensive survivorship care.
The objectives of this study were to estimate the conditional survival rates in patients who underwent alloHCT at a large transplant center during 1976–2014 stratified by time survived since transplant; to compare all-cause and cause-specific mortality risks to those of the general population; and to identify the risk factors of late nondisease-related mortality (NDRM) after alloHCT.
Methods
Study Population
The Long-term Follow-up Program supports the complete follow-up of all patients receiving HCT for hematologic diseases at City of Hope (COH) and was the primary resource of data for the current study. A uniform protocol is used to collect patients’ clinical and demographic data, health outcomes, and vital status from medical record and external data linkage (National Death Index Plus program) (8–10). The Long-term Follow-up Program protocol was approved by the Institutional Review Board at COH; written informed consent was obtained from all participants in accordance with the Declaration of Helsinki. A total of 4972 pediatric and adult patients underwent a first alloHCT for hematologic disease at COH during 1976–2014. We excluded 341 who received a prior autologous HCT, 144 who refused participation, and 2 who died the day of alloHCT. The current report comprises 4485 individuals, of whom 2667 had died during follow-up. Medical records were the source documentation for 1987 (74.5%) deaths, and NDI was the source for 570 (21.3%) deaths; cause of death was missing for 110 (4.1%) deaths.
Causes of death were prioritized as follows: external causes, primary disease, subsequent malignant neoplasms (SMN), GvHD, infection, cardiovascular disease, pulmonary disease, organ failure, others, and unknown (6,11). Cumulative disease-related mortality (DRM; due to primary disease) and NDRM (all other causes) were computed by treating NDRM and DRM respectively as competing risks.
Statistical Analysis
Conditional survival estimates were obtained using the Kaplan-Meier method for the subcohort that had already survived a given length (x) of time after alloHCT. If S(x) is the unconditional (traditional) survival probability at time x, then the conditional survival probability S at time y > x is S(y|x)=S(x + y)/S(x). Because conditional survival can vary by temporal changes in patient characteristics, transplant efficacy, and supportive care, we also calculated 5-year survival rates using average values of age and relapse risk at alloHCT, sex, donor type, GvHD, and diagnosis of 1-year survivors as covariate values in Cox regression for 1- to 16-year survivors by HCT year categories. The standardized mortality ratio (SMR), or the ratio of observed to expected number of deaths, was used to quantify all-cause and cause-specific mortality risks for the cohort compared with the general population. Person-years at risk were computed from an anchor date (date of or 1 year after alloHCT) until date of death, date last known alive, or December 31, 2016, whichever came first. The expected number of deaths was calculated using the calendar year-, sex-, race (white or nonwhite)-, and age-specific US mortality rates (12,13). The equality of SMRs was tested; two-sided 95% confidence intervals (CI) were calculated (13).
Risk factors of NDRM in 1- and 5-year survivors were determined using multivariable Fine-Gray subdistributional hazard regression treating non-NDRM deaths as competing risk (14). Regardless of statistical significance, the baseline model included sex, age at alloHCT (treated as continuous orcategorical), race (non-Hispanic white, Hispanic, Asian, African American or other or unknown race), diagnosis (acute lymphocytic leukemia, acute myeloid leukemia or myelodysplastic syndromes, chronic myeloid leukemia, non-Hodgkin lymphoma or Hodgkin lymphoma, chronic lymphocytic leukemia or multiple myeloma or other leukemias, severe aplastic anemia or other hematologic diseases), relapse risk at alloHCT (standard, high) (Supplementary Table 1, available online) (15), and calendar year of alloHCT (continuous or categorical). The effects of the following were examined univariately, adjusted for covariates in the baseline model: graft type (bone marrow, cord blood, peripheral stem cells), donor type (related, unrelated), myeloablative conditioning (yes, no), occurrence of grade II–IV acute GvHD or any chronic GvHD within one year of alloHCT (yes, no), total body irradiation (TBI; yes, no) and conditioning chemotherapy agents administered to more than 49 patients. The variable with the lowest P value less than .05 was added to the baseline model. The statistical significance of the remaining variables was assessed one at a time in the updated model. The process continued until no more variables could be added. SAS 9.4 (SAS Institute Inc., Cary, NC) was used. All tests were two-sided; two-sided P less than .05 was considered statistically significant.
Results
Cohort Characteristics
The median age at alloHCT was 38.8 years (range = 0.3–75.4 years); 56.4% were male; 53.6% were non-Hispanic white; 62.0% received stem cells from related donors; 68.5% underwent myeloablative conditioning; 58.9% were treated with TBI; 45.3% were at high risk of relapse at alloHCT; and 68.6% developed GvHD within 1 year of alloHCT (Table 1).
Table 1.
Demographic and clinical characteristics of the allogeneic hematopoietic cell transplantation cohort, by diagnosis
| Variables | Entire cohort | AML | ALL | CML | MDS | NHL | SAA | Other* |
|---|---|---|---|---|---|---|---|---|
| Patients, No. (%) | 4485 (100.0) | 1448 (100.0) | 1026 (100.0) | 685 (100.0) | 479 (100.0) | 417 (100.0) | 180 (100.0) | 250 (100.0) |
| Sex, No. (%) | ||||||||
| Male | 2531 (56.4) | 721 (49.8) | 619 (60.3) | 401 (58.5) | 267 (55.7) | 264 (63.3) | 103 (57.2) | 156 (62.4) |
| Female | 1954 (43.6) | 727 (50.2) | 407 (39.7) | 284 (41.5) | 212 (44.3) | 153 (36.7) | 77 (42.8) | 94 (37.6) |
| Age at transplantation, median (range), y | 38.8 (0.3–75.4) | 41.5 (0.8–74.1) | 27.4 (0.6–72.9) | 38 (5–71.5) | 52.7 (0.7–74.9) | 47.7 (2.1–74.5) | 24 (2.5–66.5) | 37.3 (0.3–75.4) |
| Age at transplantation, No. (%), y | ||||||||
| <18 | 568 (12.7) | 161 (11.1) | 224 (21.8) | 24 (3.5) | 21 (4.4) | 21 (5.0) | 55 (30.6) | 62 (24.8) |
| 18–29 | 965 (21.5) | 266 (18.4) | 337 (32.9) | 159 (23.2) | 35 (7.3) | 60 (14.4) | 73 (40.5) | 35 (14.0) |
| 30–39 | 812 (18.1) | 255 (17.6) | 190 (18.5) | 197 (28.8) | 57 (11.9) | 54 (13.0) | 26 (14.4) | 33 (13.2) |
| 40–49 | 904 (20.2) | 281 (19.4) | 154 (15.0) | 208 (30.4) | 90 (18.8) | 105 (25.2) | 16 (8.9) | 50 (20.0) |
| 50–59 | 741 (16.5) | 267 (18.4) | 88 (8.6) | 81 (11.8) | 136 (28.4) | 113 (27.1) | 9 (5.0) | 47 (18.8) |
| ≥60 | 495 (11.0) | 218 (15.1) | 33 (3.2) | 16 (2.3) | 140 (29.2) | 64 (15.4) | 1 (0.6) | 23 (9.2) |
| Years survived, No. (%) | ||||||||
| ≥1 | 2956 (65.9) | 896 (61.9) | 663 (64.6) | 470 (68.6) | 316 (66.0) | 287 (68.8) | 153 (85.0) | 171 (68.4) |
| ≥2 | 2578 (57.5) | 761 (52.6) | 548 (53.4) | 429 (62.6) | 285 (59.5) | 254 (60.9) | 144 (80.0) | 157 (62.8) |
| ≥5 | 1800 (40.1) | 496 (34.3) | 364 (35.5) | 375 (54.7) | 172 (35.9) | 171 (41.0) | 117 (65.0) | 105 (42.0) |
| ≥10 | 1113 (24.8) | 283 (19.5) | 217 (21.2) | 314 (45.8) | 85 (17.7) | 78 (18.7) | 86 (47.8) | 50 (20.0) |
| ≥15 | 708 (15.8) | 169 (11.7) | 139 (13.5) | 247 (36.1) | 32 (6.7) | 38 (9.1) | 54 (30.0) | 29 (11.6) |
| Race, No. (%) | ||||||||
| White | 3750 (83.6) | 1176 (81.2) | 894 (87.1) | 572 (83.5) | 409 (85.4) | 361 (86.6) | 132 (73.3) | 206 (82.4) |
| Black/African American | 156 (3.5) | 40 (2.8) | 31 (3.0) | 31 (4.5) | 9 (1.9) | 13 (3.1) | 9 (5.0) | 23 (9.2) |
| Asian | 449 (10.0) | 184 (12.7) | 76 (7.4) | 68 (10.0) | 51 (10.6) | 37 (8.9) | 19 (10.6) | 14 (5.6) |
| Hawaiian/Pacific Islander/ American Indian/Alaska Native | 35 (0.8) | 19 (1.3) | 6 (0.6) | 3 (0.4) | 2 (0.4) | 3 (0.7) | 1 (0.6) | 1 (0.4) |
| Other | 92 (2.0) | 29 (2.0) | 18 (1.8) | 11 (1.6) | 8 (1.7) | 3 (0.7) | 17 (9.4) | 6 (2.4) |
| Missing | 3 (0.1) | 0 | 1 (0.1) | 0 | 0 | 0 | 2 (1.1) | 0 |
| Ethnicity, No. (%) | ||||||||
| Hispanic | 1381 (30.8) | 382 (26.4) | 470 (45.8) | 203 (29.6) | 94 (19.6) | 100 (24.0) | 78 (43.3) | 54 (21.6) |
| Non-Hispanic | 3094 (69.0) | 1062 (73.3) | 554 (54.0) | 481 (70.2) | 384 (80.2) | 317 (76.0) | 102 (56.7) | 194 (77.6) |
| Missing | 10 (0.2) | 4 (0.3) | 2 (0.2) | 1 (0.2) | 1 (0.2) | 0 | 0 | 2 (0.8) |
| Year of transplantation, No. (%) | ||||||||
| 1976–1989 | 557 (12.4) | 188 (13.0) | 159 (15.5) | 113 (16.5) | 20 (4.2) | 15 (3.6) | 35 (19.5) | 27 (10.8) |
| 1990–1999 | 1083 (24.2) | 284 (19.6) | 221 (21.5) | 347 (50.7) | 87 (18.1) | 57 (13.7) | 42 (23.3) | 45 (18.0) |
| 2000–2009 | 1782 (39.7) | 575 (39.7) | 417 (40.7) | 190 (27.7) | 203 (42.4) | 217 (52.0) | 74 (41.1) | 106 (42.4) |
| 2010–2014 | 1063 (23.7) | 401 (27.7) | 229 (22.3) | 35 (5.1) | 169 (35.3) | 128 (30.7) | 29 (16.1) | 72 (28.8) |
| Donor type, No. (%) | ||||||||
| Related | 2783 (62.0) | 888 (61.3) | 662 (64.5) | 449 (65.6) | 234 (48.9) | 263 (63.1) | 130 (72.2) | 157 (62.8) |
| Unrelated | 1702 (38.0) | 560 (38.7) | 364 (35.5) | 236 (34.4) | 245 (51.1) | 154 (36.9) | 50 (27.8) | 93 (37.2) |
| Graft type, No. (%) | ||||||||
| BM | 2088 (46.6) | 557 (38.5) | 485 (47.3) | 515 (75.2) | 136 (28.4) | 103 (24.7) | 168 (93.3) | 124 (49.6) |
| PBSC | 2217 (49.4) | 832 (57.5) | 475 (46.3) | 160 (23.4) | 326 (68.1) | 302 (72.4) | 10 (5.6) | 112 (44.8) |
| BM + PBSC | 11 (0.3) | 4 (0.2) | 1 (0.1) | 1 (0.1) | 4 (0.8) | 0 | 0 | 1 (0.4) |
| Cord blood | 169 (3.8) | 55 (3.8) | 65 (6.3) | 9 (1.3) | 13 (2.7) | 12 (2.9) | 2 (1.1) | 13 (5.2) |
| HCT conditioning | ||||||||
| Reduced intensity/nonmyeloablative | 1399 (31.2) | 419 (28.9) | 106 (10.3) | 30 (4.4) | 293 (61.2) | 221 (53.0) | 173 (96.1) | 157 (62.8) |
| Myeloablative | 3074 (68.5) | 1028 (71.0) | 911 (88.8) | 655 (95.6) | 186 (38.8) | 196 (47.0) | 6 (3.3) | 92 (36.8) |
| Missing | 12 (0.3) | 1 (0.1) | 9 (0.9) | 0 | 0 | 0 | 1 (0.6) | 1 (0.4) |
| Relapse risk at HCT, No. (%) | ||||||||
| Standard | 2396 (53.4) | 900 (62.1) | 698 (68.0) | 312 (45.6) | 103 (21.5) | 123 (29.5) | 180 (100) | 80 (32.0) |
| High | 2031 (45.3) | 541 (37.4) | 326 (31.8) | 373 (54.4) | 346 (72.2) | 292 (70.0) | 0 | 153 (61.2) |
| Missing | 58 (1.3) | 7 (0.5) | 2 (0.2) | 0 | 30 (6.3) | 2 (0.5) | 0 | 17 (6.8) |
| TBI, No. (%) | ||||||||
| No TBI used | 1833 (40.9) | 590 (40.8) | 147 (14.3) | 187 (27.3) | 394 (82.3) | 218 (52.3) | 134 (74.4) | 163 (65.2) |
| <8 Gy | 198 (4.4) | 43 (3.0) | 18 (1.8) | 12 (1.8) | 12 (2.5) | 27 (6.5) | 41 (22.8) | 45 (18.0) |
| ≥8 Gy | 2442 (54.4) | 814 (56.2) | 852 (83.0) | 486 (70.9) | 73 (15.2) | 172 (41.2) | 4 (2.2) | 41 (16.4) |
| Missing | 12 (0.3) | 1 (0.1) | 9 (0.9) | 0 | 0 | 0 | 1 (0.6) | 1 (0.4) |
| Grade II–IV acute GvHD or any chronic GvHD within 1 year of HCT, No. (%) | ||||||||
| Yes | 3077 (68.6) | 993 (68.6) | 683 (66.6) | 469 (68.5) | 360 (75.2) | 303 (72.7) | 114 (63.3) | 155 (62.0) |
| No | 1180 (26.3) | 383 (26.4) | 297 (29.0) | 176 (25.7) | 106 (22.1) | 102 (24.4) | 45 (25.0) | 71 (28.4) |
| Missing | 228 (5.1) | 72 (5.0) | 46 (4.5) | 40 (5.8) | 13 (2.7) | 12 (2.9) | 21 (11.7) | 24 (9.6) |
CLL = chronic lymphocytic leukemia (n = 66), other leukemia (n = 15), Hodgkin lymphoma (n = 52), multiple myeloma (n = 44), and other nonmalignant hematologic disorders (n = 73). ALL = acute lymphoblastic leukemia; AML = acute myeloid leukemia; BM = bone marrow; CML = chronic myelogenous leukemia; GvHD = graft-versus-host disease; HCT = hematopoietic cell transplantation; MDS = myelodysplastic syndrome; NHL = non-Hodgkin lymphoma; PBSC = peripheral blood stem cells; SAA = severe aplastic anemia; TBI = total body irradiation.
Overall and Conditional Survival
The 5-year survival rate for the overall cohort from the time of alloHCT was 48.6% (95% CI = 47.1% to 50.0%) (Figure 1A). The most common cause of death was primary disease (50.0%), followed by GvHD (19.9%), infection (10.0%), cardiopulmonary causes (6.7%), other or unknown causes (6.6%), SMN (4.5%), organ failure (1.3%), and external causes (1.0%). The subsequent unadjusted 5-year survival rates for those who already survived 1, 2, 5, 10, and 15 years since alloHCT were 71.2% (95% CI = 69.4% to 72.8%), 78.7% (95% CI = 76.9% to 80.3%), 87.4% (95% CI = 85.7% to 89.0%), 93.5% (95% CI = 91.7% to 94.9%), and 86.2% (95% CI = 82.8% to 88.9%), respectively (Figure 1A). Supplementary Figure 1 (available online) shows the corresponding plots by primary diagnosis. Conditional survival rates were consistently higher for patients with severe aplastic anemia compared with other diagnoses. Conditional 5-year survival rates adjusted for age and relapse risk at alloHCT, sex, donor type, GvHD, and diagnosis for 1- to 16-year survivors are shown in Supplementary Table 2 (available online). The unadjusted and adjusted rates averaged across HCT year categories (73.0%, 80.4%, 89.0%, 92.5%, and 81.2% for patients who lived 1, 2, 5, 10, or 15 years after alloHCT, respectively) were consistent.
Figure 1.
A) Survival probability, and cumulative mortality of (B) disease-related and (C) nondisease-related deaths, by years since allogeneic hematopoietic cell transplantation (overall) and stratified on time survived since allogeneic hematopoietic cell transplantation (≥1, ≥2, ≥5, ≥10, and ≥15 years). alloHCT = allogeneic hematopoietic cell transplantation.
Comparison of Mortality Risk With the General Population
The mortality risk of the alloHCT cohort was 30-fold compared with the general US population (SMR = 30.3, 95% CI = 29.2 to 35.5; Table 2). For patients who survived at least 1, 2, 5, 10, or 15 years after alloHCT, the SMRs were 13.3 (95% CI = 12.5 to 14.0), 9.1 (95% CI = 8.4 to 9.7), 5.2 (95% CI = 4.7 to 5.8), 3.6 (95% CI = 3.1 to 4.1), and 3.6 (95% CI = 3.0 to 4.1), respectively. Similar decreases in the SMRs were seen across all diagnosis groups (Table 2). The SMR was statistically significantly higher for males (43.8, 95% CI = 41.6 to 45.9) compared with females (21.2, 95% CI = 20.0 to 22.5; P < .001), in white (31.1, 95% CI = 29.8 to 32.4) than in nonwhite patients (26.7, 95% CI = 24.1 to 29.3; P = .0045), and in younger than in older patients (eg, 85.8, 95% CI = 77.5 to 94.1 aged younger than 21 years vs 12.8, 95% CI = 11.3 to 14.3 aged 60 years and older; P < .001). These differences persisted as survival time increased. Although the SMR was statistically significantly higher for patients at high risk of relapse (38.0, 95% CI = 36.0 to 40.0) compared with those at standard risk (23.5, 95% CI = 22.2 to 24.8; P < .001) and for those who underwent HCT with an unrelated donor (39.3, 95% CI = 36.9 to 41.7) compared with those with a related donor (26.5, 95% CI = 25.2 to 27.8; P < .001), these group differences began to mitigate about 5 years after HCT (Table 2). Person-years at risk and the observed and expected number of deaths are available in Supplementary Table 3 (available online).
Table 2.
Standardized mortality ratios (95% confidence interval) by patient and clinical characteristics and cause of death, stratified by minimum survival time
| Standardized mortality ratio (95% CI) |
||||||
|---|---|---|---|---|---|---|
| Overall (n = 4485) | ≥1-year survival (n = 2956) | ≥2-year survival (n = 2578) | ≥5-year survival (n = 1800) | ≥10-year survival (n = 1113) | ≥15-year survival (n = 709) | |
| All cohort | 30.3 (29.2 to 35.5) | 13.3 (12.5 to 14.0) | 9.1 (8.4 to 9.7) | 5.2 (4.7 to 5.8) | 3.6 (3.1 to 4.1) | 3.6 (3.0 to 4.1) |
| By patient and clinical characteristics | ||||||
| Sex | ||||||
| Male | 43.8 (41.6 to 45.9) | 20.3 (18.8 to 21.8) | 13.9 (12.7 to 15.2) | 8.4 (7.3 to 9.4) | 6.3 (5.2 to 7.4) | 5.7 (4.5 to 7.0) |
| Female | 21.2 (20.0 to 22.5) | 8.5 (7.7 to 9.3) | 5.9 (5.2 to 6.5) | 3.2 (2.7 to 3.8) | 2.0 (1.6 to 2.5) | 2.3 (1.7 to 2.9) |
| P * | <.001 | <.001 | <.001 | <.001 | <.001 | <.001 |
| Race | ||||||
| White | 31.1 (29.8 to 32.4) | 13.7 (12.8 to 14.6) | 9.5 (8.7 to 10.2) | 5.6 (5.0 to 6.2) | 4.0 (3.4 to 4.6) | 3.9 (3.2 to 4.6) |
| Nonwhite | 26.7 (24.1 to 29.3) | 11.2 (9.6 to 12.9) | 7.3 (6.0 to 8.7) | 3.7 (2.7 to 4.7) | 1.7 (0.9 to 2.5) | 1.7 (0.7 to 2.7) |
| P | .0045 | .018 | .013 | .0082 | .0017 | .018 |
| Primary diagnosis | ||||||
| AML | 32.8 (30.7 to 34.9) | 13.5 (12.1 to 14.9) | 8.8 (7.7 to 9.9) | 4.9 (4.0 to 5.8) | 3.5 (2.6 to 4.5) | 3.8 (2.6 to 5.0) |
| ALL | 63.5 (58.7 to 68.4) | 29.3 (26.0 to 32.6) | 18.6 (15.9 to 21.3) | 9.6 (7.5 to 11.6) | 8.3 (6.1 to 10.5) | 9.1 (6.3 to 11.9) |
| CML | 18.9 (17.0 to 20.8) | 8.6 (7.3 to 9.8) | 6.6 (5.5 to 7.7) | 4.7 (3.7 to 5.6) | 2.9 (2.2 to 3.7) | 2.2 (1.5 to 2.9) |
| MDS | 21.4 (18.9 to 24.0) | 8.7 (7.0 to 10.3) | 6.4 (5.0 to 7.8) | 3.8 (2.5 to 5.0) | 1.8 (0.7 to 2.9) | † |
| NHL | 26.3 (22.9 to 29.8) | 11.9 (9.6 to 14.2) | 8.2 (6.2 to 10.1) | 4.4 (2.8 to 5.9) | 2.4 (0.8 to 4.0) | † |
| SAA | 17.0 (12.4 to 21.6) | 9.0 (5.7 to 12.3) | 6.1 (3.4 to 8.9) | 4.0 (1.7 to 6.2) | 2.2 (0.4 to 3.9) | † |
| P | <.001 | <.001 | <.001 | <.001 | <.001 | <.001 |
| Age at alloHCT, y | ||||||
| <21 | 85.8 (77.5 to 94.1) | 36.2 (30.7 to 41.7) | 23.7 (19.3 to 28.2) | 12.0 (8.7 to 15.3) | 8.2 (5.3 to 11.1) | 8.0 (5.0 to 11.1) |
| 21–39 | 47.7 (44.7 to 50.7) | 20.4 (18.4 to 22.4) | 13.9 (12.3 to 15.5) | 8.3 (7.0 to 9.6) | 5.9 (4.8 to 7.0) | 5.0 (3.9 to 6.1) |
| 40–59 | 24.6 (23.1 to 26.1) | 10.6 (9.6 to 11.6) | 7.3 (6.5 to 8.2) | 3.9 (3.2 to 4.6) | 2.0 (1.5 to 2.5) | 1.7 (1.1 to 2.3) |
| ≥60 | 12.8 (11.3 to 14.3) | 6.5 (5.5 to 7.6) | 4.6 (3.7 to 5.5) | 2.7 (1.9 to 3.5) | 1.0 (0.3 to 1.8) | † |
| P | <.001 | <.001 | <.001 | <.001 | <.001 | <.001 |
| Year of alloHCT | ||||||
| 1976–1989 | 41.8 (38.0 to 45.7) | 17.3 (14.8 to 19.8) | 13.1 (10.9 to 15.3) | 9.2 (7.4 to 11.1) | 7.7 (6.0 to 9.4) | 7.0 (5.4 to 8.7) |
| 1990–1999 | 28.6 (26.5 to 30.7) | 12.0 (10.7 to 13.4) | 9.0 (7.9 to 10.2) | 5.8 (4.9 to 6.8) | 4.1 (3.3 to 5.0) | 2.8 (2.1 to 3.5) |
| 2000–2009 | 27.0 (25.3 to 28.6) | 12.3 (11.2 to 13.4) | 8.4 (7.4 to 9.3) | 4.1 (3.5 to 4.8) | 0.9 (0.5 to 1.3) | † |
| 2010–2014 | 33.8 (30.7 to 36.9) | 15.0 (13.0 to 17.1) | 7.9 (6.4 to 9.5) | 2.0 (0.8 to 3.2) | —‡ | —‡ |
| P | <.001 | .001 | <.001 | <.001 | <.001 | <.001 |
| Relapse risk at alloHCT | ||||||
| Standard risk | 23.5 (22.2 to 24.8) | 11.8 (10.8 to 12.8) | 8.0 (7.2 to 8.8) | 4.9 (4.2 to 5.5) | 3.5 (2.9 to 4.1) | 3.3 (2.6 to 4.0) |
| High risk | 38.0 (36.0 to 40.0) | 15.2 (14.0 to 16.5) | 10.5 (9.4 to 11.6) | 5.8 (5.0 to 6.7) | 3.7 (2.9 to 4.6) | 4.0 (2.9 to 5.0) |
| P | <.001 | <.001 | <.001 | .084 | .70 | .36 |
| Donor type | ||||||
| Related | 26.5 (25.2 to 27.8) | 12.3 (11.4 to 13.2) | 8.8 (8.0 to 9.5) | 5.3 (4.7 to 5.9) | 3.8 (3.2 to 4.4) | 3.7 (3.0 to 4.3) |
| Unrelated | 39.3 (36.9 to 41.7) | 15.6 (14.1 to 17.1) | 9.8 (8.6 to 11.0) | 5.0 (4.0 to 6.1) | 2.6 (1.6 to 3.5) | 2.8 (1.3 to 4.2) |
| P | <.001 | .001 | .16 | .70 | .065 | .37 |
| By cause of death | ||||||
| Cardiovascular | 3.7 (2.8 to 4.5) | 2.8 (2.1 to 3.5) | 2.4 (1.8 to 3.1) | 1.9 (1.2 to 2.7) | 1.9 (1.2 to 2.7) | 2.0 (1.1 to 3.0) |
| Pulmonary | 16.0 (12.6 to 19.3) | 8.9 (6.3 to 11.4) | 6.3 (4.2 to 8.5) | 4.4 (2.5 to 6.4) | 4.2 (1.9 to 6.5) | 4.4 (1.5 to 7.3) |
| Subsequent malignancy | 6.4 (5.4 to 7.5) | 4.9 (4.0 to 5.8) | 4.5 (3.7 to 5.4) | 3.9 (3.1 to 4.8) | 3.4 (2.5 to 4.4) | 3.7 (2.5 to 5.0) |
Two-sided test of equality of SMRs. alloHCT = allogeneic hematopoietic cell transplantation; ALL = acute lymphocytic leukemia; AML = acute myeloid leukemia; CI = confidence interval; CML = chronic myeloid leukemia; MDS = myelodysplastic syndromes; NHL = non-Hodgkin lymphoma; SAA = severe aplastic anemia.
Standardized mortality ratio is not reported when the number of survivors is <50 and/or the number of deaths is <5.
Not applicable.
Cause-Specific Cumulative Mortality
As expected, the DRM risk increased precipitously in the first year after alloHCT and plateaued with increasing survival (Figure 1B). In contrast, the NDRM risk continued to increase with time (Figure 1C). The DRM and NDRM risks diverge sharply in 5-year survivors of alloHCT (Figure 2). The cause of death varied depending on the number of years already survived (Figure 3). Whereas primary diseases accounted for 50.0% and GvHD for 19.9% of deaths in the entire cohort, they accounted for increasingly fewer deaths among longer surviving patients (10.4% for primary disease, 2.2% for GvHD in 15-year survivors). In contrast, SMN and cardiopulmonary diseases accounted for an increasing percentage of deaths as survivors lived longer, from 4.5% for SMN and 6.7% for cardiopulmonary diseases in all survivors to 26.1% and 20.2% for the respective diseases in 15-year survivors (Supplementary Table 4, available online). Among SMN-related deaths, the proportion of solid tumors increased with longer survival (Supplementary Table 5, available online). Lastly, mortality risks of SMN and cardiopulmonary causes statistically significantly exceeded the corresponding risks in the general population even after 15-year survival (Table 2).
Figure 2.
Cumulative cause-specific mortality (95% confidence bands) for 5-year survivors of allogeneic hematopoietic cell transplantation. alloHCT = allogeneic hematopoietic cell transplantation.
Figure 3.
Distribution of causes of deaths by minimum years survived after allogeneic hematopoietic cell transplantation. GvHD = graft-versus-host disease; SMN = subsequent malignant neoplasms.
Risk Factors of Nondisease-Related Mortality in 1- and 5-Year Survivors
In 1-year survivors, NDRM was statistically significantly higher in males (hazard ratio [HR] = 1.31, 95% CI = 1.10 to 1.56), older patients at alloHCT (HR = 1.03, 95% CI = 1.02 to 1.04 per year increase in age; 40–60 years, HR = 1.78, 95% CI = 1.44 to 2.21; 60 years and older, HR = 2.33, 95% CI = 1.63 to 3.32 compared with younger than 40 years), patients with unrelated donors (HR = 1.38, 95% CI = 1.12 to 1.71), patients who developed chronic GvHD within 1 year of alloHCT (HR = 1.86, 95% CI = 1.49 to 2.32), and patients treated with TBI (HR = 1.35, 95% CI = 1.08 to 1.70) (Table 3). NDRM decreased for alloHCT performed more recently (HR = 0.95 per calendar year, 95% CI = 0.94 to 0.96). Relative to 1976–1989, the hazard ratio decreased steadily in subsequent 5-year calendar intervals of alloHCT: 0.70, 0.61, 0.42, 0.32, and 0.26 (Ptrend < .001). These results were generally similar for 5-year survivors (Table 3), including the effects of myeloablative conditioning, which were not statistically significant for 1- or 5-year survivors. We also examined the hazard ratio by calendar year in patients transplanted during 1976–2011 followed for at least 5 years. The decline in NDRM in more recent HCT eras remained evident (HR = 0.95 per calendar year increase, 95% CI = 0.94 to 0.97; HR = 0.71, 0.61, 0.43, and 0.32 for HCTs from 1990–1994, 1995–1999, 2000–2004, and 2005–2010, respectively, relative to 1976–1989; P < .001).
Table 3.
Multivariable regression results of the risk factors of nondisease-related mortality in 1- and 5-year survivors*
| Covariables | 1-year survivors | 5-year survivors | ||
|---|---|---|---|---|
| (n = 2956) | (n = 1725) | |||
| No. NDR deaths (n = 555) |
No. NDR deaths (n = 234) |
|||
| HR (95% CI) | P | HR (95% CI) | P | |
| Sex | ||||
| Female | 1.0 (Referent) | 1.0 (Referent) | ||
| Male | 1.31 (1.10 to 1.56) | .002 | 1.39 (1.06 to 1.82) | .02 |
| Age at alloHCT, y | ||||
| As continuous variable | 1.03 (1.02 to 1.04) | <.001 | 1.06 (1.04 to 1.07) | <.001 |
| As categorical variable | <.001† | <.001† | ||
| <40 | 1.0 (Referent) | 1.0 (Referent) | ||
| 40–60 | 1.78 (1.44 to 2.21) | <.001 | 2.78 (1.99 to 3.89) | <.001 |
| >60 | 2.33 (1.63 to 3.32) | <.001 | 5.39 (2.77 to 10.47) | <.001 |
| Race/ethnicity | .08† | .11† | ||
| Non-Hispanic white | 1.0 (Referent) | 1.0 (Referent) | ||
| Hispanic | 1.16 (0.96 to 1.41) | .13 | 0.93 (0.70 to 1.25) | .64 |
| Asian | 0.79 (0.57 to 1.10) | .16 | 0.57 (0.34 to 0.97) | .04 |
| African American, other race, missing | 1.30 (0.87 to 1.92) | .20 | 1.44 (0.76 to 2.74) | .27 |
| Primary diagnosis | .10† | .03† | ||
| ALL | 1.0 (Referent) | 1.0 (Referent) | ||
| AML, MDS | 0.82 (0.65 to 1.03) | .09 | 0.82 (0.57 to 1.18) | .29 |
| CML | 0.68 (0.53 to 0.88) | .003 | 0.55 (0.38 to 0.80) | .002 |
| NHL, HL | 0.76 (0.54 to 1.06) | .10 | 0.74 (0.42 to 1.29) | .29 |
| CLL, MM, other leukemia | 0.76 (0.45 to 1.29) | .31 | 0.42 (0.18 to 0.99) | .047 |
| SAA, other hematologic diseases | 0.79 (0.48 to 1.32) | .37 | 0.79 (0.34 to 1.83) | .58 |
| Donor type | ||||
| Related | 1.0 (Referent) | 1.0 (Referent) | ||
| Unrelated | 1.38 (1.12 to 1.71) | .002 | 1.35 (0.94 to 1.95) | .10 |
| Relapse risk at alloHCT | ||||
| Standard risk | 1.0 (Referent) | 1.0 (Referent) | ||
| High risk | 1.16 (0.97 to 1.40) | .11 | 1.28 (0.95 to 1.71) | .10 |
| GvHD within 1 y of alloHCT | ||||
| No | 1.0 (Referent) | 1.0 (Referent) | ||
| Yes | 1.86 (1.49 to 2.32) | <.001 | 1.33 (1.00 to 1.78) | 0.05 |
| Total body irradiation | ||||
| Unexposed | 1.0 (Referent) | 1.0 (Referent) | ||
| Exposed | 1.35 (1.08 to 1.70) | .01 | 1.77 (1.18 to 2.64) | .006 |
| Year of alloHCT | ||||
| As continuous variable | 0.95 (0.94 to 0.96) | <.001 | 0.95 (0.93 to 0.97) | <0.001 |
| As categorical variable | <.001† | <.001† | ||
| 1976–1989 | 1.0 (Referent) | 1.0 (Referent) | ||
| 1990–1994 | 0.70 (0.53 to 0.94) | .02 | 0.66 (0.46 to 0.97) | .03 |
| 1995–1999 | 0.61 (0.46 to 0.82) | <.001 | 0.53 (0.35 to 0.80) | .003 |
| 2000–2004 | 0.42 (0.31 to 0.57) | <.001 | 0.45 (0.29 to 0.71) | <.001 |
| 2005–2009 | 0.32 (0.23 to 0.45) | <.001 | 0.33 (0.19 to 0.57) | <.001 |
| 2010–2014 | 0.26 (0.18 to 0.37) | <.001 | 0.26 (0.09 to 0.80) | .02 |
Sex, age at alloHCT, race and/or ethnicity, primary diagnosis, relapse risk, and year of alloHCT were included in the model regardless of statistical significance. ALL = acute lymphocytic leukemia; alloHCT = allogeneic hematopoietic cell transplantation; AML = acute myeloid leukemia; CI = confidence interval; CLL = chronic lymphocytic leukemia; CML = chronic myeloid leukemia; GvHD = graft-versus-host disease; HL = Hodgkin lymphoma; HR = hazard ratio; MDS = myelodysplastic syndromes; MM = multiple myeloma; NHL = non-Hodgkin lymphoma; NDR = nondisease-related; SAA = severe aplastic anemia.
P value for test of homogeneity of categorical variables.
Discussion
The overall 5-year survival probability of our cohort from the time of alloHCT was 48.6%. However, the conditional probability of surviving an additional 5 years increased with elapsed survival time to 71.2%, 78.7%, 87.4%, and 93.5% after surviving 1, 2, 5, and 10 years, respectively, then declined to 86.2% after 15 years. The decline was consistent with the steep increase in NDRM among 15-year survivors. The overall all-cause mortality was 30-fold that of the general population, which also declined as patients survived longer. Nevertheless, the risk remained elevated even in 15-year survivors because of increasing NDRM, notably SMN and cardiovascular diseases. The current study findings are critically important for long-term survivors and their health-care providers, because they would help facilitate discussions regarding long-term prognosis, conditioned on survival time after HCT. Importantly, understanding the temporally changing burden of NDRM can facilitate timely screening and intervention to mitigate its risk.
The all-cause standardized mortality risk varied by patient characteristics: higher in males than in females, higher in white patients than in nonwhite patients, and higher in younger than in older patients. These differences persisted regardless of the length of survival. In contrast, the higher mortality risk for patients at greater (vs standard) risk of relapse at alloHCT and those with unrelated (vs related) donors diminished at 5 years after alloHCT. Although DRM plateaued as patients survived longer, NDRM continued to rise, with SMN and cardiopulmonary diseases increasingly accounting for these nondisease-related deaths. Solid tumors also increasingly accounted for SMN-related deaths. Notably, even in 15-year survivors the risk of dying from SMN and cardiopulmonary diseases exceeded that of the general population. Male sex, older age at alloHCT, having an unrelated donor, GvHD within 1 year of alloHCT, and TBI exposure were associated with higher NDRM. NDRM was also lower in patients receiving alloHCT more recently: the 2010–2014 rate was only one-fourth of that in 1976–1989. Overall, our results show that multiple factors explain part of the variability in mortality risk over time and that the causes of mortality shift over time, with NDRM accounting for an increasing percentage of mortality.
Because other factors might explain the improved unadjusted survival rates with time, we estimated the 5-year survival rates adjusted for treatment era, age and relapse risk at alloHCT, sex, stem cell donor type, presence of GvHD, and diagnosis. The adjusted 5-year conditional survival rates were similar to the unadjusted rates, thus these alternative factors are unlikely to explain the increase in survival probability with increased survivorship. Five-year conditional survival rates also started to decline after 15 years’ survival in both adjusted and unadjusted analyses (eg, from 93.5% to 86.2% in 10- and 15-year survivors, respectively). This is likely due to increasing NDRM, including deaths due to SMN and cardiopulmonary disease, rather than primary disease-related mortality hazard and is consistent with the increasing NDRM and plateauing DRM shown in the 5-year survivors. Similar decline and rise in mortality hazard with years after diagnosis have been reported in pediatric and adult solid cancer survivors (16,17).
Three other studies have reported the survival probabilities for 2- and 5-year survivors of alloHCT. One study reported a survival probability of 85% at 10 years from alloHCT for 2-year survivors, who were disease-free and had received a myeloablative alloHCT (4). Another study reported a survival probability of 90.2% at 5 years, 84.6% at 10 years, and 80.2% at 15 years from alloHCT for 2-year survivors who were disease-free at study entry (2). A third study reported a survival probability of 80.4% at 20 years for 5-year survivors of either allogeneic (56%) or autologous HCT (44%) who were disease-free (3). The characteristics and treatments varied among these patients. In our study, the conditional survival probability at 5, 10, and 15 years from alloHCT for 2-year survivors was 84.1%, 73.6%, and 68.8%, respectively. In 5-year survivors, the conditional survival probability at 20 years was 70.4%. Our estimates are lower than those of the other three studies, likely because our cohort included patients with and without active disease at the time of HCT. However, consistent with the prior results, we determined that the SMR was higher for patients with increased risk of relapse than for those at standard risk. Given the current study design, it was not feasible to determine the disease status at intermediate time points.
We observed that the mortality rates for alloHCT survivors with severe aplastic anemia were elevated relative to the general population in 2- and 5-year survivors. This result is different from that of a prior study, which reported that severe aplastic anemia patients free of primary disease and with resolved GvHD at 2 years did not have increased risk of death (18). However, another study found that the risk of death for patients surviving more than 2 years was greater than that of the general population (4). Once again, it is possible that differences in the patient population, including our inclusion of patients with active GvHD, may explain the differences in results.
As survival increased, the risk of death by cardiopulmonary disease and SMN increased. This is not surprising given that heart disease and cancer are the leading causes of aging-related mortality in the United States. Moreover, elevated risk of cardiovascular complications with pre-HCT treatments and post-HCT comorbidities in HCT survivors has been well documented (19,20). The increased mortality risk from these health conditions in long-term survivors highlights the excess burden of aging-related diseases in this population, which warrants ongoing clinical surveillance.
Strengths of this study include the large sample size and long duration of the study and a span across treatment eras to include contemporary changes in alloHCT practice such as increased utilization of reduced intensity conditioning (eg, in older patients), better supportive care, and GvHD management. This is the only study that examined cause-specific conditional survival across a range of survival durations. This study was limited because it was conducted at a single institution, which may not represent the general US alloHCT population. We observed that the NDRM decreased with calendar time, indicating that improvements in patient care, transplant practice, and conditioning regimens may have reduced the long-term NDRM rates. Additional investigation is needed to determine which practices may improve long-term NDRM.
There is an important clinical need to understand how mortality risk changes with survival time. This work supports the use of conditional survival to dynamically evaluate risk during survivorship. The overall survival probability, calculated conventionally from the time of HCT, does not reflect changes in prognosis with time. Conditional survival takes elapsed survival time into account, providing more relevant information for patients who already survived one or more years beyond alloHCT. The overall mortality risk for our cohort was 30-fold greater than that of the general population, indicating that ongoing clinical surveillance is warranted. Even 15-year survivors had mortality risks more than threefold greater than that of the general population. The subgroups that should be monitored closely to reduce premature mortality from NDRM include males, patients older than 40 years at alloHCT, patients with unrelated donors, patients developing GvHD within 1 year, and patients exposed to TBI. This information is important to both clinicians and patients, and it can help establish or strengthen screening recommendations.
Funding
This study was supported in part by the Lymphoma & Leukemia Society Scholar Award (Armenian).
Notes
Disclaimer: None of the authors have any conflict of interest.
The funder had no role in the design of the study; the collection, analysis, and interpretation of the data; the writing of the manuscript; and the decision to submit the manuscript for publication.
This paper was presented at the 60th American Society of Hematology Annual Meeting, December 1–4, 2018, San Diego, CA.
Supplementary Material
References
- 1. Majhail NS, Tao L, Bredeson C, et al. Prevalence of hematopoietic cell transplant survivors in the United States. Biol Blood Marrow Transplant. 2013;19(10):1498–1501. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 2. Bhatia S, Francisco L, Carter A, et al. Late mortality after allogeneic hematopoietic cell transplantation and functional status of long-term survivors: report from the Bone Marrow Transplant Survivor Study. Blood. 2007;110(10):3784–3792. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 3. Martin PJ, Counts GW Jr, Appelbaum FR, et al. Life expectancy in patients surviving more than 5 years after hematopoietic cell transplantation. J Clin Oncol. 2010;28(6):1011–1016. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 4. Wingard JR, Majhail NS, Brazauskas R, et al. Long-term survival and late deaths after allogeneic hematopoietic cell transplantation. J Clin Oncol. 2011;29(16):2230–2239. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 5. Hieke S, Kleber M, Konig C, Engelhardt M, Schumacher M. Conditional survival: a useful concept to provide information on how prognosis evolves over time. Clin Cancer Res. 2015;21(7):1530–1536. [DOI] [PubMed] [Google Scholar]
- 6. Vanderwalde AM, Sun CL, Laddaran L, et al. Conditional survival and cause-specific mortality after autologous hematopoietic cell transplantation for hematological malignancies. Leukemia. 2013;27(5):1139–1145. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 7. Sun CL, Francisco L, Kawashima T, et al. Prevalence and predictors of chronic health conditions after hematopoietic cell transplantation: a report from the Bone Marrow Transplant Survivor Study. Blood. 2010;116(17):3129–3139; quiz 3377. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 8. Wong FL, Francisco L, Togawa K, et al. Long-term recovery after hematopoietic cell transplantation: predictors of quality-of-life concerns. Blood. 2010;115(12):2508–2519. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 9. Armenian SH, Yang D, Teh JB, et al. Prediction of cardiovascular disease among hematopoietic cell transplantation survivors. Blood Adv. 2018;2(14):1756–1764. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 10. Sharafeldin N, Bosworth A, Patel SK, et al. Cognitive functioning after hematopoietic cell transplantation for hematologic malignancy: results from a prospective longitudinal study. J Clin Oncol. 2018;36(5):463–475. [DOI] [PubMed] [Google Scholar]
- 11. Copelan E, Casper JT, Carter SL, et al. A scheme for defining cause of death and its application in the T cell depletion trial. Biol Blood Marrow Transplant. 2007;13(12):1469–1476. [DOI] [PubMed] [Google Scholar]
- 12.Center for Disease Control and Prevention NCfHS. Compressed Mortality File 1999-2016 on CDC WONDER Online Database; 2017. https://wonder.cdc.gov/mortsql.html. Accessed June, 2018.
- 13. Breslow NE, Day N. Statistical Methods in Cancer Research Volume II—The Design and Analysis of Cohort Studies. Lyon: Oxford University Press; 1987. [PubMed] [Google Scholar]
- 14. Fine JP, Gray RJ. A proportional hazards model for the subdistribution of a competing risk. J Am Stat Assoc. 1999;94(446):496–509. [Google Scholar]
- 15. Sorror ML, Storb RF, Sandmaier BM, 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]
- 16. Swords DS, Mulvihill SJ, Firpo MA, Scaife CL. Causes of death and conditional survival estimates of medium- and long-term survivors of pancreatic adenocarcinoma. JAMA Oncol. 2018;4(8):1129–1130. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 17. Armstrong GT, Yasui Y, Robison LL. Reduction in late mortality after childhood cancer. N Engl J Med. 2016;375(3):290–292. [DOI] [PubMed] [Google Scholar]
- 18. Socie G, Stone JV, Wingard JR, et al. Long-term survival and late deaths after allogeneic bone marrow transplantation. Late Effects Working Committee of the International Bone Marrow Transplant Registry. N Engl J Med. 1999;341(1):14–21. [DOI] [PubMed] [Google Scholar]
- 19. Armenian SH, Sun C-L, Francisco L, et al. Late congestive heart failure after hematopoietic cell transplantation. J Clin Oncol. 2008;26(34):5537–5543. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 20. Armenian SH, Sun CL, Shannon T, et al. Incidence and predictors of congestive heart failure after autologous hematopoietic cell transplantation. Blood. 2011;118(23):6023–6029. [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.