Peri-transplant RBC Transfusion is Associated with Increased Risk of GvHD after Allogeneic Stem Cell Transplantation

Abstract

More than 90% of allogeneic hematopoietic stem cell transplant (allo-HSCT) patients receive red blood cell (RBC) or platelet transfusions in the peritransplant period. We tested the hypothesis that transfusions are associated with development of severe acute graft-versus-host disease (grade III/IV aGvHD) or mortality in allo-HSCT in a retrospective study of 322 consecutive patients receiving allogeneic bone marrow or G-CSF-mobilized blood stem cell grafts for hematological malignancies. Counting RBC and platelet units between day −7 pre-transplant and +27 post-transplant, but excluding transfusions administered after a diagnosis of aGvHD, yielded medians of 5 RBC and 2 platelet units transfused. 63 patients (20%) developed a maximal grade of III–IV aGvHD with onset up to day 150 post-transplant (median aGvHD onset of 28 days). HLA mis-match (HR 2.4 (1.2, 4.7), p=0.01), and transfusion of > median number of RBC units (HR 2.1 (1.1, 3.7), p=0.02) were independently associated with greater risk of grade III–IV aGvHD in a multivariable analysis model. Disease risk strata (HR 1.7 (1.2, 2.4) for high risk vs. low risk, p=0.005) and transfusion of > median RBC units (HR 1.4 (1.0, 2.0), p=0.054) were independently associated with inferior overall survival. These data support our hypothesis that peritransplant RBC transfusions are associated with the risk of developing severe aGvHD and worse overall survival following allo-HSCT, and suggest that strategies to reduce routine RBC transfusion may favorably reduce GvHD incidence and severity.

Introduction

Transplant-related acute graft-versus-host disease (aGvHD) is a major cause of non-relapse mortality after allo-HSCT and remains a significant barrier to the success of the transplant maneuver.^{1, 2} Alloreactive donor T-cells within the graft initiate an immune attack on recipient target tissues during GvHD pathogenesis.³ Numerous studies aimed at reducing GvHD severity and improving prognosis following allo-HSCT have been performed over the years, with only modest reductions in severe aGvHD incidence and treatment-related mortality.^{2, 4, 5} Successful clinical strategies for GvHD prevention have largely been limited to improvement in HLA matching and the use of in vivo and ex vivo T-cell depletion of the graft.^{6, 7, 8, 9} However, T-cell depletion of the graft has been associated with poor donor immune cell function and higher incidences of disease relapse and opportunistic infections.¹⁰

Risk factors for aGvHD include recipient age, donor gender, conditioning intensity, graft source, HLA match, and donor relation to patient.^{11, 12, 13} ABO mismatch between HSCT donor and recipient has been associated with increased risk of aGvHD in some studied patient populations but not in others.^14,15 Greater than 90% of allo-HSCT patients receive red blood cell (RBC) and platelet transfusions just prior to transplant and during the first month post-transplant, but transfusion has not been well studied as a risk factor for adverse outcomes.^{16, 17, 18, 19} Increased serum ferritin levels due to pre-transplant RBC transfusions were found to be associated with higher risk of GvHD by Pullarkat et al.,²⁰ but this was not seen in other studies.^{21, 22, 23} Pre-clinical studies in murine allo-HSCT models indicate that RBC transfusions can sensitize transplant recipients to minor histocompatibility antigens (miHA),^{24, 25, 26} while clinical studies of platelet transfusion found platelet-specific and anti-human leukocyte antibodies in multiple platelet-transfused patients.^{27, 28, 29} Thus, third-party transfusions might serve as a source of alloantigen that primes donor T cells, via indirect antigen presentation by donor-derived dendritic cells, to antigens mismatched between the stem cell donor and recipient,^{25, 30} and contribute to increased risk of aGvHD.^{31, 32, 33}

Herein, we present the results of a retrospective analysis of the association of RBC and platelet transfusions with GvHD and survival, and show, for the first time, that larger numbers of RBC transfusions during the pre- and post-transplant period are associated with increased incidences of maximal grade III/IV aGvHD and inferior overall survival in allo-HSCT patients.

Materials and Methods

Study endpoints and selective criteria

We conducted an IRB-approved retrospective single-center study of 322 consecutive adult patients who received allogeneic BMT or G-CSF-mobilized PBSC transplantation from sibling or unrelated donors between January 2007 and January 2013. A 10/10 allele match was considered a full match, while 7/10, 8/10, or 9/10 allele matches were considered mismatched, following molecular typing at the allele level. All patients had primary malignant hematological diseases, and patients receiving haplo-identical transplants, T-cell depleted grafts, or umbilical cord blood transplants were excluded. For patients who received a subsequent allogeneic transplant (N = 9), only transplant characteristics and transfusion and GvHD data relevant to the first allo-HSCT were included, but survival data and cause of death (if applicable) included status after the second transplant.

KPS score, transplant, and aGvHD

Karnofsky Performance Status (KPS) score was determined prior to admission.³⁴ Disease risk was classified as low, intermediate or high according to ASBMT/CIBMTR 2015 standards.³⁵ Patients were admitted a week prior to allo-HSCT for the conditioning regimen, which was classified as myeloablative (busulfan/cyclophosphamide, or cyclophosphamide/total body irradiation), reduced intensity (fludarabine/melphalan), or non-myeloablative (fludarabine/total body irradiation). Pre-transplant serum ferritin levels obtained between 5 and 90 days before transplant (median 28 days) were available for 299/322 patients. After transplantation of allogeneic bone marrow or G-CSF-mobilized PBSC, patients remained hospitalized until hematopoietic engraftment, and were monitored in the BMT clinic until day 100. Engraftment was defined according to CIMBTR criteria.³⁶ Acute GvHD onset within 150 days post-transplant was graded based on the 1994 Keystone consensus guidelines.³⁷ For our analyses, patients were assigned to aGvHD groups based on the maximal grade of aGvHD recorded after the initial onset of aGvHD. For example, patients with initial grade II aGVHD that became refractory developed grade III or IV aGVHD and were analyzed according to the higher assigned grade.

RBC and platelet transfusions

All RBC and platelet transfusions administered between day −7 and day +27 were provided by the Emory University Hospital Blood Bank after purchase from the American Red Cross Blood Services, Southern Region. RBC units were collected and stored in either CPD-AS1 or CPDA-1, and were stored for a maximum of 42 days or 35 days at 1–6°C, respectively. All RBC and platelet units were also irradiated; RBC were used within 28 days of irradiation. Standing orders specified transfusion of two RBC units when the daily hematocrit was <27%, or one irradiated platelet apheresis unit (stored for a maximum of 5 days at RT with agitation) when platelet counts were <10,000 per uL. These thresholds remained in effect throughout the timeframe of this study. Additional platelet units were given at the time of presentation of clinical bleeding, and additional RBC units were given for hemorrhage-induced anemia to maintain a hematocrit of >8%. Pre-transplant erythropoiesis stimulating agents (darbepoiten or epogen) were not used in these patients.

Statistical methods

Competing risk endpoints were: 1) grade III–IV aGvHD and 2) mortality without developing grade III/IV aGvHD (patients who died early, up to day 150, without aGvHD or having grades I–II aGvHD, including those diagnosed with chronic GvHD or overlap syndrome).³⁸ Of the 25 patients in the early-mortality competing risk group, 13 had no aGvHD, 5 had grade I aGvHD, and 7 had grade II aGvHD. Overlap syndrome chronic GvHD was diagnosed in 2 of the patients in the early mortality group, (2 of the 7 patients with grade II aGvHD). We did not exclude patients with de novo chronic GvHD from the early mortality group, however none of the patients who died up to day 150 with no aGvHD developed de novo chronic GvHD prior to this time point. The cumulative incidence of grade III–IV aGvHD and early mortality without developing grade III/IV aGvHD were estimated using the cumulative incidence function.³⁹ A competing risk analysis estimated the subdistribution hazard for grade III/IV aGvHD and non-grade III/IV aGVHD mortality using a Cox proportional hazards regression model implemented with SAS PROC PHreg (version 9.4). The subdistributional hazard ratio and its 95% confidence interval were calculated for each risk factor in univariable analysis. The covariates included in multivariable analysis was limited to baseline covariates associated with grade III/IV aGvHD at the p < 0.2 level. Similar Cox regression methods were used for analyses of risk factors potentially associated with all-cause mortality, excluding aGvHD from the multivariable analysis since this is a well-documented cause of death in allo-HSCT patients. Bootstrap bagging was used to identify stable and reliable predictors of Grade III–IV aGvHD. A dataset was constructed of equal size to the original (n=322) by random sampling of cases with replacement (bootstrap sampling). On average, approximately one-third of subjects were not sampled, whereas some subjects were sampled more than once. The bootstrap sample was analyzed using the Cox model with an automated forward stepwise algorithm with entry criterion of p <0.01 and a retention criterion of p<0.05. The result was stored. This process of sampling, automated analysis and storing was repeated 1000 times. The number of times a risk factor appeared in these 1000 analyses was taken as reflection of the reliability (signal). Following Brieman’s median rule (devised to balance type I and type II errors), risk factors were determined to be reliably associated with the outcome if they appeared in at least 50% of the models.^{40, 41} The cause-specific hazard ratio and its 95% confidence interval were calculated for each factor in the absence of others in the final model identified with bootstrap bagging. Cumulative long-term survival was estimated with the Kaplan-Meier method. Log-rank tests were used to compare survival by aGvHD grade and by numbers of RBC and platelet units transfused. 2×2 contingency tables and Fisher’s exact test were used to compare cause-of-death rates between groups.

Results

Patient characteristics summary

A total of 322 consecutive patients who underwent allo-HSCT from January 2007 to January 2013 were studied, excluding patients with non-malignant underlying disease or recipients of T-cell depleted or cord blood grafts. The median age at transplantation was 51 years old (range, 19 to 73). 55% of patients were male and 45% were female. The most common underlying disease was acute myeloblastic leukemia (41%). Underlying disease risk was low (41%), intermediate (26%) or high (32%). Allografts were obtained from 10/10 HLA-matched sibling donors (35%), unrelated donors (43%), or from donors mismatched at 1 or 2 HLA alleles (3 sibling donors and 65 unrelated donors, 1% and 20%, respectively). Graft sources were bone marrow (11%) or mobilized PBSC (89%). Please refer to Table 1 for additional patient characteristics details.

Table 1.

Transplant Patient Characteristics

Factor	Category	N=322	%
Age at transplant	Median (range)	51 (19–73)
Gender (Recipient)	Male / Female	178 / 144	55% / 45%
Gender (Donor)	Male / Female	195 / 127	61% / 39%
Combination	Female to Male / Other combinations	65 / 257	20% / 80%
Year of Transplant	Jan ’07 – Dec ’09 vs. Jan ’10 – Jan ‘13	157 / 165	49% / 51%
Graft source	Bone marrow / G-CSF mobilized PBSC	34 / 288	11% / 89%
Relationship to donor	Sibling / Unrelated	117 / 205	36% / 64%
HLA match	Sibling 10/10 match / Unrelated 10/10 match	114 / 140	35% / 43%
	Sibling <10/10 match / Unrelated <10/10 match	3 / 65	1% / 20%
ABO match	Matched / Mismatched	162 / 146	50% / 45%
	Unknown	14	4%
KPS score	≥ 90 / ≤ 80	230 / 92	71% / 29%
Disease	Acute myeloid leukemia	133	41%
	Myelodysplastic syndrome	46	14%
	Myeloproliferative syndrome	17	5%
	Acute lymphoblastic leukemia	40	12%
	Acute leukemia, biphenotypic or undifferentiated	4	1%
	Chronic lymphoid leukemia	15	5%
	Non-Hodgkin’s lymphoma	32	10%
	Hodgkin’s lymphoma	6	2%
	Multiple myeloma	7	2%
	Chronic myeloid leukemia	22	7%
Disease risk status	Low	133	41%
	Intermediate	85	26%
	High	104	32%
Conditioning Regimen	Myeloablative	134	41%
	Reduced intensity conditioning	159	50%
	Non-myeloablative	29	9%
GvHD Prophylaxis	Regimens containing Tacrolimus/Cyclosporine	268 / 54	83% / 17%
CMV antibody	Recipient and/or Donor positive	254	79%
	Both negative / Both unknown	39 / 29	12% / 9%

RBC and platelet transfusions

RBC and platelet transfusions were recorded from one week prior to transplantation to 4 weeks after transplantation. Only RBC and platelet units transfused before a diagnosis of aGvHD (any grade) were considered in univariable and multivariable analysis to avoid an indication bias from including units transfused due to GvHD-related anemia and thrombocytopenia. The median number of RBC units transfused was 5 (range 0–30) with 288 patients (89%) receiving RBC units during this observational time frame, and 34 patients (11%) who received no RBC units. The median storage age of RBC units was 19 days (average, 21 days, range 2 – 42 days). Platelet transfusions were administered to 274 patients (85%), and 48 patients (15%) received no platelet transfusions (median 2, range 0 – 34).

RBC transfusions associated with aGvHD

239 patients (74%) developed aGvHD with the following maximum grades by day 150 post-transplant: grade I (n = 106, 33%), grade II (n = 70, 22%), grade III (n = 47, 15%), and grade IV (n = 16, 5%). The calculated cumulative incidence curve for the 63 patients with grade III/IV aGvHD patients is shown in Figure 1A, along with the cumulative incidence of death in the first 150 days without grade III/IV aGVHD as the competing risk, (n = 25, including 13 patients with no GvHD, 5 patients with grade I aGvHD, and 7 patients with grade II aGvHD). The median day of onset for all grades of aGvHD was day +32 days, while the median onset days for grades I, II and III/IV aGvHD were days +34, +30, and +28, respectively (Figure 1B). For patients who developed severe grade III/IV GvHD, there was a median of 12 days between the last RBC transfusion prior to the diagnosis of aGvHD and the onset of aGvHD (range 1 to 139 days), and 92 days between the onset of GvHD and death (range 6 to 1821 days). 90% of the patients with maximal grades III–IV aGVHD had onset by day 65. While day 150 was selected as a cutoff time for aGvHD diagnosis in order to include patients with late-onset aGvHD, only 2 cases of grades III–IV acute GVHD (2 out of 63 cases, or 3%) had onset days between day 100 and day 150 (Supplemental Figure 1).

Figure 1. Development of grade III/IV aGvHD and early mortality among allo-HSCT patients.

A) Calculated cumulative incidence curves of competing risks grade III/IV aGvHD (blue) and mortality without grade III/IV aGvHD (red) within 150 days post transplantation. The cumulative incidence function (CIF) subdistribution hazards for grade III/IV aGvHD and mortality at days 30 and 60 post-transplant are shown beneath. B) Increasing frequencies of patients diagnosed with grade I, II, and III/IV aGvHD to day 150 post-transplant. p=0.022, t-test comparing onset times for grade I vs. grade III/IV aGvHD. Gray bar from day −7 to day +27 post-transplant represents the timeframe during which RBC and platelet transfusions were tallied.

Univariate analysis identified several risk factors significantly associated with developing grade III/IV aGvHD (Table 2A) including a lower hematocrit on admission (<25%, p=0.02), HLA mismatched donor (p=0.007), and > median of 5 RBC units transfused (p=0.001). Longer time to neutrophil engraftment (≥ median of 15 days) was inversely associated with grade III/IV aGvHD in univariate analyses (Table 2A). Of note, higher pre-transplant ferritin levels (> median of 1146 ng/ml) were not associated with increased risk of developing grade III/IV aGvHD. Multivariate Cox regression analysis included all variables that were potentially associated with increased grade III/IV aGvHD (p ≤ 0.2 in and HR > 1 in univariate analysis), and revealed that HLA mismatch and more RBC units transfused (but not platelet units) were independently associated with increased risk of grade III/IV aGvHD (p=0.01 and p=0.02, respectively Table 2B). Similar results were obtained when we tested a cutoff time of 100 days post-transplant for diagnosis of aGvHD (rather than 150 days) for univariate and multivariate analyses of factors potentially associated with grade III/IV aGvHD (data not shown).

Table 2A.

Risk factors for grade III/IV aGvHD and mortality using univariable competing-risks Cox regression model

Risk Factor. N = 322 unless otherwise noted	Grade III–IV aGvHD (Events=63)		Early Mortality without Grade III/IV aGvHD (Events=25)
	HR	P	HR	P
Age at BMT (per 10 yr increase)	1.1 (0.9, 1.4)	0.24	0.9 (0.6,1.2)	0.37
Transplant Jan ’07–Dec ’09 vs. Jan’10–Jan’13	0.8 (0.5, 1.3)	0.39	3.6 (1.4, 9.6)	0.01
KPS score (≤ 80 vs. ≥ 90)	0.7 (0.4, 1.3)	0.23	1.0 (0.4, 2.4)	0.98
HCT on Admission (< 25% vs. ≥25%)	2.0 (1.1, 3.5)	0.02	1.2 (0.5, 3.2)	0.66
Disease risk (Low, Intermediate, High)		0.19		0.23
Intermediate vs. Low	0.9 (0.4, 1.7)	0.63	1.2 (0.4, 3.4)	0.76
High vs. Low	1.5 (0.9, 2.6)	0.16	2.1 (0.9, 5.2)	0.10
D/R Gender (Female to Male vs. other combinations)	1.4 (0.6, 2.0)	0.66	0.8 (0.3, 2.1)	0.58
Source (PBSC vs. BM)	0.8 (0.4, 1.6)	0.47	0.4 (0.2,1.0)	0.05
HLA Match (MM vs. Full)		0.025		0.40
10/10 Unrelated donor vs. 10/10 Sibling	1.5 (0.8, 2.7)	0.210	1.7 (0.7, 4.1)	0.20
<10/10 Match vs. 10/10 Sibling	2.4 (1.3, 4.6)	0.007	1.1 (0.3, 3.5)	0.91
ABO Match (MM vs. Match, N = 308)	1.2 (0.7, 1.9)	0.59	1.6 (0.7, 3.6)	0.29
Conditioning Regimen		0.38		0.73
Reduced Intensity vs. Nonmyeloablative	2.1 (0.7, 6.8)	0.20	1.0 (0.2, 4.3)	0.96
Myeloablative vs. Nonmyeloablative	2.3 (0.7, 7.3)	0.17	1.3 (0.3, 5.9)	0.72
Immune suppression CSA vs. FK506	1.1 (0.5, 2.1)	0.90	1.4 (0.5, 4.1)	0.54
Neutrophil engraft day: Median (≥ 15 vs. <15)	0.6 (0.4, 0.9)	0.03	0.8 (0.4, 1.7)	0.50
Platelet engraft day: Median (≥ 18 vs. < 18)	1.0 (0.6, 1.7)	0.89	5.6 (1.7, 18.5)	0.005
Avg age of RBC units (10d increment, N=288)	1.0 (0.6, 1.6)	1.00	0.4 (0.2, 0.8)	0.01
Pre-HSCT serum ferritin: Median (> 1146 vs. ≤ 1146 ng/ml, N=299)	0.6 (0.4, 1.0)	0.07	1.6 (0.7, 3.9)	0.29
RBC units (−7 to 27 days): Median (> 5 vs. ≤ 5)	2.4 (1.4, 4.1)	0.001	2.0 (0.9, 4.5)	0.09
Platelets (−7 to +27): Median (> 2 vs. ≤ 2)	1.6 (1.0, 2.6)	0.08	3.1 (1.3, 7.2)	0.009

Table 2B.

Risk factors for grade III/IV aGvHD using Multivariable Cox regression model (N=322)

Risk Factor	HR	95% CI	P	Reliability
HCT on Admission (< 25% vs. ≥25%)	1.37	0.77, 2.45	0.29	26%
Disease risk (Low, Intermediate, High)			0.38	16%
Intermediate vs Low	0.79	0.40, 1.54	0.48
High vs Low	1.28	0.71, 2.28	0.41
HLA Match			0.04	67%
10/10 Unrelated donor vs. 10/10 Sibling	1.33	0.70, 2.50	0.38
<10/10 Match vs. 10/10 Sibling	2.37	1.21, 4.67	0.01
RBC units (d−7 to +27): Median (> 5 vs. ≤5)	2.06	1.14, 3.73	0.02	78%
Platelet units (d−7 to +27): Median (> 2 vs. ≤2)	0.88	0.49, 1.54	0.63	5%

In performing univariate analyses for the competing risk of early mortality without grade III/IV aGvHD, several factors were significantly associated with death in this group including earlier date of transplantation (2007 – 2009 vs. 2010 – 2012; p=0.01), platelet engraftment time equal to or greater than the median of 18 days (p=0.005), and transfusion of greater than the median number of 2 platelet units (p=0.009). We also found that mobilized PBSC grafts and increased average storage time of transfused RBC units were both inversely associated with the early mortality competing risk (HR=0.4, p=0.05, and HR=0.4, p=0.01, respectively, Table 2A). Importantly, neither HLA mismatch nor more transfused RBC units were associated with early mortality in the absence of grade III/IV aGvHD. A multivariable model for early mortality without grade III/IV aGvHD showed earlier date of transplantation (2007 – 2009) and longer time to platelet engraftment (≥ 18 days) both independently associated with death (p = 0.005 and p = 0.017, respectively, data not shown).

RBC transfusions associated with overall survival

All surviving patients were followed for at least 2 years post-transplant (range 2.0 to 8.5 years, median 4.7 years). Cumulative survival curves stratified by maximal grade aGvHD and by numbers of RBC and platelet units transfused between days −7 to day +27 (and before diagnosis of aGvHD) are shown in Figure 2. Patients with severe aGvHD or those that received > median of 5 RBC units or > median of 2 platelet transfusions had worse survival (p<0.0003, p=0.00014 and p=0.002, respectively).

Figure 2. Overall survival was associated with aGvHD grade and numbers of RBC units and platelet units transfused from day −7 to day +27 in Kaplan-Meier analyses.

A) Kaplan-Meier analyses of cumulative survival comparing patients with grade 0/I (blue), II (yellow) and III/IV (maroon). B) Cumulative survival of patients transfused with > 5 RBC units (yellow) vs. ≤ 5 RBC units (blue). C) Cumulative survival of patients transfused with > 2 platelet units (yellow) vs. ≤ 2 platelet units (blue). Plots are truncated at 7 years post-transplant.

Several factors, in univariate analysis, were significantly associated with worse overall survival (p<0.05), including hematocrit on admission, disease risk, HLA mismatch, aGvHD score, and numbers of both RBC and platelet transfusions (Table 3A). There was a trend toward inferior overall survival among pts with increased pre-transplant ferritin levels, but this was not statistically significant (p = 0.13). Increasing average age of transfused RBC was inversely associated with overall mortality. In multivariate analysis (excluding aGvHD grade), high-risk disease status and the number of transfused RBC units (but not the number of platelet transfusions) were found to be independently associated with mortality (HR 1.7 (1.2, 2.4), p=0.005, and HR 1.4 (1.0, 2.0), p=0.054, respectively, Table 3B).

Table 3.

Analysis of factors potentially associated with all-cause long-term mortality

3A. Univariate Cox regression analysis
Risk Factor (N=322 unless noted)	HR	95% CI	P
Age at BMT (per 10 year increase)	1.1	1.0, 1.2	0.21
Transplant Jan ’07–Dec ’09 vs. Jan’10–Jan’13	1.2	0.9, 1.6	0.24
KPS score (≤ 80 vs. ≥ 90)	1.2	0.8, 1.6	0.40
HCT on Admission (<25 vs. ≥25%)	1.5	1.0, 2.2	0.028
Disease risk (Low, Intermediate, High.)			0.007
Intermediate vs. Low	1.2	0.8, 1.9	0.275
High vs. Low	1.9	1.4, 2.7	0.0002
Female to Male vs. other combinations	0.8	0.6, 1.2	0.403
HLA Match			0.018
10/10 Unrelated donor vs. 10/10 Sibling	1.6	1.0, 2.7	0.06
<10/10 Match vs. 10/10 Sibling	2.2	1.3, 3.9	0.005
Conditioning Regimen			0.585
Reduced Intensity vs. Nonmyeloablative	1.1	0.6, 1.8	0.845
Myeloablative vs. Nonmyeloablative	0.9	0.5, 1.6	0.692
Immunosuppression CSA vs. FK506	1.0	0.7, 1.6	0.820
AGvHD score			<0.0001
aGvHD, grade 2 vs. 0–1	1.2	0.8, 1.7	0.463
aGvHD, grades 3–4 vs. 0–1	2.6	1.8, 3.8	<0.0001
Average age of RBC units (10d incr, N=288)	0.7	0.5, 0.9	0.0096
Pre-HSCT ferritin: (> 1146 vs. ≤1146 ng/ml, N=299)	1.3	0.9, 1.8	0.13
Transfused RBC units (> 5 vs. ≤ 5)	1.8	1.3, 2.4	0.0002
Transfused platelet units (> 2 vs. ≤ 2)	1.6	1.2, 2.2	0.0023

3B. Multivariate Cox regression analysis
Risk Factor (N=322)	HR	95% CI	P	Reliability
HCT on Admission (<25 vs ≥25%)	1.2	0.8, 1.8	0.391	16%
Disease risk (Low, Intermediate, High)			0.018	79%
Intermediate vs. Low	1.3	0.8, 1.9	0.260
High vs. Low	1.7	1.2, 2.4	0.005
HLA Match			0.413	25%
10/10 Unrelated donor vs. 10/10 Sibling	1.3	0.9, 1.8	0.190
<10/10 Match vs. 10/10 Sibling	1.2	0.8, 1.9	0.392
Transfused RBC units (> 5 vs. ≤ 5)	1.4	1.0, 2.0	0.054	71%
Transfused platelet units (> 2 vs. ≤ 2)	1.2	0.9, 1.7	0.250	34%

RBC transfusion trends over time

We examined the cumulative incidence of first RBC transfusion during the d−7 to d+27 observational window in transplant patients who were either ABO matched or ABO mismatched with their donors. As shown in Figure 3A, 40% of patients received RBC transfusions during the week prior to transplant, and by 2 weeks post-transplant the fraction of patients who had received one or more RBC transfusions had risen to nearly 90%, with no significant differences in cumulative incidence between these two groups. Quantifying the numbers of patients receiving RBC transfusions each week, the proportion of patients with ABO mismatched donors receiving RBC transfusions was significantly greater than for patients with ABO matched donors only during the 3^rd week post-transplant, p=0.0338, Figure 3B. Nonmyeloablative conditioning regimens were associated with fewer patients receiving RBC transfusions during weeks 1, 2 and 3 post-transplant compared with myeloablative conditioning (Figure 3C). Among patients who developed grade III/IV aGvHD (Figure 3D), a significantly larger fraction of patients received RBC units each week (except week 2), compared to patients who had grade 0/I or grade II aGvHD.

Figure 3. Weekly frequencies of patients receiving RBC transfusions patients differed based upon patient characteristics and aGvHD severity.

A) Cumulative incidence of patients who received RBC transfusions as grouped by ABO match between transplant recipient and donor. B–D) Percentages of patients receiving RBC transfusions each week (−1 week to +4 weeks post-transplant) as grouped by: B) ABO match between transplant recipient and donor, C) Conditioning regimen intensity (Nonmyeloablative, NMA; Reduced Intensity, RIC; Myeloablative, MAC), D) aGvHD severity. * p<0.05, ** p<0.01, *** p<0.001.

Transfusions and Transplant Outcomes

In order to analyze possible associations of RBC and platelet transfusions with different causes of post-transplant mortality, we grouped patients into 4 categories based upon the number of total RBC and platelet units received from day −7 to +27 post-transplant, excluding transfusions received after a diagnosis of aGvHD. Four groups were defined based upon fewer receipt of ≤ median or > median units transfused, resulting in a group of 120 patients who had received < median number of both platelet and RBC transfusions (A), 45 patients with ≤ median RBC units transfused but > median platelet units (B), 43 patients with > median RBC but ≤ median platelet units (C), and 114 patients that received > median number of both types of transfusions (D). Mortality was lowest for the groups with fewer RBC transfusions (A and B, 43% and 44%), and greatest for the group with more RBC and platelet transfusions (D, 66%, p=0.006 Fisher’s exact test compared with group A; Figure 4). Comparisons of the rates of deaths attributed to relapse, GvHD, organ failure, and secondary malignancy in groups B, C and D were not statistically significant from cause-of-death rates in group A. Only the rate of infection-related deaths in group B was significantly higher than that in group A (25% vs. 6%, p=0.03). It is noteworthy that patients who died of organ failure were not over-represented among groups that received more transfusions, indicating that the numbers of RBC and platelet units transfusions in the peri-transplant period were not simply surrogates for greater co-morbidities among sicker patients.

Figure 4. Causes of death in 4 patient groups based on RBC and platelet transfusions. Groups that received more RBC transfusions had higher frequencies of GvHD-related mortality.

Patients were grouped based upon the number of total RBC and platelet units received from day −7 to +27 post-transplant, excluding those received after a diagnosis of aGvHD (> 5 vs. ≤ 5 RBC units, and > 2 vs. ≤ 2 platelet units). Pie chart sizes are proportional to the % mortality in each group.

Discussion

This study is, to our knowledge, the first published demonstration of a significant association between the number of RBC units transfused in the pre- and post-transplant period with the incidence of severe aGvHD and overall survival in patients undergoing allo-HSCT. Using both univariable and multivariable Cox regression analyses, we found significant and robust associations between the number of RBC units transfused (but not platelet transfusions) and development of grade III/IV aGVHD as well as with overall survival. Use of bootstrapping in the multivariable models allowed for determinations of reliability for these associations (Tables 2B and 3B). A recent study in patients with severe aplastic anemia (SAA) found that the number of pre-transplant RBC transfusions was associated with aGvHD and overall mortality after allo-HSCT, which the authors suggested was attributable to pre-transplant iron overload in the high-transfusion group.⁴² In contrast, we excluded patients with non-malignant hematological conditions from our study. We analyzed pre-transplant serum ferritin as a surrogate for transfusion history, since accurate record of transfusions prior to admission for the conditioning regimen was not available for a patient population from a catchment area across the South-Eastern states. We did not find an association between increased pretransplant ferritin levels and grade III/IV GvHD. As in other studies^{20, 21, 22, 23}, we found a trend toward inferior overall survival among patients with higher pre-transplant ferritin levels, although this was not statistically significant in our patient population. Differences seen in association of pre-transplant ferritin levels with GvHD among different clinical studies may be related to differences in patient disease groups and conditioning regimens.²² Our results suggest that the increased risk of aGvHD among patients who received more RBC transfusions was not accounted for by increased serum ferritin, although this is not conclusive due to a lack of lifetime transfusion records for these patients, which is a limitation of the study.

This retrospective study has a number of additional limitations. First, we restricted our observational time frame for RBC transfusion data to a period in which allo-HSCT patients were being exclusively managed at the transplant center: from day −7 pre-transplant to day +27 post-transplant, and received all transfused units from a single source. Second, in order to avoid the effects of GvHD-associated gastro-intestinal blood loss on transfusion, we only considered transfusions performed prior to the initial diagnosis of aGvHD, censoring transfusions performed after a diagnosis of aGvHD from the analysis. Finally, aGvHD may be associated with the underlying anemia, such that RBC transfusions are a surrogate measure of anemia in the peri-transplant period. A recent study in immuno-incompetent pre-term infants demonstrated that severe anemia, but not RBC transfusion, was independently associated with a 6-fold increased risk of necrotizing enterocolitis, an inflammatory and necrotic condition of the intestines.⁴³

No consistent association of ABO mismatch between allo-HSCT donor and recipient with increased aGvHD has been demonstrated. ABO minor mismatch was associated with increased non-relapse mortality in one study, but there was only a non-significant trend of increased risk for grades II–IV aGvHD.¹⁴ The same study identified a significant association of ABO minor mismatch with grades II–IV aGvHD among the subset of patients who received bone marrow grafts.¹⁴ ABO incompatibility was not associated with grades II–IV aGvHD in nonmyeloablative allo-HSCT recipients in another study.¹⁵ The mechanism by which transfused RBC might increase the risk of severe acute GvHD is unknown. Although some RBCs may express trace amounts of MHC class I molecules in the form of Bennett - Goodspeed antigen,⁴⁴ as a general principle, platelets and not RBCs express MHC I.^{44, 45} Both RBCs and platelets also express minor histocompatibility antigens, with over 340 known RBC blood group antigens and 33 known platelet alloantigens.^{46, 47} Minor antigens on donor RBCs enter both class I (cross presentation) and class II pathways (indirect presentation),^{48, 49} and thus it is possible that allo-antigen present on transfused RBCs may be potently presented by recipient APCs leading to donor T cell activation. Using mouse BMT models, we have demonstrated that interactions between activated donor antigen-presenting cells and donor T cells occurring in the first week after allo-HSCT regulate the activation and proliferation of donor T cells, including aGvHD activity.³⁰ An alternative explanation for the association of RBC transfusion with aGvHD is that innate immune activation may result from transfused RBCs that increase costimulatory molecules on recipient APCs,⁴⁸ leading to increased activation of donor T cells and increased aGVHD in response to host allo-antigens presented by recipient APCs. Further studies are needed to distinguish these possible mechanisms. The distinct association of transfused RBC, but not platelets, with aGvHD is noteworthy. Transfusion of platelets is also known to induce immune activation in some recipients;³³ however the fine details of this immune activation are distinct from that found for RBC transfusion.

Another potential mechanism that could raise the risk of severe aGvHD after RBC transfusion is increased inflammation in the recipient. Pre-clinical animal models have clearly demonstrated that transfusion of stored RBCs results in systemic inflammation and immune activation,^{31, 32} while controlled trials in healthy volunteers did not observe such an effect.^{50, 51} Another area of great interest in the field of red cell transfusion is whether increased storage time of the red cell unit is associated with a higher incidence of post-transfusion complications.⁵² Neuman et al. reported that older RBC units have decreased NO-mediated vasodilation effects on the vasculature of anemic patients.⁵³ The authors discussed the possibility that the products synthesized in RBC units during preservation may be deleterious to patients undergoing allo-HSCT. In the current retrospective study, we observed the opposite trend, with increased average storage time of the RBC units associated with better survival, and no association with grade III/IV aGvHD. However, the average storage age for RBC units may include very fresh RBC units as well as older units, thus this measure may not provide a comprehensive representation of RBC unit storage times. We also saw no association of the maximum storage time for a RBC unit for each patient with survival or severe aGvHD (data not shown). Other possible factors are the gender and age of persons donating RBC units, both of which were shown to impact survival of transfusion recipients.⁵⁴ We have initiated a prospective clinical study that includes storage time of transfused RBC units, gender and ages of RBC donors, as well as measures of inflammation and immune cell activation after RBC transfusion in allo-HSCT patients.

In conclusion, we have shown that allo-HSCT patients who received more RBC transfusions between day −7 and day +27 after transplantation had a higher risk of developing severe aGvHD and had worse overall survival. A model was constructed to estimate the risk of developing severe aGvHD according to the cumulative number of transfused RBC units, assuming 1) the baseline risk of grade III/IV aGvHD is 15%, 2) the risk of early death without aGvHD is 5%, and 3) the incidence of developing grade III/IV aGvHD is increased by 2% for 7 days following each RBC transfusion. These assumptions correspond to the observed rate of grade III/IV aGvHD, and death without grade III/IV aGvHD (Figure 1). The observed incidence of grade III/IV aGvHD increased for each transfused RBC unit and corresponds to the predicted risks from this model (Supplemental Figure 2). After analysis of our retrospective institutional data presented here regarding association of RBC transfusions with increased risk of severe aGvHD, the transfusion trigger of <27% that was in place for allo-HSCT patients in this study (transplanted January 2017 – January 2013), has now been substantially lowered to <21%. Prospective studies, such as our recently-opened study at Emory, are warranted to determine causality between RBC transfusions in allo-HSCT patients and increased risk of severe aGvHD, providing further rationale for development of improved transfusion practices for these patients. Such approaches might reduce the number of transfusions by using symptom-driven criteria and limited transfusion thresholds, rather than liberal HCT-driven criteria for RBC transfusion, may favorably reduce GvHD incidence and severity.

Supplementary Material

supp

S1. Timelines for grade III/IV aGvHD patients.

For each patient who developed grade III/IV aGvHD, the timeline shows each RBC transfusion from day −7 to day 27 post-transplant, administered prior to diagnosis of aGvHD (blue diamond) and deaths (red X) up to day 150 post-transplant.

S2. The rate of Grade III/IV aGvHD among allo-HSCT patients correlates with a model based upon the numbers of RBC units transfused.

A mathematical model was developed to predict the rate of grade III/IV aGvHD among allo-HSCT patients as a function of the number of RBC units transfused. The model incorporates assumptions based on data from the current study as described in the text. Plots showing the predicted frequencies of grade III/IV aGvHD (black circles) and actual the frequencies (red squares) demonstrate the fit of the data with the estimated values.

Highlights.

• RBC and platelet transfusions were counted for 322 allo-HSCT patients day −7 to +27

• More RBC transfusions were associated with increased risk of severe aGvHD

• More RBC transfusions were associated with inferior overall survival

• Platelet transfusions were not associated with aGvHD or worse overall survival

• Strategies to reduce routine RBC transfusion may reduce GvHD incidence and severity