2.1 Definition of platelet transfusion refractoriness

PTR is the repeated failure to achieve the desired level of blood platelets in a patient following platelet transfusions. According to the published definitions, PTR in the present study was defined as a posttransfusion platelet increment (PI) that is less than 5 × 10⁹/L after receiving at least two successive daily platelet transfusions.^{2, 13, 14}

2.2 Patients enrollment

De novo AML patients diagnosed in our department were successively enrolled in this study. Patients with secondary or myelodysplastic syndromes/myeloproliferative neoplasms (MDS/ MPN)-transformed AML were excluded. Clinical and laboratory features of eligible patients were retrospectively reviewed. Risk assessment and risk stratification-guided treatment were administrated as previously ^15-17 and described in the Supplementary Methods.

This study was approved by the Institutional Review Board of Tongji Hospital, Tongji Medical College, Huazhong University of Science and Technology. Informed consent was obtained from each individual in accordance with the principles expressed in the Declaration of Helsinki.

2.3 Detection of anti-HLA class I antibodies

Whole blood samples were collected from patients with procoagulation tubes when platelet counts were lower than 20 × 10⁹/L after induction chemotherapy. Then serum was separated by centrifugation at 2500 revolutions per minute and frozen at −80℃. The storage time was less than 3 months.

Luminex single antigen beads (SAB) was used in a high-throughput HLA antibody screening assay (LABScreen Single Antigen HLA Class I, One Lambda) to evaluate the existence of antibodies targeting any donor-specific HLA class I antigens in the serum, which were the main immunological cause. ^{18, 19} This assay was conducted according to the manufacturer's instructions.^20-22 Briefly, 20 μL serum was incubated with 5μL microbeads in a 96-well V-bottomed plate in the dark at 25℃ for 30 minutes. After three times of washing, R-Phycoerythrin-conjugated goat antihuman IgG antibody was added to the plate for another 30 minutes incubation. The plate was washed for two more times. The single antigen detection beads can identify the following HLA-I specificities: HLA-A1, 3, 11, 23-24, 29-31, 33-34, 36, 66, 74, and 80; HLA-B7, 8, 13, 18, 27, 35, 37-39, 41, 42, 44-65, 67, 71-73, 75-78, 81, and 82; and HLA-Cw1, 2, 4-6, 8-10, 12, and 14-18.²³ Mean fluorescence intensities (MFI) were acquired using a Luminex analyzer. The results were considered to be positive when MFI was above 5000 and strongly positive when MFI was higher than 10 000.^{24, 25}

2.4 Statistical analysis

The patients’ characteristics were described by numbers and frequencies for categorical variables, and by medians and interquartile ranges (IQR) for continuous variables. Differences were analyzed using chi-square test (or Fisher's exact test in case of small expected numbers) for categorical variables and Student's t test for continuous variables. A logistic regression model for multivariate analysis was used to evaluate the risk factors associated with PTR in univariate analysis. P values less than .050 (two tailed) were considered to be statistically significant. All statistical analyses were performed using Statistical Package for the Social Sciences (SPSS) version 22.0 (IBM) and GraphPad Prism 6.

3.1 Occurrence of PTR in patients with de novo AML

From June 2012 through June 2018, a total of 560 patients with de novo AML were screened for eligibility. The median age was 45 years (range, 12-86) with a male-to-female ratio of nearly 4:3. After induction chemotherapy, platelet transfusions were given to patients with active hemorrhage or patients with platelet counts less than 20.0 × 10⁹/L. PTR occurred in 66 (11.8%) patients when they received platelet transfusion from random donors. The median time interval between the conduction of chemotherapy and the onset of PTR was 9.0 days (IQR: 6.0-17.0 days). The median posttransfusion increments with platelets derived from random donors in PTR patients were 0.5 × 10⁹/L (IQR: −3.0-3.5 × 10⁹/L). Among the 32 PTR patients who were transfused with HLA-matched platelets, 10 (31.3%) patients showed responses with a median posttransfusion PI of 26.0 × 10⁹/L (IQR: 17.8-36.3 × 10⁹/L). And the posttransfusion PI of HLA-matched platelets was significantly higher compared to nonmatched platelet from random donors (9.6 × 10⁹/L vs −0.7 × 10⁹/L, P < .001, Figure S1). Other management strategies, including intravenous gamma-globulin and glucocorticoid, were not effective enough to improve platelet response. Severe bleeding events during induction therapy occurred in 5 PTR patients (7.6%), involving gastrointestinal track (n = 2), cerebrum (n = 1), lung (n = 1), and multiple organs (n = 1).

The median age for patients with PTR was 46 years (range, 15-70) with a female predominance (male-to-female ratio, 2:3, P = .007). Patients with splenomegaly were more likely to develop PTR (27.3% vs 14.4%, P = .007). However, differences in history of transfusion, pregnancy, or autoimmune diseases, the attack of fever or infection, and the use of antibiotics or liposomal amphotericin B were consistent in patients with or without PTR (Table 1).

3.2 Clinical features of PTR in patients with de novo AML

The clinical features of PTR patients were summarized in Table 2. At the disease onset, the white blood cell counts, platelet counts, and the percentages of bone marrow blasts were comparable between patients with and without PTR (Table 2). However, in univariate analysis, PTR occurred predominantly in patients with M4 French-American-British (FAB) subtype (P = .001), rearrangements of RUNX1-RUNX1T1 (P = .026) or CBFB-MYH11 (P < .001), and c-Kit mutation (P = .047), but not inclined to manifest in patients with NPM1, CEBPA, and FLT3-ITD mutations (Table 2). Notably, rearrangements of RUNX1-RUNX1T1 (OR = 5.025, [1.762-14.329], P = .003) or CBFB-MYH11 (OR = 20.285, [4.121-99.848], P < .001), which were commonly referred to as core binding factor AML (CBF-AML), were independently associated with the occurrence of PTR in multivariate analysis (Table 3).

3.3 PTR frequently manifested in patients with CBF-AML

PTR manifested in 30.4% (28/92) of the patients who had CBF-AML, in contrast to 8.1% (38/468) of the others, indicating a strong correlation between PTR and CBF-AML (P < .001, Figure 1A, Figure S2A,B). Thus, when received from random donors, the median posttransfusion PI was more significantly reduced in patients with CBF-AML than in other patients (P < .001, Figure 1B). Among patients with CBF-AML, the incidence of PTR (33.3% vs 28.9%, P = 1.000, Figure 1C) and the posttransfusion PI (P = .847, Figure 1D) was comparable in patients with or without c-Kit mutations. Intriguingly, PTR was dramatically developed in patients who achieved 3-log or more MRD reduction before the second consolidation chemotherapy than in patients who did not achieved it (45.2% vs 17.5%, P = .007, Figure 1E). Accordingly, when received from random donors, the median posttransfusion PI was markedly decreased in patients who had better MRD reduction (P < .001, Figure 1F).

3.4 Outcomes

In CBF-AML patients, there were 6/92 (6.5%) who died in the first month after the commence of induction therapy, which was defined as an early death (Table S1). Among these patients, 5/28 (17.9%) early death occurred in PTR group compared to 1/64 (1.6%) in non-PTR group (P = .009). And early death secondary to bleeding was obviously higher in patients who had PTR (10.7% vs 0). In CBF-AML patients, there was no statistical difference in overall survival (OS) between PTR and non-PTR group (P = .745, Table S1, Figure S3A). When early death was excluded in both groups, the OS in PTR group showed a tendency of longer survival, but the difference was still not statistically significant (6-month, 1-year, 2-year, and 5-year OS rates: 100, 94.1, 72.1, and 46.4% in PTR group vs 93.2, 75.6, 65.9, and 40.0% in non-PTR group, P = .337, Table S1, Figure S3B).

In all PTR patients, 7/66 (10.6%) have an early death, including 5/28 (17.9%) in CBF-AML group and 2/38 (5.3%) in non-CBF-AML group (P = .125, Table S2). And early death secondary to bleeding was comparable between CBF and non-CBF-AML group (10.7% vs 5.3%, P = .643). No difference in OS was observed between PTR patients with CBF and non-CBF-AML (Figure S3C, P = .294), except patients with early death were excluded (Figure S3D, P = .040).

3.5 High-throughput HLA antibody screening

Using Luminex single antigen beads, we examined anti-HLA-I antibodies and screened their HLA specificities in the serum from 133 de novo AML patients. HLA-I antibodies were detected in 9.0% (12/133) among these patients, including in 62.5% (10/16) of patients with PTR and in 1.7% (2/117) of non-PTR patients (P < .001, Figure 2A). Among the 12 patients who had anti-HLA-I antibodies, 10 (83.3%) developed PTR. Furthermore, HLA-I antibodies were detected in 25.0% (5/20) of patients with CBF-AML, while only in 6.2% (7/113) of the non-CBF-AML patients (P = .018, Figure 2B).

3.6 Epitope HLA-B closely related to PTR

A total of 71 HLA serotypes were identified (Figure 2C), among which 14 were specific for HLA-A epitopes, 43 for HLA-B epitopes, and 14 for HLA-C epitopes. HLA-B77 was the most frequently (50.0%, 6/12) identified serotype. HLA-B35, B45, B47, B52, B56, B62, B75, B76, B78, and B82 were commonly detected as well (41.7%, 5/12) in patients with PTR. However, two HLA-I antibodies, targeting HLA-B45 and HLA-C epitopes, were detected in 2 non-PTR patients separately (Figure 2C). HLA-B45, the solely detected serotype in one non-PTR patient, was identified with marginal MFI intensity (6684.58). HLA-C epitope, identified in the other non-PTR patient, had been previously considered to be expressed on the surface of platelet but not contributing to PTR.²⁶