How I treat patients who are refractory to platelet transfusions Free

This review takes a critical look at the dilemmas of PR. Our cases illustrate practical challenges, including when to test for different causes and transfusion management. The cases also highlight the intensive nature of investigation and logistical support. However, a critical question remains: how do these practices affect clinical outcomes?

PR describes a poor response to platelet transfusion, and conventional thinking has divided etiologies into 2 broad categories: immune and nonimmune.^1-3 The presence of PR increases the complexity of managing patients with thrombocytopenia requiring transfusion.^4,5 Because the key measure defining PR is the difference between an expected posttransfusion platelet count and pretransfusion platelet count, algorithms have been developed to make calculations more “accurate” by incorporating recipient blood volume and the dose of platelets given (Box 1). The definition of PR established by the Trial to Reduce Alloimmunization to Platelets (TRAP) study is 2 consecutive corrected count increments (CCIs) at 1 hour after platelet transfusion measuring <5 × 10⁹/L.⁶ In clinical practice, due to the absence of available platelet dose information, unadjusted platelet increments (PIs) are often used. The thresholds clinicians use for defining their “poor” increments vary, with PR ranges of ≤5 × 10⁹/L, 7.5 × 10⁹/L, 10 × 10⁹/L, or 20 × 10⁹/L described in the literature. A common pragmatic approach is to consider the response suboptimal if the platelet count has not changed after 20 to 24 hours. However, it should be recognized that these algorithms do not incorporate clinical factors such as bleeding.

PR can be secondary to both product and patient factors. The influence of product factors outside of the type of platelets (eg, pathogen reduced) on PR is often not appreciated. The impact of donor and donation characteristics on outcomes remains poorly understood, despite attempts to evaluate this in vivo.⁷ Patient factors are also highly relevant, given the patients needing platelet transfusions are diverse with multiple comorbidities. As part of the Recipient and Donor Epidemiology Study III databases, Gottschall et al⁸ reported on a large, retrospective multicentered analysis of platelet transfusions, with half of the transfusions given in the setting of hematologic malignancies. Overall, increments under 20 × 10⁹/L were very common but demonstrated variability associated with pretransfusion platelet counts.

Poor PIs are observed in many populations. In adult critical care admissions, the median PI was reported as 2 (range, –1 to 8).⁹ In this study, clinical outcomes including 28-day survival, duration of mechanical ventilation, and length of stay in the intensive care unit did not differ between good or poor “responders” to platelet transfusion. In an international study of critically ill children receiving at least 1 platelet transfusion, PI data were highly variable. Similar findings were described in sick neonates with thrombocytopenia.^10,11 

Although rates of reported PR prevalence vary (4.8%-49.6%),^12-14 the burden of PR can be significant. It is often linked with illness severity, increased risk of bleeding, longer stay in hospital, and decreased survival, along with higher hospital costs, including increased demand for transfusions.^14,15 Nonimmune causes of refractoriness secondary to factors such as infection and splenomegaly are also relevant in many patients with PR.^1,16,17 

Our first patient is a 43-year-old multiparous woman with acute myeloid leukemia and no prior transfusion exposure. During consolidation therapy, she developed headaches with a subarachnoid hemorrhage on computed tomography scan. Repeated transfusions with pooled platelets were suboptimal, with minimal (≤2 × 10⁹/L) increments in platelet counts. Bilateral retinal and vitreous hemorrhaging developed. Tranexamic acid was started alongside intensive platelet transfusions. Platelet refractoriness (PR) investigations were initiated. HLA testing demonstrated multiple class I HLA antibodies (Figure 1), but due to limited suitable ethnic donors in the apheresis program, the first HLA-selected platelet was not available for several days. By this time, a total of 34 “splits” of 17 platelet pools were used during multiple episodic transfusions, including an eventual continuous infusion. Using our solid organ transplant desensitization protocol, an attempt to modify the alloimmunization profile to increase donors was undertaken with plasmapheresis, IV immune globulin, rituximab, and bortezomib. Close collaboration between the transfusion service and blood supplier ensured that sufficient donors were available to ensure HLA matched platelet support to eventual hematopoietic stem cell transplantation.

Only a minority of cases of PR have demonstrated immune etiologies. The common question is how should one identify patients with higher pretest probability for testing.¹⁸ Immune factors, if identified, may be single antibodies, or a combination of antibodies, targeting class I HLA, human platelet antigen (HPA), ABH antigens, drugs directed at platelet antigens, or autoantibodies against platelet glycoproteins (eg, GP IV/CD36). However, the most clinically relevant and studied immune etiology is HLA alloimmunization, reflecting ∼80% to 90% of immune PR cases,¹⁹ which is the focus of this article.

Some reports suggest that differential increments at 1 hour²⁰ vs 24 hours²¹ can be helpful to distinguish broad categories of immune PR from nonimmune PR, but in practice, specific testing for immune causes is required in patients with persisting evidence of PR and no evidence of nonimmune factors. Poor count increments in a woman with a history of previous pregnancy, when the use of leukocyte-reduced blood components is standard, should prompt investigations for immune refractoriness (see “Case 1”). Therefore, the diagnosis of nonimmune PR is one of exclusion.

There remains uncertainty regarding why certain patients become alloimmunized, which alloimmunized patients have clinical refractoriness, and why there is such heterogeneity in patient characteristics, antibody profiles, and longitudinal patterns.^15,22 Some patients develop single alloantibodies, whereas others have challenging broad and complex alloantibody specificities. The timing of antibody formation for those who become alloimmunized is typically within 4 to 8 weeks of starting transfusion support with random platelets but may be earlier if patients were primed through pregnancy¹⁸ or preceding transfusion events, especially if with non–leukocyte-reduced components. The data, however, do not support a simple dose-response relationship to the number of units transfused.²¹ 

Investigation of immune causes of refractoriness is most commonly performed in patients with underlying hematologic malignancies, because they are the most heavily platelet-transfused cohort.^23,24 In the TRAP study, some patients with leukemia who developed alloantibodies during induction chemotherapy lost these antibodies, with little relationship to the profile of platelet transfusion history.²² The same variable disappearance of HLA antibodies and resolution of PR has been reported elsewhere in patients with hematologic malignancies.²⁵ Because this patient group has immunosuppression due to their underlying disease and treatment, the transience and persistence of the antibodies seen in them may not be demonstrable in other patient groups. The presence of alloantibodies to HLAs does not always correlate with clinically relevant PR. HLA alloimmunization has been recorded in healthy blood donors, as shown in the Leukocyte Antibody Prevalence Study, in which HLA antibodies were detected in nearly a quarter of previously pregnant female donors.²⁶ 

Studies, including TRAP, have demonstrated the effectiveness of UV irradiation and leukocyte reduction to prevent alloimmunization,^6,27 and these studies reiterate the role of patient and product factors affecting PR etiologies. They confirm a higher overall incidence of HLA alloimmunization than PR rates.²¹ In conjunction with leukocyte reduction, ABO matching for platelet transfusions has been reported to further reduce alloimmunization risk, but this needs corroboration in further studies.²⁸ 

There are currently no guidelines for patients with documented alloimmunization identified for other reasons (eg, transplant) who have not had platelet transfusion exposures to assess clinical significance, but closer monitoring for PR may be warranted if transfusion is subsequently required (see “Case 2”).

The mechanistic pathways for HLA alloimmunization, whether direct or indirect, and the role of donor or recipient antigen-presenting cells are subjects of considerable research interest.²⁹ For discussion on these mechanisms of immune PR, readers are directed to other reviews.^30-32 

As part of a solid organ transplant workup, an adolescent boy was identified as being sensitized for HLA class I antibodies secondary to transfusions administered during prior surgical procedures (Figure 2). In S.N.’s facility, the transfusion service is not notified about transplant patients with HLA alloimmunization until the time of transplantation, at which time, they would code non–HLA-selected platelets with the need to monitor for efficacy and posttransfusion increments (PIs). However, this patient required significant non–transplant-related transfusion support, including red cells and platelets. Satisfactory increments were seen for all nonoperative platelet transfusions. At their 2-month follow-up, there was no progression in the HLA antibody profile.

Screening and testing for the presence of HLA antibodies may use techniques such as lymphocytotoxicity (LCT), enzyme-linked immunosorbent assay, flow cytometric immunofluorescence tests, or the newer multiplex flow cytometric bead (Luminex) assays. The historical cell-based complement-dependent cytotoxicity (CDC), also known as LCT, methods have been largely replaced by flow cytometry or solid-phase techniques such as enzyme-linked immunosorbent assays or Luminex assays.

CDC methods are technically challenging, complicated by low sensitivity and cross-reactivity with non-HLA antibodies when used in PR investigations. Flow cytometric methods using a panel of lymphocytes have increased sensitivity compared to CDC methods, but there are major issues with interinstitution variability as well as interaction with non-HLA antibodies and therapeutics such as IV immune globulin. Luminex technology uses HLAs (purified or recombinant) coupled to polystyrene beads, which detect antibody binding from the patient plasma/serum after fluorescently labeled anti-human immunoglobulin G is added, and the coated beads are exposed to laser irradiation for measurement of mean fluorescence intensity (MFI). This technology, as a measure of antibody avidity, can be used as a screen with beads coated with multiple antigenic specificities or as a single-antigen bead antibody-identification platform.³³ 

In transplant populations, positive thresholds are typically reflected by MFIs >1000 in the screen and >3000 for single-antigen bead testing, but appropriate thresholds for PR are controversial. There are arguments supporting the same thresholds to ease logistics if the platelet assays are being performed by the same technologists, but different laboratories may use higher MFI thresholds or perform dilutions for their refractory investigations to correlate with historical CDC testing or platelet crossmatching. Blandin et al³⁴ demonstrated no significant difference in median 1-hour CCIs between matched and nonmatched platelet components in patients when MFI cutoffs were 1000. Pedini³⁵ demonstrated a 48% effectiveness rate for platelets that expressed antigens that had antibody detected at MFIs of 1000, but this dropped to 14% at MFIs of 2000 and 5% at 5000. Another study reported a PR association of 90% when the MFI was >5000³⁶ or when there was a cumulative MFI of ≥10 000.³⁷ 

These newer semiquantitative methodologies have increased sensitivity for HLA alloantibodies, making them ideal for stem cell and solid organ transplant programs. However, further validation of these methods and their clinical predictive applicability as part of platelet investigations are awaited. Moreover, this validation is frequently limited by the lack of posttransfusion platelet counts for patients provided with antigen-negative vs antigen-positive platelet transfusion for specificities identified at low fluorescence thresholds. There have also been no studies to evaluate whether the historical antibody waning effect is still present when more sensitive semiquantitative assays are used or whether this phenomenon was secondary to lower sensitivity of the LCT assays.

The main treatment option in clinical practice for thrombocytopenia is transfusion of platelets. Although thrombocytopenia is well established to be associated with adverse outcomes,³⁸ the extent to which thrombocytopenia is a clinically relevant modifiable risk factor by platelet transfusions is unclear, as discussed in “Clinical impact.” There is increasing data to suggest that platelet transfusions may cause harm in addition to those risks generally seen for all blood components (eg, circulatory overload).³⁹ These additional risks include adverse outcomes of mortality and bleeding, as demonstrated in randomized trials in neonates and in adults taking antiplatelet drugs,^10,40 and are hypothesized to relate to the potential actions of platelets as an immune cell.^29,41-43 

There is controversy as to whether it is better to continue support of patients with PR with platelets while awaiting the results of investigations to confirm or exclude immune etiologies. Some clinicians cease prophylactic platelet transfusion due to the emerging evidence of harm with platelets. However, in patients with thrombocytopenia with features of bleeding (as in case 1), there is often a need to intervene. Chu et al⁴⁴ suggest that the provision of ABO group–specific platelet pools with more diverse HLA specificities may allow for a possible escape from the mechanisms of immune clearance. Another approach that has been described is the provision of a continuous platelet infusion or platelet drip, but the reports of clinical benefit are only published as case reports or small case series. These reports of continuous infusions vary in approach: 3 platelet units over 24 hours, with each split in half and the splits infused over 4 hours each; prolonged infusion (eg, 6 hours) of single standard unit; or high-volume infusions with 24 to 30 platelet units per day.^2,45,46 

If testing demonstrates that no anti-HLA antibodies are contributing to the PR, many question the role of attempting HLA matching prophylactically. A cohort study of refractory patients receiving HLA-matched platelets compared increments across varying degrees of HLA match/mismatch in single-donor apheresis components and confirmed little effect of HLA matched vs mismatched in those with a negative HLA antibody screen.⁴⁷ For patients with nonimmune PR, the clinician is then left with a decision regarding whether to continue repeated platelet transfusions, accepting that increments may be very poor. Given the uncertain evidence of superiority for platelet transfusions and the recognition of more risks of transfusion from some randomized trials, it may be prudent to withhold repeated transfusions in nonbleeding patients and rather apply the clinical context to the need for platelet transfusion.⁴⁸ 

Two recent randomized trials established that the use of tranexamic acid as an adjunct in addition to prophylactic platelet transfusions did not reduce death or World Health Organization grade ≥2 bleeding in unselected patients with hematologic malignancies undergoing intensive therapy. However, these trials documented the safety of tranexamic acid for periods of use up to 30 days of prolonged thrombocytopenia; and similar to our first case, clinicians may choose to offer tranexamic acid to individual patients with problematic symptoms (eg, oral bleeding).^49,50 

The provision of HLA-matched platelets for patients with HLA alloimmune refractoriness is a current standard of care.^19,51 The strategies for selection of appropriate platelet components for patients with HLA or HPA immune PR vary. For example, the strategies used by NHS Blood and Transplant (NHSBT) in England, the 2 blood suppliers in Canada, and Malaysia’s National Blood Centre are summarized in Table 1. These 3 blood suppliers use internal algorithms to find donors based on the patient’s HLA type and the occurrence of antibody-defined “unacceptable” HLAs. It is clear that considerable resources are devoted to support these processes.

There are different general approaches to platelet provision in patients with demonstrable HLA alloantibodies: serologically crossmatched platelets; HLA matched (platelets that are matched to the recipient’s HLA phenotype); or HLA selected (platelets that are selected to avoid antibody specificities).⁵² Serologically crossmatch-compatible platelets by solid-phase red cell adherence assay can use either random platelet units or those selected on the basis of their HLA phenotype. Some jurisdictions with limited availability of HLA-typed platelet donors or access to HLA antibody identification facilities perform crossmatching against units already in stock to facilitate transfusion before complete identification of HLA antibody specificities.⁵³ Small single-center studies have demonstrated that this approach may be appropriate.^54,55 

Classic HLA matching requires donors and recipients to be matched for HLA A and B antigens, with the aim of a complete 4 of 4 match, but permissive mismatches based on cross-reactive epitope groups are acceptable. The ability to find 4 of 4 HLA-matched platelet donors requires a large donor pool with a broad ethnic base. NHSBT currently has ∼10 000 active donors for HLA support. However, ∼93% of these donors are of White ethnic ancestry, but the patient population for hematologic disorders (malignancies, platelet disorders, and hemoglobinopathies) is only ∼53% White, with a much higher relative proportion of individuals from Southeast Asian ethnic backgrounds.

Given these limitations, some blood suppliers use the antibody specificity prediction method to select platelet donors who lack cognate antigen for the identified HLA class I or HPA antibody specificities to increase the number of possible “matches” (also known as HLA selection). More informative platelet crossmatching can proceed with randomly selected platelets (apheresis or whole blood derived), those that are classic HLA matches, or HLA-selected components; but this technique is technically challenging and often restricted to reference laboratories.

In individuals who have class I HLA antibodies, HLA-selected or -matched platelets often lead to higher mean increments, but this may not be consistent. The strength of the class I HLA antibody reactivity can affect the degree of PR, with some “permissivity” seen when platelets are transfused from a donor who is positive for a single specificity or one who is positive for a few antigens associated with low antibody intensity.⁵⁶ A single-center retrospective study reported on 94 HLA-sensitized patients who received 877 platelet products of different types: random apheresis platelets, pooled platelets, and HLA-selected platelets.⁵⁷ It was expected that pooled platelets would have benefits over random apheresis platelets due to this reported permissivity, but no statistically significant difference was identified. Their reported increments for random apheresis platelets, pooled platelets, and HLA-selected platelets were a median of 0.89 (range, –9.72 to 20.34), 2.60 (range, –16.66 to 46.55), and 10.09 (range, –5.51 to 47.53), respectively, indicating highly variable responses but also that HLA-selected platelets lead to greater increments in some but not all patients.

In a cohort study of patients with positive screens, HLA-matched components provided the highest PI, although there was still some benefit in increments if HLA-mismatched (to a varying degree) platelets were administered. Of note, major and minor ABO incompatibility further reduced increments,⁴⁷ consistent with findings from a systematic review of 19 studies that found evidence of higher increments for ABO-identical platelet transfusions.⁵⁸ In a retrospective study, in which antibody specificity prediction was used to identify HLA-selected platelets, the median PI was higher than non–HLA-matched platelet transfusions at 1 hour. However, this difference was not seen at 24 hours, and there were no sustained effects of HLA-matched platelet transfusions.³⁴ This reiterates the common occurrence of concomitant nonimmune factors for PR, because many patients had fever or a degree of splenomegaly.

Different techniques may help select platelets for transfusion when fully HLA-matched units are unavailable. Available inventory may be searched for crossmatch-compatible units and/or group-specific units while waiting. Recent data suggest that nonmatched platelets with a sum of donor-specific antibody (DSA) MFIs <10 000 (low DSA-MFI or least incompatible units) may have higher median CCIs than randomly selected platelet units,^37,59 so these may be an option to consider.

A 55-year-old multiparous woman was first evaluated for PR 1 month after a diagnosis of acute myeloid leukemia, with results shown in Figure 3A. HLA-selected platelets provided support for treatment and hematopoietic stem cell transplantation. At relapse 1 year later, the previous HLA donor pool was automatically reactivated. However, increments were borderline to satisfactory, ranging between 10 × 10⁹/L and 20 × 10⁹/L after each HLA-selected unit. Repeat testing demonstrated that the relapse of the hematologic malignancy did not result in a full recurrence of historical HLA antibody profiles (Figure 3B). In this case, the borderline increments to platelets were likely due to nonimmune factors for PR.

The clinical as opposed to platelet count increment benefits of offering crossmatched, low DSA-MFI, HLA-matched or HLA-selected platelets to patients is unclear. A systematic review by Pavenski et al⁶⁰ reported that higher PIs were seen with the use of HLA-selected platelets, but many studies failed to provide information on clinical outcomes, including bleeding. Much of this literature has been dominated by retrospective analyses. More recently, a double-blind, randomized, noninferiority crossover trial was reported, in which the effect of prophylactic HLA epitope-matched (HEM) platelets (using the HLAMatchmaker algorithm) was compared with standard HLA-matched platelet transfusions.⁶¹ The rationale was that HLA epitopes could be defined structurally as well as molecularly, and computer algorithms designed to evaluate donor-recipient compatibility at the epitope level may better predict the specificities of HLA antibodies than antigen or cross-reactive epitope group matching.^62,63 In this trial, 40 alloimmunized patients with thrombocytopenia randomly received 4 pairs of different platelet transfusions, either HEM or standard HLA matched. The primary outcome was 1-hour PIs, but secondary outcomes evaluated bleeding. With >100 platelet transfusions in each arm, PIs with HEM platelets were almost identical to those in patients receiving HLA standard antigen-matched platelets. There were no differences in the secondary outcomes of platelet count, transfusion requirements, or bleeding events. The results support a role for HEM platelets as an alternative approach for immune PR. Interestingly, because adequate 1-hour PIs were more frequently observed with a mean number of 3.2 epitope mismatches compared with 5.5 epitope mismatches, epitope matching may have a more specific role in supporting complex alloimmunized patients (K. Mepani and S. J. Stanworth, NHSBT, written communications, 19 June 2024).

Donor factors can potentially influence platelet increments and may be better defined in the future when selecting (apheresis) platelets for transfusion but have received less research attention. A proportion of donors manifest naturally low expression of certain class I HLAs such as HLA-B8, -B12, or -B35,⁶⁴ which is possibly determined by their zygosity, and it has been demonstrated that platelets from these donors do not undergo antibody-mediated internalization upon sensitization by cognate patient antibodies. Identification of these “low-expressor” donors may expand the donor pool for certain patients with PR, although this proposed approach requires extensive further clinical studies.

Serial investigations are commonly recommended in alloimmunized patients once HLA-matched/selected platelet support is initiated. The timing of these repeat investigations may be influenced by changes in PI and local blood supplier policies. We recommend repeat testing be performed when there are 2 sequential platelet transfusions with the appropriate platelets that demonstrate poor increments or every 6 months to a year if increments are adequate or not available. Since the TRAP study demonstrated that many newly alloimmunized patients had loss of antibody by lymphocytotoxic antibody methodology²² and another study demonstrated waning or loss of antibody in up to 30% of patients,⁶⁵ repeat investigations to evaluate changes to antibody profiles may help refine donor selection or identify progressive sensitization, including new HLA or HPA specificities.²⁵ Strategies such as antibody absorption,⁴⁶ desensitization protocols⁶⁶ (eg, using immunoglobulins, plasma exchange, and immunomodulatory drugs such as rituximab or bortezomib), novel therapies (eg, eculizumab),⁶⁷ and use of “designed” platelet components with altered HLA expression (eg, acid treatment^68,69 or generated from HLA-deficient pluripotent stem cells)⁷⁰ are areas of ongoing research and should at present be considered unproven interventions.³⁰ 

PR is a common challenge, but at the core is a “laboratory” phenomenon of presumed clinically inadequate PIs after platelet transfusion. Even in patients with identified HLA antibodies, the link to refractoriness is far from straightforward, and coexisting nonimmune causes may be more relevant to suboptimal transfusion responses. The challenge is to identify which patients with PR need detailed investigation and which patients benefit from an optimized product, such as HLA-selected platelets. A sample algorithm and clinical decision support tool are provided (Figure 4; supplemental Table 3, available on the Blood website), but we acknowledge that the best management is uncertain when a suitable platelet unit is not available. Many aspects of testing and management are supported by limited prospective evidence for clinical outcomes. Research needs to be extended to health economic evaluations^71,72 and alternative markers of bleeding risk other than platelet counts.^73,74 Randomized trials of different matching strategies can be performed but should be in addition to analyses with larger data sets, perhaps using techniques that incorporate biomarkers and information on clinical outcomes to help target investigations and better use of platelet transfusions. Strategies of preventing alloimmunization by providing prophylactic HLA selection to “at-risk” patients have been reported, but overall benefits are unclear.⁷⁵ Our cases illustrate when to maintain a high index of suspicion (eg, multiple pregnancies or previously received transfusion), the rapidity of onset of refractoriness in previously sensitized patients, the need for timely management decisions in some cases while awaiting results, follow-up testing to refine donor pools, and the importance of coordination between testing laboratories, blood suppliers, and clinicians.

The authors thank Kirti Mepani from NHS Blood and Transplant; Johnny Mack, Matthew Seftel, Nancy Hua, and Natasha Rickards from Canadian Blood Services; Nancy Robitaille and Marie-Claire Chevrier from Hema-Quebec; and Nor Hafizah Ahmad from the National Blood Centre of Malaysia, for information on current practices. They also thank Michael Murphy and Akash Gupta for review of drafts, and Kim Lacey and Kiara Crossley for assistance with formatting and figures.

Contribution: All authors participated in the development, review, and revision of the manuscript.

Conflict-of-interest disclosure: The authors declare no competing financial interests.

Correspondence: Susan Nahirniak, University of Alberta, 4B1.19 Walter Mackenzie Centre, 8440-112 St, Edmonton, AB T6G 2B7, Canada; email: susan.nahirniak@albertaprecisionlabs.ca.