Use of Human Leukocyte Antigen (HLA)–Incompatible Platelet Units in HLA Platelet-Refractory Patients With Limited Number of or Low-Level HLA Donor-Specific Antibodies Results in Permissive Transfusions

We conducted an Institutional Review Board–approved retrospective analysis of all HLA-selected platelet transfusions administered from 2016 to 2018 at a large urban academic/teaching hospital. We defined HLA-selected platelets as units intended for transfusion support of thrombocytopenic patients in the setting of HLA alloPR. Apheresis platelets were purchased from a blood vendor with an accredited in-house HLA laboratory performing HLA typing on platelet donors. HLA-selected units were deemed HLA compatible if the donor units were: (1) HLA-matched units (M-Us): donor low-resolution HLA-A and HLA-B type was identical or zero mismatched to the intended recipient (equivalent to Duquesnoy match grade A or BU¹⁵ ), or (2) HLA antigen–negative units (AN-Us): donor HLA antigen(s) are not targeted by patient HLA alloantibodies (ie, no DSAs present). Serologic crossmatch compatible platelets were excluded. Occasionally, HLA-compatible units were unavailable, and, when clinically indicated, a physician board-certified in both transfusion medicine and histocompatibility would approve the use of HLA-selected platelet units against which a patient had DSAs (ie, HLA incompatible).

All patients included were age >18 years, previously diagnosed with HLA alloPR by board-certified transfusion medicine physicians, and had received at least 1 HLA-selected platelet unit (with a posttransfusion platelet count measured within 1–4 hours). Patients had a diagnosis of alloPR when documented postplatelet transfusion CCI was <5000 following 2 routine (ie, non–HLA-selected) platelet transfusions and when solid-phase assay demonstrated HLA alloantibodies. All reported transfusion reactions were routinely investigated by board-certified transfusion medicine physicians and reported to the National Healthcare Safety Network Hemovigilance Module Surveillance Protocol.¹⁷ 

Patient HLA-A and HLA-B HLA typing was performed by molecular methods at our institution's histocompatibility laboratory or by a reference HLA laboratory. All laboratories used were duly accredited by appropriate agencies. The indications for HLA typing of patients include potential hematopoietic progenitor cell transplant or transfusion support.

Circulating class I HLA antibody identification was established by solid-phase single-antigen bead testing (LABScreen single antigen, One Lambda/ThermoFisher, West Hills, California) on dithioerythritol-treated serum and analyzed using HLA Fusion software (One Lambda/ThermoFisher, West Hills, California) in a Luminex 200 instrument (Luminex Corp, Austin, Texas). HLA antibody specificity was assigned using the software baseline MFI for each antigen; an MFI ≥1000 was considered reactive (positive), although overall reactivity pattern was also considered. The calculated panel reactive antibody (CPRA) % was determined using the Organ Procurement and Transplantation Network online CPRA calculator¹⁸  and based on positive antibody specificities. The CPRA describes the likelihood of having an incompatible donor based on the recipients' known antibodies and on donor population HLA antigen frequency.

All antibody specificities were compared against HLA-A and HLA-B typing of donors. For this portion of the study, only individual antibody specificities with MFI ≥1000 against donor HLA-A and/or HLA-B antigens were called DSAs. The DSAs targeting a homozygous HLA locus were counted once. Thus, DSA number could range from 0 to 4. HLA-compatible units (ie, M-Us and AN-Us) are all considered zero/no DSA.

Platelet units against which the recipient had any DSA in the most recent antibody test were further arbitrarily categorized into: (1) units targeted by low-level DSA (L-DSA-Us) if the cumulative MFI of DSA(s) was ≥1000 to <6000, or (2) units targeted by high-level DSA (H-DSA-Us) if cumulative MFI of DSA(s) was ≥6000. The H-DSA-U MFI cumulative threshold was chosen to reflect our institution's prior HLA-incompatible platelet unit rejection threshold. We obtained cumulative MFI by adding the baseline MFI of all HLA-A and HLA-B antibodies (against each individual donor unit), hypothesizing that the combined MFIs of multiple DSAs, even if individual antibody specificities have an MFI <1000, could produce an additive effect. We do not routinely match platelet transfusion for ABO compatibility, but for this study, we categorized platelet transfusions as ABO incompatible (ABOi) when the recipient had ABO isohemagglutinin targeting the donor unit (ie, ABO major incompatibility in the host versus graft direction).

Patient demographics, clinical history, laboratory values, and transfusion history as well as platelet product data were gathered from our medical center's blood bank and histocompatibility laboratory records and electronic medical record.

At the time of this study, our hospital employed a routine posttransfusion platelet count ordering system following all platelet transfusions. Only transfusion events that included a posttransfusion platelet count within 4 hours of platelet transfusion were further analyzed. The CCI was calculated based on the Davis formula,¹⁹  and body surface area (in m²) was calculated per the Mosteller formula.²⁰  An adequate CCI was ≥5000. Permissive HLA-incompatible transfusions were defined as those which produced adequate CCI despite the presence of DSA.

Statistical analysis included Kruskal-Wallis followed by Mann-Whitney U performed with GraphPad Prism version 8.00 for Mac OSX (GraphPad Software, La Jolla, California). P < .05 was considered statistically significant.

During the study period, 708 HLA-selected platelet units were transfused; each transfused platelet unit was considered a transfusion event. Of the 708 HLA-selected platelet units, 279 (39%) eventually met inclusion criteria for evaluation listed in Figure 1. These 279 HLA-selected units were transfused into 26 distinct patients. Detailed patient demographics are outlined in Table 1, but in summary, most of these patients were female, older, affected by hematologic neoplasm, highly HLA allosensitized (ie, high CPRA), and were newly diagnosed with alloPR during the study period. The median CCI for all 279 transfusion events was 11 800 (interquartile range [IQR], 7000–15 700).

To understand whether the number of DSAs a patient has against a platelet unit affected permissiveness of transfusion, we compared the CCI of transfusion events based on the number of individual DSAs (MFI >1000) as shown in Figure 2. In this study, 240 of 279 HLA-selected platelets (86%) were transfused in the absence of DSA, which was expected because in practice, we preferentially select M-Us or AN-Us when possible. Transfusions across no DSA produced a significantly higher median CCI compared with those across 1 or more DSAs (12 400 and 6700, respectively; P < .001). When further stratified based on the number of DSAs, 20 of 29 transfusions (69%) across a single DSA were permissive, whereas transfusions across 2 or ≥3 DSAs were unlikely to be permissive (n = 2 of 6 [33%] and n = 0 of 4 [0%], respectively; Figure 2). We performed additional pairwise comparison of posttransfusion CCI when transfusing across 1 versus 2, 2 versus 3, and 1 versus 3 DSAs. There was no statistically significant difference transfusing across 1 versus 2 DSAs (P = .25), despite a trend toward higher CCI in 1 versus 2 DSAs. Similarly, there was no statistically significant difference transfusing across 2 versus 3 DSAs (P = .36), despite a trend toward higher CCI in 2 versus 3 DSAs. When comparing 1 versus 3 DSAs, transfusion across 1 DSA yielded statistically higher CCI (P = .04). The data support an inverse relationship between CCI and the number of DSAs.

Beyond the number of DSAs, we also investigated whether the level of DSA impacted posttransfusion CCI. For this specific analysis, we used cumulative MFI as a measure of DSA level. There was no statistically significant difference in CCI achieved when M-Us, AN-Us, or L-DSA-Us (cumulative DSA MFI <6000) were transfused (P = .43). However, H-DSA-Us (cumulative DSA MFI ≥6000) produced significantly lower CCI (median, 0; P < .001) than non–H-DSA-Us (M-Us, AN-Us, and L-DSA-Us; Figure 3). Interestingly, the median CCI when transfusing across low-level DSA was 12 600, well above the adequate CCI threshold of 5000, indicating permissive transfusion. In contrast, the median CCI when transfusing across high-level DSA was statistically significantly lower at 0 (P < .001) and certainly well below adequate CCI threshold, indicating nonpermissive transfusion.

We investigated whether DSA level (mean cumulative MFI) targeting platelet units varied with the number of DSAs. We found that the cumulative MFI appeared similar when units were targeted by 1 DSA (2593; SD, 2240) or 2 DSAs (3964; SD, 5100), but that 3 DSAs had much higher cumulative MFI levels (27333; SD, 34 803).

Further, we compared the proportion of permissive transfusions with M-Us, AN-Us, L-DSA-Us, and H-DSA-Us. HLA-matched units, AN-Us, and L-DSA-Us were permissive in 232 of 267 transfusion events (87%), compared with H-DSA-Us (n = 2 of 12 [17%], rounded to the nearest whole number; Table 2). Indeed, an outlier transfusion categorized as H-DSA-U (Figure 3, in bold black/filled-in circle) is probably more correctly L-DSA-U. The platelet unit was typed as HLA-A2, HLA-A24, HLA-B15, HLA-B44. The patient had multiple antibodies targeting the low-resolution HLA-B15 antigen, which includes HLA-B62, HLA-63, HLA-75, HLA-76, HLA-77, and some HLA-70, with MFI ranging from 100 to 6800. The specific DSA could not be properly assigned, and therefore the average of all the potential DSAs was taken, resulting in this unit being categorized as H-DSA-U. Excluding this outlier transfusion event, only 1 of 11 H-DSA-Us (9%) was permissive.

It should be noted that because of the way we categorized HLA-incompatible units based on either the number or level of DSA, the number of transfusion events that comprise “none” (n = 240 of 279 [86%]) in the number of DSAs category (Figure 2) will not equal the sum of events in the level of DSA category (Figure 3) that comprise “M-U” (n = 52 of 279 [19%]) and “AN-U” (n = 147 of 279 [53%]). This is because some donor units may have 2 possible DSAs with individual MFI <1000, and thus would be categorized as “none” by number of DSAs. When the same unit is categorized based on level of DSA, the cumulative MFI may be >1000, and thus it is included in the L-DSA-U category.

We reviewed whether higher frequency donor population HLA antigens are more difficult to avoid transfusing against to determine if our findings were limited to certain HLA antigens. Using the CPRA calculator to define the frequency of individual HLA antigens, we found we transfused across DSA targeting the antigens with frequencies from <1% (eg, HLA-B67) to 48% (HLA-A2; Table 3). Notably, in units transfused across 1 DSA, units with the following HLA antigens yielded only poor CCI: HLA-A25, HLA-B27, HLA-B35, and HLA-B75. Within this subset, the DSA against HLA-B27 had the highest mean MFI (9400), whereas the MFI of the DSA targeting the remaining antigens was cumulatively <2000.

Finally, we investigated if ABO isohemagglutinins act in synergy with HLA DSA and whether platelet storage time (“age”) impacted CCI. Figure 4 shows that concomitant ABOi units do not result in significantly lower CCI compared with ABO-compatible units, regardless of coexisting HLA DSA level. Although statistically insignificant, we noted a trend toward lower CCI in the presence of both low-level DSA and ABO incompatibility (Figure 4, C) but not in the presence of both high-level DSA and ABO incompatibility (Figure 4, D). Table 4 shows the impact of storage time for 221 platelet units in this study. Overall, the CCI, based on platelet unit age, did not impact CCI significantly. But in platelets >4 days old transfused across low-level HLA DSA, concomitant ABO incompatibility resulted in significantly decreased, albeit still adequate, platelet recovery compared with similarly stored ABO compatible units (CCI, 15 300 versus 8000, respectively; P = .02).

Among all 279 transfusion events in the study period, only 1 transfusion reaction was documented, a minor allergic transfusion reaction to an L-DSA-U (incidence of 0.35%). This incidence of allergic transfusion reactions in our population appears similar to the predicted rate of allergic reactions to all transfusions (n = 14 of 4857 [0.29%])²¹  and the predicted rate of minor allergic transfusion reactions to apheresis platelets (n = 1616 of 93 737 [1.72%]).²² 

In this study, we explore the factors that contribute to permissive versus nonpermissive HLA-incompatible platelet transfusion support of HLA alloPR patients. Patients meeting full inclusion criteria were predominantly elderly, female, affected by hematologic neoplasm, not having received stem cell transplantation, and having received a new diagnosis during the study period (Table 1). Although we feel this is an accurate reflection of patients most commonly experiencing alloPR and requiring HLA-selected transfusion support, it is entirely possible that this is not the experience of other centers.

Peripheral platelet count is used to guide platelet transfusion, with a platelet count of 10 000/μL considered a reasonable threshold for bleeding prophylaxis.²³  A study investigating platelet dose efficacy showed patients with platelet counts of <5000/μL sustained an increased number of days with bleeding episodes (25%) as opposed to patients with platelet counts 6000 to 80 000/μL (17%).²⁴  Platelet dose (half-dose, standard dose, or double dose) given for a platelet count of 10 000/μL did not affect overall bleeding rate, although low-dose recipients received more platelet transfusions than those receiving a high dose. Notably in this same study, even half the standard dose of platelets yielded a similar CCI (∼10 000) compared with standard and double the standard platelet doses. Thus, the platelet dose required to prevent significant bleeding is debatable.²⁵  Nonetheless, an adequate CCI (“permissive transfusion”) is likely a reasonable surrogate for the efficacy of platelet transfusion and can be used as a metric for comparing different categories of HLA-selected platelets when supporting HLA alloPR patients.

Providing platelets in a timely manner to patients with alloPR is a pressing clinical challenge faced by transfusion medicine physicians and blood banks. Matching patient and donor HLA antigens provided the earliest successful form of transfusion support for patients.²⁶  Solid-phase single-antigen bead immunoassays have expanded transfusion support options to antigen-negative units. In our experience, our alloPR patients are highly sensitized (ie, high CPRA), and our retrospective data indicate that we were unable to provide HLA-compatible (ie, matched or antigen-negative) units in 39 of 279 transfusion events (14%). Thus, a fair number of platelet transfusion requests can only be filled using HLA-incompatible units. It is important to note that patients can present with HLA allosensitization levels lower than what we observed in this study. A high CPRA may result from numerous antibodies against many antigens or few antibodies against high(er) frequency HLA antigens. That is, a patient with an antibody against HLA-A2 would have a CPRA of almost 50% because HLA-A2 is present in approximately 45% to 48% of donors. In this situation, we would opt to respect the HLA-A2 DSA by avoiding transfusion of HLA-A2–positive platelet units when possible, regardless of the DSA MFI and the patient's relatively lower sensitization level (CPRA 48%). Platelets are a limited resource and should be directed toward patients in need and expected to benefit most from transfusion. Some patients initially present with low CPRA due to a low level of (or even barely detectable) HLA antibodies, but upon transfusion with platelets against which they have low-level DSA, memory responses are elicited, causing rapid subsequent DSA increase. This makes transfusion support more challenging in the future. Regardless of whether the DSA and CPRA are low level (MFI <6000), it is prudent to avoid further sensitization, especially if the patient will require prolonged transfusion support.

Given a choice, we prioritize M-Us and AN-Us to avoid any HLA DSA. Because our hospital inventory depends on donor recruitment and collection, it is entirely likely that more commonly available units are usually positive for more common HLA antigens. We suspect but cannot confirm our blood supplier's donor pool reasonably reflects the HLA antigen frequencies of the broader population. Additionally, our patients have varying transfusion support needs. Some require HLA-compatible platelets daily, whereas others need only a single platelet transfusion or a few platelet transfusions.

We show that HLA-selected platelet units against which a patient has a limited number or low-level DSA are viable options for transfusion support. HLA-selected platelet units against which a recipient only has low-level DSA in particular are permissive compared with HLA-compatible units (M-Us or AN-Us) and without increasing the risk of transfusion reactions. The use of L-DSA-Us in times of urgent need greatly expands the potential donor pool, expediting platelet transfusion support for highly sensitized patients. This agrees with a recent study by Karlström et al.¹⁶ 

Notably, the level of all DSA present (ie, cumulative MFI) appears more likely to dictate permissive (or nonpermissive) transfusion, rather than the number of DSAs. The cumulative MFI cutoff >6000 that in our data generally predicted nonpermissive HLA-incompatible transfusion is slightly higher than our experience in solid organ transplant recipients, where HLA class I DSA MFI of 2000 to 4000 commonly results in positive flow lymphocyte crossmatch. Because platelets express lower levels of HLA (class I) antigen than lymphocytes,²⁷  it makes sense that a higher level (quantity) of antibody and/or higher affinity is needed for detection of binding and clearance. Solid-phase immunoassays have reported variability in MFI, which limits the use of a single DSA MFI cutoff of >6000 as nonpermissive for all transfusion medicine services. More recently, standardization of testing has been shown to decrease variability,²⁸  and treatment of patient sera to remove potential interfering substances also shows improved interlaboratory variability in these assays.²⁹  This would then indicate that transfusion medicine services could probably use an MFI of 6000 as a starting point for assessing permissive transfusions for their own practices.

There are at least 28 HLA-A and 50 HLA-B antigens currently recognized, and in our study we transfused units in the presence of DSA against only a subset of all possible HLA antigens (8 HLA-A and 19 HLA-B antigens), but these antigen targets represented frequencies that were relatively uncommon to very common. Although we cannot definitely predict that transfusion across a single DSA or a cumulative MFI <6000 for every single antigen-antibody pair will result in an adequate posttransfusion count, our data suggest that one is more likely to have a permissive transfusion when selecting incompatible HLA units with these 2 parameters without regard for what the HLA antigen target is.

Undoubtedly, at times, truly HLA-compatible platelet units (M-Us or AN-Us), are nonpermissive and units with multiple or high-level DSA are permissive (Figures 2 through 4). Clearly, other factors beyond recipient HLA DSA influence CCI, including variation in donor HLA antigen expression levels.³⁰  For instance, some HLA antigens (such as HLA-B44) are poorly expressed and, despite strong anti-B44 DSA, may result in permissive transfusion.²⁷  Additionally, unmeasured recipient antibodies (eg, ABO agglutinin titer, non-HLA, or other non-ABO platelet antibodies) may contribute to nonpermissive transfusions. Because of the retrospective nature of this study, we are also unable to account for every patient disease state or drug regimen that may also impact platelet count recovery.

ABOi L-DSA-Us versus ABO-compatible L-DSA-Us produced a nonsignificant trend toward lower CCI (median CCI, 12 000 versus 12 800, respectively; P = .09), but both produced median CCI well above the adequate threshold of >5000, suggesting both produce permissive transfusion and either may be used to support the alloPR thrombocytopenic patient. However, 8 of 36 ABOi L-DSA-Us (22%) were nonpermissive compared with 0 of 18 ABO-compatible L-DSA-Us (0%). Other studies have shown that isohemagglutinins by themselves do not impact CCI.¹³  We were unable to determine ABO isohemagglutinin titers (ie, the level of ABO antibodies) for these patients; potentially, specific titers of ABO isohemagglutinins are synergistic with L-DSA-Us, but this requires further study.

Alternately, we may have lacked sufficient L-DSA-U transfusion events to detect a statistically significant difference between compatible and incompatible L-DSA-U transfusion events. Nonetheless, given that almost a quarter of ABOi L-DSA-Us (compared with none for ABO-compatible L-DSA-Us) resulted in nonpermissive transfusions, it may be reasonable to prioritize ABO-compatible L-DSA-Us versus ABOi L-DSA-Us when afforded a choice. Interestingly, when platelet units are transfused after storage of ≥4 days in the presence of low-level DSA and ABO incompatibility, there is a significant decrease in platelet recovery, but it still yields an overall adequate CCI. This suggests that although they are not the most important consideration when choosing between HLA-incompatible platelet units, factors such as ABO compatibility and shorter storage time may be of use to fully optimize transfusion support.

Although many studies have addressed the efficacy of various types of HLA-selected platelets in alloPR patients, a review showed that most were small studies and did not use current solid-phase antibody testing technology.³¹ 

We used the number and level of DSAs as a means to quantify the nonpermissive effect of HLA DSA and the apparent lack of role for concomitant ABO incompatibility. Despite comparable and adequate CCI for transfusion support using M-Us, AN-Us, and many L-DSA-Us, the use of AN-Us and L-DSA-Us merits additional care, especially in patients with chronic transfusions. In our experience, there is a potential for the transfusion of mismatched HLA antigens to theoretically induce further allosensitization.

Alternative experimental approaches to reduce recipient HLA alloimmunization have been reported, including the use of rituximab, bortezomib, intravenous immunoglobulin (Ig), IgG denaturing enzymes, and plasmapheresis.^32–36  These methods moderate alloantibody production or decrease alloantibody half-life. Although these agents may blunt alloPR, they may elevate the risks to the patient because of their side effects and provide limited benefit because of their variable efficacy. Moreover, they increase health care costs and may not provide the desired effect within a clinically relevant time frame.

Selecting platelet donors with low surface class I HLA antigen expression is another compelling solution, which introduces additional challenges of understanding and establishing HLA expression across blood donors.³⁰  Up to 70% to 90% of class I HLA molecules can be removed from the platelet surface by acid treatment, decreasing antibody-mediated phagocytosis. Acid-treated platelets appeared to retain function, but platelet recovery following treatment was decreased to approximately 60%.³⁷  Such treatment requires technician hands-on time and could impact sterility. Additionally, blood products intended for transfusion are regulated by the US Food and Drug Administration, and such manipulation of platelets opens up regulatory challenges. As such, thoughtfully chosen HLA-selected units rather than the above experimental approaches are probably the most effective manner for transfusion support of alloPR patients.

Finally, among the 279 platelet transfusion events investigated, 17 (6.1%) demonstrated variation in the antibody levels of split specificities comprising a broad antigen. This occurs because low-resolution HLA typing of donors may contribute to incorrect assignment of DSA number or level (see Figure 3 outlier unit represented by a black circle). This is important because our blood vendor routinely performs donor HLA typing at low resolution. This means that the donor HLA type may be reported as a broad or parent antigen (eg, HLA-B15), but current technology allows antigen identification to be further refined into multiple similar but distinct antigens (ie, antigenic splits) or even at high- or allele-level resolution.¹⁵  We routinely see patients whose split antibodies (eg, HLA-B62, HLA-B75, etc, are common splits of HLA-B15) have variation in MFI. And even within distinct antigens, allele-specific antibodies have been reported to be important in transfusion support.³⁸  With routine high(er) resolution donor HLA typing, selection of L-DSA-Us may be more reliably performed.

In summary, this work shows that platelet transfusion support of patients with alloPR can be accomplished by the provision of HLA-compatible units, but approximately 14% of the time such units are not available. A viable option, then, is to use HLA-selected platelets against which the recipient has few or low-level DSAs (cumulative MFI <6000) only, which not only leads to adequate CCI but also does not lead to increased frequency of transfusion reactions. HLA-incompatible units targeted by multiple or high-level DSAs (cumulative MFI >6000) should be avoided because they are likely to result in nonpermissive transfusion. Therefore, when needed at our institution, HLA-incompatible donor units against which the intended recipient has <1 DSA or the cumulative DSAs have a mean fluorescence of <6000 are prioritized. Identifying similar institution-specific MFI cutoffs at other institutions will inform HLA-incompatible donor unit choice and expand HLA-selected platelet unit pool per center; however, a DSA MFI <6000, established by solid-phase bead testing, may be used as an initial cutoff.

We would like to acknowledge the following people for their help in preparation of the manuscript: Raymond Comenzo, MD; Adam Norfolk, BS, MBA; and Matthew Sullivan, BA, MA.