There are no accepted methods to predict the development of platelet transfusion refractoriness (PTR) due to human leucocyte antigen (HLA)-alloimmunization. Hence, matched platelets are usually given only to patients demonstrating PTR, necessarily resulting in some ineffective random donor platelets (RDPLT) transfusions. To assess its utility in predicting PTR, we retrospectively tested samples from 387 patients receiving chemotherapy for acute leukaemia or autologous transplantation using a micro-bead flow cytometry assay. The average of the mean fluorescence intensities (avgMFI) of the class I beads in the screening assay was correlated with outcomes of RDPLT transfusions during a 2 week period. Antibodies were detected in 57 patients; 66 developed PTR, of whom 28 were alloimmunized. avgMFI usefully predicted the development of PTR (area under the receiver operating curve 0·87, 95% confidence interval: 0·77–0·96). A logistic regression model estimated the probability of PTR to be >90% when avgMFI >5440. These results indicate that micro-bead flow cytometry assays could inform a risk-adapted strategy for managing thrombocytopaenic HLA allo-immunized patients.
Haematology patients receiving intensive chemotherapy often require platelet support. Depending on the study population, and the definition of refractoriness used, up to a third of such patients have persistently inadequate responses to transfused platelets (platelet transfusion refractoriness [PTR]) (Hod & Schwartz, 2010; Pavenski et al, 2012). PTR can result from many factors including disseminated intravascular coagulation, splenomegaly, sepsis and some medications, as well as from clearance mediated by allo-antibodies against class I HLA antigens on donor platelets. However, not all HLA-alloimmunized patients will develop PTR. Most published studies on PTR have utilized either complement-dependent cytotoxicity or enzyme-linked immunosorbent assays to detect antibodies (Bishop et al, 1988; Doughty et al, 1994; Legler et al, 1997; The Trial to Reduce Alloimmunization to Platelets Study Group, 1997; Slichter et al, 2005). Positive predictive values of these assays have been inadequate for clinical decision making.
Because of this, and also due to difficulties in sourcing them, HLA-matched platelets are usually reserved for patients with confirmed PTR. This practice sees some patients initially receiving random donor platelet (RDPLT) transfusions, which will not achieve target platelet count increments. Transfusion practice could be improved if it were possible to further risk stratify allo-immunized patients. High risk patients could be considered for pre-emptive use of HLA-matched platelets, cryopreserved autologous platelets, or possibly thrombopoietin analogues.
Sensitive and specific micro-bead flow-cytometric methods are now in widespread use for the detection and quantification of HLA antibodies. Mean fluorescence intensities (MFI) obtained from these assays have been shown to broadly correlate with antibody titres (Carrick et al, 2011). We investigated whether MFIs could be used to stratify the risk of PTR among allo-immunized patients.
Patient and methods
Since 2005, patients undergoing intensive chemotherapy in our unit were routinely screened for HLA antibodies, but given matched platelets only if diagnosed to have PTR. We retrospectively identified 501 patients who were treated with induction therapy for acute leukaemia, or received an autologous stem cell transplant, between January 2005 and March 2012. Patients were excluded if they did not have a serum sample taken within 6 weeks of commencing cellular blood product transfusions (n = 92), received intravenous immunoglobulin in the preceding 3 months (n = 4), were scheduled for prophylactic HLA matched platelet transfusions due to prior PTR (n = 8), were randomized to the active arm of the Trial Of Prophylactic Platelets study (Stanworth et al, 2012) (n = 7), or had missing platelet counts preventing ascertainment of the primary endpoint (n = 3). All patients received at least one cellular blood product transfusion.
Serum samples were taken during a 6-week period prior to receiving the first cellular blood product related to the index admission for intensive therapy. A 1 week ‘grace period’ for testing was allowed for patients (103/387) who received emergency transfusions. However, to minimize the influence of sensitization occurring after screening, only outcomes during the immediate 2 weeks from commencing cellular blood product transfusions were considered. This also applied to those cases where sampling was performed within 1 week of receiving a blood product.
The primary endpoint was the development of PTR, which was defined as having received ≥2 consecutive RDPLT transfusions associated with a corrected count increment (CCI) of <2·5 at 18–24 h. The standard definition of CCI was used (Hod & Schwartz, 2010): (Post-transfusion platelet count/l – Pre-transfusion platelet count/l) × (Body surface area in m2)/(Platelets transfused × 10−11). Calculation of the CCI was based on the following morning's platelet count because 1-h counts were not routinely performed. Due to the retrospective nature of this study, we did not require demonstration of poor increments following fresh (<48-h old) platelets.
Antibody testing was performed using a micro-bead flow cytometry assay (Lifecodes LifeScreen Deluxe, with positive results confirmed by Lifecodes Class I ID assay, Gen-Probe Transplant Diagnostics, Stamford, CT, USA) either during the treatment period, or on serum samples stored at −30°C. As this was a retrospective study, multiple batches of the screening assay were used. However, calibration beads were used according to the manufacturer's guidelines to minimize inter-lot variability.
The screening assay uses seven class I beads, each carrying different HLA antigens with significant overlap of serologically cross-reactive groups. Mean fluorescence intensities (MFI) were acquired using a Luminex 100 analyser (Luminex Corporation, Austin, TX, USA), and analysed using Lifecodes Quicktype v2.5.5 (Gen-Probe Transplant Diagnostics, Stamford, CT, USA). The software determines the test result by comparing the MFI of individual beads against MFIs of a panel of three different negative control beads supplied with each kit. A standard adjustment for variable background signal was also applied. The Lifecodes Class I ID assay used to confirm screen positive results has a manufacturer reported sensitivity of 97%. The laboratory maintains current accreditation with the American Society for Histocompatibility and Immunogenetics.
We defined the predictor variable avgMFI to be the average MFI of the seven individual beads, weighted by whether the presence of antibodies was confirmed or not:
where w = 1 if the presence of antibodies is confirmed, and 0 otherwise; and the subscript i refers to the ith class I bead.
The institutional Research Ethics Committee approved the study and waived the requirement for individual consent. All authors had access to the entire dataset. Statistical analysis was performed using R (version 2.15.2) (R Foundation for Statistical Computing, Vienna, Austria) including the pROC (Robin et al, 2011) and MKmisc (Kohl, 2012) packages. P < 0·05 was considered statistically significant.
The final cohort comprised 387 patients, median age 56 years and included 160 (41%) females and 231 (60%) patients undergoing autologous transplantation. Antibodies were detected in 57 (14·7%) patients of whom 45 (78·9%) were female. A total of 1443 RDPLT transfusions (mean platelet count 2·4 × 1011/unit) were studied, of which 896 (62%) were single donor apheresis units, and only 40 (2·8%) were ABO-mismatched. The median age of transfused platelets was 4 d.
Sixty-six (17%) patients reached the primary endpoint of PTR, of whom 28 (23 females) had positive screen tests; 29 (22 females) of 321 patients who did not develop PTR also tested positive. The MFIs of the 7 class I beads showed significant correlation (Spearman's ρ for pairwise comparisons ranging from 0·79 to 0·98), justifying the use of an average as a predictor variable. avgMFI values ranged between 0 and 17 056 for males, and 0 and 14 861 for females. Among antibody positive patients, avgMFI values were significantly different between the refractory and non-refractory groups (Fig 1). The area under the receiver operating characteristic curve for avgMFI as a predictor of PTR was 0·86 (95% confidence interval: 0·77–0·96) (Fig 2).
A logistic regression model was constructed as follows: logit (probability of PTR) = α + β × (avgMFI). Estimates (standard errors) for the 2 parameters were: −1·9663 (0·1598) and 0·000784 (0·000186) for α and β respectively, both with P < 0·0001 of being zero. (Hosmer-Lemeshow H statistic 11·516, P = 0·1741, indicating acceptable model fit.) Fig 3 shows the predicted probability of PTR based on this model.
Consistent with previous reports (Carrick et al, 2011), higher avgMFIs correlated with an increased range of target antigens, probably due to increasingly avid binding to cross-reactive epitopes. (Spearman's ρ = 0·77 for correlation between avgMFI and panel reactive antibody percentages (cPRA) calculated for the American population, and used here as a surrogate for the range of target antigens.) However, cPRA was >80% in 25/27 patients with avgMFI >1000, indicating poor ability to further discriminate between patients with varying risks of refractoriness. As HLA antigen frequencies in the Australian platelet donor pool are not documented, we could not directly relate antibody specificities to PTR risk. cPRA did not improve the ability to predict PTR when used in addition to avgMFI.
The aim of this study was to assess if there was a correlation between the quantity of antibodies present at the time of transfusion, and the development of refractoriness. Although it is possible that sensitization occurred due to transfusions after the screening sample was taken, we restricted attention only to outcomes within the first 2 weeks commencing from the first cellular blood product transfusion, limiting the time available for de novo antibody production.
Intuitively, the breadth of antibody specificities would be expected to correlate with the risk of PTR by increasing the probability of a cognate antigen being present in a given RDPLT transfusion. Whilst our data support this, we show that cPRAs soon reached a ceiling of 80–100% when avgMFIs increased over 1000, limiting further discriminatory capability. This phenomenon, which is probably due to increasingly avid binding to cross-reactive epitopes, may partly explain why PRAs based on complement-dependent cytotoxicity assays have not proved useful in predicting PTR. In contrast, avgMFI showed a wide range of values, reflecting both the amount of antibodies as well as, via its correlation with cPRA, the breadth of antibody specificities. By using an average of MFIs we necessarily lost some fine detail regarding specific antibody-antigen interactions. However, this would be mitigated by the significant correlations between MFIs among the seven class I screening beads due to cross reactive epitopes.
We demonstrate that the avgMFI refines the ability to predict PTR. The odds ratio of PTR between the first and third quartiles of avgMFI was 33·8 (95% confidence interval: 6·6–173·1). In this data set, for avgMFI values >5440 the estimated probability of PTR was >90%. These results show that MFIs could be used to prospectively identify a subgroup of patients at high risk of developing PTR. Whilst not all patients developing PTR will be identified in this way, an area under the receiver operating characteristic curve of 0·86 indicates a clinically useful predictive ability, especially considering that multiple causes of refractoriness are operational in this study population.
However, for a number of reasons we have refrained from specifying cut-off values. Firstly, several different assay platforms are in use, and at present there is no validated method to harmonize results between them. Secondly, an optimal cut-off is dependent on the risk level that a clinical unit is prepared to accept, taking into account local costs of testing and of sourcing alternate methods for supporting allo-immunized thrombocytopaenic patients. Thirdly, we studied patients with acute leukaemia undergoing induction therapy as well as patients undergoing autologous transplantation. Individual laboratories may wish to include other high-risk categories (e.g. allogeneic stem cell transplant recipients, patients receiving long-term platelet transfusions) in their screening strategy, while avoiding prospectively testing autologous transplant recipients due to their relatively brief period of thrombocytopenia. Cut-offs are likely to be affected by the populations being considered. Finally, further work is needed to explore the dose–response relationship beyond a 2-week period, with serial testing in patients continuing to require platelet transfusions.
In conclusion, we provide evidence for the concept that PTR risk due to HLA allo-immunisation is usefully predicted by the MFIs of antibodies detected using micro-bead flow cytometry. If validated by studies using different assay platforms and screening protocols, this should enable haematology units to develop pre-emptive strategies for supporting allo-immunized thrombocytopaenic patients based on the estimated risk of refractoriness.
The authors would like to acknowledge the assistance of Dr Ben Saxon, Dr Magda Teague, and Ms Rhonda Holdsworth, of the Australian Red Cross Blood Service.
A.B, P.B, K.R, and K.D designed the study; E.T, I.H, and G.B performed antibody testing; A.B collected clinical data and performed statistical analysis; A.B, P.B, and K.R wrote the paper.
All authors declare that they have no conflict of interest. These results were previously presented in abstract form at the annual meetings of the Asia-Pacific Histocompatibility & Immunogenetics Association (Adelaide, Australia, 2012) and the American Society of Haematology (Atlanta, GA, USA, 2012).