Eleven years of alloimmunization in 6496 patients with sickle cell disease in France who received transfusion

Sickle cell disease (SCD), affecting an estimated 19 800 to 32 400 patients in 2016,¹ is the most frequent genetic disease in France. Sixty percent of patients with SCD (pSCDs) reside near Paris,¹ an administrative region with >12 million residents and 200 hospitals. Transfusion is a major therapy, but patients are at risk of forming antibodies to red blood cell (RBC) antigens.^2-4 Phenotype mismatch is frequent when patients are mainly of Afro-Caribbean descent but donors mainly of European descent: rare phenotypes differ⁵ as does the distribution of common antigens, with a lower prevalence of C, E, Fy^a, Fy^b, Jk^b, and S antigens in Afro-Caribbeans.⁶ However, alloimmunization remains high when donors are of African descent.^7,8 Partial antigens, that is, variations of common antigens, primarily D, C, and e in the Rh blood group system, present a risk of immunization if carriers are exposed to the conventional antigen,^9-11 and low-frequency antigens (LFAs) in reference to the European population have a higher prevalence among Afro-Caribbeans.^5,12,13 Beyond this diversity, inflammatory status of pSCD, especially during acute episodes, promotes erythrocyte alloimmunization.¹⁴ Alloimmunization is a well-known risk factor for additional antibody formation¹⁵ and can lead to delayed hemolytic transfusion reactions (DHTRs), sometimes with life threatening hyperhemolysis.¹⁶ 

The American Society of Hematology recommends selecting RBC units (RBCus) matched for D, C, E and K antigens for all pSCDs and suggests (with low certainty) extending compatibility to the FY, JK, and MNS systems for alloimmunized patients (also called high responders¹⁰) when there is a high risk of hyperhemolysis.¹⁷ Because of the limited treatment options for DHTR, an increasing number of clinicians request prophylactic matching for the extended phenotype for patients who are nonimmunized and at high risk of DHTR because evanescent antibodies may be missed by antibody screening tests. In most settings, such matching is not possible for all patients, and a major challenge in blood bank inventory management is reserving rare extended phenotype RBCus from Afro-Caribbean donors for patients who would benefit from them the most. Patients under transfusion exchange programs require a large number of units but rarely experience hyperhemolysis.^16,18 Little is known about the chronology of antibody formation in patients, which limits the ability to adapt the unit selection strategy on a case-by-case basis.

Since 1998, pSCDs in France received transfusion with leuco-reduced RH-K–matched RBCus (including c and e), after serological crossmatch.^19,20 Antigen-negative units are selected prophylactically to heed partial RH antigens,^9-11 and immunized patients receive units compatible for the cognate antibodies. Matching for LFAs such as C^w, Lu^a, Js^a, V/VS, and Kp^a is not mandatory.^19,20 Patients who are considered at a high risk of hyperhemolysis because of a history of hyperhemolysis or a history of immunization to a high frequency antigen (HFA) or an antigen of the extended phenotype (AEP), FY, JK, and Ss, receive RBCus matched for that AEP. Since 2008, all transfusions are recorded in a single, centralized, regional database by Etablissement français du sang. Thanks to universal health care and national guidelines,²¹ patient management is relatively homogenous.

The study aimed to study the evolution of RBCu consumption for pSCD, frequencies of antibodies and antibody associations, identify risk factors for additional antibody formation using the chronology of appearance of antibodies, and identify characteristics of high and low-responder patients.

We conducted a retrospective analysis of patients (SS, SC, or Sβ) who received transfusion with at least 1 RBCu between 2008 and 2018 in the Paris area. Previously known patients and patients entered in the database for the first time during the period were eligible. We collected data on patient demographics, patient and donor unit antigen phenotypes, transfusions (unit count, dates, and number during the period and units received before 2008), and antibodies (dates and specificity).

Within the network of laboratories using the database, antibody screens were performed by gel testing in the indirect antiglobulin test at 37°C, with additional testing with enzyme-treated panel cells when necessary. Antibody screens are mandatory in the 72 hours preceding each transfusion, at each patient visit or hospitalization, and prescribed 1 to 3 months after each transfusion. Individual antibody screens were not available to be reviewed but interpretation is expected to be consistent between the laboratories because they follow the same standard operating procedures. Warm autoantibodies (WAAs) are typically panagglutinations in the absence of recent transfusion, with positive autocontrols and a positive direct antiglobulin test or autoreactive antibodies directed at an antigen expressed by the patient (such as auto–anti-D or auto–anti-Jk^a), which can be eluted from or adsorbed onto the patient’s own RBCs. Because pSCDs often receive transfusions, the percent of hemoglobin A, when available, can be helpful for the interpretation of WAAs. Antibodies of undetermined specificity (AUSs) may be reported either for previously identified antibodies not characterized to save resources (eg, weakly reactive anti-E) or because of a lack of appropriate panel cells or for emerging antibodies. Cold autoantibodies are not routinely detected but are not predictive of additional or clinically significant antibody formation.²² 

Antibody subgroups were defined to facilitate the analysis and interpret data in broader terms. RH-K antibodies were anti-D, anti-C, anti-E, anti-c, anti-e, and anti-K. Antibodies to AEP were anti-Fy^a, anti-Fy^b, anti-Jk^a, anti-Jk^b, anti-S, anti-s, and anti-N. Anti-M, anti-C^w, and antibodies to Lewis antigens are generally naturally occurring antibodies (NOAs) and were considered as a group. Antibodies to HFAs and LFAs, AUSs, and WAAs (with or without an identified specificity) were considered as separate groups.

Transfusion indications were not available. To approximate patients who received transfusion in acute situations and during transfusion exchange programs, we defined 2 groups: patients who occasionally received transfusions were those who received a total of ≤12 units over the study period (occacionally transfused group) whereas patients who received multiple transfusions (multitransfused group) were those who received ≥50 units.

The analysis was performed in 3 steps, as shown in the study flowchart (Figure 1). Statistical analysis was conducted with SAS software 9.4, and significance set at P = .05 for this exploratory study.²³ Step 1 focused on RBCu consumption of all patients over the study period.

Step 2 investigated patients with antibodies to RBC antigens, antibody distribution and prevalence, order of antibody formation, and association between antibody subgroups. Patients who received transfusion or were immunized before universal leukoreduction (1998) were excluded. We compared immunization based on sex at 3 moments in life: childhood (aged 0-18 years), childbearing years (aged 18-50 years), and later (aged >50 years) using log-rank test. Because previous studies found that receipt of their first transfusion after the age of 5 was associated with higher immunization rates,²⁴ we compared immunization in patients who received their first transfusion at different ages. χ² with Yate correction for continuity, when small numbers required it, or exact Fisher tests were used to compare the order in which antibodies (or groups of antibodies) appeared and the risk of subsequent antibody formation, determine whether new specificities were influenced by previous specificities, and compare patients who occasionally received transfusions and those who received multiple transfusions.

Step 3 was a longitudinal analysis of patients who made their first recorded antibody after 2008 and after their first transfusion to reduce bias introduced by older data and antibodies unrelated to transfusion. The number of units that patients received before their first positive antibody screen was analyzed as Kaplan-Meier survival curves, in which each event was the first immunization of a patient.

The protocol was submitted and approved by Etablissement français du sang. Registration on the Health Data Hub platform was not required for this fully anonymized data set, in accordance with French legislation and as confirmed by the Health Data Hub. The study was conducted in accordance with the Declaration of Helsinki.

From 2008 to 2018, 6496 patients received transfusion with a total of 239 944 RBCus. Patients received from 1 to 1372 units. Each year, an average of 1806 patients received transfusion with a median of 7 units (Figure 2). The 3 main pediatric and 5 main adult hospitals transfused 78.9% of units, whereas 59.6% (n = 106) of the remaining hospitals had <1 patient per year receive transfusions, and 22.5% (n = 40) had <10 patients per year receive transfusions. Patients were committed to their transfusion center, with 87.7% patients receiving transfusions at a single or 2 facilities over the 11-year period (supplemental Data). Mean age at first transfusion was 15.3 years (range, 0-84 years), and the median age was 10.4 years (interquartile range [IQR], 3-25). Sex ratio was 0.8 (male to female). Comparison of 2008 and 2018 as per the age at transfusion shows an increase in the total RBCu count for all age groups and an important increase in mean units per patient for adults aged from 26 to 55 years (Figure 2).

Patient phenotypes were as expected, with a high prevalence of O (48.1%), B (23.3%), D⁺C^–E^–c⁺e⁺K^– (59.4%), Fy (a^–b^–) (90.6%), Jk (a⁺b^–) (58.2%), and S^–s⁺ (73.7%) and comparatively high prevalence of S^–s^– (U^– or Uvar⁺; 1.6%).^6,25 Definite exposures to Fy^a, Fy^b, Jk^a, and Jk^b antigens for patients who were antigen-negative ranged from 23.2% to 32.0%, with an additional unknown number of exposures for ∼55% of RBCus with untested extended phenotype (Table 1). Patients were less likely to be exposed to the cognate antigen if they had S^– phenotype (18.3% of the units transfused to patients with S^– phenotype were of S⁺ phenotype) but more likely to be exposed if they had s^– phenotype (40.2%). Patients (n = 476) who received transfusion or were immunized before universal leukoreduction (1998) were excluded from subsequent steps of the analysis.

In total, 1742 patients made antibodies (Figure 3). The median age at first antibody formation was 12.2 years (IQR, 5.8-26.1). The distribution of the total number of units received per patient was different between immunized and nonimmunized patients (P < .0001; Figure 2D): as expected, immunized patients had received more RBCus. Figure 4 shows that age and sex had an effect on immunization. Women of childbearing age formed antibodies earlier than males (P = .0049), but there was no difference between male and female during childhood (P = .1690) or later in life (P = .7352). Age <5 years at first transfusion was weakly associated with earlier immunization (P = .0395).

Patients who made any antibody were more likely to make additional antibodies and 1742 of 6020 patients were immunized (29.8%). Of these, 900 patients made 1 antibody, and 842 (48.3%) made multiple antibodies (n = 412 made 2 different antibodies, n = 224 made 3, n = 107 made 4, and n = 99 made ≥5 different antibodies). Patients made antibodies with between 1 and 14 different antibody specificities. (Figure 3; supplemental Results). The most frequent antibodies were WAAs (n = 873), AUSs (n = 800), anti-C (n = 289), NOAs (280 anti-M, 247 anti-Le^a, and 131 anti-Le^b), anti-E (n = 188), anti-S (n = 181), anti-Jk^b (n = 126), anti-D (n = 102), and antibodies to LFAs (n = 108 anti-Kp^a, n = 61 anti-C^w, and n = 55 anti-Lu^a).

Figure 5 shows that the first antibody to appear is predictive of subsequent immunization. Patients who made a WAA or NOA at their first immunization were less likely to develop a second antibody compared with patients who made another type of antibody first. Patients whose first immunization included an antibody to RH-K, an AEP, or HFA were at a higher risk than other patients of developing a second antibody. The distribution of antibodies after a WAA or RH-K antibody was not random (P < .0001). After a WAA, the second antibody was more likely to be an AUS than when the first immunization did not include a WAA; an RH-K antibody was more likely to be followed by an AUS or a WAA than when the first immunization was not to RH-K. Other associations did not reach statistical significance. The immunization risk when the first immunization included an AUS or antibody to an LFA was not significantly different from that of patients who made another type of antibody first, but the most frequent second antibody after an AUS was a WAA (n = 79).

Chronology of transfusion and immunization analyzed in 643 patients is shown in Table 2. Female patients made their first antibody slightly earlier than males, after 6 units (IQR, 2-14), vs 7 (IQR, 3-21) for males (supplemental Data). Antibodies to HFA and AEPs were made earlier than other antibody specificities, after 6 (IQR, 3-10) and 5 (IQR, 3-20) units, respectively (Table 2; Figure 6). The 1255 patients who had not made an antibody by the 20th unit received an additional 118 796 RBCus, ranging from 1 to 1277 units per patient (mean, 94.7; median, 33) in addition to the 20 units already received; only 71 of 1255 patients (5.7%) subsequently developed clinically significant antibodies (to RH-K, HFA, LFA, or AEP).

Between 2008 and 2018, alloimmunization prevalence was higher in patients who received multiple transfusions (≥50 units; 156 of 660 patients; 23.6%), likely to be in exchange transfusion programs, than in patients who only occasionally received transfusion (≤12 units; 259 of 3451 patients; 7.5%; P < .0001), likely for acute situations. The distribution of antibodies was different between the 2 groups, with more WAAs, AUSs, or antibodies to LFAs but fewer NOAs and antibodies to an AEP or HFA (P < .0001) in the multitransfused group (supplemental Data). The number of immunization occurrences (dates when new antibody specificities were detected) and number of immunized patients by the 12th RBCu were not significantly different (supplemental Data).

This, to our knowledge, is the largest retrospective study on alloimmunization of pSCDs to date,^7,24,26-29 in a population consistently receiving prophylactically RH-K–matched units from a predominantly European donor population. The aim of the study was to study chronology and number of units received before antibody formation and trends for similar antibodies.

Units transfused to pSCDs in 2022 reached twice the number transfused in 2008, accounting for nearly 10% of all RBCus transfused in the Paris area. The notable increase in units transfused to patients aged ≥26 years suggests that the rise in unit consumption is driven by an extended life expectancy and more intensive treatment in these age groups. The major increase in RBCu consumption after 2015 follows 2 studies that showed that chronic organ damage, strokes, and cardiovascular–related complications were a major cause of mortality among pSCDs from 1979 to 2010 in France.^30,31 French guidelines published in 2015 promoted transfusion programs²¹ to prevent these complications.^32-34 Limitation of avoidable transfusions such as preoperative transfusions³⁵ does not counterbalance the number of units required for transfusion exchange programs, with automated exchanges requiring even more units per exchange.^34,36 

Patients form their first antibodies after only a few transfusions.^24,29,37 When all antibodies or antibodies with a determined specificity (excluding WAAs and AUS) were considered, 75% of patients made their first antibody by the 17th RBCu. Excluding only WAAs, it was by the 20th RBCu. Antibodies to HFAs and AEP were made even earlier, with 90% of first antibodies to HFAs developed by the 16th RBCu, suggesting that these antigens are highly immunogenic. This range is similar to observations in smaller cohorts²⁴ and is only a few units more than the previously established threshold of 12 units for increased DHTR risk.¹⁸ Patients not immunized after the 20th RH-K–matched unit were likely to be low responders, that is, at a low risk for alloimmunization. Until the 20th unit, patients and their antibody screens should be monitored closely. Immunization events followed an exponential decay curve, and only in rare cases did patients appear to be less sensitive but capable of alloimmunization.

Many variables may influence alloimmunization risk, such as age and sex. Women of childbearing age have increased risks of alloimmunization due to pregnancies.^38,39 Contrary to young age being a protective effect, as previously reported,²⁴ patients aged ≤5 years at first transfusion appeared to make antibodies earlier. However, the similarity between the curves for the 2 age groups, a P value close to .05, and the absence of difference when other age limits were tested (P = .2283 for patients whose first transfusion was before or after the age of 3; supplemental Data) suggests that the effect may not be major.

Future, prospective studies may allow for adjusting the threshold of 20 units observed in this cohort, and propose an individual risk score with additional variables. Factors that affected the detection of antibodies in this cohort and could be better accounted for in a prospective, controlled study include antibody evanescence (failure to detect the antibody in the optimal time frame) and true exposures to each antigen (eg, a random unit may or may not be Fy(a⁺)). Innate and external factors have been shown to influence the risk of alloimmunization^40,41 and could be considered as additional variables for a personalized score, including inflammation¹⁴ and transfusion indications. Transfusions in acute situations promote alloimmunization more than those in transfusion programs^7,24,26,28 while being at the highest risk for DHTR with hyperhemolysis^18,42 (DHTR events are recorded in another national database and could not be extracted for this study).

Our findings also suggest a link between transfusion in acute situations and clinical significance of antibodies: although patients who received multiple transfusions have higher cumulative RBC antigen exposure, patients who received only occasional transfusions made more antibodies with specificities that are considered clinically significant. The difference was not due to the immunization rate, which was similar after the same number of units had been transfused. Multiple RBCus transfused together can expose a patient to multiple RBC antigens, but the alloimmunization risk associated with the transfusion of 4 Jk(b+) RBCu, for example, may not be very different from the transfusion of a single Jk(b+) unit.⁴³ We found no evidence in this cohort for an increased risk of alloimmunization when multiple units were transfused within a short time or, conversely, spread over time (supplemental Data).

Antibodies to RH-K, LFAs, and AEPs in the cohort were interesting. Despite French¹⁹ and international¹⁷ recommendations match RH-K for pSCDs, n = 515 (29.6%) of the immunized patients (7.9% of the entire cohort) made RH-K antibodies. This is lower than that of other cohorts^7,26 and represents a decrease compared with our earlier observations,^9,11 probably partly because of the relatively recent generalization of prophylactic selection of antigen-negative RBCus for partial antigens (except for RHCE∗ce733G),^9-11 which have a high prevalence in pSCDs.^7,26 Some anti-C may be unrecognized, mimicking antibodies and not true alloantibodies.^44,45 

Few antibodies to LFA typically expressed by individuals of African descent (V, VS, Js^a, etc)¹³ were found. Such antibodies are evanescent, only exceptionally represented on panel cells in France and may go undetected^13,46 but may represent up to one-third of alloantibodies in pSCDs who received transfusions with units from African donors.¹² Conversely, many anti-C^w, -Kp^a, and -Lu^a were detected. Although anti-C^w can be naturally occurring, these numbers probably result from the presence of these antigens, typically expressed by European individuals, on routine panel cells and frequent transfusion of units from European donors. Donor ethnicity cannot be recorded or used for unit selection. The rarity of D⁺C^–E⁻ donors and prophylactic use of D^– units for partial D often require resorting to D^–C^–E⁻ units (likely from European donors) for the transfusion of patients who have D⁺C^–E⁻ phenotype.⁴⁷ Patient who have D⁺ phenotype with partial C may also receive D^–C^–E⁻ for a lack of D⁺C^–E⁻ units. In this study, 29.0% of the RBCus transfused to patients with D⁺ phenotype were of D^– phenotype. This may contribute to the lower proportion of anti-D in this cohort than in others^7,26 but put a strain on inventory.

The observation of antibodies to AEP were of particular interest to us, because clinicians are increasingly requesting units matched for nonimmunized patients’ extended phenotypes. AEPs that pSCDs most often lack are Fy^a, Jk^b, and S, but units negative for all 3 are rare. Comparing exposures to immunizations in this cohort appeared consistent with a higher immunogenicity for S and Jk^b, although relative immunogenicity of all and the most immunogenic antigens is difficult to establish and varies depending on the methodology.^48,49 In this study, patients with Fy(a^–) or Jk(b^–) phenotype were exposed to the cognate antigen more than patients with S^–, although extended phenotypes were not known for half the units (a previous study determined that when the extended phenotype was determined for all units transfused to nonimmunized pSCDs, the likelihood of patients who were Fy^a–, Jk^b–, or S^– being exposed to each of the cognate antigens was ∼55% per RBCu and per antigen).⁵⁰ However, anti-S was the most frequent antibody to an AEP (some of which may be explained be naturally occurring antibodies²⁵), as reported in previous SCD cohorts,²⁷ whereas anti-Fy^a has been detected more often than anti-Jk^b in other cohorts.^7,26 This difference compared with our cohort could be due to evanescence or different donor populations. Antibodies to Jk antigens are known to be evanescent^46,47,51 but common causes of DHTR.^6,25 Anti-Fy^a may be a precursor of anti-Fy3 or anti-Fy5,²⁵ but this did not seem to be the case in this cohort. One in 3 patients with anti-Fy^a made anti-Fy3 or anti-Fy5; n = 66 made anti-Fy^a but no anti-Fy3/anti-Fy5; n = 40 made anti-Fy3/anti-Fy5 but no anti-Fy^a; and n = 33 made anti-Fy^a and anti-Fy3/anti-Fy5, among which n = 18 made the antibodies concomitantly or within days or weeks of one another, but only n = 7 made anti-Fy^a months or years before anti-Fy3/anti-Fy5. Overall, anti-Fy^a is only rarely a precursor to anti-Fy3 or anti-Fy5. Our observations encourage matching primarily for S and Jk^b antigens.

All immunized patients had an increased risk of subsequent antibody formation and antibody specificities at first immunization were predictive of future antibody formation. Pathophysiology to explain why some antibody specificities may precede others is an active field of fundamental research. A recent study in mice showed that immune priming of T cells to an intracellular antigen can influence subsequent alloantibody formation to a surface RBC antigen.⁵² Whether the mechanism could be the same in humans remains to be investigated. In our study, the immunization risk after HFA was the highest. First antibody specificities associated with a lower risk (but still higher than pSCD with no antibodies) were WAAs²⁸ and NOAs. We found a link between AUS and WAAs in our cohort, likely because of AUS being reported when a WAA has become too weakly reactive to agglutinate all panel cells. Similarly, RH-K antibodies were likely to be followed by an AUS, probably reported as an RH-K antibody with decreased reactivity (regulations require 3 reactive cells to confirm specificity). NOAs, especially anti-M, are not uncommon in children with SCD,^53,54 often unrelated to transfusion and stimulated by infections.^25,55 These antibodies are considered of low clinical significance in most patients but may rarely cause DHTR,^6,56 and caution is warranted when such antibodies are reactive at 37°C or patients have a history of hyperhemolysis.⁵⁷ Any antibody has the potential of causing DHTR with hyperhemolysis in pSCDs.^6,56,58 

In practice, our findings support adjusting transfusion strategies (in addition to RH-K matching) based on the number of units received and antibody specificities in immunized patients. Table 3 illustrates our proposed strategy for phenotype matching in pSCDs. Although prophylactic RH-K matching will always be necessary and decisions must also be taken on a case-by-case basis, we would generally discourage introducing extended phenotype matching after the 20th RBCu or when only WAAs, anti-M, anti-Lewis, antibodies to RH-K, or AUSs have been identified, in the absence of DHTR history. History of DHTR, regardless of antibodies identified^17,59 and risk factors of DHTR,¹⁸ should always be taken into account. When extended phenocompatibility is required but inventory strained, we suggest primarily matching for S and Jk antigens.

This macroscopic, retrospective study of alloimmunization in a real-life setting has limitations. First, RBCu counts extracted from the regional database: some patients might have received additional transfusions in other regions, during travel or when abroad.

Second, in the detection of antibodies: all transfusions, but not all antibody screenings, are recorded in the regional database, and some antibody screens might have been performed outside of the network. This is likely a relatively minor issue, because most positive screens will be forwarded for antibody identification or crossmatch to larger laboratories using the regional database. The network’s laboratories use the same standard operating procedures for antibody detection and reporting, but variability in interpretation may exist, and specificities of AUSs and antibodies to LFAs are not systematically identified. The dates of screening tests were not controlled, and some antibodies might have been missed because of evanescence or if patients missed the recommended screening test 1 to 3 months after transfusion. The analysis spans 11 years of data, but patients may develop additional antibodies subsequently.

Third, findings are influenced by French practices that go beyond current international guidelines: the selection of units matched for c and e, prophylactic selection of antigen-negative units for most partial Rh antigens, and extension of phenotype matching in nonimmunized patients with a high risk for hyperhemolysis, whenever possible. Conversely, patients might have exceptionally undergone transfusion, not following the guidelines, for example, for urgent care at a new facility that is not aware that the patient has SCD. Thresholds found here may vary depending on donor and recipient populations. When less extensive matching is provided, adhering exactly to international guidelines, lower thresholds for immunization are expected,²⁴ and thresholds that we propose here are likely to be conservative.

Lastly, the statistical analysis for this exploratory study was not corrected for multiple testing,²³ and some of the associations might have occurred by chance. Limitations are expected in a retrospective analysis, and further monitoring of alloimmunization in pSCDs in different settings will continue to be necessary.

The authors thank Pavlos-Aimilios Marinatos and Brigitte Bonneaudeau (Etablissement français du sang, Saint-Denis, France) for assistance with the ethical requirement procedures. The authors thank Connie M. Westhoff, Christine Lomas-Francis (New York Blood Center, NY), and Stella Chou and Stacey Uter (Children’s Hospital of Philadelphia, PA) for discussion of their experience of anti-Kp^a alloimmunization in pSCDs. The authors also thank Corinne Pondarré (Centre Hospitalier Intercommunal de Créteil, France) and Bérengère Koehl (Hôpital Robert Debré, Paris, France) for discussion of their experience in performing transfusions for children with SCD and Suella Martino (Referral Center for Sickle Cell Disease, Henri Mondor Hospital, Créteil, France). Moreover, the authors thank Christophe Tournamille (Etablissement français du sang Créteil, France) for discussion of Rh variants and management of pSCDs; and the late Catherine Collet (Etablissement français du sang Ile-de-France) for discussion of study design.

Contribution: A. Floch, S.V., A. Francois, and F.P. designed the research; S.V., S.P., A.J., and A. Francois collected and curated the data; A. Floch, S.V., L.M., A.H., F.G., and F.P. analyzed and interpreted the data; and A. Floch and F.P. prepared the original draft of the manuscript, which was critically reviewed and approved by all authors.

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

Correspondence: France Pirenne, Etablissement français du sang, Hôpital Henri Mondor, 51 Ave du Maréchal de Lattre de Tassigny, 94000 Créteil, France; email: france.pirenne@efs.sante.fr.