Standard Compared With Extended Red Blood Cell Antigen Matching for Prevention of Subsequent Hemolytic Disease of the Fetus and Newborn: A Systematic Review

Abstract

Objective:

To systematically review and meta-analyze alloimmunization among recipients of red blood cells (RBCs) matched for ABO blood type and Rhesus D (ABO+D) antigen compared with those also matched for c, E and Kell (cEK).

Data Sources:

Four online databases (Medline, Scopus, Embase, ClinicalTrials.gov) were searched from March 28th 2023 to April 1^st 2024. The search protocol was peer-reviewed and published on PROSPERO (CRD42023411620).

Methods of Study Selection:

Studies reporting alloimmunization as primary outcome among recipients of RBCs matched for either ABO+D versus additional cEK matching were included. Patients transfused with unmatched RBCs or a mixture of matching regimens were excluded. Risk of bias was assessed using Cochrane ROBINS-I and ROB-2 tools. Random effects meta-analysis was used to combine effect estimates.

Tabulation, Integration, and Results:

Ten studies met criteria. Risk of bias was low. Overall 91,221 patients were transfused, of whom 40,220 (44.1%) received additional cEK-matched RBCs. The overall rate of alloimmunization was 6.2% (95% confidence interval [CI]: 2.5%, 14.9%) for ABO+D only matching and 1.9% (95% CI: 0.7%, 5.1%) when adding cEK. Time of follow-up antibody testing ranged from 6 to 18 months following transfusion. Additional cEK-match was associated with significantly less alloimmunization when compared with standard ABO+D match (OR: 0.37, 95% CI: 0.20, 0.69). This association remained when excluding chronically transfused patients (OR 0.65; CI: 0.54, 0.79), and for alloimmunization to c, E, or K antigens only (OR 0.29, CI: 0.18, 0.47).

Conclusion:

Additional cEK red blood cell matching protocols were associated with lower odds of recipient alloimmunization. Given severe sequelae of alloimmunization in pregnancy, routine cEK-matching for transfusion of people with pregnancy potential under age 50 in the United States merits consideration.

Systematic Review Registration:

PROSPERO, CRD42023411620.

PRÉCIS

Extending routine red blood cell antigen matching to include high-risk antigens Rhesus c, E, and Kell significantly reduces recipient alloimmunization.

INTRODUCTION

Alloimmunization can be associated with serious fetal and neonatal sequelae in pregnancy, including hemolytic anemia, thrombocytopenia, hydrops and death (1). Maternal sequelae can include intravascular hemolytic transfusion reactions, limited availability of compatible blood products for future transfusion needs, and an increase in healthcare costs (2). Globally, alloimmunization is estimated to affect between 0.36–3.4% of all pregnancies (3–9). Using data from 9.9 million pregnancies in the United States (U.S.) over a 12-year period (10), the prevalence of maternal alloimmunization was estimated at 1.5% of all pregnancies, an estimate higher than in Europe, United Kingdom (U.K.), Canada or Australia.

Of 360 known red blood cell antibodies, more than 50 are associated with HDFN (11). The most clinically significant in pregnancy are Rhesus D, c, E and Kell (12, 13). The vast majority of Rhesus D alloimmunization is thought to arise from transplacental feto-maternal hemorrhage, and implementation of programs for antenatal and postnatal anti-D immune globulin prophylaxis have led to significant reduction (14, 15). By contrast, the strongest risk factor for Rhesus c, E and Kell alloimmunization is allogeneic RBC transfusion, for which there is no similar and widely available prophylaxis (16).

U.S. national guidelines recommend that all blood products be routinely phenotyped for ABO blood type and Rhesus D. Additional antigen matching to include Kell, Rhesus c and E (cEK) has been typically reserved for patients at highest risk for chronic allogenic transfusion, such as those with sickle cell, thalassemia and myelodysplasia (17–19). Females under age 50, however, are at elevated risk of alloimmunization, and the consequences of alloimmunization if pregnancy occurs can be substantial (20, 21), prompting several high-income countries to offer routine cEK-matching in this population (22–28).

The authors hypothesize that alloimmunization with antigens conferring high risk for development of HDFN may be significantly reduced by using extended prophylactic matching with the addition of RBC antigens c, E and Kell. The objective of this study was to systematically review and meta-analyze incidence of alloimmunization in recipients of allogeneic blood transfusions who received exclusively red blood cells matched for ABO blood type and Rhesus D antigen (ABO+D) versus those receiving blood matched additionally for Rhesus c, Rhesus E, and Kell antigens.

SOURCES

We designed our study based on the Cochrane Guidelines for Systematic Reviews of Interventions and the Preferred Reporting Systematic Review and Meta-analysis (PRISMA) guideline for reporting systematic reviews (29, 30). A search protocol (CRD42023411620 – available on request) was registered with the International Prospective Register of Systematic Reviews (PROSPERO) database housed at the National Institute for Health and Care Research in the United Kingdom (31). Search terms were devised in English by the primary authors and an experienced medical librarian [LAS]. These search terms were validated against pre-selected articles, and searches were conducted, with input on keywords from content experts on the author team. Medical subject heading (MeSH) terms representing the concepts of red blood cell antigens, blood transfusion, and alloimmunization including autoantibodies, isoantibodies, and hemolytic disease of the newborn were used (32, 33). Details of MeSH terms used are available in Appendix 1, available online at http://links.lww.com/xxx. Database filters were used to remove publication types such as editorials, letters, and comments, and animal-only studies, as was appropriate for each database.

Searches were conducted using a computerized system on March 28^th 2023 and repeated on April 1^st 2024. The search strategy and results were independently peer-reviewed by a second librarian [SC] using a modified PRESS Checklist (34). All citations were imported into Covidence, a systematic review matching software (Veritas Health Innovation, Melbourne, Australia). Four scientific research databases (MEDLINE via PubMed, Embase, Scopus and ClinicalTrials.gov) were queried, using our independently-validated search terms. The full, reproducible search strategies and search results for all databases are included in Appendix 1.

STUDY SELECTION

Studies eligible for inclusion in our analysis included prospective or retrospective cohort studies reporting recipient alloimmunization as the primary outcome in patients receiving allogeneic red blood cell transfusion matched either for ABO+D only or additionally for c, E and Kell. Studies were excluded if they were non peer-reviewed, were systematic reviews, literature reviews or case reports, included non-human subjects, cohorts of less than 10 patients in each arm, patients transfused without a standardized transfusion matching protocol, patient transfused with both ABO+D and cEK matched blood products, patients in the cEK arm previously transfused with ABO+D unmatched for cEK without reported antibody testing in the interim, or if the study described a standard (control) match protocol extending beyond ABO blood type and Rhesus D, described an extended (intervention) protocol that did not include cEK, or duplicated reported outcomes from an already included patient cohort (35). For studies in which individual patient-level data were published, patients known to be alloimmunized prior to or by mechanisms other than RBC transfusion were excluded. For articles not published in English that met inclusion criteria during the title/abstract matching, abstracts were reviewed for usable data. Due to restrictions in funding, we chose not to translate non-English papers and they were excluded at the full text review phase.

After exclusion of duplicates, publication titles and abstracts were screened by four reviewers [RS, JO, MA, AS] working independently in teams of two, using Covidence software. The same four reviewers reviewed full texts, selecting those meeting criteria for inclusion in the systematic review and meta-analysis. Conflicts at each stage were resolved by an independent reviewer [JF]. Quality assessment was conducted by two independent reviewers [JO, MA] using the Cochrane Tool to Assess Risk of Bias in Cohort Studies (ROBINS-I) for non-randomized data and Cochrane Tool for Risk of Bias (RoB-2) for randomized data (36, 37). The ROBINS-I used predefined criteria to describe studies as low, unclear or high risk of study design bias in each of 8 distinct categories: Confounding, selection of participants, classification of interventions, deviation from intended intervention, missing data, outcome measurement, selection of reported results, and overall risk of bias. Conflicts were resolved by an independent reviewer [RS].

Two reviewers [RS, MA] performed data extraction using a standardized template on Microsoft Excel v2311 (Microsoft Corporation) and disagreements were resolved by consensus discussion. No automation tools were used in the data extraction process. Data extraction included descriptive data on country, year of publication, number of recipients assigned to each protocol, whether transfusion episodes were single or multiple, and frequency of occurrence of individual red blood cell antibodies. Data from studies reporting alloimmunization but not reporting antigen-specific results were included for the outcome of overall alloimmunization only. The primary outcome was alloimmunization among recipients of red blood cell transfusion from either an ABO-D only matching versus the addition of cEK matching. Secondary outcomes were alloimmunization arising from cEK alone and Kell alone. Data from included studies were manually tabulated into a Microsoft Excel template format in preparation for meta-analysis.

Because there was substantial heterogeneity in the incidence of alloimmunization across included studies, estimates of the proportion of patients affected by alloimmunization were estimated by a generalized linear mixed model, using a logistic link function (38). Odds Ratio was used as the reported effect measure for comparison of outcomes between ABO-D only matching compared with the addition of cEK matching; confidence intervals were used for each study and for the overall outcome to assess degree of certainty in the body of evidence as an outcome. Study heterogeneity was assessed using Higgins I² index, providing an estimate of results variability not due to chance. Combined effect estimates of donor match protocol were meta-analyzed using a restricted maximum likelihood (REML) random effects model to minimize type 1 error and account for study heterogeneity resulting from variation in study design and geography (39). Publication bias was examined graphically using a funnel plot and Egger’s test, and study design bias was assessed by two independent reviewers. Subgroup analyses to explore study heterogeneity included analysis of outcomes when excluding studies of chronically transfused cohorts, exclusion of the only randomized trial, analysis of alloimmunization to antigens with high-risk for HDFN (Rhesus c, Rhesus E and Kell), and alloimmunization to Rhesus D. Given the discrepancy in size between the largest cohort and the remaining 9 study cohorts, a further subgroup analysis was undertaken of the remaining 9 studies. All data synthesis and statistical analyses were performed using Stata Statistical Software, Version 18 (StataCorp, College Station, TX) and R version 4.4.1 (R Foundation for Statistical Computing, Vienna, Austria). This study was considered exempt from review by Duke University Health System Institutional Review Board.

RESULTS

Database searches yielded a total of 6,586 publication records. After exclusion of 2,788 duplicates, 3,798 publication titles and abstracts were screened. 139 texts were reviewed in full, of which 10 met criteria for systematic review and meta-analysis. Several studies initially appearing to meet inclusion criteria were excluded due to reportage of alloimmunization among cEK-matched recipients in the absence of a control arm of ABO+D-recipients, differing selection criteria for those undergoing ABO+D matching versus extended cEK matching, or extended match recipients who may have received K-match only (17, 18, 35, 40–43). A schematic flow diagram details the search as described in Figure 1.

Figure 1:

Flow diagram of study selection.

Studies included represented single- or multi-center patient cohorts performed in 8 countries on 4 continents, published between 1994 and 2017 (all studies bar one published since 2009). Nine represented retrospective cohorts and there was one prospective, randomized control trial. Six studies represented single-center cohorts and four described multi-center data. Seven studies included predominantly chronically transfused patients with hemoglobinopathies, such as patients with sickle cell disease, myelodysplasia or thalassemia; three studies described cohorts without hemoglobinopathies, the vast majority of whom were transfused in a single episode, including all patients in the only included randomized control trial. Two studies did not report sex of recipients, including the largest cohort, which also did not report age of recipients.

Overall risks of both publication bias and study design bias were low. Moderate bias was noted among several studies in the domains of confounding and selection of participants. Descriptive characteristics and risk of bias assessment of included studies are shown in Table 1 and Appendix 2 (Appendix 2 is available online at http://links.lww.com/xxx) (40, 44–52).

Table 1 –

Descriptive Characteristics of Included Studies

Source	Country	Years	Subjects	Mean Age in Years	Characteristics	Risk of Bias
Ameen	Kuwait	2000–09	233	28.5	Sickle cell disease	Moderate
Baia	Portugal	2011–13	633	69	Hematologic disease	Low
Cohen	USA	2001–15	87	7.5	Children with bone marrow failure	Moderate
Delaney	HICs	2007–16	631	56.8	Warm autoantibodies	Moderate
Hmida	Tunisia	1990–93	364	10.25	Children with hemoglobinopathy	Moderate
Lin	Canada	2001–14	176	72	Myelodysplasia	Low
Makarovska	Macedonia	2005–15	83,013	-	All patients	Low
Oud	Netherlands	2005–19	5,505	30.7	All patients	Low
Pujani	India	2009–14	484	2.5	Children with thalassemia	Low
Schonewille	Netherlands	2008–11	333	69.5	All surgical patients	Low

USA: United States of America; HICs: High-income countries; HDFN: Hemolytic Disease of Fetus and Newborn; IUT: Intrauterine Transfusion

In total, there were 91,221 unique transfusion recipients, of whom 40,220 (43.3%) received exclusively blood that was additionally screened for CEK and 51,001 (56.7%) who received blood screened for ABO+D only. At least 91.4% of ABO+D recipients and 88.8% of additional cEK-matched recipients were transfused on a single occasion only. Among studies reporting sex and age, 44.9% were male, 55.1% female and mean age was 35.8 years. Time of follow-up antibody testing ranged from 6 to 18 months following transfusion.

Using a generalized linear mixed model to account for study heterogeneity, the overall rate of alloimmunization was 6.2% (95% confidence interval [CI]: 2.5%, 14.9%) for ABO+D only matching and 1.9% (95% CI: 0.7%, 5.1%) when adding cEK. Comparing ABO+D matching protocol to those additionally matching cEK, the incidence of any alloimmunization was significantly reduced with addition of cEK match (OR 0.37; 95% Confidence Interval [CI]: 0.20, 0.69; Figure 2). Significant study heterogeneity was observed (Higgins I2 index of 84.18%, p<0.01). A funnel plot of effect estimates versus study size demonstrated mild asymmetry (Figure 3, Egger test p-value of 0.35), indicating low potential for publication bias, with most studies falling within the expected range.

Figure 2:

Forest plot of all alloimmunization associated with additional Rhesus D, c, E and Kell (cEK) match compared with ABO blood type and Rhesus D (ABO+D) match protocols.

Figure 3:

Funnel plot for primary study outcome.

In subgroup analyses, reduction in overall alloimmunization associated with additional cEK-matching remained significant when excluding cohorts of chronically transfused populations (OR 0.65; CI: 0.54, 0.79; Figure 4). The incidence of alloimmunization with antibodies known to be high-risk for development of HDFN (Rh c, E, Kell) alone was also reduced significantly with additional cEK matching compared with ABO+D (OR 0.29, CI: 0.18, 0.47; Figure 5) and for Kell alloimmunization alone (OR 0.36; CI: 0.21, 0.62). These effects on cEK alloimmunization were again maintained when excluding studies of chronically transfused populations (OR 0.37, CI: 0.19, 0.72) (Appendix 3, available online at http://links.lww.com/xxx). There was no statistical difference noted in incidence of alloimmunization to Rhesus D antigen between ABO+D and addition of cEK-matching protocols (OR 0.19, CI: 0.02, 1.67) (Appendix 4, available online at http://links.lww.com/xxx).

Figure 4:

Forest plot of alloimmunization associated with additional Rhesus D, c, E and Kell (cEK) match compared with ABO blood type and Rhesus D (ABO+D) match protocols when excluding chronically transfused patient cohorts.

Figure 5:

Forest plot of Rhesus D, c, E and Kell (cEK) alloimmunization associated with additional cEK match compared with ABO blood type and Rhesus D (ABO+D) match protocols.

DISCUSSION

In this systematic review and meta-analysis comparing an ABO+D donor-recipient match protocol versus extended antigen matching protocols that also included Rhesus c, E and Kell, cEK-matching was associated with lower odds of new alloimmunization. Reductions in alloimmunization were particularly marked among alloantibodies conferring the highest risk for future development of hemolytic disease of the fetus and newborn (HDFN). It is notable that these effects were maintained when excluding studies of chronically transfused patient populations. Strengths of our review include our study selection criteria, a clear measurable outcome of recipient alloimmunization, low publication and design bias, and a large, geographically diverse, final pooled cohort.

Our study has several significant limitations, including limited available peer-reviewed original studies that met inclusion criteria, differences in antigen matching protocols, testing protocols, and timing of follow-up antibody testing. Several studies included chronically transfused populations, results of which may have limited applicability to routine extended cEK matching. A subgroup analysis that excluded these studies, however, demonstrates a similarly significant reduction in alloimmunization associated with cEK matching in general and singly transfused populations only (Appendix 3, http://links.lww.com/xxx).

Only one randomized controlled trial was eligible for inclusion (Schonewille et al), a large, multicenter trial in the Netherlands of ABO+D versus an additional matched cEK protocol in patients scheduled for a single elective blood transfusion and followed up with antibody tests at 3 fixed timepoints over 6 months (51). This population has inherent dissimilarities to a nationwide population in the United States, demographically, in terms of healthcare access, and with resources for routine cEK matching readily available in each center of study associated with a centralized national blood bank. Nonetheless, we believe the findings of this well-designed, prospective study retain significance for patients outside the Netherlands. We also note that conducting a population-based, randomized, controlled trial in larger countries where blood banks are decentralized and centers inadequately resourced to provide routine cEK matching for the purposes of study may not be feasible.

Moderate bias was noted in the domains of both confounding and participant selection in several studies. These biases, however, were not present in either the largest cohort studies nor studies of non-chronically transfused patients, in whom a significant reduction in alloimmunization was noted in subgroup analysis. Noting the potential for heterogenous study design or sample size variance to distort our effect estimates, we were reassured to find that the significant reduction in alloimmunization was maintained both when excluding the only randomized control trial (OR 0.39, CI: 0.20, 0.77) (Appendix 5, available online at http://links.lww.com/xxx), and when excluding the largest patient cohort (OR 0.32, CI: 0.16, 0.67) (Appendix 6, available online at http://links.lww.com/xxx) (51, 52).

A limitation of our review process was lack of individual patient-level data, limiting our ability to account for potentially missing data. Though this prevented us from stratifying alloimmunization by sex and age, the results of several individual studies were notable. In the only study in which individual patient data were available, the randomized controlled trial by Schonewille et al. demonstrated that the reduction in alloimmunization associated with cEK was primarily significant in females (51). Only one study – a retrospective study of patients with myelodysplasia at a single center – suggested a negative effect of cEK-matching (46). Aside from being the smallest cohort with less than 20 patients in the cEK arm, new alloimmunization occurred in only 1 patient in each group. In addition, it was noted that the mean number of units transfused to the ABO+D group was 18 (median 9 units), versus 134 (median 20 units) among the cEK group. Given that these findings suggested that provision of cEK-matched blood was effective for preventing alloimmunization, the authors instituted a policy that all such patients be transfused with cEK-matched blood in future.

The most notable study that did not meet eligibility criteria for inclusion in our study was a frequently cited, multinational, retrospective study of females with offspring with severe HDFN by Delaney et al. (35). In this study, the authors compared maternal alloimmunization in centers matching for ABO+D only versus centers offering an extended antigen-negative protocol. No significant difference in alloimmunization was seen. It was noted, however, that the majority of patients included (83%) were thought to have been sensitized by a prior pregnancy rather than transfusion. In the minority attributed to transfusion, alloimmunization was significantly lower in those receiving cEK-matched blood (6.1%) versus ABO+D (35.3%). This study was ultimately excluded from our final analysis due to a large proportion of patients previously receiving transfusions outside the centers of study, and only a minority of the participating centers offering extended antigen-match protocols that included all three high-risk antigens Rhesus c, Rhesus E and Kell.

The prevalence of high-risk antibodies for HDFN in the U.S. including Rhesus c, E and Kell has consistently risen since 2010 (10). While alloimmunization with Rhesus c, E and Kell does not always result in development of HDFN, the consequence when it does occur can be devastating (53). U.S. transfusion guidelines recommend routine ABO+D antigen match only, with prophylactic cEK-match reserved for patients with hemoglobinopathies (19). This now contrasts sharply with most other high-income countries: In Europe, where routine cEK-matching is recommended for all blood donors in several countries (26), in Canada, where all donors are routinely phenotyped for Kell and approximately 40% are additionally matched for Rhesus c and E (28), and in Australia, where all are phenotyped for Kell (54). These policies have been associated with significant reductions in overall alloimmunization, transfusion-associated cEK alloimmunization, and Kell-mediated HDFN in reproductive age females (40, 41, 55).

The most likely reason routine cEK matching has not yet been adopted in the U.S. are concerns regarding cost-effectiveness. Formal analyses of cost-effectiveness are lacking. Though expense has not been a barrier to recommending this approach in other high-income countries, many of these countries have several important differences with respect to the U.S. These include common funding for healthcare institutions and blood banks, economy of scale, and centralized bargaining power for equipment and staffing (56–58). It is also notable that most of these countries additionally recommend routine transfusion of Kell-negative blood to female people under age 50, and several have recently piloted routine genotyping for cEK, any of which may have contributed to observed reductions in alloimmunization. (27, 28, 54, 59–62). Nonetheless, the United States is now an outlier among high-income countries in not recommending routine extended antigen matching in people with pregnancy potential.

In conclusion, our systematic review and meta-analysis of the literature suggests that extending routine donor-recipient matching to include red blood cell antigens Rhesus c, E and Kell (cEK) was associated with decreased odds of recipient alloimmunization when compared with standard ABO+D matching protocol. This was particularly true for alloantibodies high risk for development of hemolytic disease of the fetus and newborn such as Rhesus c, Rhesus E and Kell. Given the rising prevalence of maternal alloimmunization with high-risk alloantibodies, a comprehensive cost-effectiveness analysis of routine donor-recipient cEK-matching in this population in the United States is now warranted. Additional studies of Kell-only matching, universal Kell-negative blood transfusion, or RBC antigen genotyping may also be considered. These findings may ultimately inform changes to national guidelines on protocols for donor-recipient matching of blood products prescribed to people with pregnancy potential in the United States and elsewhere.

Data Availability:

For template data collection forms, original data extracted and used for analysis, and analytic code please contact corresponding author as above.