Use of an anti-D-alloimmunization kinetics model to correct the interval censored D-alloimmunization rate following red blood cell transfusions

1 INTRODUCTION

The rate of D-alloimmunization following the transfusion of RhD-positive red blood cells (RBC) to RhD-negative individuals has been studied for many years. Early investigations that involved administering small quantities of RhD-positive RBCs to healthy volunteers found D-alloimmunization rates of between approximately 50% and 90%.¹ These studies were performed largely to determine the optimal dose and frequency of administering RhD-positive RBCs to volunteers for the purpose of producing Rh immunoglobulin (Ig). Although these high D-alloimmunization rates were derived from healthy volunteers, the routine clinical practice at many hospitals was to administer only RhD-negative RBCs to RhD-negative patients and those whose RhD-types were unknown when urgent transfusions were required. This was especially true for females of childbearing potential because of the possibility that D-alloimmunization could cause hemolytic disease of the fetus and newborn (HDFN) in a future pregnancy.²

However, the practice of administering RhD-positive RBCs to RhD-negative individuals, despite the risk of alloimmunization, can occur in massive or emergency transfusion situations.^3-5 Recent interest in the consequences of transfusing RhD-positive RBCs to RhD-negative patients has intensified, particularly with the resurgence of civilian low titer group O whole blood (LTOWB) transfusions.^6-9 It has been recently recognized that the rate of D-alloimmunization following the transfusion of RhD-positive RBCs to hospitalized RhD-negative patients is much lower than the seroconversion rate amongst healthy volunteers.¹ Many of the studies of hospitalized transfusion recipients reported D-alloimmunization rates between 20% and 30%.¹ However, in many of these studies the recipients were not serially examined for D-alloimmunization following their RhD-positive exposure. The term serological follow up is used to describe the time from the index RhD-positive RBC transfusion in an RhD-negative recipient until the last antibody detection test (herein referred to as an antibody screen) that did not demonstrate anti-D (the antibody against the RhD-antigen). Thus, in studies of the D-alloimmunization rate, RhD-negative recipients who had one antibody screen performed a few weeks after their RhD-positive exposure that did not show anti-D, that is, short serological follow up, were treated with the same weight in the study as patients who also did not become D-alloimmunized but who had multiple antibody screens performed over a longer period of time, that is, longer serological follow up. As the patient's serological follow up increases without the detection of anti-D, the greater the probability becomes that the patient will not go on to develop anti-D.

When studying the rate of D-alloimmunization amongst RhD-negative patients who received RhD-positive RBCs or LTOWB (herein collectively referred to as RBCs), a patient who produces anti-D has clearly achieved the endpoint of the study. However, there is uncertainty about how to handle those recipients who had not become D-alloimmunized during the study. Including a participant amongst the non-alloimmunized study participants who became D-alloimmunized after the study period had ended would lead to underestimating the true rate of D-alloimmunization because that participant was effectively a false negative. Since it is less certain that a patient with shorter serological follow up will truly remain non-D-alloimmunized compared to a patient with longer serological follow up, adjusting for differences in the length of serologic follow up could be helpful in obtaining a more accurate D-alloimmunization rate.

To address this issue, a regression model was developed to weight the contribution of non-D alloimmunized patients to the overall D-alloimmunization rate according to the length of their serological follow up. This model was then applied to adjust the D-alloimmunization rate amongst transfused but not D-alloimmunized RhD-negative patients for the length of each patient's serological follow up in order to obtain a potentially more accurate D-alloimmunization rate.

2 STUDY DESIGN AND METHODS

3 RESULTS

3.1 Weighted model of the kinetics of D-alloimmunization

Figure 1 illustrates the results of the literature search. Of the 31 identified studies, 5 studies contained patient data that could be used to build the model. In addition, patient datasets were obtained from of the RADAR study participants and from a separate single center (Table 1).^{10, 11} In total, 419 recipients were included in the model including 126 who were D-alloimmunized.

TABLE 1. Summary of the studies used in building the weighted model.

Year	Reference	Study population	Exposure	Study design	Immunized subjects, n (% alloimmunized in each study)	Total number of immunized and non-immunized recipients (n)
1970	12	Healthy volunteers	0.5 or 5 mL RBC	Prospective	9 (81)	11
1971	13	Healthy volunteers	0.5 mL RBC	Prospective	8 (89)	9
1971	14	Healthy volunteers	≥1 RBC units	Prospective	18 (82)	22
1981	15	Healthy volunteers	1 mL RBC	Prospective	4 (33)	12
1981	16	Healthy volunteers	≥1 RBC units	Prospectivea	21 (75)	28
Per. Com.	Single center dataset with serially screened recipients	Patients	≥1 RBC units	Prospectiveb	41 (13)	312
Per. Com.	RADAR study sites	Patients	≥1 RBC or LTOWB units	Retrospectivec	25d	25 D-alloimmunizedd

• Abbreviations: LTOWB, low titer group O whole blood; Per. Com., Personal communication; RBC, red blood cell.
• ^a To fit data from this study to the weighted model, the monthly distribution was randomly sampled using an R-simulation, assuming a normal distribution around the reported average time to first detection of anti-D, with a standard deviation estimated as ¼ of the reported range.
• ^b Recipients not tested for alloimmunization at month 4 and 5.
• ^c Monthly frequencies based on logistic regression analysis of data.
• ^d The total number of RhD-negative recipients is not known because the data supplied from the RADAR study sites only included D-alloimmunized patients, therefore, the rate of D-alloimmunization cannot be calculated.

Figure 2 presents the results of the weighted model, which demonstrates the timing of anti-D production amongst RhD-negative recipients of RhD-positive RBCs who became D-alloimmunized. For example, the model predicts that a patient who is destined to become D-alloimmunized has an approximately 70% chance that their anti-D will develop within the first 3 months after the RhD-positive transfusion. Note that the model does not predict which patients will become D-alloimmunized, rather it demonstrates the timing of D-alloimmunization amongst those who will make the antibody.

The aggregate regression of the three RhD-negative recipient subgroups, that is, 0.5–5 mL RhD-positive RBC (volunteer; n = 21), at least one RhD-positive RBC unit (volunteer; n = 39), and at least one RhD-positive RBC unit (patients; n = 66) fit the data well (R² = 0.99). Individual regressions for each of these three recipient subgroups were also performed, and all subgroup regressions fit the data well (R² = 0.99; Figure 2). The 95% confidence intervals of the variables in the subgroup regressions overlapped indicating that they were not statistically different. The results from the logistic regression analysis using a non-linear least squares method are shown in Data S1.

4 DISCUSSION

In this study, the application of a weighted model for D-alloimmunization increased the uncorrected rate of D-alloimmunization by an absolute value of 6.3% points. It was expected that applying the weighted model would raise the D-alloimmunization rate because it accounts for the potential for D-alloimmunization to occur amongst recipients with short serological follow up lengths. The weighted model is itself probabilistic based on the timing of anti-D detection in recipients who seroconverted. Thus, the actual rate of D-alloimmunization in this cohort likely lies somewhere between 20.0% and 26.3%. Applying this weighted model to account for the likelihood that recipients with relatively short serological follow up lengths might have become D-alloimmunized in the future can improve the accuracy of future studies of the seroconversion rate.

In building the weighted model it was found that the kinetics of anti-D detection between the three subgroups of recipients were quite similar, despite differences in the doses of RhD-positive RBCs administered and the recipients' underlying health status. This is interesting because it is recognized that the rate of D-alloimmunization differs considerably between healthy volunteers and hospitalized patients¹ as well as the dose of RhD-positive RBCs administered (increasing rates of alloimmunization up to approximately 30 mL of RhD-positive RBCs, after which the rate becomes constant¹⁷). Thus, it seems that the volume of the RhD-positive RBCs administered and the recipient's underlying health status modulate the likelihood of D-alloimmunization but not the kinetics of anti-D formation in a recipient who seroconverts.

The weighted model was not applied to the patients in Group 1 whose last antibody screen that did not show anti-D was performed >180 days after the index RhD-positive transfusion because it was assumed that the absence of anti-D beyond 180 days reflected their true immunohematological status. While it is possible that patients with a negative antibody screen that long after the index RhD-positive transfusion could have been D-alloimmunized with an evanescent antibody, anti-D has a low evanescence rate.^18-20 Thus, the assumption that the absence of anti-D on their last antibody screen reflects a truly not D-alloimmunized state is reasonable.²¹

This study had several other limitations. The retrospective nature of its design prevented the investigators from taking a more proactive role in obtaining serial antibody screens on the patients who were exposed to RhD-positive RBCs, hence the need for the weighted model for those with short serological follow up lengths. As is common for retrospective studies of alloimmunization rates, the date on which the anti-D was detected was taken to be the day on which the patient actually produced the antibody, although these dates might not perfectly coincide. For those who seroconverted, knowing the amount of time between the date of the last antibody screen that did not contain anti-D and the date of the screen that first demonstrated the anti-D allowed for a more accurate determination of the month in which the anti-D was produced. RhD-positive platelets were not considered to be an alloimmunizing stimulus because of the much lower rate of D-alloimmunization following RhD-positive platelet exposure compared to RhD-positive RBC exposure and none of the patients or volunteers in these studies only received RhD-positive platelets.²²

The hospitalized patients who were included in the rate of D-alloimmunization part of this study might have received RhD-positive transfusions at hospitals at which the investigators did not have access to transfusion histories, and they could also have been sensitized during pregnancy. However, any patient whose anti-D was detected within 14 days of the index RhD-positive transfusion at this hospital was excluded due to the possibility that the anti-D was due to an anamnestic response thereby reducing the potential confounding effect of antibody evanescence. The clinical diagnoses and treatments of the RhD-negative recipients who received RhD-positive transfusions were not known, however they were drawn from a large, regional hospital over a long period of time thereby likely reflecting a pragmatic sample of the population of hospitalized patients. Also, at this hospital, once a patient has a positive antibody screen, additional antibody screens are not performed before future blood products are issued. Instead, the RBC units are serologically crossmatched with recipient plasma and an incompatibility would be further investigated for the presence of additional antibodies before the unit is issued. Thus, it is possible that some patients with anti-D might not have been detected if RhD-negative RBCs were issued after a non-anti-D antibody had been detected.

Some of the limitations of building the weighted model included the fact that at the single center that contributed data, RhD-negative recipients of RhD-positive RBCs underwent serial antibody screening at months 1, 2, 3, and 6 following the index RhD-positive transfusion but not at months 4 or 5. Thus, the data used to build the model for months 4 and 5 came solely from the patients in the previously published studies and from the sites that participated in the RADAR study. Also, it was assumed that all recipients who were destined to produce anti-D would do so by 6 months following the RhD-positive index transfusion. This assumption might overestimate the weight of a negative antibody screen result >180 days post RhD-positive exposure. However, setting an upper time limit for D-alloimmunization was necessary for building the model. Six months was chosen as the upper limit because the dataset from the single center that featured serially screened recipients contained only one patient who developed an anti-D between months 6 and 12. A published study also found that primary D-alloimmunization after 6 months was rare.¹⁶ Therefore, it is unlikely that a recipient would become D-alloimmunized beyond 6 months.

These data suggest that if RhD-negative patients are to have an antibody detection test performed after RhD-positive transfusion to determine if they have become D-alloimmunized, the optimum timing of the test is around 6 months after the exposure. This timing has been adopted for screening RhD-negative recipients of RhD-positive RBCs in Germany because it helps to reduce the collection of samples at times that are less likely to demonstrate seroconversion.²³ Furthermore, knowing the optimal time to screen recipients could have implications for screening RhD-negative females of childbearing potential following an RhD-positive transfusion during resuscitation from life threatening bleeding for antenatal counseling about her risk of HDFN and on recipients who might require future transfusions as part of their ongoing therapy to ensure their safety from delayed hemolytic reactions.

The weighted model can be applied to future studies of D-alloimmunization to help correct for the uncertainty that accompanies the immunohematological status of patients with relatively short lengths of serological follow up. While several weeks of serological follow up might be sufficient to detect a new antibody, weighting the results of those recipients who did not produce anti-D within 180 days post exposure offers a novel method of accounting for those with a relatively short length of serological follow up.

CONFLICT OF INTEREST

MHY has given paid lectures/reimbursed travel/consulting fees for: Grifols, Hemanext, Terumo, Legacy Innovations. MHY owns equity in Velico Medical, Inc.