Abstract
Chronic red blood cell (RBC) transfusions are essential for sickle cell disease (SCD) management, but the cumulative demand for donor RBCs strains blood supplies and increases risks of iron overload and alloimmunization. Isovolemic hemodilution red cell exchange (IHD-RCE) is a 2-step procedure designed to reduce RBC requirements, but its effectiveness remains unclear, particularly in pediatric patients. We conducted a retrospective cohort study of 45 children and young adults with SCD, each receiving standard RCE followed by IHD-RCE, comparing within-patient RBC use between RCE types. Inclusion required ≥6 consecutive standard RCEs and ≥6 consecutive IHD-RCEs within a year of each other. Across 2667 RCE procedures, paired analyses demonstrated both standard and IHD-RCE required a median of 1.2 RBC units per liter of total blood volume (TBV; P = .3017). Adjusted analyses showed that IHD-RCE increased RBC requirements for patients with a TBV of ≤3500 mL. Although not statistically significant, IHD-RCE offered modest RBC savings in patients with a TBV of >4500 mL. Stratified analysis showed IHD-RCE related savings only for patients requiring >4 RBC units per transfusion before transitioning from standard RCE to IHD-RCE (P = .0186). These findings challenge the assumption that IHD-RCE universally reduces donor blood needs. In patients with low TBV or RBC needs, IHD-RCE does not reduce, and may even increase, RBC use. Our findings underscore the importance of individualized transfusion strategies, especially during growth and puberty, because not all patients benefit from IHD-RCE.
Introduction
Chronic red blood cell (RBC) transfusion therapy is fundamental to sickle cell disease (SCD) treatment with indications including stroke prophylaxis, acute chest syndrome, and preparation for curative therapy.1–4 RBC transfusion decreases a patient’s proportion of sickle hemoglobin (HbS) RBCs, improves anemia, and suppresses endogenous erythropoiesis. However, chronic RBC transfusion also carries significant risks including alloimmunization and iron loading.2,5–9
Automated red cell exchange (RCE) can effectively reduce HbS levels with less iron loading compared with simple transfusions, and it is the recommended approach for chronic transfusion therapy in the 2020 American Society of Hematology guidelines for SCD transfusion support.10 However, RCE requires more RBC units per transfusion compared with simple transfusions,11 straining limited resources and increasing donor exposures. Additionally, identifying sufficient RBC units for patients who are alloimmunized to multiple antigens or require donors with uncommon RBC antigen profiles can be challenging.12
Isovolemic hemodilution (IHD) RCE is a 2-step procedure that was introduced as a method intended to decrease donor RBC requirements.13 Initially, the patient undergoes an RBC depletion and concurrent volume replacement with normal saline or an albumin solution until the target intraprocedure hematocrit is achieved. Next, a standard RCE with donor RBCs is performed at the lower hematocrit, and at the end of the procedure, the patient’s hematocrit is typically raised back to the preprocedure level. In current apheresis devices, IHD-RCE is an integrally programmed procedure. Theoretically, performing the RCE at a lower hematocrit achieves the same therapeutic goal while modestly reducing the amount of donor blood required or reaching a lower fraction of cells remaining (FCR); a lower FCR could then lengthen the transfusion interval and reduce blood usage over time.
The potential benefits of IHD-RCE include a decrease in RBC unit requirements and/or procedure frequency compared with an otherwise identical standard RCE procedure. However, due to very limited evidence, the American Society of Hematology guideline panel for transfusion support concluded that IHD-RCE could not be recommended over standard RCE, and the need for high-quality studies was highlighted.10 Their systematic review identified only 5 comparative observational studies with a total of 122 patients.10 Whether IHD-RCE affects RBC unit usage remains unclear, with several studies identifying a decrease of ~1 unit per procedure,14–17 and 2 other studies published only as abstracts finding no difference.18,19 An international survey of 117 apheresis practitioners found only 40% used IHD-RCE at all, further highlighting the clinical equipoise surrounding the practice.20 Of note, most studies of IHD-RCE focus on adult patients. Although no consensus exists regarding a minimum age for IHD-RCE, 30% of surveyed practitioners using IHD-RCE reported a minimum initiation age, and 40% reported a minimum weight.20
For at least 3 decades, our institution has used IHD-RCE as standard of care for patients with SCD who require chronic RCE and have completed at least 6 months of standard RCE. This practice is used without a minimum patient age, size, or RBC unit requirement, as patients with small RBC needs (eg, 2–3 units per transfusion) may reach a lower FCR and theoretically have improved control of HbS even if full unit savings were not possible. Here, using a single-center, retrospective cohort and paired, within-patient comparisons, we evaluated whether IHD-RCE, compared with standard RCE, reduces the number of RBC units needed in pediatric and young adult patients with SCD requiring chronic transfusion therapy.
Patients and methods
Study design and population
We conducted a retrospective cohort study of patients with SCD, genotype SS, receiving chronic RCE transfusions at the Children’s Hospital of Philadelphia from 2003 to 2022. We compared outcomes in patients who received standard RCE followed by a period of IHD-RCE (Figure 1). An RCE “cluster” was defined as ≥6 consecutive procedures of the same RCE type (standard or IHD) with ≤45 days between transfusions. Patients were included if they had a standard RCE cluster ending within 1 year of the start of an IHD-RCE cluster, with a consistent HbS goal (<30% or <50%) across all included transfusions. Most patients at our institution have a target HbS goal of <30%, in accordance with SCD guidelines.1,21 A HbS goal of <50% is used for certain patients for indications outside of stroke prophylaxis, or, per a historical institutional practice, after several years of clinical stability on chronic transfusions with HbS goal of <30%.1 The longest pair of IHD-RCE and standard RCE clusters for each patient was analyzed.
Figure 1. Details of RCE clusters and transfusions.
(A) Schema detailing RCE cluster requirements for each individual patient. To be eligible, a patient required an IHD-RCE cluster starting within 1 year of the end of a standard RCE cluster, with a constant HbS goal across all procedures. Green indicates RCE transfusions outside the study clusters, whereas yellow and blue indicate standard RCE and IHD-RCE transfusions, respectively, included in the study. Duration between individual transfusions within each cluster was no longer than 45 days, and a maximum of 1 year was allowed between the end of the standard RCE cluster and the IHD-RCE clusters. HbS goal was consistent across the 2 clusters. The longest combination of standard RCE and IHD-RCE clusters meeting this requirement was chosen for each qualifying patient. (B) Box plots of IHD-RCE hemodilution parameters by pre-RCE hematocrit. The percentage hematocrit change (y-axis) represents the pre-RCE hematocrit minus programmed nadir RCE hematocrit, that is the lowest hematocrit after the hemodilution step, for IHD-RCE. RCE transfusions are categorized by pre-RCE hematocrit, shown in the x-axis. Lines within IQR box plots indicate the median and whiskers equal minimum or maximum up to 1.5× IQR. Outliers (>1.5× IQR from edges of box) are marked as dots. The number of RCE procedures included in each pre-RCE hematocrit category is specified under each box. Additional details, including mean values, are provided in supplemental Table 1. Figure 1A created with biorender.com. Dain A. (2026) https://biorender.com/pqqzwo4.
Patients received standard RCE and IHD-RCE based on institutional standard operating procedures (supplemental Methods, RCE Procedures). Patients were transitioned from simple to exchange transfusions during late childhood or early adolescence, and IHD-RCE was introduced after at least 6 months of standard RCE. IHD-RCE was not used for patients with severe cerebral vasculopathy, overt ischemic stroke in the previous 6 months, or baseline hematocrit of <23%.
RBC units were preordered based on previous requirements. Partial units were not given; the actual number used for each RCE was based on pre-RCE laboratory data and unit volumes to achieve a target FCR and postprocedure HbS level. Additional details are provided under RCE Procedures in the supplemental Methods. For patients with a goal of maintaining an HbS level of <30% and <50%, most required a postprocedure HbS level of 12% to 15% and 20% to 25%, respectively. For all RCEs, the end-procedure hematocrit was programmed to equal the greater of 27% or the pre-RCE hematocrit.
Data source
We collected data using a FileMaker Pro Advanced database (Claris International Inc, Cupertino, CA) with all RCE encounters since 2003, supplemented with the electronic health record when specified. Missingness for all covariates and outcome variables ranged from 0% to 2.5%.
Exposure, outcomes, and covariates
RCE procedures with a documented programmed nadir hematocrit were classified as IHD-RCE, the primary exposure. The primary outcome was RBC units per RCE procedure. Secondary outcomes included annualized RBC use (365 × [total RBC units in cluster/cluster duration in days]) and RBC units per liter of total blood volume (TBV). Additional outcomes included procedure time (minutes), programmed and actual FCR (%), pretransfusion HbS level (%), interval between individual transfusions (days), net RBC mass transfused per kilogram body weight, and change in cerebrovascular imaging results. Actual FCR was defined as immediate post-RCE HbS (%) divided by the pre-RCE HbS (%) measured within 3 days of RCE and calculated only for patients with a HbS goal of <30%. We calculated the net RBC mass transfused per kg using machine manufacturer instructions (Table 2). Iron loading was not included as an outcome because many other variables, such as chelation regimen(s) and iron overload measurements, were expected to vary considerably between each patient’s 2 RCE clusters and between individual patients. Standard RCE was the first method used when transitioning from simple transfusions to RCE; patient iron status was therefore expected to skew higher during standard RCE vs IHD-RCE, creating potential bias.
Table 2.
Transfusion characteristics and RBC requirements
| Standard RCE | IHD-RCE | P value* | |
|---|---|---|---|
| Transfusions in cluster, n | 15 (10–28) | 18 (11–59) | .0257† |
| Days between transfusions‡ | 23 (21.7–28) | 24 (21.6–27.3) | .5687 |
| FCR, % ‡ | |||
| Programmed | 42.9 (34.4–50.7) | 39.1 (34.2–45.2) | .0071† |
| Actual§ | 46.9 (37.6–51.6) | 42.4 (37.2–47.6) | .0118† |
| Pretransfusion HbS, % ‡ | |||
| Goal of <30% | 29.5 (27.8–32.0) | 29.3 (28.1–30.9) | .2800 |
| Goal of <50% | 45.1 (42.5–46.3) | 44.8 (40.8–46.1) | .8658 |
| Pre-RCE hematocrit, %‡ | 27.0 (26.1–28.4) | 29.0 (27.8–30.5) | <.001† |
| Net RCE hematocrit change, %‡,‖ | |||
| Programmed | 1.4 (0.9–2.8) | 0.2 (−0.0 to 0.7) | <.001† |
| Actual | 2.0 (1.0–3.0) | 0.9 (0.4–1.5) | <.001† |
| Net RBC mass transfused, mL/kg‡,¶ | 1.0 (0.1–1.3) | −0.2 (−0.7 to 0.6) | <.001† |
| Transfusion duration, min‡ | 91.2 (76.3–139.5) | 114.5 (94.4–155.0) | <.001† |
| RBC units per TBV, U/L | |||
| Per transfusion‡ | 1.2 (1.0–1.5) | 1.2 (1.0–1.4) | .3017 |
| Annual | 19.2 (15.0–26.2) | 18.3 (16.8–21.1) | .1501 |
| RBC units | |||
| Per transfusion‡ | 3.0 (2.5–4.0) | 4.0 (3.3–5.0) | .0001† |
| Annual | 50.5 (40.7–68.4) | 63.2 (54.1–81.1) | .0032† |
Data are presented as median (IQR).
Paired Wilcoxon signed-rank test.
P < .05.
Mean values first calculated for individual patients then used to calculate median and IQR across all patients.
Post-RCE HbS divided by pre-RCE HbS.
End RCE hematocrit minus pre-RCE hematocrit.
The net RBC mass transfused per kilogram was calculated using the following assumptions and calculations: RBC mass = RBC volume × hematocrit. Net RBC mass transfused = [(volume of RBC transfused × average hematocrit of transfused units) − (volume of RCE waste bag × hematocrit of waste bag)] − 70 (manufacturer provided constant to account for RBC mass left in tubing). The net RBC mass transfused was divided by patient body weight at each RCE. For each patient, net RBC mass per kilogram body weight was averaged for all transfusions in each cluster. Average values across patients were used to calculate the median and IQR shown in the table.
Brain magnetic resonance imaging and magnetic resonance angiography reports were reviewed by 1 physician (A.S.D.) to identify cerebrovascular disease progression during IHD clusters, because cerebral ischemia is a potential risk of hemodilution.20 Cerebrovascular disease progression was defined as new or progressive silent cerebral infarct or steno-occlusive cerebral vasculopathy, or new or recurrent overt stroke (supplemental Methods, Cerebrovascular Outcomes).
Baseline patient and RCE characteristics were assessed at the first transfusion of the standard RCE and IHD-RCE clusters. Mean percentage HbS, hematocrit, and RBC units required in the 90 days before the start of each cluster were calculated for baseline values.
Statistics
We used descriptive statistics to summarize characteristics of patients and RCE procedures included in the study. RBC units and RBC units per liter of TBV for each transfusion were averaged within each patient’s standard RCE and IHD-RCE transfusion “clusters” and used for paired comparisons between RCE types using the Wilcoxon signed-rank test.
RBC requirements by RCE type were estimated using a generalized multivariable linear mixed Poisson model with a log link, random effects for patients, and a first-order autoregressive correlation structure. This model accounts for repeat measures per patient and the longitudinal nature of the data. Possible confounders, including age and calendar year at start of follow-up, sex, machine type (Terumo Optia [Terumo BCT, Shibuya, Tokyo, Japan] vs COBE Spectra [Terumo BCT]), and TBV at each transfusion were assessed for inclusion in our multivariate analysis. HbS goal was included a priori given its known influence on transfusion requirements. TBV at each transfusion was assessed as an interaction term to examine whether IHD had a differential impact on RBC use based on patient size. Exponentiated estimates were reported as rate ratios, representing the ratio of the expected RBC units for IHD-RCE compared with standard RCE (reference group), with 95% confidence intervals.
To examine the procedure duration of IHD-RCE vs standard RCE, independent of units required, we used a mixed-effects linear model of RCE duration in minutes, adjusted for RBC units per procedure as a fixed effect, and individual patient as a random effect. P values were 2-sided with a significance threshold of P < .05. Proportions of patients with progressive cerebrovascular disease in standard vs IHD-RCE was compared using the Fisher exact test.
Sensitivity analyses
We performed several preplanned sensitivity analyses to assess the robustness of results. To assess the impact of HbS goal and machine type, we applied alternative inclusion criteria requiring (1) HbS goal of <30% or (2) RCE using only the Terumo Optia machine, which has been used since 2012. To assess whether hemodilution’s impact occurs only after several procedures, a sensitivity analysis excluded the first 4 transfusions in each cluster. We expected patients to have more IHD-RCE vs standard RCE transfusions. To assess the impact of longer cluster duration on RBC utilization, we redefined the 2 clusters to be approximately equal in duration. Finally, to assess potential misclassification of RCE type, we used an alternative definition of IHD-RCE requiring a programmed hematocrit decrease of ≥3% during hemodilution and excluded transfusions not meeting this threshold before cluster creation. Further methodologic details are provided in supplemental Methods, Additional Details of Sensitivity Analyses. Analyses were performed using Stata version 17 (StataCorp, College Station, TX).
Post hoc analyses
Initial results suggested that the effect of IHD on RBC use varied by TBV. Because patients with higher TBV typically require more RBC units per transfusion, we stratified patients into 2 groups based on their mean unit usage per RCE in the 90 days before the first transfusion of the IHD-RCE cluster. A range of thresholds was evaluated to identify an RBC unit threshold at which IHD-RCE provided a statistically significant unit savings of ≥1 unit per transfusion compared with standard RCE. Differences in RBC unit requirements during IHD-RCE vs standard RCE clusters were assessed using the paired Wilcoxon signed-rank test.
The Children’s Hospital of Philadelphia institutional review board approved this study, which was performed in accordance with the Declaration of Helsinki.
Results
Cohort and RCE cluster characteristics
Among 280 patients with SCD-SS with ≥1 RCEs in our database, 45 patients met inclusion criteria and collectively received 2667 RCE procedures (Table 1). Patients were included if they had a standard RCE cluster ending within 1 year of the start of an IHD-RCE cluster with a consistent HbS goal (<30% or <50%) across all included transfusions (Figure 1A). Most patients (38/45 [84%]) had a target maintenance HbS level of <30%, and nearly all (42/45 [93%]) received chronic RCE for primary or secondary stroke prophylaxis. The median year of the first study RCE was 2008 (interquartile range [IQR], 2006–2013). Patients received a median of 20 RCE transfusions (IQR, 5–55) before study inclusion and required a median of 3 units (IQR, 2.3–4.5) at baseline, with a range of 2 to 9 units across all patients.
Table 1.
Patient and transfusion characteristics
| Variable | N = 45 (%) | P † | |
|---|---|---|---|
| Sex * | |||
| Female | 20 (44.4) | ||
| Male | 25 (55.6) | ||
| Hemoglobin S goal, % * | |||
| <30% | 38 (84.4) | ||
| <50% | 7 (15.6) | ||
| RCE indication * | |||
| Primary stroke prophylaxis | 23 (51.1) | ||
| Secondary stroke prophylaxis | 19 (42.2) | ||
| Other | 3 (6.7) | ||
| Machine type * | |||
| COBE Spectra | 36 (80) | ||
| Terumo Optia | 9 (20) | ||
| Number of RCE transfusions before study entry, median (IQR)* | 20.0 (5.0, 55.0) | ||
| RCE transfusion cluster type | |||
| Standard RCE | IHD-RCE | ||
| Year of first procedure in RCE cluster, median (IQR) | 2008 (2006, 2013) | 2010 (2008, 2014) | |
| Baseline RBC unit requirement at start of cluster, median (IQR)‡ | 3 (2.3, 4.5) | 3.3 (3, 5) | .005§ |
| Age at first RCE in cluster, y | 12.4 (9.1, 14.9) | 13.8 (11.4, 17.2) | < .001§ |
| TBV at first RCE in cluster, mL | 2369.0 (2044.0, 3264.0) | 3241.0 (2479.0, 3738.0) | < .001§ |
| Baseline pre-RCE hematocrit in cluster, %‡ | 26.6 (26.0, 28.1) | 27.6 (26.9, 29.2) | .0050§ |
| Baseline HbS in cluster, % ‡ | |||
| Goal S < 30% | 31.8 (27.7, 36.7) | 31.1 (28.3, 32.6) | .2800 |
| Goal S < 50% | 43.4 (39.4, 44.4) | 44.1 (41.6, 45.9) | .1763 |
Data are displayed as median (IQR) for continuous measures and n (%) for categorical measures.
Values constant between RCE types.
Paired Wilcoxon signed-rank test.
Mean value in 90 days prior to first RCE type cluster transfusion calculated for each patient. Median value and IQR across patients reported.
P < .05.
The median age and TBV at the start of standard RCE were 12.4 years (IQR, 9.1–14.9) and 2369 mL (IQR: 2044–3264), respectively, coinciding with a time of growth and pubertal change (Table 1). Expectedly, median age and TBV were higher at the start of IHD-RCE at 13.8 years (IQR, 11.4–17.2) and 3241 mL (IQR, 2479–3738), respectively, because by clinical practice and study design standard RCE clusters occurred before IHD-RCE clusters, and patients’ TBV increased with growth over time. Most patients (38/45 [84%]) were aged ≤18 years at the first IHD-RCE. The median baseline hematocrit at the start of standard RCE was 26.6% (IQR 26.0–28.1) vs 27.6% (IQR, 26.9–29.2; P = .0050) at the start of IHD-RCE. Baseline percentage HbS was not significantly different between the start of standard RCE or IHD-RCE clusters for patients with a target maintenance HbS level of <30% or <50%.
Standard RCE clusters included a median of 15 transfusions (IQR, 10–28; Table 2). IHD-RCE clusters were significantly longer, because once a patient transitioned to IHD-RCE, it was continued indefinitely if tolerated and no contraindication (ie, moyamoya disease) developed over time, with a median of 18 transfusions (IQR, 11–59; P = .0257) per cluster. The median interval between transfusions was 23 and 24 days for standard RCE and IHD-RCE, respectively. The degree of hemodilution in IHD-RCE rose with pre-RCE hematocrit, with a median 3% hematocrit decrease with pre-RCE hematocrits of ≤25% and a 6% decrease with pre-RCE hematocrits of ≥31% (Figure 1B; supplemental Table 1).
To provide Rh- and K-matched RBCs, units were ordered in advance based on the preprocedure HbS levels and the number of units used for the preceding RCE. On the day of the procedure, the actual number of units to be used was recalculated based on the patient’s preprocedure HbS level and the volume of red cells required to reach the target FCR and postprocedure HbS. The median programmed FCR was significantly higher in standard RCE (42.9%; IQR, 34.4–50.7) than IHD-RCE (39.1%; IQR, 34.2–45.2; P = .0064), as was the actual FCR (Table 2). Notably, this FCR difference did not result in a significantly different pre-RCE HbS between the standard and IHD-RCE clusters or a difference in the time between transfusions. For those with a target maintenance HbS of <30%, the median pretransfusion HbS level was 29.5% (IQR, 27.8–32.0) and 29.3% (IQR, 28.1–30.9) for standard RCE and IHD-RCE, respectively. For those with a goal HbS of <50%, the median HbS was 45.1% (IQR, 42.5–46.3) and 44.8% (IQR, 40.8–46.1) for standard RCE and IHD-RCE, respectively. Pretransfusion hematocrit was lower in standard RCE clusters (27.0% [IQR, 26.1–28.4]) vs IHD-RCE clusters (29.0% [IQR, 27.8–30.5]; P < .001) likely, in part, due to the physiologic rise in hematocrit after puberty. Hematocrit values were, on average, minimally higher after RCE compared with before RCE, with a median 2.0% (IQR, 1.0–3.0) hematocrit increase in standard RCE and a 0.9% (IQR, 0.4–1.5) increase in IHD-RCE (P < .001).
IHD-RCE procedure duration was longer than standard RCE in paired, unadjusted analysis (Table 2). After adjusting for the number of RBC units needed per transfusion, IHD-RCE was associated with a significant but small increase in procedure time, with an estimated 5.9 additional minutes per procedure (95% confidence interval, 1.8–10.0; P = .005).
RBC requirements
Requirements for individual RCE transfusions ranged from 2 to 10 units for both standard RCE and IHD-RCE. When accounting for TBV (in liters) at each transfusion, RBC use was similar between standard and IHD-RCE procedures, with a median of 1.2 units per liter (U/L; IQR, 1.0–1.5) in standard clusters and 1.2 U/L (IQR, 1.0–1.4) in IHD clusters (P = .3017; Table 2). The annualized unit per liter of TBV requirement was slightly higher for the standard RCE clusters, with a median 19.2 U/L per year (IQR, 15.0–26.2) vs 18.3 U/L per year (IQR, 16.8–21.1). This difference was not statistically significant. As expected, standard RCE required fewer absolute units vs IHD-RCE, because the transition from standard RCE to IHD-RCE typically occurred during the growth and pubertal changes of adolescence.
RCE procedure type (standard or IHD), HbS, and TBV were included as covariates in the multivariable model estimating per-transfusion RBC requirements. The association between RCE type and RBC units transfused varied significantly by TBV (interaction term P < .05). The remaining covariates and interaction terms tested did not contribute significantly to the model. Multivariable analysis estimated a statistically significant increase in unit requirements with IHD-RCE compared with standard RCE at TBV values of ≤3500 mL (Table 3; supplemental Table 2). As TBV rose, there was a trend toward RBC savings with IHD-RCE (Figure 2); however, this difference was not statistically significant. For all TBV values, the difference in estimated RBC requirements between standard RCE and IHD-RCE was <1 unit.
Table 3.
Model-estimated, per-transfusion RBC requirements at example TBV values
| Example TBV, mL | Standard RCE estimated units (95% CI) | IHD-RCE estimated units (95% CI) | P value |
|---|---|---|---|
| 2000 | 2.8 (2.6–3.0) | 3.2 (3.0–3.5) | .0012* |
| 2500 | 3.2 (3.0–3.4) | 3.6 (3.3–3.8) | .0018* |
| 3000 | 3.7 (3.5–39) | 3.9 (3.7–4.2) | .0055* |
| 3500 | 4.2 (3.9–4.4) | 4.4 (4.2–4.6) | .044* |
| 4000 | 4.8 (4.5–5.1) | 4.8 (4.6–5.1) | .4073 |
| 4500 | 5.4 (5.1–5.8) | 5.4 (5.1–5.7) | .7526 |
| 5000 | 6.2 (5.7–6.7) | 6.0 (5.6–6.3) | .2687 |
| 5500 | 7.0 (6.4–7.7) | 6.7 (6.1–7.1) | .1101 |
| 6000 | 8.0 (7.2–8.9) | 7.3 (6.7–8.0) | .0565 |
RBC units per transfusion estimated from a multivariable mixed-effects Poisson regression model with random intercept for patient, adjusted for HbS goal and TBV, and including an interaction term for TBV × RCE type. Estimated RBC unit values assume goal HbS of <30%. Rate ratios from this model are provided in supplemental Table 2.
CI, confidence interval.
P < .05.
Figure 2. RBC units required for standard RCE vs IHD-RCE as estimated by the mixed-effects generalized linear model with Poisson distribution, accounting for repeat measures per patient and longitudinal data structure.
IHD-RCE is predicted to save RBC units compared with standard RCE only in patients with larger TBV. The model was adjusted for HbS goal and TBV (at each RCE), with a TBV × RCE category interaction term. RBC units were estimated across TBV values, assuming an HbS goal of <30%. Dashed line indicates estimated RBC units for standard RCE, and solid line indicates estimated RBC units for IHD-RCE. Shaded areas represent 95% confidence intervals, with yellow corresponding to standard RCE and blue corresponding to IHD-RCE.
Stratified analysis by RBC requirements at time of transition to IHD-RCE
We further analyzed RBC unit requirements by stratifying patients into 2 groups based on average RBC unit requirements in the 90 days before transition to IHD-RCE (Table 4). In paired analyses, patients needing ≤4 RBC units per transfusion before switching to IHD-RCE (n = 31) had similar RBC units per liter TBV requirements with standard RCE compared with IHD-RCE (P = .5435; Figure 3). Similar to our primary paired analysis, FCR was significantly higher in the standard RCE group, but time between transfusions and pretransfusion HbS levels did not differ between standard RCE and IHD-RCE (Table 4). Patients initially needing >4 units per procedure had a statistically significant units per liter TBV savings with IHD-RCE (P = .0186; Figure 3). For example, for a patient with a TBV of 3.6 L, IHD-RCE would require 1 less unit per procedure compared with standard RCE (5 vs 6 units). Similarly, only patients in the >4-unit group saw IHD-related unit per liter TBV savings in annualized analyses (P = .0313; Table 4).
Table 4.
Patient characteristics, transfusion characteristics, and RBC outcomes, stratified by RBC requirements before IHD-RCE
| ≤4 U (n = 31) | >4 U (n = 14) | |||||
|---|---|---|---|---|---|---|
| Baseline characteristics at transition to IHD-RCE † | ||||||
| Age, y | 11.9 (8.2, 13.7) | 14.6 (11.9, 19.5) | ||||
| TBV, mL | 3033 (2201, 3479) | 3620 (3025, 4412) | ||||
| HbS goal, % | ||||||
| < 30% | 25 (80.6) | 13 (92.9) | ||||
| < 50% | 6 (19.4) | 1 (17.1) | ||||
| RBC units, range‡ | 2–4 | 4.3–9.6 | ||||
| Standard RCE | IHD-RCE | P * | Standard RCE | IHD-RCE | P * | |
| Days between transfusions§ | 22.6 (21.5, 28.0) | 23.4 (21.4, 28.0) | .9687 | 24. (22.0, 26.1) | 25.5 (22.6, 27.3) | .3967 |
| FCR, % § | ||||||
| Programmed | 43.5 (38.5, 50.7) | 39.1 (34.2, 44.1) | .0003‖ | 33.9 (26.4, 50.7) | 40.2 (32.8, 45.8) | .4326 |
| Actual¶ | 48.9 (43.2, 55.7) | 43.8 (40.1, 49.5) | .0013‖ | 31.8 (27.8, 48.4) | 37.5 (30.9, 44.4) | .9750 |
| Pre-RCE HbS, % § | ||||||
| HbS goal < 30% | 29.2 (27.6, 30.0) | 28.7 (27.5, 30.0) | .4926 | 32.0 (28.9, 34.7) | 30.8 (29.8, 33.4) | .3824 |
| HbS goal < 50% | 45.2 (44.7, 46.3) | 43.8 (40.8, 45.6) | .6002 | 42.5 (NA) | 46.2 (NA) | NA |
| Pre-RCE hematocrit, %§ | 26.7 (26.2, 27.9) | 28.8 (27.8, 30.1) | <.001‖ | 28.0 (25.0, 29.7) | 29.4 (26.9, 30.7) | .0157‖ |
| Net RCE hematocrit change, % § # | ||||||
| Programmed | 1.4 (1.0, 2.8) | 0.1 (−0.0, 0.5) | <.001‖ | 1.5 (0.7, 2.8) | 0.5 (−0.2, 1.0) | .0063‖ |
| Actual | 2.1 (1.0, 2.7) | 0.7 (0.2, 1.5) | <.001‖ | 1.6 (0.2, 3.4) | 1.2 (0.5, 1.7) | .0219‖ |
| RBC units/TBV, U/L | ||||||
| Per transfusion§ | 1.1 (0.9, 1.3) | 1.2 (1.0, 1.3) | .5435 | 1.7 (1.2, 2.0) | 1.4 (1.0, 1.6) | .0186‖ |
| Annual | 18.8 (14.6, 21.2) | 18.3 (16.7, 20.8) | .8446 | 28.6 (19.1, 36.7) | 19.6 (17.4, 26.6) | .0303‖ |
| RBC units | ||||||
| Per transfusion§ | 2.9 (2.3, 3.0) | 3.7 (3.0, 4.3) | <.001‖ | 5.6 (3.8, 6.9) | 5.2 (4.9, 7.0) | .9250 |
| Annual | 43.4 (37.7, 53.9) | 58.1 (52.5, 70.8) | <.001‖ | 96.6 (65.9, 113.5) | 85.7 (70.4, 108.2) | .3003 |
Data are presented as median (IQR) and n (%) unless otherwise specified.
Paired Wilcoxon signed-rank test.
Baseline characteristics at first included IHD-RCE procedure, stratified by RBC requirements before IHD-RCE
Mean RBC units per procedure required by patient in the 90 days before the first IHD-RCE cluster transfusion. Range of values across patients reported.
Mean values calculated within each patient cluster and used to calculate median across patients.
P < .05.
Post-RCE HbS divided by pre-RCE HbS.
Post-RCE hematocrit minus pre-RCE hematocrit.
Figure 3. RBC units per liter TBV required for standard RCE vs IHD-RCE, stratified by unit requirements at start of the IHD-RCE cluster.
In patients requiring on average ≤4 RBC units per RCE in the 90 days before the first IHD-RCE cluster transfusion, IHD-RCE and standard RCE did not differ in RBC units required per liter TBV. In patients requiring >4 RBC units before transition to IHD-RCE, IHD-RCE required less RBC units per liter of blood volume. Standard RCE results are shown in yellow and IHD-RCE results are shown in blue. The box indicates the IQR and inside line represents the median. The whiskers equal minimum or maximum up to 1.5× IQR. Outliers (>1.5× IQR from edges of box) are marked as dots. P values were calculated using the paired Wilcoxon rank-sum test.
Sensitivity analyses incorporating minimum hemodilution requirements and excluding initial RCE procedures in each cluster showed results similar to those of our primary analyses, with no significant difference in units per milliliter TBV required between IHD-RCE and standard RCE (supplemental Tables 3 and 4). Analyses restricting the study population to those with HbS goal of <30% and those using the Terumo Optia device also showed similar results, as did analyses requiring clusters to be similar in duration.
Cerebrovascular outcomes
No overt strokes occurred during study follow-up. New or progressive cerebral vasculopathy developed in 3 patients during standard RCE, 3 patients during IHD-RCE, and 1 patient during both the standard and IHD-RCE clusters (supplemental Table 5). One patient developed a new silent cerebral infarct while undergoing IHD-RCE. There was no significant difference in cerebrovascular disease progression between the standard and IHD-RCE clusters (P = .714).
Discussion
This single-center cohort of patients with SCD-SS undergoing chronic RCE is, to our knowledge, the largest cohort of patients in which IHD-RCE has been directly compared with standard RCE in the same individuals. Sample sizes for previous studies range from 6 to 30 patients, who were primarily adults.14,16,17,22–25 We found that performing RCE with IHD was associated with RBC savings only in patients requiring at least 5 RBC units per procedure at the time of transition to IHD-RCE. In the overall cohort, neither unadjusted paired analyses nor a multivariable model showed significant RBC unit savings for IHD-RCE compared with standard RCE, likely due to greater representation of small TBV and low RBC unit requirements in our patient sample. For patients with a TBV of ≤3500 mL, IHD-RCE was instead estimated to require more RBC units compared with standard RCE. These findings have important practice implications for pediatric and young adult patients with SCD requiring chronic RCE transfusions and for the larger health care system, in which Rh- and K-matched RBC units are limited.12
By first removing patient blood without donor red cell replacement, and subsequently performing an exchange at a lower hematocrit, IHD-RCE should theoretically require less blood compared with standard RCE. There are several reasons why we may not have observed statistically significant differences in unit requirements when examining our entire study sample. Transfusion of whole units is typically preferred to optimize resource use and minimize donor exposures. Because we use only whole units for RCE, for patients requiring 2 to 4 units of RBC per transfusion, IHD-RCE would need to reduce RBC requirements by 25% to 50% to save 1 unit. In contrast, for patients requiring 6 to 8 RBC units per transfusion, a 12% to 16% reduction in RBC requirement would be sufficient to save 1 RBC unit. Indeed, RBC volume savings of ~8% to 11% have previously been attributed to IHD-RCE compared with standard RCE.14,24 Interestingly, in our data, although IHD-RCE led to a lower FCR in procedures with low unit requirements, no significant improvement in HbS control or increase in interprocedure interval was seen, possibly because the magnitude of FCR reduction was insufficient to cause a meaningful change in HbS levels at the next transfusion date. Our model-predicted analysis suggested RBC unit savings may occur with hemodilution once a patient’s TBV is greater than ~4500 mL (an individual with a body weight of ~64 kg).
Several factors likely led to our findings of significant IHD-RCE unit savings only with the stratified analyses. The peak age of abnormal transcranial Doppler ultrasound, the most common reason for chronic transfusions, is at or before age 7 years.26 Our typical institutional practice is to transition from simple RBC transfusions to automated RCE when RCE can be done peripherally, generally at ~7 to 9 years of age. Therefore, our cohort was skewed toward younger patients with small TBV values and RBC unit requirements, limiting our overall model’s power at higher TBV values. Our sample size did not allow for multivariable stratified analysis; however, paired unadjusted analyses stratified by baseline RBC units demonstrated clear IHD-RCE–related savings in patients requiring >4 units per transfusion.
The use of IHD-RCE for patients who are unlikely to benefit from RBC unit savings has important implications. First, programming an additional hemodilution step adds operational complexity, leaving additional opportunity for error. The acute anemia induced by hemodilution poses a theoretical neurologic risk, particularly in patients with SCD who have decreased cerebrovascular reserve and are at risk of stroke. The negative impacts of acute, severe anemia on neurologic and cognitive outcomes are well established27–29 however, it is unclear whether the transient and relatively milder anemia of IHD-RCE is sufficient to cause harm. Reassuringly, we did not find an increase in cerebrovascular complications associated with IHD-RCE, but our sample size was limited, and a prospective study with regularly scheduled magnetic resonance imaging and angiography, as well as neurocognitive testing, would be needed to definitively determine the impact of IHD-RCE on neurologic outcomes. Finally, our results demonstrated that IHD-RCE requires more time compared with standard RCE, although this difference is primarily related to the larger unit requirements with IHD-RCE.
Previous studies in adult patients have shown RBC unit savings with IHD-RCE. Sarode et al reported an average of 1 RBC unit saved per procedure and fewer procedures annually with IHD-RCE, whereas Hequet et al reported an average of 2 RBC units per procedure saved with IHD-RCE.14,16 Similar to our findings, Hequet et al found that patients with larger TBV had more IHD-related savings. Our study differed from these aforementioned cohorts with patients having smaller TBVs and requiring fewer red cell units per procedure. In these previous studies, donor units required ranged from 6 to 13 units per transfusion, compared with a median of 3 and 4 units in standard RCE and IHD-RCE, respectively, in our cohort.14–16,22,24,25 Patients included in our analysis primarily targeted an HbS goal of <30%, whereas published studies of IHD-RCE had greater representation of patients with higher HbS goals.14,16,25
Variation in institutional transfusion practices may also contribute to RBC unit savings observed. Some apheresis practices specify a constant FCR across transfusions, with adjustment of RBC units as needed to meet HbS goals,14,16,22,23,25,30 whereas others individualize the goal FCR between patients and procedures, as is our practice.17,24,31 Certain institutions give greater importance to long intertransfusion intervals, which require increased red cell units per procedure and lower overall efficiency.14,16,22 Approaches to hemodilution differ as well, including the minimum age or TBV of patients eligible for IHD-RCE and the nadir hematocrit used.20,31 Further investigation into RCE practices and opportunities for standardization should therefore be pursued.
Our study has several limitations, including those inherent to our retrospective study design. Our single-center cohort limits generalizability to centers using different RCE protocols, and our sample size of 45 patients does not represent the full spectrum of patients requiring RCE. Our inclusion criteria resulted in a largely adolescent cohort that experienced significant growth and underwent puberty during the study period, leading to potential selection bias. We performed preliminary analyses using broader inclusion criteria, with fewer restrictions regarding timing or consistency between consecutive transfusions. These analyses included additional patients and procedures with adult TBVs but were not informative due to numerous confounders. We used within-patient paired analyses and a multivariable, mixed-effects model to minimize confounding in the comparison between standard RCE and IHD-RCE; however, we could not account for unmeasured confounders, particularly those related to older patient age at time of IHD-RCE. We could not explore possible IHD-RCE–related adverse effects outside of cerebrovascular disease, because our database did not record comprehensive clinical variables such as vital signs. Finally, differential iron loading between RCE types was outside the scope of our study because there are many contributing factors, including the increment from pre- to post-RCE hematocrit, age, pubertal status, degree of iron overload at time of chronic RCE initiation, and chelation adherence. Moreover, the use of serum ferritin to estimate iron burden in SCD has limitations that were evidenced in a number of studies,6,32,33 and, for many, the transition to IHD-RCE occurred before liver R2 magnetic resonance imaging assessment was common.
In summary, this study provides important practice implications for the use of IHD-RCE to potentially decrease red cell unit requirements for patients with SCD requiring chronic transfusion therapy. Our study is, to our knowledge, the first to examine, using intra-patient comparisons, IHD-RCE in children and adolescent patients who require fewer RBC units per RCE vs adults. Our results suggest that, for patients with a TBV of ≤3500 mL and/or those requiring ≥4 RBC units per RCE, IHD-RCE will not provide unit savings and, in fact, requires more blood per transfusion. Thus, the hemodilution step may be reserved for patients with a TBV of >3500 mL requiring at least 5 RBC units per procedure. An individualized approach, particularly during times of growth and puberty, is required to determine whether IHD-RCE will provide benefit to the patient. Future work comparing approaches to RCE, with and without hemodilution, will be important to standardize care and improve transfusion outcomes for all patients with SCD.
Supplementary Material
The full-text version of this article contains a data supplement.
Key Points.
IHD-RCE can reduce donor red cell requirements when >4 RBC units per RCE are needed.
IHD-RCE may increase donor red cell requirements in pediatric patients with TBVs of ≤3500 mL compared with standard RCE.
Acknowledgments
The authors thank the patients who enrolled in the study, blood donors, and members of the Children’s Hospital of Philadelphia (CHOP) Blood Bank and Apheresis Service who contributed to the study. They especially thank Haewon C. Kim who established the Apheresis Service at CHOP, contributed to the advent of the isovolemic hemodilution protocol decades ago, and has been very generous with sharing her expertise in apheresis to the international community.
This work was supported, in part, by the National Institutes of Health/National Heart, Lung, and Blood Institute HL147879 (S.T.C. and D.F.F.), HL169401 (S.T.C. and D.F.F.), T32 HL0007150 (A.S.D.), and a Distinguished Chair in Pediatrics (S.T.C.).
Conflict-of-interest disclosure:
J.L.K. has a consultancy role with Agios, Forma Therapeutics, Vertex Pharmaceuticals, Bio-Marin, Chiesi, Bristol Myers Squibb, bluebird bio Inc, Silence Therapeutics, and Novo Nordisk; and receives research funding from Forma Therapeutics, Agios, bluebird bio Inc, Editas Medicine, Pfizer, and Vertex Pharmaceuticals. The remaining authors declare no competing financial interests.
Footnotes
Presented in abstract form at the 66th annual meeting of the American Society of Hematology, San Diego, CA, 7 December 2024 (https://doi.org/10.1182/blood-2024-207479)
References
- 1.DeBaun MR, Jordan LC, King AA, et al. American Society of Hematology 2020 guidelines for sickle cell disease: prevention, diagnosis, and treatment of cerebrovascular disease in children and adults. Blood Adv. 2020;4(8):1554–1588. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 2.Linder GE, Chou ST. Red cell transfusion and alloimmunization in sickle cell disease. Haematologica. 2021;106(7):1805–1815. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 3.Sharma A How I treat sickle cell disease with gene therapy. Blood. 2024;144(26):2693–2705. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 4.Zakieh A, Mercure-Corriveau N, Lanzkron S, et al. Chronic automated red cell exchange therapy for sickle cell disease. Transfusion. 2024;64(8): 1509–1519. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 5.Vichinsky EP, Luban NL, Wright E, et al. Prospective RBC phenotype matching in a stroke-prevention trial in sickle cell anemia: a multicenter transfusion trial. Transfusion. 2001;41(9):1086–1092. [DOI] [PubMed] [Google Scholar]
- 6.Adamkiewicz TV, Abboud MR, Paley C, et al. Serum ferritin level changes in children with sickle cell disease on chronic blood transfusion are nonlinear and are associated with iron load and liver injury. Blood. 2009;114(21):4632–4638. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 7.Chou ST, Jackson T, Vege S, Smith-Whitley K, Friedman DF, Westhoff CM. High prevalence of red blood cell alloimmunization in sickle cell disease despite transfusion from Rh-matched minority donors. Blood. 2013;122(6):1062–1071. [DOI] [PubMed] [Google Scholar]
- 8.Hudson KE, Fasano RM, Karafin MS, Hendrickson JE, Francis RO. Mechanisms of alloimmunization in sickle cell disease. Curr Opin Hematol. 2019; 26(6):434–441. [DOI] [PubMed] [Google Scholar]
- 9.Rollins MR, Chou ST. Adverse events of red blood cell transfusions in patients with sickle cell disease. Transfus Apher Sci. 2022;61(5):103557. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 10.Chou ST, Alsawas M, Fasano RM, et al. American Society of Hematology 2020 guidelines for sickle cell disease: transfusion support. Blood Adv. 2020;4(2):327–355. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 11.Logistics Kelly S., risks, and benefits of automated red blood cell exchange for patients with sickle cell disease. Hematology Am Soc Hematol Educ Program. 2023;2023(1):646–652. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 12.Leiva-Torres GA, Cigna M, Constanzo-Yanez J, et al. Transfusing children with sickle cell disease using blood group genotyping when the pool of Black donors is limited. Transfusion. 2024;64(4):716–726. [DOI] [PubMed] [Google Scholar]
- 13.Kim HC, Dugan NP, Silber JH, et al. Erythrocytapheresis therapy to reduce iron overload in chronically transfused patients with sickle cell disease. Blood. 1994;83(4):1136–1142. [PubMed] [Google Scholar]
- 14.Sarode R, Matevosyan K, Rogers ZR, Burner JD, Rutherford C. Advantages of isovolemic hemodilution-red cell exchange therapy to prevent recurrent stroke in sickle cell anemia patients. J Clin Apher. 2011;26(4):200–207. [DOI] [PubMed] [Google Scholar]
- 15.Quirolo K, Bertolone S, Hassell K, et al. The evaluation of a new apheresis device for automated red blood cell exchange procedures in patients with sickle cell disease. Transfusion. 2015;55(4):775–781. [DOI] [PubMed] [Google Scholar]
- 16.Hequet O, Poutrel S, Connes P, et al. Automatic depletion with Spectra Optia allows a safe 16% reduction of red blood cell pack consumption in exchanged sickle cell anemia patients. Transfusion. 2019;59(5):1692–1697. [DOI] [PubMed] [Google Scholar]
- 17.Uminski K, Perelman I, Tinmouth AT, Mack J. Use of automated isovolemic hemodilution red-cell exchange in patients with sickle cell disease: a Canadian single center experience. Transfusion. 2025;65(2):325–332. [DOI] [PubMed] [Google Scholar]
- 18.Su L, Bowie A, Baker K, Dent K, Morgan K, Smith-Fields C. Surveillance of post-procedure red cell gain/loss following red cell exchange procedures with and without depletion in pediatric sickle cell patients: a single institution experience. J Clin Apher. 2016;31(2):96–97. [Google Scholar]
- 19.Castro AE, Dwyre DM, Medina MA, Fernando LP. Developing and growing a chronic red cell exchange program: a single institution experience. J Clin Apher. 2016;31(2):100. [Google Scholar]
- 20.Karafin MS, Hendrickson JE, Kim HC, et al. Red cell exchange for patients with sickle cell disease: an international survey of current practices. Transfusion. 2020;60(7):1424–1433. [DOI] [PubMed] [Google Scholar]
- 21.National Heart, Lung, and Blood Institute. Evidence-Based Management of Sickle Cell Disease: Expert Panel. 2014. Accessed 29 December 2022. https://www.nhlbi.nih.gov/health-topics/evidence-based-management-sickle-cell-disease
- 22.Poullin P, Sanderson F, Bernit E, Brun M, Berdah Y, Badens C. Comparative evaluation of the depletion-red cell exchange program with the Spectra Optia and the isovolemic hemodilution-red cell exchange method with the COBE Spectra in sickle cell disease patients. J Clin Apher. 2016;31(5):429–433. [DOI] [PubMed] [Google Scholar]
- 23.Kim J, Joseph R, Matevosyan K, Sarode R. Comparison of Spectra Optia and COBE Spectra apheresis systems’ performances for red blood cell exchange procedures. Transfus Apher Sci. 2016;55(3):368–370. [DOI] [PubMed] [Google Scholar]
- 24.Ziemba Y, Xu C, Fomani KM, et al. Safety and benefits of automated red cell depletion-exchange compared to standard exchange in patients with sickle cell disease undergoing chronic transfusion. Transfusion. 2021;61(2):526–536. [DOI] [PubMed] [Google Scholar]
- 25.Reddy MS, Alkashash A, Nord A, Tetrick A. Revisiting the benefits of isovolemic hemodilution red cell exchange for sickle cell disease. J Clin Apher. 2023;38(5):522–528. [DOI] [PubMed] [Google Scholar]
- 26.Kanter J, Phillips S, Schlenz AM, et al. Transcranial doppler screening in a current cohort of children with sickle cell anemia: results from the DISPLACE study. J Pediatr Hematol Oncol. 2021;43(8):e1062–e1068. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 27.Weiskopf RB, Kramer JH, Viele M, et al. Acute severe isovolemic anemia impairs cognitive function and memory in humans. Anesthesiology. 2000; 92(6):1646–1652. [DOI] [PubMed] [Google Scholar]
- 28.Dowling MM, Quinn CT, Rogers ZR, Buchanan GR. Acute silent cerebral infarction in children with sickle cell anemia. Pediatr Blood Cancer. 2010; 54(3):461–464. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 29.Scothorn DJ, Price C, Schwartz D, et al. Risk of recurrent stroke in children with sickle cell disease receiving blood transfusion therapy for at least five years after initial stroke. J Pediatr. 2002;140(3):348–354. [DOI] [PubMed] [Google Scholar]
- 30.M Ross J, Forté S, Mercure-Corriveau N, et al. Automated red blood cell exchange with a post-procedure haematocrit targeted at 34% in the chronic management of sickle cell disease. Br J Haematol. 2024;205(4):1556–1564. [DOI] [PubMed] [Google Scholar]
- 31.Cunard R, Gopal S, Kopko PM, Dang MU, Hazle KM, Sanchez AP. Comprehensive guide to managing a chronic automated red cell exchange program in sickle cell disease. J Clin Apher. 2022;37(5):497–506. [DOI] [PubMed] [Google Scholar]
- 32.Puliyel M, Sposto R, Berdoukas VA, et al. Ferritin trends do not predict changes in total body iron in patients with transfusional iron overload. Am J Hematol. 2014;89(4):391–394. [DOI] [PubMed] [Google Scholar]
- 33.Tsitsikas DA, Nzouakou R, Ameen V, Sirigireddy B, Amos RJ. Comparison of serial serum ferritin measurements and liver iron concentration assessed by MRI in adult transfused patients with sickle cell disease. Eur J Haematol. 2014;92(2):164–167. [DOI] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.