Real‐World Use of Rituximab in the Treatment of Cold Agglutinin Disease in the United States: A Retrospective Study

ABSTRACT

Introduction

Rituximab is used off‐label for treating cold agglutinin disease (CAD) in the United States; however, real‐world data on its use are limited.

Methods

This study examined (1) treatment patterns, (2) prespecified safety outcomes, and (3) changes in hemolytic markers post rituximab infusion (threshold change from baseline: hemoglobin [Hb]: ≥2 g/dL increase; bilirubin or lactate dehydrogenase [LDH]: ≥50% decrease). Of 611 patients with CAD (Cohort 1) identified using Optum's de‐identified Market Clarity database (2007‐2021), 94 who received rituximab (Cohort 2) were included.

Results

Rituximab was predominantly used as monotherapy; the median (interquartile range; IQR) number of rituximab courses per patient was 1.0 (1.0–2.0); 24 (25.5%) patients had ≥1 incomplete course and 37 (39.4%) required blood transfusion. The incidence rate (IR) of serious infections per 1000 patient‐years (95% CI) post rituximab initiation was three times higher than before rituximab initiation (637.1 [242.2–1032.0] vs. 245.4 [125.1–365.6]). Cohort 2 had a similar IR of thromboembolic events and slightly higher hospitalization rates and deaths than Cohort 1. After the first course of rituximab, 69.5% (41/59), 59.6% (28/47), and 31.3% (10/32) of patients reached thresholds for Hb, bilirubin, and LDH, respectively. The median (IQR) duration of improvement was 44 (13.3–90.8), 98 (29.0–257.0), and 93 (21.3–161.8) days for Hb, bilirubin, and LDH, respectively. Reversal from threshold occurred in 68.3% (28/41), 53.6% (15/28), and 80.0% (8/10) of patients for Hb, bilirubin, and LDH, respectively.

Conclusion

Rituximab showed a limited durability of response and an increased risk of infection, suggesting the need for more effective and safer treatment options.

Trial Registration: The authors have confirmed clinical trial registration is not needed for this submission.

1. Introduction

Cold agglutinin disease (CAD) is a rare autoimmune hemolytic anemia (AIHA) mediated by autoantibodies that bind with antigens on red blood cells when temperatures are at or below 37°C, resulting in erythrocyte agglutination. This condition leads to the activation of the classical complement pathway, triggering a cascade of events, including complement‐mediated hemolysis and chronic inflammation [1]. CAD is associated with a bone marrow clonal B‐cell lymphoproliferative disorder without any apparent clinical or radiological evidence of malignancy, which is distinct from well‐characterized B‐cell lymphoma types [2, 3, 4]. This distinction was evident in a study involving 232 patients with CAD, reporting a transformation to diffuse large B‐cell lymphoma in only 3.4% of cases over an 8‐year period [5].

CAD accounts for 15% to 30% of AIHAs, with an estimated prevalence of 5 to 20 people per million and an incidence of 0.5 to 1.9 cases per million every year [5]. Patients with CAD experience uncontrolled complement‐mediated hemolysis, with most remaining anemic because of the inability to compensate for their hemolysis [6]. Prominent fatigue, caused by both anemia and the release of anaphylatoxins during complement activation, is a key symptom of the disease [7, 8, 9]. Moreover, patients with CAD are at an increased risk of early mortality and thromboembolic events (TEs), which can substantially impact their quality of life (QoL) [10, 11, 12, 13, 14]. The heightened risk of mortality, as emphasized by Bylsma et al. [11], may encompass various factors beyond TEs, including chronic anemia and end‐organ damage [5]. Thus, the primary treatment goal in CAD is to control the underlying hemolysis, which predominantly causes anemia, fatigue, and a decline in QoL, and to manage circulatory symptoms [7, 15, 16].

The pharmacological treatment of CAD mainly involves two approaches: inhibiting complement‐mediated hemolysis and limiting clonal B‐cell lymphoproliferation (off‐label) [17]. Until recently, no approved therapies were available to treat chronic hemolysis in patients with CAD. Sutimlimab is a humanized, monoclonal antibody that selectively inhibits the complement pathway at C1s and is the first approved treatment for CAD, currently approved in the United States, the European Union, Japan, Israel, and Switzerland [18].

Rituximab, a B‐cell‐targeting monoclonal antibody (anti‐CD20), is approved for the treatment of B‐cell lymphoma and severe rheumatoid arthritis [19]. Because of its ability to target B cells, rituximab has been used off‐label as a first‐line treatment for AIHAs, including CAD [20]. Two prospective studies with 20 and 27 patients showed that rituximab monotherapy at a weekly dose of 375 mg/m² for 4 weeks led to a complete response in 5% and 3% of patients and a partial response in 40% and 51% of patients, respectively; the median duration of response was 6.5 and 11 months (range: 2–42 months), respectively [21, 22]. For patients without contraindications to cytotoxic therapy, rituximab can be used in combination with bendamustine or fludarabine [7].

Despite the off‐label use of rituximab in CAD, real‐world evidence on the patterns of rituximab use, its safety, its effect on the levels of hemolytic biomarkers, and the duration of benefit in patients with CAD is limited in the United States. Most studies evaluating the effectiveness of rituximab in CAD are small and open‐label, with 30 or fewer patients, and are primarily based in Europe [21, 22, 23, 24]. The present observational study aimed to characterize patients with CAD from the United States who initiated rituximab and to describe treatment outcomes.

2. Materials and Methods

2.1. Objectives

(1) To characterize patients with CAD who initiated rituximab and examine their treatment patterns; (2) To assess prespecified safety outcomes of interest in patients with CAD overall and those initiating rituximab, which included comparing serious infections (SIs) prior to/post rituximab treatment within Cohort 2 (rituximab users) and the rates of hospitalizations, TEs, and mortality between Cohort 2 and Cohort 1 (the overall CAD cohort); and (3) To examine changes in the levels of hemolytic biomarkers (hemoglobin [Hb], bilirubin, and lactate dehydrogenase [LDH]) in Cohort 2 post rituximab infusion.

2.2. Study Design

This real‐world, retrospective, observational, cohort study identified adult patients (aged ≥18 years) diagnosed with CAD from Optum's de‐identified Market Clarity database (dataset 2007–21). The diagnosis of CAD was associated with a combination of the International Classification of Diseases, Tenth Revision (ICD‐10) code: D59.12, and Optum's natural language processing of electronic health records (EHRs) and physicians’ notes for CAD diagnosis. The exact algorithm used is described in Appendix S1.

Adult patients with a CAD diagnosis were excluded if they (1) did not have continuous medical activity during the baseline period (i.e., 365 days prior to and including the index date), (2) developed comorbidities associated with cold agglutinin syndrome (CAS; secondary CAD) from the start of all available data to the index date (inclusive), (3) had a previous diagnosis of mycoplasma, cytomegalovirus, or Epstein–Barr virus within 2 weeks prior to the index date (inclusive), or (4) used rituximab prior to the index date.

The index date for Cohort 1 was the date of the first CAD algorithm event during the patient identification period, i.e., the estimated CAD diagnosis date. The index date for the subset of Cohort 1 who received ≥1 rituximab infusion on or after their CAD diagnosis date (Cohort 2) was the date of the first rituximab infusion (Figure 1).

FIGURE 1.

Study cohorts. CAD, cold agglutinin disease; CAS, cold agglutinin syndrome; N, total number of patients; n, number of patients under the specified category.

For Cohort 1, the follow‐up period started from the index date and continued until a gap of >365 days* in medical activity (*defined as a gap of >365 days between EHR events), death, the end of patient data, or the end of the study period (September 30, 2021). The follow‐up was divided into specific time periods relative to the index date, with censoring applied to each period; the periods included 0 to 90, 91 to 180, 181 to 360, 361 to 540, and 541 to 720 days after the index date.

The follow‐up period for Cohort 2 started from the first initiation of rituximab treatment and had the same ending rules as Cohort 1. In addition, course‐specific follow‐up periods were defined for Cohort 2 with a censoring condition added if another rituximab treatment regimen was initiated. All Cohort 2 follow‐up periods were segmented into specific time intervals with censoring applied, ranging from 0 to 90, 91 to 180, 181 to 360, 361 to 540, and 541 to 720 days after the index.

2.3. Outcomes

Baseline demographics and clinical characteristics were reported at the index date (or in the baseline period) for both cohorts. In Cohort 2, rituximab treatment patterns were assessed from the date of rituximab initiation until censoring, including the number of courses, treatment type (monotherapy or combination therapy), complete or incomplete courses, the receipt of blood transfusion (BT), and corticosteroid use (oral or intravenous [IV]). The monotherapy was a 4‐week treatment with one administration of rituximab weekly; the course was considered incomplete if fewer than four administrations were received. The combination therapy was a 4‐month treatment with one administration of rituximab per month, along with the administration of a combination drug (fludarabine, bendamustine, cyclophosphamide [±dexamethasone], doxorubicin hydrochloride, or vincristine sulfate) within 3 days of rituximab administration; the course was considered incomplete if fewer than four administrations were received. Please refer to Table S1 for more details on the treatment description.

The number and percentage of patients with defined outcomes and incidence rates (IRs) were calculated for prespecified safety outcomes of interest, i.e., SIs, all‐cause hospitalizations, TEs, and mortality, during the full follow‐up period (January 1, 2007 to September 30, 2021). The outcomes of interest were identified using ICD‐9/ICD‐10 codes, as detailed in Appendix S2. The IRs of SIs per 1000 patient‐years before and after (within 3–6 months of the first infusion) rituximab treatment were assessed in Cohort 2. The period over which SIs were assessed before rituximab infusion varied in length across patients, with the assessment starting at the CAD diagnosis. The period of 3 to 6 months following rituximab infusion was selected to exclude SIs that started before rituximab initiation and continued after rituximab initiation. To avoid double counting, any SIs recorded within 60 days of one another were considered the same or continuing infections. For example, if a patient had a diagnosis of sepsis on January 1st and then another on January 30th, it would have been considered the same infection as both diagnoses occurred within 60 days. However, if the diagnosis had been for a different type of infection (e.g., pneumonia and skin abscess), the total number of infections would have been counted as 2.

The changes in the levels of hemolytic markers (Hb, bilirubin, and LDH) from pre‐ to post‐rituximab treatment were assessed in Cohort 2. Laboratory values were assessed >5 weeks following the start of a rituximab course until censoring, defined as the earliest of (1) relapse; (2) an increase of ≥2 g/dL in the Hb level compared with the index date levels and a decrease of ≥50% in the bilirubin or LDH index date levels were considered the threshold changes for Hb, bilirubin, and LDH. The number of patients reaching the threshold level of hemolytic markers, the time from rituximab initiation, and the duration for which the levels of hemolytic biomarkers did not reverse after reaching the threshold were assessed. Table S2 presents the definitions of the terms used to describe the treatment patterns and the threshold levels of hemolytic biomarkers. Given that rituximab typically begins to show its effects after approximately 5 weeks, the assessment of the response was started on the 29th day after the first rituximab injection [22].

3. Results

A total of 611 patients with CAD were identified (Cohort 1). Of the 611 patients, 124 received rituximab. However, 30 patients were excluded because of CAS‐related diagnoses occurring after the initial CAD diagnosis, indicating a possible correction of the initial CAD diagnosis or an evolution of the disease. Therefore, a total of 94 (15.4%) patients were included in the rituximab cohort (Cohort 2) (Figure 1). Overall, patients in Cohort 2 had more severe disease at the time of treatment initiation than those in Cohort 1 at diagnosis, as evidenced by the lower levels of Hb (mean [standard deviation, SD]: 9.0 [2] vs. 10.8 [2.6] g/dL); the higher levels of bilirubin (mean [SD]: 2.3 [2.0] vs. 1.7 [3.7] µmol/L); the higher Charlson Comorbidity Index scores (mean [SD]: 2.0 [2.0] vs. 1.5 [2.1]); increased healthcare utilization, as reflected by patients with one or more hospitalizations (n [%]: 47 [50.0] vs. 140 [22.9]); and a greater prevalence of corticosteroid use (n [%]: 54 [57.4] vs. 152 [24.9]) (Table 1).

TABLE 1.

Baseline demographics and clinical characteristics.

Parameters	All patients with CAD (Cohort 1) N = 611	Patients with CAD initiating rituximab (Cohort 2) N = 94
Age, median (IQR) a (years)	70 (59.0–77.0)	72 (58.0–78.0)
Females, n (%) a	384 (62.8)	62 (66.0)
Season during index date, n (%)
Spring	163 (26.7)	23 (24.5)
Summer	110 (18.0)	15 (16.0)
Fall	163 (26.7)	24 (25.5)
Winter	175 (28.6)	32 (34.0)
Charlson Comorbidity Index scores, mean (SD) b	1.5 (2.1)	2.0 (2.0)
Patients with ≥1 hospitalization, n (%) b	140 (22.9)	47 (50.0)
Patients with ≥1 emergency room visit, n (%) b	131 (21.4)	26 (27.7)
Patients with corticosteroid use, n (%) b	152 (24.9)	54 (57.4)
Patients with Hb data available, n (%) b	462 (75.6)	91 (96.8)
Hb, mean (SD) (g/dL)	10.8 (2.6)	9.0 (2)
Patients without anemia (Hb ≥12 g/dL), n (%)	156 (33.8)	5 (5.5)
Patients with mild anemia (Hb ≥10 to <12 g/dL), n (%)	129 (27.9)	25 (27.5)
Patients with moderate anemia (Hb ≥8 to <10 g/dL), n (%)	110 (23.8)	30 (33.0)
Patients with severe anemia (Hb <8 g/dL), n (%)	67 (14.5)	31 (34.1)
Patients with bilirubin data available, n (%) b	421 (68.9)	87 (92.6)
Bilirubin, mean (SD) (µmol/L)	1.7 (3.7)	2.3 (2.0)
Patients with elevated bilirubin (>1.2 mg/dL), n (%)	170 (40.4)	62 (71.3)
Patients with LDH data available, n (%) b	230 (37.6)	81 (86.2)
LDH, mean (SD) (U/L)	410.8 (463.1)	535.9 (648.3)
Patients with elevated LDH (>250 U/L), n (%)	127 (55.2)	60 (74.1)

Abbreviations: CAD, cold agglutinin disease; Hb, hemoglobin; IQR, interquartile range; LDH, lactate dehydrogenase; N, total number of patients; n, number of patients under the specified category; SD, standard deviation.

^a At the index date.

^b Within 1 year before the index date. Biomarker value used for the calculations is the last reported value over the 1‐year baseline period.

3.1. Rituximab Treatment Patterns in Cohort 2

The median time from the CAD diagnosis to rituximab initiation was 43.5 days (interquartile range [IQR]: 13.0–308.3). The median (IQR) number of rituximab courses (Table S1) per patient and that of rituximab infusions per course were 1.0 (1.0–2.0) and 4.0 (4.0–5.3), respectively. The median (IQR) duration of rituximab treatment (the number of days from the first rituximab infusion to the last rituximab infusion) was 21.0 (21.0–64.3) days. These data reflect that for the majority of patients, the standard rituximab dosing schedule was used. Of the 94 patients, about 80% completed the first course of rituximab treatment, while 20% did not. Of the 24 patients who were re‐treated with rituximab, about 63% completed the second course of rituximab treatment, while 38% did not (Figure 2). The median (IQR) time from the initiation of the first course of rituximab to the second course was 493.5 (234.3–799.5) days. In Cohort 2, 7 (29%) of 24 patients who received a second course of rituximab as additional therapy had an incomplete prior course.

FIGURE 2.

Patients who completed the first and the second courses of rituximab treatment. ^∗Three patients who received atypical monotherapy were excluded from the calculation of duration. Standard monotherapy: rituximab infusions, without the occurrence of any combination drugs within ±3 days of the rituximab administration, ≤21 days apart but all within a window of 42 days and without qualifying for atypical monotherapy. Atypical monotherapy: rituximab infusions, without the occurrence of any combination drugs within ±3 days of the rituximab administration, with each infusion ≤42 days apart but all within a window of 126 days. IQR, interquartile range; n, number of patients under the specified category.

A majority of the patients (n = 88; 93.6%) received the rituximab monotherapy in the first course; the combination therapy was rarely used (data were suppressed in compliance with the Centers for Medicare & Medicaid Cell Size Suppression Policy). About 39% (n = 37) of rituximab initiators required BT rescue therapy. The median (IQR) time from rituximab initiation to the first BT was 31 (1.5–110.5) days. Most of the patients receiving rituximab (n = 54; 57.4%) concomitantly received corticosteroids (Table S2). The proportion of patients using corticosteroids remained the same in the year prior to rituximab initiation and while receiving rituximab treatment (n = 54; 57.4%, for both). During the follow‐up, 33.3% (n = 18 of 54) of patients treated with corticosteroids developed an infection compared with 46.2% (n = 18 of 39) of those not treated with corticosteroids.

3.2. Safety

The most common prespecified safety outcome of interest reported in Cohort 1 was SI, with roughly half of the cohort suffering at least one SI over the total follow‐up period (n = 298; 48.8%), with a mean (SD) rate of 1.3 (6.1) infections per person‐year and an IR (95% CI) of 275.8 (244.5–307.2) per 1000 person‐years. In Cohort 2, about 40% of patients had at least one infection over the whole follow‐up (n = 37; 39.4%), with a mean (SD) rate of 0.9 (2.3) infections per person‐year and an IR (95% CI) of 280.3 (190.0–370.6) per 1000 person‐years. More notably, the IR (95% CI) of SIs (per 1000 patient‐years) in Cohort 2 was ∼3 times higher in 4 to 6 months after rituximab initiation (637.1 [242.2–1032.0]) than that before rituximab initiation (245.4 [125.1–365.6]). The most common infection after rituximab initiation was sepsis, with 12% (n = 11) of patients experiencing a sepsis event, followed by pneumonia due to an unspecified organism, with 7% (n = 7) of patients experiencing such events.

About 44% (n = 266) of patients in Cohort 1 and 40.4% (n = 38) of patients in Cohort 2 were hospitalized during the follow‐up period. In patients with at least one hospitalization, during the entire follow‐up, the median (IQR) number of hospitalizations per person was 2.0 (1.0–4.0) in Cohort 1 and 2.0 (1.0–3.3) in Cohort 2. The mean (SD) rate of hospitalizations per person‐year in Cohort 1 was 1.2 (3.9) and Cohort 2 was 1.5 (4.0). The median (IQR) time to the first hospitalization was 84.5 days (0.5–429.5) in Cohort 1 and 83.5 days (19.0–265.5) in Cohort 2. The IR (95% CI) of hospitalizations per 1000 person‐years in Cohort 1 was 219.1 (192.7–245.4) and Cohort 2 was 263.3 (179.6–347.0). During the overall follow‐up periods (Appendix S3), TEs occurred in 27.7% of patients in both Cohort 1 (n = 169/611) and Cohort 2 (n = 26/94). In patients in Cohort 1 with at least one TE, the median (IQR) number of TEs was 2.0 (1.0–3.0), and for those in Cohort 2, it was 1.0 (1.0–2.0). The mean (SD) rate of TEs (index to censor) in Cohort 1 and Cohort 2 was 0.5 (2.1) and 0.6 (1.5) events per person‐year, respectively. The IR (95% CI) per 1000 person‐years for TEs (index to censor) in Cohort 1 was 120.1 (102.0–138.2) and Cohort 2 was 150.8 (92.8–208.7). The median (IQR) time to the first TE (in days) in Cohort 1 was 205.0 (56.0–621.0) and Cohort 2 was 129.0 (28.8–320.3).

During the entire follow‐up period, the same percentage of patients, 10.6% (n = 65) in Cohort 1 and (n = 10) in Cohort 2, died with a Kaplan–Meier‐estimated survival probability (95% CI) of 1.0 (0.9–1.0) and 0.9 (0.9–1.0) at 1 year and 0.9 (0.9–1.0) and 0.9 (0.8–1.0) at 2 years, respectively. In those who died, the median (IQR) number of days to death was 697.0 (150.0–1362.5) in Cohort 1 and 335.0 (66.5, 612.5) in Cohort 2. The IR (95% CI) of death per 1000 person‐years in Cohort 1 was 34.4 (26.1–42.8) and Cohort 2 was 46.7 (17.8–75.7).

3.3. Changes in Hemolytic Biomarkers Post Rituximab Treatment in Cohort 2

Biomarker data (for at least one biomarker) were available at the index and during the follow‐up for 66% (n = 62/94) of patients in Cohort 2. About 70% (n = 41/59) of patients with Hb data available had an increase of ≥2 g/dL (threshold value of Hb for change from baseline). The median (IQR) time to reach this Hb level was 48.0 (33.5–85.0) days (Figure 3A); this improvement was maintained above the threshold for a median (IQR) of 44.0 (13.3–90.8) days. The Hb levels in 68.3% (n = 28/41) of these patients reversed below the threshold (Figure 3B).

FIGURE 3.

Changes in hemolytic biomarker levels in patients receiving rituximab. (A) Patients reaching the threshold biomarker levels, and (B) patients with subsequent reversal of biomarker levels from the threshold. Hb, hemoglobin; IQR, interquartile range; LDH, lactate dehydrogenase; n, number of patients under the specified category.

About 60% (n = 28/47) of patients with bilirubin laboratory data had a ≥50% decrease in the bilirubin level; the median (IQR) time to reach this threshold bilirubin level was 68.0 (33.3–199.8) days (Figure 3A). The bilirubin levels remained below the threshold for a median (IQR) of 98.0 (29.0–257.0) days; 53.6% (n = 15/28) of these patients had a reversal in the bilirubin levels above the threshold (Figure 3B).

About 31% (n = 10/32) of patients with LDH laboratory data had a ≥50% decrease in the LDH level; the median (IQR) time to reach this threshold LDH level was 67.5 (45.3–211.3) days (Figure 3A). The LDH levels remained below the threshold for a median (IQR) of 93.0 (21.3–161.8) days; 80.0% (n = 8/10) patients had a reversal in the LDH levels above the threshold (Figure 3B).

4. Discussion

Previous Europe‐based studies focusing on rituximab treatment for CAD have been prospective in design with a limited sample size, yielding minimal insights into the real‐world use of rituximab.

Despite the data suggesting the enhanced efficacy of rituximab with combination therapy, this study provides real‐world evidence that rituximab was largely administered as monotherapy (>95%) [22, 25, 26]. A possible explanation could be that rituximab monotherapy is preferred over combination therapy in patients with primary CAD (i.e., who do not have lymphoma or other cancers), particularly, as combination therapy might not be suitable for frail or elderly patients who form the majority of CAD cases [12]. About 25.5% (n = 24) of patients treated with rituximab had at least one incomplete course of rituximab during the follow‐up; the reasons for rituximab discontinuation are unknown but may be related to intolerance issues. It is also possible that this could reflect the heterogeneous treatment approach in real‐world settings, where patient compliance with weekly infusions is not always 100%, and discontinuation may be related to challenges in adhering to a 4‐week regimen of IV infusions.

A retrospective analysis of data from 89 patients with CAD reported a response rate (i.e., improvement in clinical symptoms) of 83% with rituximab monotherapy [27], which is consistent with this study, wherein 85.5% (n = 53) of patients with available data reached a threshold change from baseline in one or more biomarker measures. This finding is consistent with that of a previous study by Jia et al., wherein a persistent decrease in the levels of bilirubin or LDH was observed within 12 months of rituximab monotherapy [27, 28].

In the patients who reached the threshold change in hemolytic markers from baseline, the responses lasted a median of 64 days, and two‐thirds of patients with any improvement experienced a reversal of biomarkers’ levels from the threshold, with a median time to reversal of 40 days from reaching the threshold. It is crucial to highlight the specific data related to the time to reach the threshold for hemolytic markers, particularly Hb, which showed a relatively longer time (∼6 weeks) to respond. Furthermore, the duration of response was notably short, i.e., 6 weeks. These data underscore a delayed onset and limited sustainability post rituximab treatment, potentially restricting their real‐world utility. The duration of sustained improvement post rituximab treatment was comparatively shorter than that in previous studies, which estimated a median response duration of 6.5 to 11 months; this discrepancy may be attributable to the variable definition of response in previous studies [21, 22]. Interestingly, this is quite similar to the findings of a study by Mullins et al. that provides one of the few longitudinal reports of Hb levels in patients with CAD treated with various therapies. Mullins et al. highlighted that 67% (12 of 18 patients) had severe anemia in the first 6 months of follow‐up after their initial therapy; 14 patients received rituximab treatment, with 86% (12 of 14) of these patients experiencing severe (7, 50%) or moderate (5, 36%) anemia following treatment [29]. The short duration of sustained improvement in hemolytic biomarkers and a high rate of the reversal of biomarkers to the threshold level in the current study suggest that rituximab monotherapy may not be very effective in achieving a durable response. It is also important to note that many of the previous studies of rituximab in patients with CAD were conducted in a clinical trial setting, compared with the real‐world setting of the current study, in which laboratory values, healthcare interactions, and measurements may vary.

Previous research had estimated a TE prevalence of 17% to 31% in patients with CAD [10]. The findings of the current study are in line with these reports, with 27.7% of patients experiencing a TE during the follow‐up in both cohorts (Cohort 1: n = 169; Cohort 2: n = 26). Concerning the observed lack of reduction in TE rates with rituximab, exploring potential therapeutic mechanisms that could effectively control complement‐mediated hemolysis in patients with CAD becomes crucial, and it could underscore the limited ability of rituximab to deplete B cells sufficiently to mitigate complement activation. However, it is important to consider that Cohort 2 patients, being more severe, could have started with a higher risk of TEs, and rituximab treatment may have reduced their risk to the level of the full CAD cohort (Cohort 1). Further research is warranted to address these issues [10].

Serious infection and hospitalization were the most frequently reported safety outcomes of interest in both cohorts. Previous studies have reported the occurrence of SIs post rituximab treatment [30, 31]. In the current study, over one‐quarter (n = 27 [28.7%]) of rituximab initiators experienced an SI in the 0 to 180 days after rituximab initiation. Notably, this was prevalent in a population predominantly treated with monotherapy; the risk would be potentially higher with combination therapy, an important aspect to consider while treating the elderly population. The most common infection was sepsis (n = 11 [12%]). Ezhuthachan et al. had shown that rituximab lowers the level of immunoglobulin M (IgM) in patients with CAD and reported significant associations between the lower levels of IgM and the risk of sepsis [32, 33].

While the majority of rituximab initiators showed an improvement in the levels of biomarkers, the duration of sustained improvement was limited, and a reversal from the threshold was common (observed in over 75% of patients). However, these observations are aligned with the mechanism of action of rituximab.

4.1. Limitations

This study provides important real‐world evidence on the characteristics of patients receiving rituximab, treatment patterns, and the safety of rituximab in patients with CAD, which can help in making informed treatment decisions. Although the current study had a relatively small sample size, given the rarity of CAD, this sample size was larger than that of most of the published studies.

EHR data are primarily collected for medical purposes and are prone to incomplete or inaccurate coding of diagnoses. This may lead to potential misclassification or underreporting of baseline characteristics or outcomes of interest. Also, no information is available on whether the prescriptions recorded in medical records were actually filled. A measurement error in medication use can be a potential source of exposure misclassification and residual confounding. However, as rituximab is identified through medication administrations and procedure codes, the records are more likely to reflect the actual use. Further, medication administrations were limited to the dates when rituximab was administered, rather than ordered, providing greater confidence that the medication was actually administered.

Another potential limitation is the presence of a diagnosis code in some instances may reflect a “rule‐out” diagnosis, a diagnosis related to a family history, or an invalid diagnosis. To minimize such instances, an algorithm for identifying CAD through Signs, Diseases, and Symptoms (SDS) terms was used, which required at least three SDS mentions of CAD (refer to the algorithm in Appendix S1 for more details). The requirement of diagnosis status in the EHR portion of the data was specifically for diagnoses prefaced with “diagnosis of,” excluding options such as “family history of” or “history of,” and thus, minimizing potential limitations caused by SDS terms.

Another limitation of this analysis was the large amount of missing laboratory data, but this may actually reflect clinical practice as laboratory tests would be expected to be more frequently conducted for rituximab‐treated patients, than for non‐treated patients. The biomarker analysis was limited to Cohort 2 where a comparison was done before and after rituximab treatment. No such analyses could be conducted for Cohort 1 as only a subset of patients were treated with rituximab.

Another limitation was the variation in rituximab administration protocols. While our dataset captured rituximab doses administered in both inpatient and outpatient settings, including the first dose, differences in treatment protocols or missed doses might be the reason that some patients received only one or two administrations. However, the number of rituximab administrations was accurately recorded, and the missing doses reflected individual treatment plans or clinical decisions rather than incomplete data capture. Rituximab treatment regimens may vary based on individual patient's needs, especially after the initial 4‐week cycle, which is a known practice in chronic disease management.

It is important to note that Cohort 2 was a subset of Cohort 1; therefore, any comparison should be interpreted with caution. Regarding safety outcomes, it is important to recognize that the safety outcomes of interest in patients with CAD (both overall and those receiving rituximab) may not necessarily be related to their CAD diagnosis or rituximab initiation. Given the potential severity of CAD and the difficulty in diagnosing the disease, hospitalizations occurring closely after the CAD index date may not always reflect hospitalizations for CAD diagnosis or an exacerbation of symptoms. Similarly, an SI within 180 days after rituximab initiation may or may not be related to treatment. To mitigate this, we excluded infections occurring from the index date to Day 90.

Importantly, an underestimation of mortality is possible as more severe patients are more likely to be excluded from the study based on the time required to meet the CAD eligibility criteria (i.e., at least three CAD mentions recorded in physicians’ notes) or two ICD‐10 codes. Hence, patients who died after the first and/or second CAD mentions (or after the first ICD‐10 code) could not be included in the study. The observed associations with secondary infections, TEs, hospital admissions, and deaths in Cohort 2 should be interpreted with caution. Rituximab treatment did not reduce the risk of adverse outcomes in the severe CAD sub‐population to the level of the overall CAD population, with baseline disease severity possibly contributing to this outcome.

5. Conclusion

Overall, patients who initiated rituximab had a more severe disease than all patients with CAD. Off‐label, rituximab was predominantly given as monotherapy for the treatment of CAD, with a median of 1.0 courses per patient. About 25% of patients did not complete their rituximab course, and rescue therapy with BTs was required in 39% of patients. Patients with CAD receiving rituximab experienced three times more SIs in the period after rituximab initiation (Day 91 to Day 180) than before starting rituximab treatment. Although patients treated with rituximab had some favorable changes in their biomarkers’ levels (69% for Hb, 59% for bilirubin, and 31% for LDH), the durability of response was short, and most of the patients had a reversal in biomarkers’ levels to pre‐rituximab levels within less than 3 months. At most, 22% (13/59) of treated patients showed sustained Hb benefit (beyond ∼6 weeks). Despite the treatment with rituximab, patients with CAD continue to experience complement‐ mediated hemolysis, and anemia, and remain at risk for thrombosis, underscoring the persistent unmet needs. This situation highlights the critical need for additional effective therapies, as rituximab alone may not always provide consistent outcomes in the treatment of CAD.

Author Contributions

C.P. contributed to the study design/conception, data acquisition, and data analysis/interpretation. I.M. contributed to data interpretation. A.K. contributed to the study design, analyzing the outcomes, and structuring the communication on the outcomes. G.M. contributed to the study design/conception, and interpretation of the results. S.B. contributed to the study design, protocol review, and the review and interpretation of the study outcomes. C.H. contributed to the study design/conception, data acquisition, and data analysis/interpretation. J.L. led implementation and conducted all analyses on the Aetion platform, including developing relevant variables; defining treatment patterns, cohorts, and analysis plans; and drafting the protocol and final report. R.P. contributed to the study design/conception, data acquisition, and data analysis/interpretation. B.W. contributed to the study design/conception, data acquisition, and data analysis/interpretation. R.Y. contributed to the study design/conception, data acquisition, and data analysis/interpretation. M.G. contributed to the study design/conception, data acquisition, and data analysis/interpretation. All authors provided critical feedback, shaped the research, and contributed to writing the manuscript.

Ethics Statement

Optum's de‐identified Market Clarity database is fully compliant with the Health Insurance Portability and Accountability Act of 1996 (HIPAA). All data were de‐identified prior to acquisition, and therefore, the Institutional Review Board's approval was not required. The study was performed in accordance with ethical principles that are consistent with the Declaration of Helsinki, International Conference on Harmonisation (ICH) Good Clinical Practice (GCP), Good Pharmacoepidemiology Practice (GPP), and the applicable legislation on non‐interventional studies and/or observational studies.

Consent

The authors have nothing to report.

Conflicts of Interest

C.P. is a consultant at AstraZeneca/Alexion and Rigel; a member of a board or advisory committee at Apellis, AstraZeneca/Alexion, Rigel, Sanofi, Sobi, and Novartis and speaker's bureau at Sobi and Apellis; and receives research funding from AstraZeneca/Alexion, Rigel, Sanofi, Incyte, and Alpine Immune Sciences. I.M. is an adviser to Alpine and a member of a steering committee for Janssen and received consultancy honoraria and speaker's bureau fees from Alexion, Apellis, Alpine, Janssen, Novartis, Rigel, and Sanofi and grant/research support from Sanofi, Novartis, Alexion, Janssen, Rigel, and Incyte. A.K. is a consultant to Sanofi. G.M., S.B., C.H., R.P., B.W., and R.Y. are employees of Sanofi. G.M. and S.B. own stocks in Sanofi. C.H. was an employee of Aetion at the time this study was conducted.

J.L. is an employee of Aetion. M.G. receives personal fees from Ionis/Akcea, Prothena, Sanofi, Janssen, Aptitude Health, Ashfield, Physicians’ Education Resource, Research to Practice, Johnson & Johnson, and Celgene for the development of educational materials for i3Health; AbbVie for Data Safety Monitoring board; and Juno and Sorrento for meetings and received a research grant from the National Cancer Institute (NCI SPORE, MM SPORE 5P50 CA186781‐04).

Supporting information

Supporting Appendix 1: Description of the algorithm.

Supporting Appendix 2: Clinical outcomes and Infections codelist.

Supporting Appendix 3: ^aPatients were required to have at least 365 days of continuous medical activity and at least 1 medical encounter in baseline. Continuous medical activity was defined as a period of time where the gap between healthcare events was ≤365 days. Continuous medical activity ended upon a gap of >365 days.

^bPatients were followed until censoring defined as death, disenrollment, end of data, end of study period; patients starting a new course of rituximab, starting rescue therapy, or starting non‐rescue therapy were assessed as the first event within rituximab Course 1, Course 2 (beginning 5 weeks after the initiation of each course), and 5 to 8, 9 to 12, 13 to 16, 17 to 20, 21 to 24, 25 to 32, 33 to 40, 41 to 48, and 49 to 52 weeks after index.

AIHA, autoimmune hemolytic anemia; CAD, cold agglutinin disease; SDS, Signs, Diseases, Symptoms.

Supplementary Table 1: Treatment description.

Hb, hemoglobin; LDH, lactate dehydrogenase.

Supplementary Table 2: Definitions of the terms used to describe the treatment patterns, threshold change, and reversal

Data Availability Statement

Source data is not available for sharing due to patient data privacy regulations surrounding source data.