The Clinical Burden of Hypereosinophilic Syndrome in a Large United States Cohort

Abstract

Background:

There are limited real-world analyses of patients with hypereosinophilic syndrome (HES) in the United States.

Objective:

To describe and compare treatment patterns and disease burden between patients with diagnosed or predicted HES and those without HES with elevated blood eosinophil count (BEC).

Methods:

Open claims data were used to identify patients with ≥2 BEC >1,000 cells/μL, who were classified into 3 cohorts: patients with an HES diagnosis code (Group 1), patients identified as having HES by a claims-based prediction model (Group 2), and patients without HES with elevated BEC (Group 3). HES-related treatments, disease manifestations, HES flares, and all-cause healthcare resource utilization (HRU) were evaluated during the 12 months following a randomly selected elevated BEC. Group 3 was compared to Groups 2 and 1, separately, using Wilcoxon rank sum test for continuous variables and Chi-squared test for categorical variables.

Results:

The study included 212 patients in Group 1, 8,089 in Group 2, and 132,945 in Group 3. Approximately 62.3% of Group 1 received ≥1 HES-related treatment, with corticosteroids being the most common (59.0%). The most common disease manifestations were those related to the upper airway/pulmonary (61.8%), constitutional (46.2%), dermatologic (35.8%), and gastrointestinal systems (34.4%). Among patients in Group 1, 22.2%, 97.2% and 25.9% had ≥1 inpatient, outpatient, and emergency department visit, respectively. Compared to Group 3, Groups 1 and 2 had more corticosteroid use and HRU (all p<0.05).

Conclusion:

Patients with HES had a substantial clinical and HRU burden versus those without HES with elevated BEC.

INTRODUCTION

Hypereosinophilic syndrome (HES) is a rare and complex group of blood disorders characterized by a disproportionately high blood eosinophil count (BEC) greater than 1,500 cells/μL.(1) There is considerable heterogeneity in disease manifestations that can range from mild symptoms to potentially life-threatening disease.(1) While all organ systems can be impacted, the most commonly affected organs include the skin, respiratory tract, and gastrointestinal tract, with symptoms such as skin rash, cough, shortness of breath, fatigue, fever, weight loss, muscle aches, and joint pain, depending on the affected organs.(2, 3) Some symptoms, such as constitutional symptoms or fatigue, are not directly related to an organ-system but have effects on quality of life.

Medical management of HES aims to prevent disease progression and organ damage by reducing eosinophil levels in the blood and tissues.(1) Oral corticosteroids (OCS) are the mainstay of treatment and can induce reductions in eosinophil levels in many patients.(4) Additional therapy options include immunosuppressive and cytotoxic agents, while the tyrosine kinase inhibitor imatinib can be used to manage myeloid HES with the FIP1L1::PDGFRA mutation.(1, 4) More recently, biologic therapies, such as benralizumab, mepolizumab, and reslizumab, have shown potential benefits in reducing or depleting circulating eosinophils.(1) Mepolizumab was approved for the treatment of HES by the United States (US) Food and Drug Administration (FDA) in 2020.(5)

The epidemiology of HES in the US is not well-documented, with one older analysis estimating a prevalence of 0.3-6.3 cases per 100,000 based on data from 2001-2005.(6) However, the true prevalence of HES may exceed these estimates given the challenges associated with HES diagnosis, as its symptoms overlap with many other eosinophilic disorders.(2) Identification of patients with HES within secondary databases (e.g., insurance claims and electronic medical records) is even more challenging because of the historical lack of an International Classification of Diseases (ICD) code for the condition until October 2020, when the HES-specific ICD Tenth Revision (ICD-10) code D72.11x was introduced.(7)

As such, there are limited analyses of patients with HES in real-world clinical practice. Nevertheless, the existing real-world evidence in international cohorts has shown a substantial disease burden among patients with HES.(8–10) Most of these prior studies included relatively small sample sizes. Knowledge on HES burden at a populational level in the US has not been reported.

A prediction model was developed using a machine learning approach and least absolute shrinkage and selection operator (LASSO) variable selection to identify patients with HES in claims data, as previously described.(11) The current study leveraged this prediction model and an ICD-10 code for HES to describe patient characteristics, treatment patterns, and disease burden (including disease manifestations and HES flares/exacerbations) in a large population of patients with diagnosed or predicted HES. Additionally, the incremental disease burden was quantified relative to patients with elevated BEC but without HES.

METHODS

Data source

This study used open medical and pharmacy claims from May 2017 to December 2021 supplied from the PatientSource^® database of Source Healthcare Analytics, LLC, a Symphony Health Solutions Corporation. The database contains healthcare information on approximately 274 million patients annually throughout the US, with insurance coverage provided by a range of payers (e.g., Medicaid, Medicare, commercial insurance, and self-pay). Information is aggregated from 3 sources: pharmacy point-of-service sales, switch/network (i.e., clearing house) transactions, and additional direct prescriptions (e.g., medical and hospital claims data feeds), and includes approximately 90% of all retail prescription claims in the US. Additionally, BEC data were pulled from May 2017 to October 2021 for the study sample.

The study was considered exempt research under 45 CFR § 46.104(d)(4) as only secondary de-identified data were used in compliance with the Health Insurance Portability and Accountability Act (HIPAA), specifically, 45 CFR § 164.514.

Study design and population

A retrospective longitudinal cohort study design was used. Patients were selected based on prespecified eligibility criteria defined prior to the analysis. Eligible patients with ≥2 elevated BEC (>1,000 cells/μL) at any time in the Symphony database were identified and dates of BEC laboratory results not overlapping with any OCS use or within 30 days of end of the OCS treatment were considered as candidate index dates. Although a BEC >1,500 cells/μL is typically the cutoff used for an initial HES diagnosis, a threshold of >1,000 cells/μL was selected as our study sought to include prevalent patients who may have lower BEC related to treatment. Patients were classified into 3 cohorts based on the presence/absence of an HES diagnosis code and predicted probability of having HES based on the prediction model:(11)

• Patients with an HES diagnosis code (Group 1) were required to have ≥1 claim with an HES diagnosis (ICD-10 code: D72.11x) after October 2020 and ≥1 elevated BEC after or within 6 months prior to the initial HES or eosinophilia diagnosis. Patients with conditions that may be in the differential diagnosis of HES who received no additional HES or eosinophilia diagnoses after the first differential diagnosis were excluded.

• Patients with predicted HES (Group 2) were required to have ≥1 elevated BEC and included those without an HES diagnosis code at any time in the data but were identified as having HES by the prediction model (i.e., predicted probability of having an HES diagnosis >0.7 at any of the candidate index dates).(11) A threshold of 0.7 was chosen to maximize the positive predictive value while maintaining a reasonable negative predictive value. Further details on the threshold are detailed in the publication by Khoury et al. These patients had ≥6 months of continuous data activity after October 2020 (i.e., washout period to ensure that any use of the HES diagnosis code after the introduction of the new ICD-10 code would be captured in the data).

• Patients without HES with elevated BEC (Group 3) were required to have ≥1 elevated BEC and were defined as those without an HES diagnosis code at any time in the data, who were predicted as not having HES (i.e., predicted probability of having an HES diagnosis ≤0.7 at all candidate index dates), and who had ≥6 months of continuous data activity after October 2020.

The index date for all patients was defined among candidate index dates as the date of a randomly selected elevated BEC (Figure E1 in Online Repository). A random date was selected as there was no diagnosis to anchor on for patients without an HES diagnosis and the observation that patients with HES diagnoses had elevated BEC prior to their index. Patients were required to have ≥12 months of continuous data activity before and after the index date, where a patient was considered active in the data when consecutive encounters were dated less than 12 months apart. The baseline period comprised the 12 months prior to the index date, while the follow-up period spanned the 12 months after the index date.

Prevalence of HES was assessed among a broader population of patients with ≥1 elevated BEC and continuous data activity from October 2020 to October 2021. Prevalence of HES was assessed for the following 2 cohorts, defined similarly to Group 1 and/or Group 2 above but with criteria tailored specifically for prevalence estimation:

• Prevalence Group 1: Patients with ≥1 HES diagnosis code between October 2020 and October 2021.

• Prevalence Group 2: Patients with ≥1 HES diagnosis code between October 2020 and October 2021 AND patients with predicted HES (i.e., without an HES diagnosis code but with ≥1 elevated BEC at any time in the whole data period that met the probability threshold [>0.7] to be classified as HES, as estimated from the prediction model). A minimum of 12 months of continuous data activity around the elevated BEC was required to ensure sufficient data for inclusion in the prediction model.

Study outcomes

Pre-defined outcomes assessed during the 12-month follow-up period included HES-related treatments, diagnoses and disease manifestations by organ system, HES flares/exacerbations, and all-cause HRU per-person-per-year (PPPY; including outpatient [OP] visits, inpatient [IP] hospitalizations, length of stay per hospitalization, emergency department [ED] visits, and other visits). The definition of HES flare/exacerbation (in Group 1 and Group 2 patients only) employed in this study was similar to the definition used in the NATRON (NCT04191304) trial,(12) and was characterized by HES clinical manifestations or laboratory abnormalities that necessitated a treatment change or resulted in a hospitalization or ED visit (Figure E2 in Online Repository). Outcomes were evaluated separately in Group 1, Group 2, and Group 3. Relevant diagnosis and procedure codes are in Tables E1–E5 in Online Repository.

In addition, the prevalence of HES was evaluated in the 2 groups identified for the prevalence analysis.

Statistical analysis

Patient characteristics during the baseline period and treatment patterns, clinical outcomes, and HRU during the follow-up period were summarized using means, standard deviations (SDs), and medians for continuous variables and counts and percentages for categorical variables.

Patients in Group 3 were compared to patients in Group 2 and patients in Group 1, separately, using Wilcoxon rank sum test for continuous variables and Chi-squared or Fisher’s exact test for categorical variables. No comparisons between Group 1 and Group 2 were conducted in this study.

Prevalence of HES was calculated as the number of patients in Prevalence Group 1 and Prevalence Group 2 divided by the total number of patients with ≥1 elevated BEC from October 2020 to October 2021.

All statistical analyses were conducted using R version 3.6.3.

RESULTS

Study population

Overall, 141,246 patients met the inclusion criteria (Figure 1 and Figure E3 in Online Repository). A total of 8,301 patients were classified as patients with an HES diagnosis code (Group 1, n=212) or predicted HES (Group 2, n=8089). Group 3 comprised 132,945 patients without HES with an elevated BEC.

Figure 1. Sample selection flowchart.

Abbreviations: BEC, blood eosinophil count; HES, hypereosinophilic syndrome; OCS, oral corticosteroid.

Notes:

[1] Index dates that overlapped with any OCS use or within 30 days of end of OCS were excluded.

[2] Patients with differential diagnoses after the index diagnosis who had no additional HES diagnosis occurring after the first differential diagnosis were excluded from Group 1. Patients were further required to have at least 1 BEC >1,000 cells/μl after or within 6-month prior to the index diagnosis to be eligible for inclusion in Group 1. Patients with no HES diagnosis code were further required to have at least 6 months of continuous data activity after October 2020. Detailed sample selection flow chart is available in Figure E3 in the Online Repository.

Detailed comparisons between patients in Group 1 with an HES diagnosis code and patients in Group 2 with predicted HES identified by the machine learning prediction model are discussed in a prior publication.(11) In general, patients in Group 2 exhibited similar clinical characteristics, treatment patterns, and economic burden as the patients in Group 1. However, patients in Group 2 had a higher prevalence of allergic diseases, solid tumors, autoimmune diseases, a higher prevalence of upper airway/pulmonary, constitutional, and gastrointestinal disease manifestations, and a lower median BEC at index.

Baseline characteristics

Patients in Group 1 had a mean ± SD (median) age at index of 53.9 ± 19.6 (58.0) years, with approximately half being female (50.9%; Table 1 and Table E6 in Online Repository). Common differential diagnoses included allergic diseases (41.0%), solid tumors (27.8%) and autoimmune diseases (13.2%). Common diagnoses and disease manifestations included upper airway/pulmonary (54.2%), constitutional (37.7%), gastrointestinal (35.4%), and dermatologic conditions (25.5%; Table 2). Complete blood count with differential (43.9%), CT scan (chest, pelvis, abdomen; 17.5%) and echocardiogram (15.1%) were the most common diagnostic procedures. The mean ± SD (median) BEC at the index date was 2,991 ± 3,596 (1,694) cells/μL, with a range of 1,008–29,880 cells/μL and 58.9% of patients having a BEC>1,500 cells/μL at index.

Table 1.

Baseline demographics and differential diagnoses for patients with and without HES¹

	Group 1 Patients with an HES diagnosis code (N = 212)	Group 2 Patients with predicted HES (N = 8,089)	Group 3 Patients without HES with elevated BEC (N = 132,945)	P-value (Group 1 vs. Group 3)	P-value (Group 2 vs. Group 3)
Demographics
Age, years	53.9 ± 19.6 [58.0]	54.9 ± 20.1 [60.0]	56.7 ± 20.1 [62.0]	<0.05	<0.001
Sex				0.937	<0.001
Female	108 (50.9%)	4,605 (56.9%)	67,053 (50.4%)
Male	104 (49.1%)	3,484 (43.1%)	65,892 (49.6%)
Differential diagnoses 2
Allergic disease	87 (41.0%)	4,275 (52.8%)	23,840 (17.9%)	<0.001	<0.001
Solid tumors	59 (27.8%)	2,469 (30.5%)	18,339 (13.8%)	<0.001	<0.001
Autoimmune diseases	28 (13.2%)	1,854 (22.9%)	9,718 (7.3%)	<0.01	<0.001
Immune deficiency/dysregulation	12 (5.7%)	529 (6.5%)	1,547 (1.2%)	<0.001	<0.001
Hematological malignancy	11 (5.2%)	728 (9.0%)	2,141 (1.6%)	<0.001	<0.001
Adrenal insufficiency	4 (1.9%)	91 (1.1%)	338 (0.3%)	<0.01	<0.001
Drug or toxin hypersensitivity	3 (1.4%)	137 (1.7%)	496 (0.4%)	<0.05	<0.001
Parasitic infection	1 (0.5%)	68 (0.8%)	431 (0.3%)	0.498	<0.001
Viral infection	1 (0.5%)	55 (0.7%)	606 (0.5%)	0.621	<0.01
Cholesterol embolization	1 (0.5%)	1 (0.0%)	26 (0.0%)	<0.05	1.000
Bacterial infection	0 (0.00%)	11 (0.1%)	97 (0.1%)	1.000	0.075
Systematic mastocytosis	0 (0.00%)	10 (0.1%)	6 (0.0%)	1.000	<0.001
Fungal infection	0 (0.00%)	20 (0.2%)	113 (0.1%)	1.000	<0.001

Values are reported as mean ± SD [median] or n (%).

Abbreviations: BEC, blood eosinophil count; HES, hypereosinophilic syndrome; N, number; SD, standard deviation.

Notes:

^[1]Demographics were summarized at the index date and baseline differential diagnoses were summarized during the 12-month pre-index period.

^[2]Specific conditions included in each category and associated ICD-10 codes are provided in Online Repository Table E1.

Table 2.

Baseline clinical characteristics for patients with and without HES¹

	Group 1 Patients with an HES diagnosis code (N = 212)	Group 2 Patients with predicted HES (N = 8,089)	Group 3 Patients without HES with elevated BEC (N = 132,945)	P-value (Group 1 vs. Group 3)	P-value (Group 2 vs. Group 3)
Selected diagnoses and disease manifestations 2,3
Upper airway/pulmonary	115 (54.2%)	4,929 (60.9%)	37,997 (28.6%)	<0.001	<0.001
Constitutional4	80 (37.7%)	3,219 (39.8%)	29,495 (22.2%)	<0.001	<0.001
Gastrointestinal	75 (35.4%)	2,919 (36.1%)	23,966 (18.0%)	<0.001	<0.001
Dermatologic	54 (25.5%)	2,333 (28.8%)	18,487 (13.9%)	<0.001	<0.001
Cardiovascular	36 (17.0%)	1,066 (13.2%)	12,825 (9.6%)	<0.001	<0.001
Hematologic	35 (16.5%)	2,333 (28.8%)	19,148 (14.4%)	0.438	<0.001
Neurologic	33 (15.6%)	1,367 (16.9%)	14,766 (11.1%)	0.051	<0.001
Patients with disease manifestations in more than 1 organ system	123 (58.0%)	5,295 (65.5%)	43,389 (32.6%)	<0.001	<0.001
Select comorbidities and psychological conditions 5
CCI score6	1.2 ± 1.6 [1.0]	1.6 ± 1.8 [1.0]	0.9 ± 1.5 [0.0]	<0.01	<0.001
Anxiety disorders	33 (15.6%)	1,281 (15.8%)	11,774 (8.9%)	<0.001	<0.001
Cognitive functioning	25 (11.8%)	996 (12.3%)	10,204 (7.7%)	<0.05	<0.001
Sleep disorders	23 (10.8%)	1,266 (15.7%)	12,555 (9.4%)	0.561	<0.001
Depressive disorders	19 (9.0%)	1,132 (14.0%)	11,845 (8.9%)	1.000	<0.001
Select diagnostic procedures 7
Complete blood count with differential	93 (43.9%)	3,386 (41.9%)	27,039 (20.3%)	<0.001	<0.001
CT scan (chest, pelvis, abdomen)	37 (17.5%)	1,631 (20.2%)	11,051 (8.3%)	<0.001	<0.001
Echocardiogram	32 (15.1%)	1,083 (13.4%)	10,257 (7.7%)	<0.001	<0.001
Troponin test	27 (12.7%)	880 (10.9%)	7,612 (5.7%)	<0.001	<0.001
Lymphocyte phenotyping	26 (12.3%)	728 (9.0%)	761 (0.6%)	<0.001	<0.001
Bone marrow biopsy	16 (7.5%)	583 (7.2%)	3 (0.0%)	<0.001	<0.001
BEC
BEC on the index date (cells/μL)	2,991.4 ± 3,595.7 [1,694.0]	2,137.6 ± 2,207.7 [1,568.0]	1,430.7 ± 759.0 [1,218.0]
>1,000-1,500	87 (41.0%)	3,637 (45.0%)	100,163 (75.3%)	<0.001	<0.001
>1,500-5,000	98 (46.2%)	4,059 (50.2%)	31,935 (24.0%)
>5,000	27 (12.7%)	393 (4.9%)	847 (0.6%)
Highest BEC (cells/μL)	4,610.5 ± 6,499.7 [2,173.0]	2,629.8 ± 2,945.9 [1,820.0]	1,518.1 ± 923.9 [1,267.0]
>1,000-1,500	50 (23.6%)	2,455 (30.3%)	92,934 (69.9%)	<0.001	<0.001
>1,500-5,000	111 (52.4%)	4,982 (61.6%)	38,776 (29.2%)
>5,000	51 (24.1%)	652 (8.1%)	1,235 (0.9%)

Values are reported as mean ± SD [median] or n (%).

Abbreviations: BEC, blood eosinophil count; CCI: Charlson Comorbidity Index; CT, computed tomography; HES, hypereosinophilic syndrome; N, number; SD, standard deviation.

Notes:

^[1]Baseline clinical characteristics were summarized during the 12-month pre-index period.

^[2]Conditions or comorbid conditions that may be associated with an HES presentation.

^[3]Specific conditions included in each category and associated ICD-10 codes are provided in Online Repository Table E2.

^[4]Constitutional conditions refer to a group of symptoms or manifestations indicating a systemic or general effect of a disease and that may affect the general well-being or status of an individual (e.g., muscle or join pain, malaise and fatigue and fever).

^[5]Specific conditions included in each category and associated ICD-10 codes are provided in Online Repository Table E3.

^[6]Based on Quan, H., et al. (2005). Coding algorithms for defining comorbidities in ICD-9-CM and ICD-10 administrative data. Medical Care, 43(11): 1130–39.

^[7]Procedures most frequently associated with making an HES diagnosis. Specific procedures included in each category and associated procedure codes are provided in Online Repository Table E4.

Compared with patients in Group 3, patients in Group 1 were younger at index (p<0.05) and had higher prevalences of many differential diagnoses and disease manifestations (p-values reported in Table 1 and Table 2), higher utilization of diagnostic procedures (all p<0.001), higher Charlson Comorbidity Index (CCI) score(13) which correlates with risk of morbidity due to concomitant diseases (p<0.01), and higher BEC at index (p<0.001; Table 1 and Table 2). Similar findings were observed in the comparison between Group 2 and Group 3 (Table 1 and Table 2).

Treatment patterns during the follow-up period

More than half of the patients in Group 1 received ≥1 HES-related treatment (62.3%), with OCS being the most common (59.0%), followed by immunomodulatory/monoclonal antibody (15.1%) and cytotoxic agents (4.2%; Table 3).

Table 3.

Treatment patterns during follow-up for patients with and without HES^1,2

	Group 1 Patients with an HES diagnosis code (N = 212)	Group 2 Patients with predicted HES (N = 8,089)	Group 3 Patients without HES with elevated BEC (N = 132,945)	P-value (Group 1 vs. Group 3)	P-value (Group 2 vs. Group 3)
No treatment	80 (37.7%)	2,449 (30.3%)	80,255 (60.4%)	<0.001	<0.001
Any corticosteroids	125 (59.0%)	5,255 (65.0%)	50,656 (38.1%)	<0.001	<0.001
OCS	93 (43.9%)	3,731 (46.1%)	29,888 (22.5%)	<0.001	<0.001
Number of days on OCS	111.0 ± 107.7 [70.0]	96.2 ± 109.8 [41.0]	41.2 ± 71.0 [12.0]	<0.001	<0.001
Cumulative OCS dosage (prednisone equivalent) (mg)	2,131.0 ± 2,753.8 [1,349.0]	1,878.0 ± 3,468.0 [857.8]	776.9 ± 1,636.9 [300.0]	<0.001	<0.001
Topical corticosteroids	47 (22.2%)	1,580 (19.5%)	16,470 (12.4%)	<0.001	<0.001
Immunomodulatory/monoclonal antibody	32 (15.1%)	1,007 (12.4%)	1,926 (1.4%)	<0.001	<0.001
Cytotoxic agents3	9 (4.2%)	386 (4.8%)	2,105 (1.6%)	<0.01	<0.001
Immunosuppressive agents4	6 (2.8%)	483 (6.0%)	2,104 (1.6%)	0.239	<0.001
TKI	0 (0.00%)	163 (2.0%)	370 (0.3%)	1.000	<0.001
JAK inhibitors	0 (0.00%)	62 (0.8%)	259 (0.2%)	1.000	<0.001
Interferon-alpha	0 (0.00%)	2 (0.0%)	3 (0.0%)	1.000	<0.05
Patients with more than 1 treatment category	32 (15.1%)	1,497 (18.5%)	4,461 (3.4%)	<0.001	<0.001

Values are reported as mean ± SD [median] or n (%).

Abbreviations: BEC, blood eosinophil count; HES, hypereosinophilic syndrome; JAK, Janus kinase; N, number; OCS, oral corticosteroid; SD, standard deviation; TKI, tyrosine kinase inhibitor.

Notes:

^[1]Follow-up period spanned 12 months after the index date.

^[2]Specific treatment included in each category and associated medication and/or procedure codes are provided in Online Repository Table E5.

^[2]Cytotoxic agents included hydroxyurea, methotrexate, and vincristine.

^[3]Immunosuppressive agents included azathioprine, busulfan, chlorambucil, cyclophosphamide, 2-chlorodeoxyadenosine alone or in combination with cytarabine, etoposide, mycophenolate mofetil, sirolimus, and tacrolimus.

During the 12 month follow-up period, compared to patients in Group 3, patients in Group 1 more frequently had OCS use (59.0% vs. 38.1%) and had longer duration of OCS treatment (mean ± SD [median]: 111.0 ± 107.7 [70.0] vs. 41.2 ± 71.0 [12.0] days) and higher cumulative OCS dosage (mean ± SD [median]: 2,131.0 ± 2,753.8 [1,349.0] vs. 776.9 ± 1,636.9 [300.0] mg prednisone equivalent; all p<0.001; Table 3). Similar findings were observed in the comparison between Group 2 and Group 3 (Table 3).

Disease manifestations and HES flares during the follow-up period

Among patients in Group 1, the most common disease manifestations were those related to the upper airway/pulmonary (61.8%), constitutional (46.2%), dermatologic (35.8%), gastrointestinal (34.4%), hematologic (28.3%), and cardiovascular systems (24.1%; Figure 2); the full breakdown of conditions within each organ system can be found in Table E7 in the Online Repository. Patients in Group 1 and 2 had similar frequencies of common disease manifestations, while both groups had significantly higher frequencies compared to patients in Group 3 (Figure 2 and Table E7 in Online Repository). Patients in Group 1 had a mean ± SD (median) of 0.63 ± 0.85 (0.00) HES flare episodes during the follow-up period; patients in group 2 had a similar number of HES flare episodes (0.64 ± 0.85 [0.00]) (Table 4).

Figure 2. Disease manifestations during follow-up for patients with and without HES¹.

Abbreviations: BEC, blood eosinophil count; EGPA, eosinophilic granulomatosis with polyangiitis; HES, hypereosinophilic syndrome.

Notes:

[1] *** indicates p-values <0.001 for comparisons between Group 1 or Group 2 and Group 3; * indicates p-values <0.05 for comparisons between Group 1 or Group 2 and Group 3.

[2] Constitutional conditions refer to a group of symptoms or manifestations indicating a systemic or general effect of a disease and that may affect the general well-being or status of an individual (e.g., muscle or join pain, malaise and fatigue and fever).

Table 4.

HES flares/exacerbations during follow-up for patients with HES

	Group 1	Group 2
	Patients with an HES diagnosis code (N = 212)	Patients with predicted HES (N = 8,089)
Number of flare episodes during follow-up	0.63 ± 0.85 [0.00]	0.64 ± 0.85 [0.00]
0	122 (57.5%)	4,481 (55.4%)
1	55 (25.9%)	2,415 (29.9%)
≥2	35 (16.5%)	1,193 (14.7%)

Values are reported as mean ± SD [median] or n (%).

Abbreviations: BEC, blood eosinophil count; HES, hypereosinophilic syndrome; N, number; SD, standard deviation.

All-cause HRU during the follow-up period

Among patients in Group 1, 22.2%, 97.2% and 25.9% had ≥1 IP, OP, and ED visit, respectively, during follow-up, with mean ± SD (median) HRU rates PPPY of 0.49 ± 1.29 (0.00), 14.73 ± 26.89 (10.00), and 0.53 ± 1.22 (0.00; Figure 3). The mean ± SD (median) length of stay per hospitalization was 5.8 ± 7.3 (3.0) days.

Figure 3. All-cause HRU during follow-up for patients with and without HES¹.

Abbreviations: BEC, blood eosinophil count; ED, emergency department; HES, hypereosinophilic syndrome; HRU, healthcare resource use; IP, inpatient; N, number; OP, outpatient; PPPY, per-person-per-year; SD, standard deviation.

Notes:

[1] *** indicates p-values <0.001 for comparisons between patients with an HES diagnosis code (Group 1) or predicted HES (Group 2) and patients without HES with elevated BEC (Group 3); ** indicates p-values <0.01 for comparisons between patients with an HES diagnosis code (Group 1) or predicted HES (Group 2) and patients without HES with elevated BEC (Group 3); * indicates p-values <0.05 for comparisons between patients with an HES diagnosis code (Group 1) or predicted HES (Group 2) and patients without HES with elevated BEC (Group 3).

Compared with patients in Group 3, higher proportions of patients in Group 1 had ≥1 IP, OP, and ED visit, along with higher rates PPPY (all p<0.05; Figure 3). In addition, patients in Group 1 had a longer length of stay per hospitalization than patients in Group 3 (p<0.05). Consistent findings were observed in the comparison between Group 2 and Group 3 (Figure 3).

Prevalence of HES

The prevalence of an HES diagnosis code (Prevalence Group 1) among patients with elevated BEC from October 2020 to October 2021 was 0.35% (270 among 76,957 patients; Figure 4). The prevalence of an HES diagnosis code or predicted HES (Prevalence Group 2) was 5.65%.

Figure 4. Prevalence of HES among patients with ≥1 assessment of BEC >1,000 cells/μL from October 2020 to October 2021¹.

Abbreviations: BEC, blood eosinophil count; HES, hypereosinophilic syndrome.

Note:

[1] Prevalence was calculated as the number of patients with an HES diagnosis, predicted HES, and total HES divided by the total number of patients with at least one assessment of BEC >1,000 cells/μL from October 2020 to October 2021 and continuous data activity during this period.

DISCUSSION

This real-world study is among the first and largest in the US to comprehensively assess the clinical and HRU burden for patients with HES. The introduction of the ICD-10 code for HES in October 2020 provided the opportunity to identify patients with physician-confirmed HES diagnoses in real-world clinical practice.(7) However, given the limited years of data available since 2020, a laboratory- and claims-based prediction model with good performance (area under the receiver operating characteristic curve of 0.82, area under the precision recall curve of 0.83) was additionally leveraged to identify patients with likely undiagnosed HES. As such, this study uniquely included both patients with a diagnosis of HES and patients predicted to have HES but who remained undiagnosed. Importantly, both these cohorts were found to experience a larger clinical and HRU burden compared to control patients without HES with elevated BEC, further supporting the prediction model in defining an HES population. Hypereosinophilia is a diagnostic observation that is often associated with considerable HRU due to workup for eosinophilia;(2) however, the differentiation of Groups 1 and 2 from Group 3 contextualizes the burden of HES among a population with elevated BEC and suggests that the burden may be related to symptoms and treatment of HES rather than to the high BEC alone.

Estimating the prevalence of HES is challenging given the heterogeneous manifestations, numerous differential diagnoses, and imperfect diagnostic criteria.(2, 14) Prior studies have shown that patients with elevated BEC have many different conditions, such as eosinophil-associated disease, bacterial infection, asthma, and hematologic malignancies.(15, 16) Our findings reveal that patients diagnosed with HES, as well as those predicted to have HES, often receive other diagnosis codes that may either be in the differential diagnosis for HES, or may represent a comorbid condition or disease manifestation, such as allergic diseases, malignancy, and autoimmune diseases. This suggests there is high potential for misdiagnosis and there remains a significant challenge in correctly diagnosing HES in claims database assessments. These diagnoses were also prevalent in Group 3 patients with elevated BEC but who did not receive an HES diagnosis code, but they were less frequently observed in confirmed HES cases. While not directly examined in this study, these conditions may represent diagnoses that have been reported to be associated with hypereosinophilia as represented in Group 3 patients. In the current study, the prevalence of HES among patients with elevated BEC was estimated at 0.35% using an HES diagnosis code in the claims data alone. However, when HES was additionally identified through the prediction model, the prevalence rate approached 5.65%. Notably, patients with an HES diagnosis code and patients with predicted HES had similar disease profiles, both of which differed significantly from Group 3 patients without HES with elevated BEC (>1,000 cells/μL), suggesting the model performs robustly in identifying HES. This finding indicates that patients with predicted HES may represent an undiagnosed population with HES, highlighting the current gap in recognition of HES. Furthermore, this finding emphasizes the need for alternate approaches or clinical decision support systems integrated into electronic health record systems to assist with recognition of HES to minimize delays in diagnosis and treatment. Although inappropriate coding is a possibility, the lack of recognition of HES ICD codes is less likely a strong cause of any misclassification in this study as patients in Group 3 did not receive the same degree of workup and testing to pursue the elevated BEC as compared with Groups 1 and 2.

The HES disease burden characterized in the current study is largely consistent with the limited real-world literature currently available.(8, 9, 10) For instance, the most common clinical manifestations among patients with an HES diagnosis code or predicted HES were those affecting the upper airway/pulmonary, constitutional, gastrointestinal, and dermatologic organ systems, which are aligned with findings from prior real-world studies of patients with HES.(8, 9, 10) With regards to treatment patterns, OCS was most commonly used, while immunomodulatory agents/monoclonal antibodies, and cytotoxic agents were used more sparingly.

While the types of treatment used in the present analysis were aligned with the literature, unlike in prior studies,(8, 10) a considerable proportion of patients with an HES diagnosis (37.7%) or predicted HES diagnosis (30.3%) did not receive any treatment during the follow-up period. This observation may be explained by the inclusion of patients at different disease stages, or the potential that patients were being observed prior to or in-between treatments. For example, OCS may be tapered off upon symptom control and reduction of BEC, and some patients with HES are treated intermittently given the long-term side effects associated with OCS therapy.(4) Misclassification of patients with HES either due to errors in the use of diagnosis codes or from the prediction model may also contribute to this finding. The PatientSource^® database is an open claims data source with data aggregated from multiple sources; as such, treatment use, diagnostic procedure use, and laboratory test results may have been underreported or incomplete if patients received care outside the covered services. Though we have implemented measures to define continuous data activity (i.e., gap of <12 months between consecutive healthcare encounters), ultimately, the design of this study limits the ability to interpret reasons for lack of treatment in this population of HES patients, and deserves further study.

Compared to patients with elevated BEC (Group 3), those with an HES diagnosis code or predicted HES had significantly higher all-cause HRU during the follow-up period, with 0.49-0.50 IP visits, 14.73-15.73 OP visits, and 0.53-0.72 ED visits PPPY. However, differences between the groups should be considered when comparing patients with HES to patients with elevated BEC given the lack of adjusted analysis. For example, Group 3 may have been diagnosed with fewer disease manifestations as a result of lower prevalence of testing in that population. Similar rates for HES-related IP (0.4) and ED (0.3) visits per year have been reported in other studies, though the rate of HES-related OP visits was lower (4.3 visits per year).(8) Whether the excess OP visits observed in the current analysis reflect an increased awareness of disease management that leads to more frequent follow-up, or represent non–HES-related visits in addition to HES management, or reflects the lower burden of disease manifestations for patients with elevated BEC (Group 3), is unclear.

Taken together, this study highlights the substantial clinical and HRU burden associated with HES, a condition that appears to affect a larger proportion of patients who are potentially undiagnosed than previously recognized. Improved awareness of the disease can lead to better diagnosis and more optimal patient management to reduce the considerable disease burden associated with HES.

The study results should be considered in the context of some limitations. First, this study supplements the original prediction model publication and was not an external validation of the prediction model as the same data source used to build the prediction model was used to describe the patient characteristics in this study. Additionally, diagnoses were not individually adjudicated by experts in the field. Further research validating the prediction model in separate databases would strengthen the validity of the findings and potentially uncover additional characteristics of patients with HES. Second, some patients with HES may not have been given an HES diagnosis code if diagnosis occurred close to the introduction of the ICD-10 code for HES, as physicians and health systems were likely still adapting to the introduction of the new code, and misclassification may occur. Exploring the reason why patients did or did not receive an HES diagnosis code was not possible in this study. However, our study identified undiagnosed patients with HES using the prediction model, which considered a large number of predictors and addressed the potential for underdiagnosis. Third, the prediction model results were sensitive to the chosen conservative threshold, though sensitivity analyses using different thresholds demonstrated similar findings. Fourth, although the requirement of continuous data activity might suggest that the results may be less generalizable to patients with less frequent encounters, this is less likely as patients with HES are anticipated to have more frequent medical encounters as was seen in the higher rates of OP visits in Group 1 and Group 2 patients in this study. Fifth, the interpretation of HES flare/exacerbation in this claims database analysis was limited due to the reliance on coding of disease manifestations or diagnoses combined with abnormal BEC values, HES-related medication use, and hospitalizations or ED visits to approximate the clinical definition used in clinical trials. It may be possible that these features were confounded or misclassified when patients presented for non–HES related hospitalizations or ED visits on or after the disease manifestation. Lastly, the current data cut (through the end of 2021) may not have been sufficient follow-up time to fully capture the uptake of recently approved treatments (e.g., mepolizumab, which was approved for HES in September 2020), or other biologics and JAK inhibitors being used off-label for HES.(5) The data were first available in 2017, so left censoring in of the data may have also restricted our ability to describe the initial diagnosis of HES, including procedures and treated prior to 2017, for patients included in this study.

CONCLUSIONS

This real-world study underscores the likelihood that HES is underdiagnosed in the US despite the clear clinical and HRU burden patients experience. The chronicity of HES was shown to have a large impact on patients, as observed through the large burden of symptoms and manifestations, as well as the high rate of OP visits. Furthermore, a sizable percentage of patients with HES were not receiving treatment in this study. These findings highlight the pressing need for improved recognition of HES and appropriate treatment and management of HES. Future studies may aim to adjudicate HES diagnosis, classify HES subtypes, provide information on prognosis or outcomes, as well as provide more comprehensive data on clinical management strategies. Further patient-centered research evaluating the impact of HES symptom burden, healthcare resource utilization, and disease impacts on quality of life is needed.

Supplementary Material

2

Figure E1. Study schema

Abbreviations: BEC, blood eosinophil count; HES, hypereosinophilic syndrome; ICD-10: International Classification of Diseases, Tenth Revision, Clinical Modification.

3

Figure E2. Claims-based algorithm to define HES flares/exacerbations¹

Abbreviations: BEC, blood eosinophil count; ED, emergency department; HES, hypereosinophilic syndrome; ICD: International Classification of Diseases; IP, inpatient; OCS, oral corticosteroid.

Notes:

[1] A 30-day washout period following the index date was implemented, during which flares were not assessed to avoid misclassifying treatment initiation or treatment change associated with the initial diagnosis as an indicator for HES flare. To be considered a discrete flare, the onset date of a new flare must have been ≥45 days after the onset of most recent flare.

4

Figure E3. Detailed sample selection flowchart

Abbreviations: BEC, blood eosinophil count; HES, hypereosinophilic syndrome; OCS, oral corticosteroid.

Highlights.

What is already known about this topic?

Given the lack of an International Classification of Diseases code for Hypereosinophilic Syndrome (HES) until October 2020, there are limited analyses of patients with HES in real-world clinical practice.

What does this article add to our knowledge?

In this study, patients with diagnosed or algorithm-predicted HES had a significantly larger clinical and healthcare resource burden than those without HES with elevated blood eosinophil count, including more corticosteroid use.

How does this study impact current management guidelines?

These findings highlight the pressing need for better algorithms and guidelines to more accurately diagnose HES. Accurate diagnosis of HES may reduce the substantial clinical burden associated with the condition.

Abbreviations

BEC

Blood eosinophil count

CCI

Charlson Comorbidity Index

ED

Emergency department

EGID

Eosinophilic gastrointestinal disorder

EGPA

Eosinophilic granulomatosis with polyangiitis

HES

Hypereosinophilic syndrome

HIPAA

Health Insurance Portability and Accountability Act

HRU

Healthcare resource use

ICD

International Classification of Diseases

IP

Inpatient

JAK

Janus kinase

LASSO

Least absolute shrinkage and selection operator

OCS

Oral corticosteroid

OP

Outpatient

PPPY

Per-person-per-year

SD

Standard deviation

TKI

Tyrosine kinase inhibitor

US

United States

Data availability statement

The data that support the findings of this study are available from Source Healthcare Analytics, LLC, a Symphony Health Solutions Corporation. Restrictions apply to the availability of these data, which were used under license for this study. Data are available directly from Source Healthcare Analytics, LLC, a Symphony Health Solutions Corporation.