Use of a claims database to characterize and estimate the incidence rate for Castleman disease

Abstract

Castleman disease (CD) is a rare lymphoproliferative disorder affecting single (unicentric; UCD) or multiple (multicentric; MCD) lymph nodes. The incidence of this difficult to diagnose disease is poorly understood, as no International Classification of Diseases, Ninth Revision (ICD-9) code is available. This study utilized a unique strategy to estimate its incidence using two commercial claims databases, IMS LifeLink™ and Truven Health Analytics MarketScan^®. Patients with an index diagnosis of lymphadenopathy (ICD-9 code 785.6) were followed longitudinally for 1 year prior to and 2 years post-index diagnosis date. An algorithm that identifies potential patients with CD was developed to determine the incidence rate in person-years. The incidence rate for CD was calculated as 21 (IMS LifeLink™) and 25 (MarketScan^®) per million person-years. Additionally, 23% of patients with CD were identified as potentially suffering from MCD. These results are consistent with the definition of an orphan disease, and the low incidence of the disease estimated in the literature.

Keywords::

• Castleman disease
• multicentric Castleman disease
• incidence

Introduction

Castleman disease (CD), named after Benjamin Castleman who first described it in 1954, is a rare lymphoproliferative disorder [1]. CD can manifest either as unicentric CD (UCD), a non-systemic lymphadenopathy in a single lymph node or single chain of lymph nodes, or as multicentric CD (MCD), a recurrent diffuse lymphadenopathy (in multiple lymph nodes or multiple chains of lymph nodes) with systemic symptoms such as fever, fatigue, anorexia, anemia and wasting [2,3]. Hepatosplenomegaly, lymph node enlargement and multiple laboratory abnormalities (e.g. hypoalbuminemia, thrombocytopenia and transaminitis) are also common in MCD [3–6].

Disease classification

Among patients with CD, UCD is more commonly encountered than MCD [3,7]; however, there is a lack of consensus regarding the proportion of UCD versus MCD in the published literature [6–12]. Histologically, there are also different variants of CD: hyaline vascular variant (HVV), plasma cell variant (PCV) and mixed variant. HVV is found in 80–90% of UCD and nearly one-third of MCD, whereas PCV and mixed variant are each seen in 20–40% of MCD and much less frequently in UCD [3,5,13,14]. Plasmablast variant is a subtype of PCV and is an aggressive form of multicentric disease, often associated with human immunodeficiency virus (HIV) and human herpes virus-8 (HHV-8) infection [3,15]. Recently, Dispenzieri et al. have classified certain patients with MCD as having a co diagnosis of POEMS syndrome [6]. POEMS disease constitutes a syndrome with a combination of Polyneuropathy, Organomegaly, Endocrinopathy, Monoclonal protein and Skin abnormalities (including hyperpigmentation and hypertrichosis). In a retrospective analysis of 113 patients with CD, the authors found that most of the risk factors for death co segregated with diagnostic criteria for POEMS syndrome [6].

Prognosis and treatment

UCD has a good prognosis as it can be treated with surgery and local radiotherapy, with reported 5-year overall survival of 91% [6,11,16,17]. Prognosis for patients with MCD remains poor, as there is no accepted treatment for non-viral MCD, and disease becomes refractory to the systemic treatments provided. In the study by Dispenzieri et al., a 5-year survival of 65% was reported for patients with MCD, with the presence of HIV and POEMS as negative prognostic factors [6]. Others have reported a median survival for patients with MCD ranging from only a couple of weeks post-diagnosis to as long as 20 years [3,11,18–20]. Sepsis, systemic inflammation leading to multiorgan failure or development of malignancy (most commonly lymphoma), has been reported to be a major cause of death in patients with MCD [3].

There is limited information on treatment options for patients with MCD. Published literature on treatment results in HIV- and HHV-8–negative MCD is limited to case reports and small retrospective case series [3,15,21–34], in which the clinical response parameters were often not well defined and limited information was provided on safety or durability of the response. Importantly, these case reports cannot be taken as representative of the magnitude of treatment effects, because a bias toward publishing positive results cannot be excluded [35].

Epidemiology

Being a rare disorder, the prevalence of CD has not been well established. There are no diagnostic codes associated with CD, UCD or MCD in national databases. More than 400 cases have been reported in the literature [36]. Based on the proportion of patients presenting to a large cancer center with lymphadenopathy of undetermined origin, later diagnosed with CD, it was estimated that the number of cases in the United States ranges from 30 000–100 000 [37,38].

In this study, we propose and investigate the utility of an algorithm that identifies and characterizes potential patients with CD in two US health insurance claims databases (IMS LifeLink™ and MarketScan^®), on which the national incidence of the disease is proposed to be estimated. The study further characterizes patients from the IMS LifeLink™ database in terms of their comorbidities and drug treatment choices.

Methods

Data source

Anonymized patient longitudinal data were sourced from the IMS LifeLink™ Health Plan Claims Database (IMS LifeLink™) and Truven Health Analytics MarketScan^® Commercial Claims and Encounters Database (MarketScan^®). The IMS LifeLink™ health plan includes inpatient and outpatient claims, diagnoses and procedures based on the International Classification of Diseases, Ninth Revision, Clinical Modification (ICD-9-CM) and Current Procedural Terminology (CPT-4) codes, as well as retail and mail order pharmacy claims for more than 80 million members from more than 100 commercial health plans across the United States. The MarketScan^® database includes claims from commercial health insurance plans sponsored by more than 100 large- and medium-sized employers in the United States. Enrollment data include information on age, gender and periods of service eligibility. Prescription claims include the National Drug Code, quantity of units dispensed and days of supply.

Study sample (CD and MCD)

The study sample began with identifying patients that had a diagnosis of enlargement of lymph nodes (ICD-9 code 785.6) and had undergone a lymph node biopsy (procedure code 38505). The first date of diagnosis of ICD-9 code 785.6 was tagged as the index diagnosis date (IDD). From this sample, patients were identified as potentially suffering from CD if they met all inclusion and exclusion criteria based on those described by Casper [3], described in Table I. From this potential CD cohort, patients who were exposed to ≥ 1 of the medications described in Table II (and in its footnote) after IDD were identified as patients with MCD.

Table I. Diagnostic criteria for Castleman disease.

Criterion	Rationale for criterion
Patients must have a diagnosis of enlargement of lymph nodes, ICD-9 code 785.6	CD is a lymphoproliferative disorder that leads to enlargement of lymph nodes
Patients must have continuous enrollment of 1 year prior to and 2 years after the IDD, with enlargement of lymph nodes	To allow a sufficient time period for identification of additional critical inclusion and exclusion criteria
Patients must have had lymph node biopsy within the 2-year follow-up period	A diagnosis of CD is usually not considered until its characteristic features are noted on an excision lymph node biopsy (CPT-4 code 38505); hence, this additional eligibility condition within 2 years of index date was required [3]
Patients should not have any of the following comorbidities in the 1 year prior to IDD: RA, lupus, cancer (including lymphoma) and HIV	Patients with lymph node histology consistent with CD should first have other causes of similar-appearing reactive lymphadenopathy ruled out; patients with a recent history of comorbidities associated with lymphadenopathy in the 1 year prior to the index date were excluded
Patients should not have any of the following comorbidities in the 1 year post-IDD: RA, lupus, cancer and HIV; or have developed NHL/HD in the second year post-IDD	To accurately and conservatively identify patients with a potential diagnosis of CD, while ruling out other diseases, those with a comorbidity associated with lymphadenopathy in the 1 year post-IDD were excluded

CD, Castleman disease; ICD-9, International Classification of Diseases, Ninth revision; CPT-4, Current Procedural Terminology; IDD, index diagnosis date; RA, rheumatoid arthritis; HIV, human immunodeficiency virus; NHL, non-Hodgkin lymphoma; HD, Hodgkin disease.

Table II. Distribution of potential patients with CD exposed to treatment.*

	IMS LifeLink™		MarketScan^®
Treatment	n	%	n	%
Doxorubicin	3	0.3	3	0.2
Dexamethasone	120	10.6	152	12.3
Prednisone	178	15.7	170	13.7
Rituximab	1	0.1	1	0.1

CD, Castleman disease.

*Other treatments evaluated included thalidomide, cyclophosphamide and tocilizumab, but no patients were identified in either database.

Calculation of incidence rate and prevalence

The annual incidence rate for CD and MCD was calculated by dividing the number of identified potential patients with CD and MCD, respectively, by the product of (a) the total number of patients that had 3 years of continuous enrollment and (b) 2 years (to account for the follow-up period from the first diagnosis).

Calculation of prevalence

The calculated incidence rate was multiplied by the average duration of the disease (2–4 years) [3,6,10,39] to estimate the prevalence range for MCD [40].

Comorbidities and comedication analysis

The potential CD cohort identified using the IMS LifeLink™ database was analyzed for associated comorbidities and comedication. Comorbidities were analyzed for a period of 6 months pre- and post-IDD to determine the association of other diseases with CD. Comorbidities were classified as “POEMS-related” and “others” (not related to POEMS). For POEMS-related comorbidities, those selected had either a p-value < 0.1 or the number of patients in pre- or post-IDD was ≥ 10. For other comorbidities, the top 20 were selected. Prevalence of comorbidities was also compared between patients with UCD and MCD in the 2 years post-IDD. Drug classes were analyzed for 1 year pre- and post-IDD to identify drug classes prescribed to potential patients with CD.

Statistical analysis

The mean age of males and females was compared using the t-test for difference of means, with level of significance fixed at 5%. This analysis was carried out for the IMS LifeLink™ and MarketScan^® databases separately. The difference in proportions was used to analyze the difference in parameters of patients before and after being diagnosed with CD, or between patients diagnosed with UCD and MCD. The parameters that were compared were comorbidities and comedications. The significance level was fixed at 10% for the difference in proportion of patients identified with POEMS-related comorbidities 6 months pre- and post-IDD.

Role of the funding source

The study was designed, funded and conducted by Janssen Research and Development. SmartAnalyst Inc. was contracted by Janssen to carry out some of the data analysis and to contribute to manuscript development.

Results

Identification of patients with potential diagnosis of CD using the IMS LifeLink™ and MarketScan^® databases

Of the nearly 83 million (122 million) patients contained in the IMS LifeLink™ (MarketScan^®) databases, 16 967 (22 218) patients were identified with an incident diagnosis of lymphadenopathy (ICD-9 code 785.6) and a lymph node biopsy (CPT-4 code 38505); all IDDs occurred in the 2001–2009 period. Of these patients, 1131 (1240) (Table III) were identified as potential patients with CD after accounting for the criteria listed in Table I [3]. The process of selecting the potential patients with CD is depicted in Figures 1(a) and 1(b).

Figure 1. Algorithm used to identify potential Castleman disease (CD) and multicentric CD (MCD) cohorts in (a) IMS LifeLink™ and (b) MarketScan^® databases. Patients were also checked for histiocytosis or post-transplant lymphoproliferative disorder within 6 months of index diagnosis date (IDD) to exclude other confounding factors; no patient was found with these indications.

Table III. Patients excluded from cohort 1.

	IMS LifeLink™		MarketScan^®
	Count: occurrence	%	Count: occurrence	%
NHL	254	12.1	295	12.9
Hodgkin	8	0.4	23	1.0
RA	7	0.3	10	0.4
Lupus	5	0.2	10	0.4
Other cancers	325	15.5	374	16.3
HIV	3	0.1	16	0.7
Combination (of above)	384	18.3	365	16.0
No occurrence	1131	54.0	1240	54.2

NHL, non-Hodgkin lymphoma; RA, rheumatoid arthritis; HIV, human immunodeficiency virus.

Demographics of potential CD cohort identified using the IMS LifeLink™ and MarketScan^® databases

In the potential CD cohort, the majority of patients were female, 64.1%, n = 725 (61.8%, n = 766). The mean age at diagnosis for males was lower at 41.0 years (44.0 years, p < 0.05) than that for females, 43.9 years (47.0 years) (Table IV). Fifty-two percent (58%) of the potential CD cohort were between the ages of 40 and 64 years; fewer than 10% (10%) of patients were older than 65 years [Figures 2(a) and 2(b)].

Figure 2. Age distribution of CD cohort identified using the (a) IMS LifeLink™ and (b) MarketScan^® databases. In the IMS LifeLink™ database, N = 1129 as two patients were not included due to non-availability of data. In the MarketScan^® database, N = 1240 patients.

Table IV. Mean and median age of potential patients with CD.

	Age at first CD diagnosis
	IMS LifeLink™			MarketScan^®
	n	Mean, years (SD)	Median, years	n	Mean, years (SD)	Median, years
All	1129*	42.9 (17.6)	44	1240	45.8 (17.2)	48
Male	404	41.0 (18.9)	43	474	44.0 (18.6)	47
Female	725	43.9 (16.7)	45	766	47.0 (16.1)	48

CD, Castleman disease; SD, standard deviation.

*Two patients were not included due to non-availability of data.

Identification of potential MCD cohort identified using the IMS LifeLink™ and MarketScan^® databases

As described in the “Methods” (Table I), 273 (24.1%) (283 [22.8%]) out of 1131 (1240) potential patients with CD who were exposed to chemotherapy were classified with potential MCD. A review of the ICD-9 codes of these patients did not identify any cases with other benign lymphoproliferative disorders (such as histiocytosis or post-transplant lymphoproliferative disorder).

Demographics of potential MCD cohort identified using the IMS LifeLink™ and MarketScan^® databases

Among the potential patients with MCD, the average age was 44 (50) years (Table V). In this subset of potential patients with MCD based on the IMS LifeLink™ data, the average age of males (45 years) was slightly higher than that of females (44 years); however, this difference in age was not statistically significant (p > 0.05).

Table V. Mean and median age of potential patients with MCD.

	Age at first CD diagnosis
	IMS LifeLink™			MarketScan^®
	n	Mean, years (SD)	Median, years	n	Mean, years (SD)	Median, years
All	273	44.3 (15.5)	46	283	49.6 (15.4)	51
Male	89	45.2 (17.3)	48	94	49.0 (17.6)	52
Female	184	43.9 (14.6)	45	189	49.8 (14.3)	51

MCD, multicentric Castleman disease; CD, Castleman disease; SD, standard deviation.

Estimation of incidence rate for CD and MCD in person-time based on IMS LifeLink™ and MarketScan^® data

In the IMS LifeLink™ database, there were 26 982 399 unique individuals with ≥ 3 years of continuous enrollment. The individuals were followed for 2 years post-IDD of lymphadenopathy to confirm CD, yielding 53 964 798 person-years. Based on 1131 new cases of CD and 53 964 798 person-years, the incidence rate for CD was calculated as 21 per million person-years. According to the literature, 30–50% of incident cases of CD have MCD [6,10]; however, as discussed previously in our analysis, the proportion of patients with MCD is closer to 22–24% in both the IMS LifeLink™ and MarketScan^® datasets. Therefore, under the assumption that 23% of the potential patients with CD have MCD, the incidence of MCD and UCD was calculated as 5.1 and 15.9 cases per million person-years, respectively. Based on the US population of 310 232 863 in 2010, this translates to an annual incidence of 1569 patients with MCD and 4932 patients with UCD in the United States (Table VI).

Table VI. Estimated incidence rate of CD and MCD.

	IMS LifeLink™		MarketScan^®
	Incidence rate (per million person-years)	Annual incidence (no. of cases)	Incidence rate (per million person-years)	Annual incidence (no. of cases)
CD	21.0	6502	24.8	7696
UCD	15.9	4932	19.1	5940
MCD	5.1	1569	5.7	1756

CD, Castleman disease; MCD, multicentric Castleman disease; UCD, unicentric Castleman disease.

Using this methodology, the incidence rate for CD and MCD was also calculated using the MarketScan^® data; the corresponding values obtained are listed in Table VI.

Estimation of MCD prevalence based on IMS LifeLink™ and MarketScan^® data

Approximate prevalence was calculated, as discussed in the “Methods” section, by taking the product of incidence and duration. Assuming the median duration of the disease ranges from 2–4 years, the prevalence of MCD was calculated as 10–20 (IMS LifeLink™) or 11–23 (MarketScan^®) cases per million person-years. Based on the US population of 310 232 863 in the year 2010, this translates to 3139–6278 (IMS LifeLink™) or 3513–7026 (MarketScan^®) prevalent cases in the United States.

Comorbidities associated with potential patients with CD identified using IMS LifeLink™ data

Comorbidities were classified as potential POEMS-related comorbidities and “others” that were not related to POEMS (Tables VII and VIII). Within the potential POEMS-related comorbidities, lymphadenitis, thyroid-related disorders, neoplasms of unspecified nature of the endocrine glands and nervous systems, swelling in limbs and hemangioma showed a significant change in the proportion of patients diagnosed pre- and post-IDD (p < 0.05). Among the top 20 other comorbidities (not related to POEMS) affecting potential patients with CD (Table VIII), there was a significant increase pre- and post-IDD (p < 0.001) in the number of patients suffering from comorbidities in the craniocervical area (e.g. salivary gland–related disorders, swelling in the head and neck, cellulitis/abscess of the neck, hypertrophy of tonsils and chronic maxillary sinusitis).

Table VII. POEMS-related comorbidities associated with potential patients with CD based on IMS LifeLink™ database.

POEMS-related comorbidities	6 months pre-IDD (n = 1131)	6 months post-IDD (n = 1131)	p-value
Acute lymphadenitis	36	67	< 0.002
Thyroid-related disorders*	42	84	< 0.001
Chronic lymphadenitis	6	67	< 0.001
Hemangioma of other sites	0	3	0.084
Nonspecific mesenteric lymphadenitis	1	5	0.101
Peripheral neuropathy	8	15	0.142
Other diseases of the spleen	0	2	0.159
Unspecified disorder of the adrenal glands	2	0	0.159
Other non-infectious disorders of the lymphatic channels	0	2	0.159
Unspecified non-infectious disorder of the lymphatic channels	0	2	0.159
Hypocalcemia	0	2	0.159
Corticoadrenal insufficiency	0	2	0.159
Hemangioma of skin and subcutaneous tissue	1	4	0.180
Alopecia, unspecified	6	11	0.222
Unspecified pleural effusion	8	13	0.271
Splenomegaly	10	15	0.313

POEMS, polyneuropathy, organomegaly, endocrinopathy, monoclonal protein, skin syndrome; CD, Castleman disease; IDD, index diagnosis date.

*Includes unspecified non-toxic nodular goiter, non-toxic multinodular goiter, non-toxic uninodular goiter, benign neoplasm of the thyroid glands and other specified and unspecified disorders of the thyroid.

Table VIII. Other comorbidities (not related to POEMS) associated with potential patients with CD based on the IMS LifeLink™ database.

Other comorbidities (not related to POEMS)	6 months pre-IDD (n = 1131)	6 months post-IDD (n = 1131)	p-value
Swelling, mass or lump in head and neck	148	433	<0.001
Salivary gland–related disorders*	29	132	<0.001
Cough	77	120	<0.001
Neoplasm of uncertain behavior of other specified sites	10	51	<0.001
Sarcoidosis	6	37	<0.001
Neoplasm, of an unspecified nature, of other specified sites	5	37	<0.001
Chronic maxillary sinusitis	9	28	<0.002
Neoplasm, of an unspecified nature, of the digestive system	2	34	<0.001
Benign neoplasm of other specified sites	3	30	<0.001
Cellulitis and abscess of the neck	7	30	<0.001
Other non-specific, abnormal histological findings	1	24	<0.001
Hypertrophy of tonsils alone	5	21	<0.002
Other voice and resonance disorders	5	22	<0.001
Lipoma of other skin and subcutaneous tissue	3	24	<0.001
Other benign neoplasm of connective and other soft tissue of the head, face and neck	2	18	<0.001
Branchial cleft cyst	2	20	<0.001
Neoplasm, of an uncertain behavior, of the lip, oral cavity and pharynx	1	14	<0.001
Lipoma of other specified sites	0	10	<0.002
Swelling, mass or lump in the chest	15	35	0.004
Localized superficial swelling, mass or lump	44	74	0.005
Neoplasm, of an unspecified nature, of the endocrine glands and other parts of the nervous system	1	6	0.059

POEMS, polyneuropathy, organomegaly, endocrinopathy, monoclonal protein, skin syndrome; CD, Castleman disease; IDD, index diagnosis date.

*Includes benign neoplasm of major salivary glands, sialoadenitis, neoplasm of uncertain behavior of the major salivary glands and other specified diseases of the salivary glands.

The prevalence of several comorbidities, and systemic symptoms known to be associated with MCD, were higher among potential patients classified as having MCD as compared to potential patients with UCD in this study, in the 2 years post-IDD. Thrombocytopenia, hepatosplenomegaly, abdominal pain, malaise and fatigue, and edema were found to have significantly higher association with potential patients with MCD as compared to potential patients with UCD (p < 0.05; Table IX), which is in agreement with the known symptoms of MCD.

Table IX. Prevalence of comorbidities and systemic symptoms in 2-year post-IDD based on IMS LifeLink™ database.

	MCD (%)	UCD (%)	Total (%)	p-value*
Total number of patients	273	858	1131
Thrombocytopenia	4.0	1.3	1.9	0.004
Hepatosplenomegaly	9.9	4.2	5.6	< 0.001
Ascites	1.8	1.0	1.2	0.308
Edema	13.6	5.8	7.7	< 0.001
Pulmonary/pleural effusion	2.2	2.6	2.5	0.728
Pericardial effusion	0.7	0.7	0.7	0.952^†
Hypoalbuminemia	0.7	0.3	0.4	0.406^‡
Hypergammaglobulinemia	0.4	0.1	0.3	0.380^‡
Malaise/fatigue	39.6	20.7	25.3	< 0.0001
Abdominal pain	27.1	16.9	19.4	0.0001
Swelling in head and neck	46.5	42.3	43.3	0.197
Any of the above	83.9	66.1	71.0	< 0.0001

IDD, index diagnosis date; MCD, multicentric Castleman disease; UCD, unicentric Castleman disease.

*p-Value for comparison of MCD versus UCD cohorts.

^†25% of cells have expected counts < 5. Chi-square may not be a valid test.

^‡50% of cells have expected counts < 5. Chi-square may not be a valid test.

Comedication analysis in potential CD cohort identified using IMS LifeLink™ data

Drug classes prescribed 1 year pre- and post-IDD were analyzed, and those with a significant change (p < 0.05) are listed in Table X. Post-IDD, lincosamides, cephalosporins and antihistamines had the highest usage increase among the drug classes analyzed in potential patients with CD, followed by opioid combinations and glucocorticosteroids potentially prescribed to treat pain, infection and other systemic symptoms caused by CD and associated comorbidities (Table X).

Table X. Drug classes prescribed to potential patients with CD based on IMS LifeLink™ database.

	From index diagnosis
Drug classes that see an increase	1 year pre-IDD (n = 1131) count	1 year post-IDD (n = 1131) count	p-value
Opioid combinations	258	461	<0.001
Cephalosporins, first generation	107	209	<0.001
Lincosamides	33	75	<0.001
Antihistamines, phenothiazines	30	56	0.004
Glucocorticosteroids	139	182	0.010
Otic anti-infectives	2	12	0.007
Penicillin combinations	130	170	0.013
Ophthalmic anti-infectives	34	53	0.083
Oil-soluble vitamins	6	16	0.033
Oxazolidinones	0	4	0.046
Ophthalmics, miscellaneous	23	38	0.026
Viral vaccines	1	6	0.051
Influenza agents	7	16	0.059
Anaphylaxis therapy agents	9	17	0.114
Local anesthetics, topical	9	17	0.114
Hematorheologic agents	0	3	0.084
Opioid agonists	55	74	0.085
Coumarin anticoagulants	21	32	0.126

CD, Castleman disease; IDD, index diagnosis date.

Discussion

CD is a very rare disorder affecting lymph nodes, with no diagnostic code or procedure code to identify patients suffering from this disease. This study has attempted to define a strategy for estimating the incidence and prevalence of CD and MCD, based on the characteristics of the disease and its pattern of diagnosis, using two health insurance claims databases: IMS LifeLink™ and MarketScan^®.

Based on the article by Casper [3], the criteria used to identify potential patients with CD were: (1) diagnosis of enlargement of lymph nodes; (2) biopsy within 2 years of diagnosis; and (3) no prior history of rheumatoid arthritis (RA), lupus, cancer (including lymphoma) and HIV. To exclude patients with diseases that have comparable diagnostic patterns, patients with a diagnosis of RA, lupus, cancer (including lymphoma) and HIV within 1 year of IDD, and non-Hodgkin lymphoma (NHL) and Hodgkin disease (HD) within 2 years of IDD, were excluded. This cohort of patients was identified as the potential CD cohort. Utilizing this strategy, the incidence rate for CD was calculated to be approximately 21 and 25 per million person-years using the IMS LifeLink™ and MarketScan^® databases, respectively. This translated to about 6502 (IMS LifeLink™) and 7696 (MarketScan^®) incident CD patients annually, which is in agreement with orphan disease status and the low incidence of the disease estimated in the literature [9,37,38].

As shown in Table III, there was a preponderance of females in the potential CD cohort identified using the IMS LifeLink™ and MarketScan^® databases. In the published literature, there is varied gender distribution among patients with CD. Although there was a slight preponderance of females in a study published by Dispenzieri et al. [6], other published studies have shown either equal distribution of patients with CD across both genders [10,12] or a preponderance of males [7,8,39]. This variation in the published literature suggests that the epidemiology of CD is not well understood, and gender distribution in patients with CD requires further investigation.

Out of the potential patients with CD, potential patients with MCD were identified based on exposure to systemic chemotherapy. Using these criteria, approximately 23% of patients with CD were identified to be patients with MCD. The common use of corticosteroids, low usage of chemotherapy (Table II) and low incidence of MCD among patients with CD is in agreement with the published literature [10]. Also, the incidences of MCD calculated using the two datasets were in agreement with each other (Table VI), which further validates the proposed method.

The method employed in this study is a relatively quick and cost-effective, strategic approach for identifying and characterizing the epidemiology and medical needs of a rare disease. All these considerations make this an important strategy to consider for estimating the incidence and prevalence of other rare diseases as well.

However, there are limitations that must be kept in mind when using the proposed method. In addition to the drawbacks associated with any claims data analysis, which include coding inaccuracies and non-availability of clinical and diagnostic information, this method may overestimate the true incidence of CD as: (1) results from biopsies were not available and therefore not taken into account, (2) there were no laboratory data available to support a diagnosis of CD or MCD and (3) patients may have had other conditions with similar clinical presentation or diagnosis patterns. This method may also underestimate the true incidence as: (1) patients with CD may not have an ICD-9 code of 785.6, (2) stringent exclusion criteria were imposed by excluding patients with comorbid conditions (e.g. NHL and HIV) with similar diagnostic patterns to CD, (3) the datasets used in this study underrepresent the older Medicare and Medicaid US population and (4) patients who did not meet the 1 year pre- and 2 years post-continuous enrollment criteria and thus were excluded from this analysis may include a higher proportion of patients with CD. The substantially lower estimation of the percentage of patients with MCD estimated in this analysis (23% of patients with CD), as compared to published literature (30–50% of patients with CD) [6,10], could be on account of this estimation being based on commercial claims data, which results in the HIV-positive patients being excluded. This is corroborated by the fact that only four out of 320 patients with MCD (MarketScan^®) were HIV positive and none were HHV-8 positive (Table XI). In another study, the incidence of MCD in patients with HIV was calculated as 4.3 new cases per 10 000 person-years [41]. In 2008, the prevalence of HIV in the United States was 682 668. Thus, total expected cases of MCD in the HIV population per year would be 294 cases. In this study, the annual incidence of MCD was calculated to be 1569 (IMS LifeLink™) to 1756 (MarketScan^®) patients (Table VI). By including these HIV-positive cases that, as explained previously, may not have been accounted for, the true incidence of MCD is likely to be 1863 (IMS LifeLink™) to 2050 (MarketScan^®) patients.

Table XI. Distribution of potential patients with CD with other comorbid conditions in second year of lymphadenopathy diagnosis.

	IMS LifeLink™		MarketScan^®
Comorbid condition	n	%	n	%
Cancer	46	4.1	54	4.4
Lupus	4	0.4	7	0.6
RA	10	0.9	5	0.4
HIV	4	0.4	4	0.3
Herpes (HHV-8)	0	0.0	0	0.0
Kaposi sarcoma	0	0.0	0	0.0

CD, Castleman disease; RA, rheumatoid arthritis; HIV, human immunodeficiency virus.

Examination of comorbidities associated with potential CD demonstrated that the most common comorbidities were swelling of the head and neck, neoplasms of the lymph nodes, salivary gland–related disorders and cough. These were classified under “other” comorbidities (not related to POEMS). In addition to these, there were potential POEMS-related comorbidities (e.g. polyneuropathy, hepatomegaly, splenomegaly, lymphadenitis and thyroid-related disorders) reported in potential patients with CD, although to a lesser extent. No evidence of other benign lymphoproliferative disorders (e.g. histiocytosis or post-transplant lymphoproliferative disorder) was found. Although these identified comorbidities (potential POEMS-related and others) are consistent with the symptoms and sequelae of CD, some comorbidities (e.g. sarcoidosis [37 patients post-IDD], salivary gland–related disorders [132 patients post-IDD] and thyroid-related disorders [84 patients IDD]) were reported in a rather high number of potential patients with CD in this study. While these comorbidities have been described in conjunction with CD [42–44], the possibility that not all of these cases were part of a CD syndrome cannot be excluded. Therefore, a sensitivity analysis was performed where patients with thyroid-related disorders, salivary gland–related disorders and sarcoidosis were excluded from the potential CD cohort (193 patients). This lowered the incidence rate for potential CD from 21 to 17 per million person-years, and from 5 to 4 per million person-years for potential MCD. Based on these data, we surmise that the incidence rate calculated with a larger potential CD cohort (1131 patients) provides an overestimation or an upper range, while the incidence rate calculated with a smaller potential CD cohort (1131 − 193 = 938 patients [as per sensitivity analysis]) provides an underestimation or a lower range.

Since patients with a prior history of cancer and patients diagnosed with cancer in the first year post-IDD (2 years post-IDD for NHL and HD) were excluded from the potential CD cohort, the comorbidities identified in this study (e.g. swellings, neoplasms, hypertrophy and organomegaly) are not expected to be confounded by malignant diagnoses.

As mentioned earlier, UCD is generally treated using surgery and radiotherapy. For MCD, some of the treatments used are steroids, rituximab, cytotoxics, interferons, antivirals and targeted therapies. It is interesting to note that in the drug classes analyzed during 1 year pre- and post-IDD, there was a predominant use of steroids, while the usage of cytotoxic chemotherapy (antineoplastics) was evident in fewer than 5% of patients (Table X).

Conclusion

CD is a rare disease, with limited clinical information to unequivocally identify patients suffering from this disorder based on a diagnostic or a procedure code. This study outlines a methodology to estimate the incidence and prevalence of CD and MCD. The method was tested and validated using two commercial claims databases, IMS LifeLink™ and MarketScan^®. The incidence and prevalence calculated from the two datasets were very similar, and therefore in agreement with each other as well as with the literature estimates. The results here highlight the need to have a specific disease code for CD. Moreover, this method is arguably a relevant, quick and cost-effective strategy to calculate the incidence and prevalence of very rare diseases, and characterize the disease and treatment experiences of those patients with the diseases.

Acknowledgements

This study was funded by Janssen Research and Development. Alanna Franchetti of MedErgy provided copy-edit support funded by Janssen.

Potential conflict of interest

Disclosure forms provided by the authors are available with the full text of this article at www.informahealthcare.com/lal.