Comparative treatment-related adverse event cost burden in immune thrombocytopenic purpura

Abstract

Aims: Real-world evidence on the safety profile and costs associated with immune thrombocytopenic purpura (ITP) treatment in adults is lacking. This study quantifies and compares adverse event (AE) crude rates and costs associated with ITP treatments as found in claims data.

Materials and methods: A retrospective claims-based analysis was conducted using IMS Pharmetrics Plus database. Included patients were ≥18 years old, with a diagnosis of ITP (2007–2012); an ITP-related claim for anti-D, intravenous immunoglobulin (IVIG), rituximab, romiplostim, or eltrombopag; and 1-year continuous enrollment (3-years for rituximab) during follow-up. AEs and event costs were identified during active treatment, defined from the first claim of each drug to a pre-defined treatment gap or end of study period. Descriptive statistics were reported with Wilcoxon rank-sum significance tests.

Results: A total of 2,518 patients were identified (mean age = 50.8 (±16.3 years); 55.8% male). Of all patients, 22.8% experienced any AE. Significantly fewer anti-D patients had any AE (13.8% vs IVIG: 21.1%, rituximab: 29.4%, romiplostim: 28.1%, eltrombopag: 22.4%). Nausea/vomiting and arthralgia/musculoskeletal pain were most common across treatments, and hemolytic events did not differ significantly across treatments. Most costly AEs were urinary tract infection, aseptic meningitis, and fever ($5000+/case); headache, nasal congestion, and hemolytic event were $4,000–5,000/case. Cost per AE did not differ by treatment.

Limitations and conclusions: Although lower than trial-based AE rates, claims for ITP treatment-related AEs are common, with higher numbers for rituximab and lower numbers for anti-D. This disparity suggests a possible differential cost burden overall that future analysis should explore.

Keywords:

• ITP
• immune thrombocytopenia purpura
• cost
• adverse event
• claims analysis

Introduction

Immune (idiopathic) thrombocytopenic purpura (ITP) is an autoimmune disorder characterized by immunologic destruction of normal platelets, as well as relatively decreased rate of platelet production^1,2. ITP is classified by duration into newly-diagnosed (within 3 months of diagnosis), persistent (3–12 months from diagnosis), and chronic (>12 months from diagnosis)³ disease status. The prevalence of chronic adult ITP in the US occurs in the range of 9.5–23.6 per 100,000 persons^4,5. A gender differential may also exist. Although the prevalence has been found to be higher among females than males, with an overall prevalence rate ratio of 1.9 for females to males⁵, additional evidence suggests that ITP may predominate in men aged up to 30 years, after which women are more often affected, with the difference subsiding around the 70-years mark⁶.

Even though ITP is a rare disease, it is associated with substantial economic burden. Retrospective database studies have reported that the average ITP-attributable hospitalization cost per year in the US ranges from $4,986–$16,476^7–9. Furthermore, ITP-associated medical resource utilization and the corresponding expenditures remain substantive and constant over time, due to the chronic nature of the condition⁷. In addition to direct cost burden, indirect cost burden for ITP has also been reported¹⁰. A large survey-based study reported that ITP is associated with a higher number of physician visits, sick leave from work, more missed time from chores, and worse scores on productivity when compared to patients without ITP¹⁰.

The treatment paradigm for ITP has been constantly shifting. Initial treatment for ITP often includes corticosteroids, and intravenous immunoglobulin (IVIG) and anti-D are also indicated for initial therapy¹¹. These may be followed by other options such as rituximab (off-label), thrombopoietin (TPO) receptor agonists romiplostim and eltrombopag, and splenectomy¹¹. While each of the treatments is associated with significant adverse events (AEs) reported in clinical studies^11–15, real-world evidence on the safety outcomes as well as associated costs is lacking in the published literature. However, patient-reported treatment burden of ITP therapies, and particularly AEs, has been reported to be bothersome, leading to discontinuation or switching of treatments¹⁶. While decision-makers must consider treatment efficacy, it will also be valuable to understand the burden of AEs associated with ITP therapies when making therapeutic decisions. The intent of the current study is to quantify and compare pharmaceutical treatment-related AE-related costs and rates relative to anti-D, as found in real-world claims data among patients with ITP.

Methods

Data source

Data was derived from the IMS PharMetrics Plus^TM claims database, which contains adjudicated medical and pharmacy claims for more than 87 million health plan members in ∼55 health plans across the US from 2006 onwards. The data are longitudinal, with ∼22 million patients having 4 or more years of continuous enrollment in their health plan at the time of this study. The database is representative of the US commercially insured population, including patients from each three-digit zip code area of the US, and data from 90% of US hospitals and 80% of all US doctors. Within the database, inpatient and outpatient diagnoses were recorded as International Classification of Diseases, Ninth Revision, Clinical Modification (ICD-9-CM) codes. Other data recorded included inpatient and outpatient procedures, dates of service, retail and mail-order prescription claims, and detailed information on pharmacy and medical benefit (co-payment/coinsurance amount, deductible, and in-network vs out-of-network), inpatient stay and provider details (specialty, zip code and attending, referring, rendering, prescribing and primary care provider). Amounts charged by providers, and amounts allowed and paid by health plans, were available for all services rendered, as were the dates of service for all claims. Other data elements included demographic variables (e.g. patient age, gender, and geographical region), product type (e.g. health maintenance and preferred provider organizations), payer type (e.g. commercial and self-pay), and start and stop dates of health plan enrollment.

Patient selection

This study was a retrospective cohort analysis. Initially, all patients with a diagnosis of ITP (ICD-9-CM: 287.31) during the study period from January 1, 2007 through December 31, 2012 were selected. Patients with a claim for anti-D, IVIG, rituximab, romiplostim, or eltrombopag on or after the diagnosis of ITP between January 1, 2007 and December 31, 2012 were retained; the first such claim was defined as the index date. Note that corticosteroids were omitted from this analysis, due to difficulty confirming usage for ITP management in claims data; additionally, due to well-established inclusion of corticosteroids in the course of ITP therapy¹¹, the decision problem facing ITP management may focus on other therapeutic options. Patients were excluded from the study if they had a claim for ITP diagnosis in the 3-month period prior to the first ITP diagnosis; a claim for any of the index drugs in the 3-month pre-index period; who had a continuous enrollment of <3 months in the pre-index period and <12 months in the post-index period; or had incomplete records of age, gender, and enrollment. Since a longer follow-up time is required to observe a rituximab treatment episode and any subsequent episodes, continuous enrollment of 36 months in the post-index period was used to identify rituximab treatment episodes¹⁷. To retain the focus of primary ITP, patients with a diagnosis of other platelet defects, non-thrombocytopenic purpuras, or thrombocytopenia due to other disorders (i.e. ICD-9 codes 287.1 [qualitative platelet defects], 287.2 [other nonthrombocytopenic purpuras], 287.4 [secondary thrombocytopenia], 279.12 [Wiskott-Aldrich syndrome], 279.11 [DiGeorge’s syndrome], 286.4 [von Willebrand’s disease], and 288.2 [genetic anomalies of leukocytes]) were excluded from the study. Finally, the adult patient population (≥18 years) was retained for the primary analysis.

Identification of treatment episodes

To assess AEs during active treatment, treatment episode was used as the unit of observation. Treatment episodes were defined by the dates of first and last administration of a given study drug. The start of the first treatment episode was defined by the first claim of the study drug. This first treatment episode was terminated and a subsequent treatment episode created if a pre-defined gap between two administrations of study drugs was observed, or if the end of continuous enrollment or the study period was reached. Episode-ending gaps were identified based upon one of two factors: (1) drug dosing schedules (TPOs)^18,19 or (2) a minimal extension of average duration of response obtained from published studies^12,17: anti-D: 3 months, IVIG: 2 months, rituximab: 1 year and 2 months, romiplostim: 2 weeks, and eltrombopag: 60 days. Up to three treatment episodes were considered per patient, to increase available sample size while remaining tractable for data analysis. As the purpose was to increase sample size, episodes were not stratified by line of therapy.

Identification of AEs and costs

Treatment-related AEs (Table 1) were selected based on events reported in package inserts^18–22. AEs were identified by medical claims with diagnosis codes (see Supplemental Appendix for coding details) within treatment episodes. Hemolytic anemia event was included in the study as a proxy for intravascular hemolysis, for which anti-D has a black box warning, since the latter event does not have a designated diagnosis code. To attribute events to treatments, AEs were identified within a pre-defined period of 3 days or 1 week of each dose during a treatment episode, depending on type of event (Table 1). The period from each dose was based on clinical opinion. Disease-specific costs (allowed amount) per event were estimated for each of the drug cohorts during the period of AE identification (3 days or 1 week). Disease-specific costs include those identified from claims which had an associated AE diagnosis claim in primary or secondary position.

Table 1. AEs of interest.

AE	Identification period after dose
Arthralgia/Musculoskeletal pain	1 week
Aseptic meningitis	1 week
Asthenia/Fatigue	3 days
Back pain	3 days
Contusion	1 week
Cough	3 days
Diarrhoea	3 days
Dyspnea	3 days
Fever	3 days
Headache	1 week
Hemolytic anemia	3 days
Infusion reactions	3 days
Nasal congestion	1 week
Nausea/Vomiting	3 days
Upper respiratory tract infection	1 week
Urinary tract infection	1 week

Measures and analyses

Patients’ baseline characteristics such as age, gender, types of health plans and payers, geographic region, Charlson Comorbidity Index (CCI) scores, comorbidities, and all-cause pre-index healthcare costs were summarized descriptively. The CCI is a widely-used numeric index that predicts mortality for patients with a range of comorbid conditions; each comorbidity is assigned a score associated with risk of dying, and the series of comorbidity-specific scores are summed to provide a mortality-predictive total score^23–25. The AE rate per treatment episode and AE cost per event were compared across the drugs, with anti-D as the reference drug. Descriptive statistics such as frequency and percentage were reported for categorical variables, with Chi-square tests. Means, standard deviations (SD), and medians were reported for numeric variables, with Wilcoxon rank-sum tests. Multivariable analysis was performed to account for baseline differences between the cohorts of interest. A generalized estimating equation (GEE) model was run to assess AE rates, while controlling for key demographic and clinical characteristics; this type of model is a semi-parametric approach to testing for group or treatment differences, and is particularly useful for multivariate analyses with potentially correlated outcomes and an unknown covariance structure²⁶. A p-value of <.05 was considered to be statistically significant.

All statistical analyses were conducted using SAS version 9.2 (SAS Institute, Cary, NC). Results are presented in 2014 US dollars.

Results

Demographic and clinical characteristics

A total of 2,518 patients met all the inclusion and exclusion criteria. Figure 1 shows the attrition of patient selection using the inclusion and exclusion criteria. Overall, the mean (SD) age of patients was 50.8 (16.3) years, and 55.8% were male (Table 2). The majority of patients were from the Midwest (33.5%) and South (30.1%). Most patients were covered by commercial insurance (61.6%) or were self-insured (33.2%). Overall, the mean (SD) number of treatment episodes was 1.4 (0.6) (Table 3).

Figure 1. Patient attrition.

Table 2. Patient demographic characteristics.

	Overall (n = 2,518)		Anti-D (n = 364)		IVIG (n = 879)		Rituximab (n = 1,025)		Romiplostim (n = 175)		Eltrombopag (n = 75)
Age: (years)
Mean	50.8		46.7		49.3*		52.7**		54.6**		53.8*
SD	16.3		17.1		16.8		15.6		15.3		13.5
Median	52		49		51		54		56		54
Gender (n, %)
Male	1,405	55.8%	239	65.7%	483	54.9%	570	55.6%	74	42.3%	39	52.0%
Female	1,113	44.2%	125	34.3%	396	45.1%	455	44.4%	101	57.7%	36	48.0%
Geographic Region (n, %)
Northeast	619	24.6%	103	28.3%	197	22.4%	256	25.0%	42	24.0%**	21	28.0%
Midwest	844	33.5%	121	33.2%	324	36.9%	329	32.1%	45	25.7%**	25	33.3%
South	759	30.1%	95	26.1%	253	28.8%	306	29.9%	79	45.1%**	26	34.7%
West	296	11.8%	45	12.4%	105	11.9%	134	13.1%	9	5.1%**	3	4.0%
Health Plan Type (n, %)
Consumer-directed	12	0.5%	1	0.3%	8	0.9%	3	0.3%	0	0.0%	0	0.0%
HMO	293	11.6%	45	12.4%	100	11.4%	127	12.4%	14	8.0%	7	9.3%
Indemnity	172	6.8%	23	6.3%	67	7.6%	65	6.3%	11	6.3%	6	8.0%
POS	167	6.6%	29	8.0%	52	5.9%	72	7.0%	8	4.6%	6	8.0%
PPO	1,857	73.7%	262	72.0%	647	73.6%	752	73.4%	140	80.0%	56	74.7%
Unknown	17	0.6%	4	1.1%	5	0.6%	6	0.6%	2	0.6%	0	0.0%
Payer Type (n, %)
Commercial	1,551	61.6%	229	62.9%	531	60.4%	652	63.6%	94	53.7%	45	60.0%
Medicaid	18	0.7%	4	1.1%	6	0.7%	7	0.7%	1	0.6%	0	0.0%
Medicare Risk	47	1.9%	8	2.2%	11	1.3%	20	2.0%	6	3.4%	2	2.7%
Self-insured	837	33.2%	111	30.5%	310	35.3%	320	31.2%	69	39.4%	27	36.0%
SCHIP	3	0.1%	1	0.3%	1	0.1%	0	0.0%	1	0.6%	0	0.0%
Medicare Cost	50	2.0%	8	2.2%	15	1.7%	22	2.1%	4	2.3%	1	1.3%
Unknown	12	0.4%	3	0.8%	5	0.6%	4	0.3%	0	0.0%	0	0.0%

*p < .05; **p < .001.

Table 3. Patient clinical characteristics.

	Overall (n = 2,518)		Anti-D (n = 364)		IVIG (n = 879)		Rituximab (n = 1,025)		Romiplostim (n = 175)		Eltrombopag (n = 75)
Number of treatment episodes (n, %)
1	1,836	72.9%	266	73.1%	593	67.5%*	865	84.4%**	59	33.7%**	53	70.7%
2	455	18.1%	81	22.3%	207	23.5%*	113	11.0%**	37	21.1%**	17	22.7%
3	227	9.0%	17	4.7%	79	9.0%*	47	4.6%**	79	45.1%**	5	6.7%
Mean (SD)	1.4 (0.6)		1.3 (0.6)		1.4 (0.7)		1.2 (0.5)		2.1 (0.9)		1.4 (0.6)
Median	1.0		1.0		1.0		1.0		2.0		1.0
Number of doses per treatment episode
Mean (SD)	4.0 (5.4)		1.9 (2.2)		2.4 (3.1)		5.1 (4.3)**		9.5 (14.0)**		5.6 (4.8)
Median	3.0		1.0		1.0		4.0		4.0		4.0
Index Year (n, %)
2007	298	11.8%	70	19.2%	90	10.2%**	138	13.5%**	0	0.0%**	0	0.0%**
2008	384	15.3%	84	23.1%	138	15.7%**	161	15.7%**	1	0.6%**	0	0.0%**
2009	416	16.5%	81	22.3%	139	15.8%**	176	17.2%**	4	2.3%**	16	21.3%**
2010	488	19.4%	54	14.8%	153	17.4%**	196	19.1%**	74	42.3%**	11	14.7%**
2011	479	19.0%	44	12.1%	181	20.6%**	181	17.7%**	48	27.4%**	25	33.3%**
2012	453	18.0%	31	8.5%	178	20.3%**	173	16.9%**	48	27.4%**	23	30.7%**
CCI Score (n, %)
0	1,279	50.8%	244	67.0%	483	54.9%	436	42.5%	83	47.4%	33	44.0%
1	432	17.2%	57	15.7%	133	15.1%	195	19.0%	26	14.9%	21	28.0%
2–3	561	22.3%	49	13.5%	185	21.0%	270	26.3%	39	22.3%	18	24.0%
4+	246	9.8%	14	3.8%	78	8.9%	124	12.1%	27	15.4%	3	4.0%
Median	1.0		1.0		1.0		2.0		2.0		2.0
Pre-index total healthcare costs
Mean (SD)	$19,751 ($36,207)		$8,900 ($16,901)		$16,176 ($37,356)		$23,433 ($35,802)		$33,089 ($53,253)		$11,921 ($17,583)
Median	$6,926		$2,812		$5,848*		$11,400**		$11,000*		$3,773

*p < .05; **p < .001.

AE rates

Across all patients, a total of 2,698 treatment episodes were reported. Overall, 22.8% of these episodes were with an AE (Table 4). Nausea/vomiting (5.3%) and arthralgia/musculoskeletal pain (5.3%) were the most common AEs across treatments, followed by headache (3.8%) and upper respiratory infection (3.7%). Anti-D episodes had a significantly lower overall AE rate than that among other drugs (p < .05). When examining differences in unadjusted rates aligning with reimbursed claims, TPO episodes were associated with higher rates of pain-related AEs, including arthralgia/musculoskeletal pain (anti-D: 2.4% vs romiplostim: 9.4%, eltrombopag: 7.2%; p < .05 for both comparisons) and back pain (anti-D: 1.5% vs romiplostim: 3.1%, eltrombopag: 2.6%; p > .05 for both comparisons) as well as upper respiratory tract infection (anti-D: 1.3% vs romiplostim: 6.4%, eltrombopag: 7.2%; p < .05 for both comparisons) as compared to anti-D episodes. The occurrence of headache events was statistically significantly higher among IVIG episodes than that among anti-D episodes (5.5% vs 2.6%; p < .05). Hemolytic anemia rates were similar across treatments.

Table 4. AE rates by drug episodes.

	Overall (n = 2,698)^a	Anti-D (n = 465)^a	IVIG (n = 1,061)^a	Rituximab (n = 564)^a	Romiplostim (n = 456)^a	Eltrombopag (n = 152)^a
Overall	22.8%	13.8%	21.1%*	29.4%*	28.1%*	22.4%*
Arthralgia/Musculoskeletal pain	5.3%	2.4%	4.7%*	5.1%*	9.4%**	7.2%**
Aseptic meningitis	0.1%	0.0%	0.2%	0.0%	0.0%	0.0%
Asthenia/Fatigue	2.8%	0.9%	2.1%	3.0%*	6.4%**	2.0%
Back pain	1.9%	1.5%	1.6%	1.8%	3.1%	2.6%
Contusion	0.5%	0.0%	0.8%	0.4%	0.4%	1.3%
Cough	1.0%	0.0%	1.0%*	0.9%	1.5%*	2.6%*
Diarrhoea	0.6%	0.2%	0.5%	0.7%	1.1%	0.0%
Dyspnea	1.6%	0.6%	0.8%	2.3%*	3.3%*	3.3%
Fever	1.0%	1.3%	0.4%	1.2%	1.8%	1.3%
Headache	3.8%	2.6%	5.5%*	2.3%	3.1%	3.3%
Hemolytic anemia	1.8%	2.2%	1.1%	3.2%	1.5%	1.3%
Infusion reactions	0.3%	1.1%	0.1%*	0.2%	0.0%	0.0%
Nasal congestion	0.4%	0.0%	0.4%	0.4%	0.9%	0.0%
Nausea/Vomiting	5.3%	3.9%	4.1%	12.2%**	2.6%	0.0%*
Upper respiratory tract infection	3.7%	1.3%	3.4%*	3.4%*	6.4%**	7.2%**
Urinary tract infection	1.9%	1.5%	1.7%	1.6%	3.3%	0.7%

*p < .05; **p < .001.

^a n represents the total number of treatment episodes.

Controlling for demographic and clinical characteristics, results from the GEE model suggested that anti-D episodes had a significantly lower overall AE rate than that among other drugs (p < .05) (Table 5). Anti-D episodes had significantly lower occurrences of arthralgia/musculoskeletal pain and upper respiratory tract infection as compared to other drugs (p < .05; except IVIG for arthralgia/musculoskeletal pain, where p > .05; data not shown) and significantly lower occurrences of asthenia/fatigue as compared to rituximab and romiplostim episodes (p < .05; data not shown). Similar to the univariate results, the occurrence of headache events was significantly higher among IVIG episodes than that among anti-D episodes in the multivariate model (p < .05; data not shown). Hemolytic anemia rates were similar across all treatments.

Table 5. GEE model results for overall AE rates.

	Odds Ratio	95% Confidence Interval		p value
		Lower Limit	Upper Limit
Comparators
IVIG vs Anti-D	1.40	1.09	1.80	.009
Rituximab vs Anti-D	1.98	1.53	2.57	<.0001
Romiplostim vs Anti-D	1.84	1.39	2.43	<.0001
Eltrombopag vs Anti-D	1.52	1.06	2.19	.024
Age	1.00	0.99	1.00	.867
Gender
Male vs Female	0.84	0.72	0.98	.030
Geographic region
Midwest vs Northeast	0.98	0.81	1.20	.866
South vs Northeast	0.99	0.81	1.21	.930
West vs Northeast	0.98	0.75	1.28	.901
CCI score
1 vs 0	1.16	0.93	1.43	.180
2–3 vs 0	1.44	1.18	1.75	.0003
≥ 4 vs 0	1.80	1.42	2.27	<.0001

AE costs

Overall, costs associated with the majority of the AEs are in the range of $2,000–$7,000 per case (Figure 2). Urinary tract infection ($7,200) is associated with the highest cost per case, followed by aseptic meningitis ($5,802) and fever ($5,080). These high cost items had more inpatient care-related costs. The majority of the costs per AE did not significantly differ by treatment (Table 6). Mean back pain and headache costs associated with rituximab episodes were significantly lower than those with anti-D ($1,052 (SD = $3,224) vs $5,788 ($5,053); p = .008, and $505 ($959) vs $3,668 ($5,502); p = .027).

Figure 2. Overall mean AE costs per case.

Table 6. AE costs per event by drug.

	Anti-D (n = 106)^a		IVIG (n = 402)^a		Rituximab (n = 513)^a		Romiplostim (n = 469)^a		Eltrombopag (n = 56)^a
	Mean (SD)	Median	Mean (SD)	Median	Mean (SD)	Median	Mean (SD)	Median	Mean (SD)	Median
Arthralgia/Musculoskeletal  pain	$1,361 ($2,816)	$160	$3,812 ($12,081)	$163	$692 ($1,926)	$144	$613 ($916)	$259	$462 ($726)	$98
Aseptic meningitis	$0 ($0)	$0	$5,802 ($5,037)	$5,802	$0 ($0)	$0	$0 ($0)	$0	$0 ($0)	$0
Asthenia/Fatigue	$1,747 ($3,265)	$165	$2,672 ($4,335)	$150	$4,910 ($8,282)	$135	$2,923 ($2,385)	$2,447	$1,804 ($3,378)	$180
Back pain	$5,788 ($5,053)	$6,167	$2,231 ($6,116)	$101	$1,052* ($3,224)	$66	$2,455 ($12,566)	$115	$299 ($320)	$176
Contusion	$0 ($0)	$0	$2,323 ($4,063)	$120	$11,816 ($16,414)	$11,816	$236 ($127)	$236	$138 ($14)	$138
Cough	$0 ($0)	$0	$1,605 ($3,998)	$78	$263 ($377)	$146	$116 ($130)	$72	$73 ($93)	$31
Diarrhea	$11,812 ($0)	$11,812	$10,203 ($14,032)	$4,496	$509** ($549)	$198	$1,171* ($3,065)	$110
Dyspnea	$2,721 ($4,332)	$431	$2,511 ($2,812)	$1,052	$2,438 ($6,670)	$200	$3,786 ($3,741)	$4,910	$132 ($158)	$93
Fever	$2,738 ($2,707)	$1,672	$27,629 ($38,548)	$12,261	$3,445 ($5,462)	$1,303	$300* ($369)	$163	$191 ($132)	$191
Headache	$3,668 ($5,502)	$602	$6,477 ($25,864)	$1,246	$505* ($959)	$128	$1,282 ($1,639)	$419	$317 ($460)	$106
Hemolytic anemia	$6,424 ($4,981)	$5,119	$5,589 ($10,298)	$4,789	$4,129 ($3,999)	$4,182	$3,360* ($3,566)	$3,038	$219 ($47)	$234
Infusion reactions	$2,273 ($2,055)	$1,993	$9,703* ($0)	$9,703	$4,162 ($0)	$4,162	$0 ($0)	$0	$0 ($0)	$0
Nasal congestion	$0 ($0)	$0	$4,029 ($7,700)	$226	$228 ($197)	$342	$6,665 ($4,382)	$7,691	$0 ($0)	$0
Nausea/Vomiting	$3,461 ($3,560)	$1,669	$3,955 ($4,302)	$2,814	$3,670 ($3,367)	$4,172	$1,255** ($2,166)	$424	$0 ($0)	$0
Upper respiratory tract  infection	$1,489 ($2,652)	$92	$927 ($2,795)	$123	$681 ($1,884)	$129	$612 ($2,070)	$101	$1,479 ($4,651)	$68
Urinary tract infection	$5,262 ($11,289)	$201	$13,263 ($24,673)	$705	$4,434 ($9,420)	$130	$3,037 ($6,498)	$437	$149 ($35)	$149

*p < .05; **p < .001.

^a n represents the total number of cases.

Discussion

This study presents real-world evidence on comparative treatment-related AE costs associated with ITP therapies among adults. Based on crude or adjusted AE rates, anti-D episodes were associated with the lowest overall AE rate compared with that among other comparators. Nausea/vomiting and arthralgia/musculoskeletal pain were the most commonly found AEs across treatments, followed by headache and upper respiratory infection. Hemolytic anemia events did not differ significantly across treatments.

Overall, the majority of the rates across treatments reported in this study are lower than the clinical trial-based rates^12–15. The differences may be attributable to differences between clinical prospective study and retrospective observational study design. Clinical studies are conducted in controlled environments, while claims databases lack clinical information. Nonetheless, these reported AE rates suggest and confirm that claims for ITP treatment-related AEs are common in real-world experience. Across treatments, urinary infection, aseptic meningitis, and fever were associated with the highest costs per event. Cost to manage each AE did not differ by treatment. Due to gaps in the literature estimating AE costs among ITP patients, results could not be compared to the published literature. These cost estimates and rates provide insight into the real-world economic burden of AEs among patients with ITP, and can also be used in economic modeling studies for ITP treatments.

Results of this study should be interpreted in light of certain limitations, however. Data reflect a largely commercially- or self-insured population, and is, therefore, not generalizable to a Medicare or Medicaid population. As with all administrative claims databases, it should be noted that claims are collected for reimbursement rather than research purposes, and, thus, the analysis is constrained by the diagnosis codes, which may include incomplete or incorrect diagnosis on medical records. The study assumed all information required for case classification was present across the cohorts of interest. The causal relationship among study drugs, AEs, and AE costs cannot be fully explained by the data source. In the absence of clinical data, as is true in claims data, it may not be feasible to determine if the observed AEs were directly related to the treatment, ITP condition severity, AE severity, or pre-existing conditions exacerbated by current treatments, or residual effects from prior treatments. Interpretation of results should be done with recognition that such clinical markers were not addressed. However, this analysis attempted to address this by limiting the observation period of AEs of within 3 days or 1 week of a treatment dose. GEE modeling also controlled for CCI in an attempt to adjust for underlying health differences between patients when estimating rates, and no significant difference in AE costs was found by treatment, as expected given a hypothesis that AE costs relate to the AE itself and not the background therapy.

Additionally, due to small sample sizes in the data for this rare condition, multiple treatment episodes were identified to increase the sample size; it was not feasible, therefore, to stratify the analysis and identify any differential burden by line of therapy in the current analysis. Similarly, due to a lack of stratification by line, it was not possible to account for timing of any splenectomies that occurred in analyses. However, GEE modeling was performed only on overall AE rates because of this sample size consideration. Also, generalizability of results should be carefully considered in light of the limited sample size. As additional data become available over time, additional analysis by line of therapy could be valuable to understand whether therapeutic sequence impacts AE burden.

The data are also limited to direct costs and do not report AE severity. As some AEs may be associated with indirect costs, the total AE costs in the current study may under-estimate the overall burden of treatment-related AEs associated with ITP therapies. Moreover, without a severity indicator, it is unclear whether high AE costs are due to the AE severity, or to uncoded concomitant diagnoses; however, because we anticipate that coding will follow the item with the highest reimbursement, we assume that the costs can be largely attributed to the specified code. Furthermore, AEs in this study reflect only those identified via FDA prescribing information, and, therefore, some AEs may be omitted from consideration; however, this implies that the burden of AEs as represented in the current evaluation may be a low estimate, and should be interpreted accordingly.

Lastly, since both the TPO agents were approved in the US in the second half of 2008 (with a mandatory Risk and Evaluation Mitigation strategy program until December 2011^27,28), sample sizes for both romiplostim and eltrombopag are lower than that for other drugs, because of the availability of data at the time when the analysis was conducted. Subsequent analysis with more recent data may enable larger sample sizes for these drugs, and more robust comparisons.

Although lower than trial-based AE rates, this claims analysis provides novel real-world evidence that AEs related to ITP treatment, whether on- or off-label, are common, and associated costs may become burdensome, especially for patients receiving chronic therapy. This is also important to consider in the context of the primary ITP drug costs. Although absolute values vary based on publication methodology, studies suggest that therapy costs may range from $2,000–$3,000 per dose for anti-D, and somewhat higher values for IVIG^29,30, to as much as $10,000–$50,000 for a course of rituximab³¹. Taken together, the overall burden related to ITP treatment and AE management may be substantial, and decision-makers may need to consider all components when choosing therapies. This study indicates that future analyses would be of value, including validation of the high AE burden in other US and ex-US databases, exploring whether differential burden exists by line of therapy, and calculating whether the overall cost burden differs across available treatment options.

Transparency

Declaration of funding

Funding for this study was provided by Emergent BioSolutions.

Declaration of financial/other relationships

TB and GL are employed by Emergent BioSolutions. The other authors have no relationships to be declared. JME peer reviewers on this manuscript have no relevant financial or other relationships to disclose.

Previous presentation

Aspects of this study were presented at the International Society on Thrombosis and Haemostasis (ISTH) 2015 Congress, Toronto, June 20–25, 2015.

Acknowledgments

No assistance in the preparation of this article is to be declared.