Indirect costs associated with metastatic breast cancer

Abstract

Objective:

To compare the indirect costs of productivity loss between metastatic breast cancer (MBC) and early stage breast cancer (EBC) patients, as well as their respective family members.

Methods:

The MarketScan® Health and Productivity Management database (2005–2009) was used. Adult BC patients eligible for employee benefits of sick leave and/or short-term disability were identified with ICD-9 codes. Difference in sick leave and short-term disability days was calculated between MBC patients and their propensity score matched EBC cohort and general population (controls) during a 12-month follow-up period. Generalized linear models were used to examine the impact of MBC on indirect costs to patients and their families.

Results:

A total of 139 MBC, 432 EBC, and 820 controls were eligible for sick leave and 432 MBC, 1552 EBC, and 4682 controls were eligible for short-term disability (not mutually exclusive). After matching, no statistical difference was found in sick leave days and the associated costs between MBC and EBC cohorts. However, MBC patients had significantly higher short-term disability costs than EBC patients and controls (MBC: $6166 ± $9194 vs EBC: $3690 ± $6673 vs Controls: $558 ± $2487, both p < 0.001). MBC patients had more sick leave cost than controls ($2383 ± $5539 vs $1282 ± $2083, p < 0.05). Controlling for covariates, MBC patients incurred 47% more short-term disability costs vs EBC patients (p = 0.009). Older patients (p = 0.002), non-HMO payers (p < 0.05), or patients not receiving chemotherapy during follow-up (p < 0.001) were associated with lower short-term disability costs. MBC patients’ families incurred 39.7% (p = 0.06) higher indirect costs compared to EBC patients’ families after controlling for key covariates.

Conclusion:

Productivity loss and associated costs in MBC patients are substantially higher than EBC patients or the general population. These findings underscore the economic burden of MBC from a US societal perspective. Various treatment regimens should be evaluated to identify opportunities to reduce the disease burden from the societal perspective.

Keywords::

• Indirect costs
• Metastatic breast cancer
• Productivity loss
• Employer

Introduction

Breast cancer is the most common cancer among women in the western world. According to the American Society of Cancer (ACS), ∼288,130 women were diagnosed with breast cancer in 2011 as new cases, comprising 30% of all cancers in women1. An estimated 39,520 breast cancer deaths in women were expected in 2011 according to ACS data1, making it the second leading cause of cancer death in women (after lung cancer). Moreover, breast cancer is the leading cancer killer for women between the ages of 20–59 years in high-income countries2.

Metastatic breast cancer represents a small proportion of all breast cancers. A study identified 4.9% of all breast cancer cases as distant metastatic disease using the Surveillance, Epidemiology and End Results (SEER) data from 1990–20033. Compared to early stage breast cancer, which is usually treatable, metastatic breast cancer (MBC) has a much poorer prognosis. Once breast cancer metastasizes to the distant parts of the body, typically bones, lungs, brain, and/or liver4, it becomes essentially not curable. A woman typically survives for 2–3 years following the diagnosis of MBC in the US5. The 5-year overall survival rate of MBC is ∼26%2. Treatment for MBC is, thus, generally palliative, with the goal of maximizing survival time and quality-of-life (QoL)6. Although surgery and radiation are effective for achieving local control, MBC treatment requires systemic therapy that may include chemotherapy combinations, hormonal therapy, and molecular targeted therapies. These aggressive therapies mean that MBC patients are often exposed to multiple agents with inherent neurotoxic side-effects7.

Breast cancer is a costly disease to treat. It has been estimated that ∼$8.1 billion per year is spent in BC treatment-associated direct medical costs in the US8. For each patient, the estimated lifetime treatment cost ranges from $20,000–$100,0009,10. Although MBC accounts for a small proportion of all breast cancer cases, the economic burden of treating MBC is significant given the growing number of treatment options11. Additionally, since breast cancer is mostly prevalent in women of working age, indirect costs such as productivity loss should not be overlooked8,9. However, there has been little research related to detailing the indirect costs of MBC/BC.

Fu et al.12 assessed the incremental indirect costs associated with breast cancer by quantifying the impact of absenteeism (paid time when absent from work) and short-term disability in the US. The authors found that, for the first year of disease post-diagnosis, breast cancer patients incurred nearly twice the amount of absenteeism-related indirect costs and more than 10-times the short-term disability related indirect costs vs a matched control group. However, the cost estimates were not stratified by breast cancer stage and no indirect costs specifically associated with MBC were reported. To the best of our knowledge, no population-based studies have directly assessed the indirect costs associated with metastatic breast cancer in the US patient population. Lidgren et al.’s13 study, which was conducted in Sweden, is the only one that has examined the indirect costs associated with MBC. It suggests that the total per-patient indirect costs of MBC due to absentiseem and early retirement is similar to the direct medical costs for treating MBC, ranging from $17,535–$21,086 (2005 US$) annually, depending on patient age.

In addition to missed work days of the patients with the disease, the economic burden of their family caregivers are often overlooked and poorly understood. Family members are often playing a vital role in the end-of-life care, suggesting income security and job protection of family members who care for dying patients14. A survey study in Canada15 found that 5% of the family caregivers of advanced breast cancer patients had to quit their job or decline advancement, along with a significant decrease in their work hours as well. However, no details were reported in terms of indirect costs associated with productivity loss in this Canadian study. No published studies have examined the economic burden of MBC from the family members’ perspective in the US, which could be substantial given that caregivers also incur the economic hardships related to lost work productivity and/or wages.

The primary objective of this study was thus to examine the indirect costs of MBC incurred by employers in the US using the data from a large US national, commercially insured claims database. In order to better understand the full economic burden of MBC, indirect costs associated with MBC patients’ family members are also evaluated. The indirect costs measured in this study refer to employer-incurred expenses that affect workplace productivity including absenteeism/sick leave and short-term disability.

Methods

Study design and data source

Data source

To estimate indirect costs associated with MBC, we conducted a retrospective matched-cohort study using longitudinal integrated medical and pharmacy claims data from the MarketScan® Commercial Claims and Encounters (CCEA) and Health and Productivity Management (HPM) databases. The CCEA database consists of the health service claims of ∼110 million employees, dependents, and retirees with health coverage through privately-insured fee-for-service or capitated health plans in the US. Data are included from both salaried and hourly employees. These data represent the US national, commercially-insured population, and they capture the full continuum of care in all settings, including physician office visits, hospital stays, and outpatient pharmacy claims. The HPM database is a sub-set of the CCEA database, including information about absenteeism or short-term disability program services contributed by large employers who are contracted with Thomson for healthcare information, consulting, and research services.

The data confirmed to the Health Insurance Portability and Accountability Act of 1996 confidentiality requirements such that neither informed consent nor Institutional Review Board approval was necessary for this study.

Sample selection

The study sample included employees who had at least two inpatient or outpatient claims with a diagnosis of BC (International Classification of Disease, 9th Revision, Clinical Modification (ICD-9-CM) 174.xx) within 1 year (n = 326,903), as shown in Table 1. The study sample was limited to women between 18–64 years of age to ensure that the absenteeism and short-term disability outcomes were more related to working-aged people (n = 326,810). Patients were excluded if they had a diagnosis of other cancers (ICD-9-CM 140.xx–173.xx, 175.xx–195.xx, 200.xx–208.xx) other than breast cancer. A total of 269,372 employees were included from the analyses by excluding other cancer patients. The study sample was further divided into early stage breast cancer (EBC) and MBC cohorts. Employees with an ICD-9 diagnosis of metastasis (196.xx–198.xx, 199.0x, 199.1x) after their first observed claim of breast cancer diagnoses were identified as MBC cohort. Patients with only breast cancer diagnosis, but no metastasis diagnosis codes during the study period, were defined as EBC cohorts.

Table 1. Sample selection.

Criteria for BC/MBC patients	Number of patients
[0] Employees with diagnosis of breast cancer from 2005–2009	326,903
[1] Employees 18 years or older at diagnosis of breast cancer	326,810
[2] Employees without other primary malignant neoplasm diagnosis from 2005–2009	269,372
[3a] Employees with health productivity data and diagnosed with early stage breast cancer	8,421
[3b] Employees with health productivity data and diagnosed with metastatic breast cancer	1,587
[4a] Employees with continuous medical insurance enrolment 6 months before and 12 months after the index date (early stage breast cancer)	2,248
[4b] Employees with continuous medical insurance enrolment 6 months before and 12 months after the index date (metastatic breast cancer)	661

The service date of the first observed claim with a diagnosis of BC or metastasis had occurred between 2005–2009; eligible EBC (n = 8421)/MBC (n = 1587) patients must have had the health productivity data (sick leave/short-term disability) during the year following the index year. In addition, the study cohorts were also restricted to patients having no diagnosis of BC within the 6 months before the index date. In total, 2248 EBC and 661 MBC patients with continuous enrollment for 6 months before and 12 months after the index date were included in the analysis cohort. Out of the 661 MBC patients, 432 patients were eligible for short-term disability and 139 patients were eligible for sick leave during the follow-up (not mutually exclusive). Out of the 2248 EBC patients, 1552 patients were eligible for short-term disability and 432 patients were eligible for sick leave during the follow-up (not mutually exclusive).

MBC and EBC patients’ family members were also identified using ‘ENROLID’ (a patient unique identification ID). Enrollees from the same family were assigned an ENROLID with the same first eight digits. The last two digits of ENROLID were different for individuals from the same family. For example, if a family member had an ENROLID of ‘1234567812’, other family members could have ENROLID such as ‘1234567810’. A total of 1166 EBC and 209 MBC patients’ adult family members were selected. Eligible family members were those who had continuous enrollment of medical insurance and productivity data for 1 year after the index date.

Measurements

Patient characteristics, including demographics (age, geographic regions, health insurance plan) at the time of index date were measured. Health status variables included the Charlson Comorbidity Index (CCI), a numeric scale reflecting the risk of death or serious disability in the next year based on the presence of a diagnosis for one of 19 conditions (e.g., diabetes, heart disease, cancer) in the 6-month pre-index period.

Work loss is represented by paid time off (including absentiseem and short-term disability). The absentiseem data contained detailed information about each employee’s absence from work, including hours missed from work, and the dates and type of absence. More specifically, the type of absence includes sick leave, personal leave, vacation, disability, and Family and Medical Leave Act (FMLA, that entitles eligible employees to take unpaid, job-protected leave for their family members and medical reasons)16. The short-term disability data contained information about the start and end date of leave due to disability. In our study, patients’ indirect costs were comprised of two components: sick leave from the absentiseem data and short-term disability. Family members’ indirect costs were defined as the sum of costs due to personal leave and leave under FMLA, excluding other types of absentiseem such as sick leave and vacation. Because cost estimates for absence from work were not included in the employers’ absentiseem and short-term disability files, the costs of absentiseem and short-term disability were calculated by multiplying the 2012 BLS average hourly rate and benefit for all the industries17.

Statistical analysis

Characteristics of study subjects that were examined, included age, geographic region, index year, insurance plan, Charlson Comorbidity index, baseline chemotherapy, and chemotherapy during the follow-up. Continuous measures (e.g., age) were summarized by mean and standard deviation and comparisons of the differences in continuous measures between study groups were made using the student t-tests. Categorical variables were summarized as proportions of the sample with the characteristic and compared using chi-square tests between study groups.

The primary objective of the study was to measure the extent of absenteeism and short-term disability between female employees with MBC and EBC, and to use these values to determine the incremental differences in indirect costs. Propensity score matching technique was employed using greedy matching algorithm for MBC and EBC patients based on a 1:1 ratio18. A series of variables as described above were included in logistic regressions for generating a propensity score for each patient. Student t-test was used to compare the work-loss days and the associated costs before and after matching19.

When estimating the cost difference between MBC and EBC patients, generalized linear models (GLMs) with log link and gamma distributions were used. Linear regression analyses were conducted to estimate excess indirect costs for employees with MBC compared with employees with EBC/without BC, controlling for the effects of socio-demographic, health status, and chemotherapy use. It has been recommended to use GLMs with a gamma distribution and log link for the multivariate analysis of cost data since they are likely to be skewed20,21.

All analyses were conducted using SAS version 9.2 (SAS Institute Inc., Cary, NC). p-values ≤ 0.05 were considered statistically significant.

Results

Indirect costs associated with MBC patients

Sick leave

Based on the inclusion/exclusion criteria, 139 MBC patients, 432 EBC patients, and 820 patients without diagnosis of breast cancer (controls) were selected for the sick leave outcome, with a mean age of 49, 51, and 50 years, respectively. Table 2 displays the mean values for patient characteristics, health status measures, and sick leave outcomes comparing MBC patients to propensity score-matched EBC patients and unmatched controls. Matching was successful, as all covariates were not significantly different between MBC and EBC patients (all p > 0.05).

Table 2. Patient characteristics and outcomes of sick leave cohorts.

Variable	EBC (n = 139)		MBC (n = 139)		Control (n = 820)		p-value (MBC vs EBC)	p-value (MBC vs Control)
Age in years, Mean (SD)	49.32	(7.39)	49.65	(7.07)	49.83	(6.76)	0.71	0.42
Age categories, n (%)							0.69	0.37
1: 18–34	5	(3.6%)	2	(1.4%)	12	(1.5%)
2: 35–44	28	(20%)	31	(22%)	168	(20%)
3: 45–54	72	(52%)	71	(51%)	430	(52%)
4: 55+	34	(24%)	35	(25%)	210	(26%)
Diagnosis index year, n (%)							0.96	<0.001
2005	18	(13%)	17	(12%)	333	(41%)
2006	32	(23%)	35	(25%)	143	(17%)
2007	41	(29%)	38	(27%)	215	(26%)
2008	48	(35%)	49	(35%)	129	(16%)
Region, n (%)							0.96	0.09
Northeast	20	(14%)	17	(12%)	131	(16%)
North Central	38	(27%)	40	(29%)	244	(30%)
South	67	(48%)	68	(49%)	305	(37%)
West	14	(10%)	14	(10%)	135	(16%)
Unknown					5	(0.61%)
Health Plan, n (%)							0.46	0.14
POS	65	(47%)	71	(51%)	288	(35%)
HMO	19	(14%)	11	(7.9%)	146	(18%)
PPO	50	(36%)	53	(38%)	341	(42%)
Comprehensive	5	(3.6%)	4	(2.9%)	35	(4.3%)
Other					9	(1.1%)
Unknown					1	(0.12%)
Comorbidity index, Mean (SD)	0.11	(0.33)	0.08	(0.30)	0.11	(0.37)	0.45	0.84
Patients w/non-zero sick leave days (n, %)	107	(77%)	108	(78%)	578	(70%)	0.84	0.12
Sick leave days	15.86	(36.61)	11.88	(17.58)	8.58	(13.97)	0.25	<0.001
Indirect costs associated with sick leave ($)	2383.3	(5539.2)	1775.4	(2628.8)	1282.3	(2083.0)	0.24	<0.001

Two regressions (Tables 3 and 4) were performed to compare the sick leave associated costs for the MBC vs EBC group and the MBC vs Control group. The key covariates adjusted in the regressions include age, region, health insurance, index diagnosis year, baseline chemotherapy, chemotherapy during follow-up, and Charlson comorbidity index. There were no significant differences in sick leave cost between MBC and EBC patients after controlling for the covariates (Table 3). Compared to the control group/general population, MBC patients incurred 56.1% ($1584 [CI = $1160–$2164] vs $1015 [CI = $899–$1146]; p = 0.01) more sick leave cost in the adjusted model (Table 4). Other cost drivers include region and health insurance plans: patients residing in the South/Northeast had higher costs vs the Northwest (all p < 0.05), and patients with insurance of POS/comprehensive incurred higher costs vs HMO (all p < 0.05).

Table 3. GLM model of indirect costs associated with patients’ short-term disability/sick leave (MBC vs EBC).

Variable	Indirect cost of sick leave			Indirect cost of short-term disability
	Exp(B)	95% CI	p-value	Exp(B)	95% CI	p-value
Intercept	1246.072	(344.71, 4504.31)	<0.001	6073.558	(2277.61, 16,195.96)	<0.001
MBC vs EBC	1.06	(0.74, 1.53)	0.753	1.477	(1.1, 1.98)	0.009
Age	1.003	(0.98, 1.02)	0.793	0.973	(0.96, 0.99)	0.002
Index year (ref = 2005)
2008	1.377	(0.87, 2.19)	0.175	1.889	(1.31, 2.72)	<0.001
2007	1.218	(0.76, 1.95)	0.413	1.842	(1.28, 2.65)	<0.001
2006	1.18	(0.71, 1.96)	0.524	1.627	(1.12, 2.37)	0.011
Region (ref = Northwest)
West	1.057	(0.69, 1.63)	0.800	0.758	(0.53, 1.08)	0.127
South	1.991	(1.39, 2.86)	<0.001	0.939	(0.69, 1.28)	0.689
Northeast	1.509	(0.94, 2.41)	0.086	1.259	(0.89, 1.78)	0.193
Unknown	9.267	(0.91, 94.35)	0.060	0.001	(0, 0.09)	0.003
Health insurance plan (ref = HMO)
PPO	0.841	(0.52, 1.35)	0.474	0.63	(0.47, 0.84)	0.002
POS	0.281	(0.17, 0.45)	<0.001	0.681	(0.47, 0.99)	0.043
Other	0.771	(0.11, 5.26)	0.790	0.368	(0.22, 0.63)	<0.001
Comprehensive	0.284	(0.1, 0.78)	0.015	0.756	(0.43, 1.32)	0.326
Unknown				1.25	(0.24, 6.53)	0.792
Baseline Chemotherapy use (ref = No)	0.58	(0.27, 1.23)	0.155	0.839	(0.43, 1.63)	0.606
Chemotherapy use during follow-up (ref = No)	1.247	(0.9, 1.72)	0.180	3.773	(2.95, 4.82)	<0.001
Charlson Comorbidity index	1.216	(0.82, 1.81)	0.334	1.06	(0.79, 1.42)	0.700

Table 4. GLM model of indirect costs associated with patients’ short-term disability/Sick leave (MBC vs Controls).

Variable	Indirect cost of sick leave			Indirect cost of short-term disability
	Exp(B)	95% CI	p-value	Exp(B)	95% CI	p-value
Intercept	645.211	(252.38, 1649.51)	<0.001	782.055	(422.95, 1446.05)	<0.001
MBC vs Control	1.561	(1.11, 2.2)	0.010	12.584	(9.55, 16.58)	<0.001
Age	1.011	(0.99, 1.03)	0.198	0.992	(0.98, 1)	0.172
Index year (ref = 2005)
2008	1.298	(0.92, 1.82)	0.132	0.838	(0.68, 1.03)	0.094
2007	1.537	(1.13, 2.08)	0.005	0.847	(0.69, 1.05)	0.122
2006	1.619	(1.14, 2.3)	0.007	0.772	(0.61, 0.98)	0.033
Region (ref = Northwest)
West	1.156	(0.8, 1.67)	0.441	1.181	(0.94, 1.48)	0.152
South	1.374	(1.02, 1.84)	0.034	1.111	(0.91, 1.35)	0.287
Northeast	1.704	(1.18, 2.45)	0.004	0.963	(0.77, 1.2)	0.736
Unknown	2.139	(0.45, 10.14)	0.338	0.084	(0.02, 0.43)	0.003
Health insurance plan (ref = HMO)
PPO	0.88	(0.63, 1.22)	0.444	1.231	(1.02, 1.49)	0.031
POS	0.276	(0.2, 0.39)	<0.001	0.742	(0.56, 0.99)	0.040
Other	0.302	(0.09, 1.04)	0.057	0.67	(0.45, 0.99)	0.047
Comprehensive	0.193	(0.1, 0.36)	<0.001	0.84	(0.58, 1.21)	0.354
Unknown	0.001	(0, 0.03)	<0.001	0.998	(0.54, 1.83)	0.995
Charlson Comorbidity index	1.319	(0.97, 1.8)	0.079	1.423	(1.19, 1.7)	<0.001

Short-term disability

Similarly, 432 MBC patients, 1552 EBC patients, and 4682 controls were selected for the short-term disability outcome with a mean age of 49, 51, and 50 years, respectively. Table 5 revealed that indirect cost burden results of the MBC group are presented for short-term disability patient samples compared with the propensity score matched EBC groups and unmatched control groups. After employing propensity score matching for MBC and EBC patients based on a 1:1 ratio, there were 431 MBC and EBC patients included in the analysis (one MBC patient was dropped due to a lack of the exact match using the greedy matching algorithm). Their matching was successful since there were no significant differences in demographic and clinical variables between MBC and EBC patients in the short-term disability samples (Table 5). Employees with MBC had significantly higher short-term disability days and the associated cost in the 12 months after diagnosis than their matched EBC patients (41.2 ± 61.4 vs 24.7 ± 44.7 days; $6166 ± $9194 vs $3690 ± $6673; both p < 0.001) short-term disability (Table 5). In addition, MBC patients had higher short-term disability/costs than the control population (41.2 ± 61.4 vs 3.7 ± 16.6 days; $6166 ± $9194 vs $558 ± $2487; both p < 0.001).

Table 5. Patient characteristics and outcomes of short-term disability cohorts.

Variable	MBC (n = 431)		EBC (n = 431)		Control (n = 4682)		p-value (MBC vs EBC)	p-value(MBC vs Control)
Age in years, Mean (SD)	49.22	(7.18)	49.83	(7.41)	50.06	(7.01)	0.22	0.017
Age categories, n (%)							0.45	0.21
18–34	10	(1.2%)	5	(2.3%)	89	(1.9%)
35–44	99	(21%)	89	(23%)	907	(19%)
45–54	216	(52%)	222	(50%)	2380	(51%)
55+	106	(27%)	115	(25%)	1306	(28%)
Diagnosis index year, n (%)							0.68	<0.001
2005	54	(13%)	57	(13%)	1173	(25%)
2006	82	(16%)	71	(19%)	730	(16%)
2007	143	(32%)	138	(33%)	1223	(26%)
2008	152	(38%)	165	(35%)	1556	(33%)
Region, n (%)							0.98	0.34
Northeast	95	(22%)	96	(22%)	1058	(23%)
North Central	110	(25%)	107	(26%)	1245	(27%)
South	157	(38%)	162	(36%)	1506	(32%)
West	69	(15%)	66	(16%)	863	(18%)
Unknown					10	(.21%)
Health Plan, n (%)							0.99	<0.001
POS	72	(18%)	76	(17%)	490	(10%)
HMO	97	(23%)	97	(23%)	1071	(23%)
PPO	212	(49%)	211	(49%)	2592	(55%)
Comprehensive	23	(5.6%)	24	(5.3%)	268	(5.7%)
Other	24	(4.6%)	20	(5.6%)	184	(3.9%)
Unknown	3	(0.70%)	3	(0.70%)	77	(1.6%)
Comorbidity index, Mean (SD)	0.17	(0.43)	0.15	(0.44)	0.12	(0.39)	0.53	0.013
Patients with non-zero short-term disability days, n (%)	211	(49%)	162	(38%)	516	(11%)	<0.001	<0.001
Short-term disability days	41.24	(61.43)	24.70	(44.67)	3.73	(16.58)	<0.001	<0.001
Indirect costs associated with short-term disability ($)	6165.8	(9194.2)	3689.7	(6673.1)	558.34	(2487.4)	<0.001	<0.001

These results were followed by the multivariate analysis for indirect costs using generalized linear models, which are presented in Tables 3 and 4. MBC patients incurred 47.7% more short-term disability cost compared with EBC patients ($3953 [CI = $3072–$5086] vs $2676 [CI = $2360–$3034]; p = 0.009) and 11.6-times ($6683 [CI = $5140–$8689] vs $531 [CI = $492–$573]; p < 0.001) more short-term disability cost than controls. Older patients (p = 0.002), patients diagnosed in 2005 (p < 0.05), non-HMO payers (p < 0.05), or patients not receiving chemotherapy during follow-up (p < 0.001) were associated with lower short-term disability costs.

Indirect costs associated with patients’ family members

A total of 209 MBC and 1166 EBC patients’ adult family members were identified using the inclusion/exclusion criteria. The mean age of the family population was 51 years, and 99.9% of them were male (expected to be the patient’s spouse). There were no significant differences in the social-demographic and clinical characteristics between the study groups except for age (49.7 ± 6.6 vs 51.1 ± 6.5 years, p = 0.007), as shown in Table 6

Table 6. Family members’ characteristics and outcomes (MBC vs EBC).

Variable	MBC’s family members (n = 209)		EBC’s family members (n = 1166)		p-value
Age, Mean (SD)	49.69	(6.60)	51.01	(6.48)	0.007
Age, n (%)					0.023
1: 18–34	3	(1.4%)	13	(1.1%)
2: 35–44	44	(21%)	158	(14%)
3: 45–54	110	(53%)	627	(54%)
4: 55+	52	(25%)	368	(32%)
Sex, n (%)					0.55
Male	209	(100%)	1164	(100%)
Female	0	(0.00%)	2	(0.17%)
Diagnosis index year, n (%)					0.19
2005	84	(40%)	509	(44%)
2006	40	(19%)	166	(14%)
2007	45	(22%)	226	(19%)
2008	40	(19%)	265	(23%)
Region, n (%)					0.84
Northeast	43	(21%)	234	(20%)
North Central	37	(18%)	227	(19%)
South	93	(44%)	487	(42%)
West	36	(17%)	218	(19%)
Health Plan, n (%)					0.17
POS	78	(37%)	365	(31%)
HMO	28	(13%)	148	(13%)
PPO	100	(48%)	614	(53%)
Other	3	(1.4%)	39	(3.3%)
Occupation type, n (%)					0.32
Oil Gas Extraction, Mining	36	(17%)	173	(15%)
Manufacturing Durable Goods	78	(37%)	492	(42%)
Manufacturing Nondurable Goods	29	(14%)	125	(11%)
Transportation Communications Utilities	60	(29%)	321	(28%)
Services	6	(2.9%)	55	(4.7%)
Charlson Comorbidity index, Mean (SD)	0.19	(0.46)	0.18	(0.51)	0.82
Family members with non-zero leave days*, n (%)	92	(44%)	502	(43%)	0.80
Leave days among all families, Mean (SD)	2.81	(6.06)	2.07	(4.75)	0.047
Leave days among families with non-zero leave days	6.39	(7.80)	4.81	(6.27)	0.033
Cost of leave days among all families ($)	473.2	(1019.0)	348.1	(799.1)	0.047
Cost of leave days among families with non-zero leave days ($)	1075.1	(1311.1)	808.6	(1054.4)	0.033

*Leave days = FMLA leave days plus personal leave days.

MBC patients’ family members had higher leave days (including the leave under FMLA and other personal reasons) and associated costs than EBC (leave days: 2.8 ± 6.1 vs 2.1 ± 4.8, cost: $473 ± $1019 vs $348 ± $799, both p < 0.05). A similar proportion of families had non-zero leave days (FMLA plus personal leave days) between the MBC and EBC groups (44% vs 43%, p = 0.80). Among families who had non-zero leave days, MBC patients’ family members took higher leave days and the associated costs than EBC patients’ families (Table 6). After controlling for the covariates, including family members’ age, occupation type, Charlson comorbidity index, health insurance plan, and region, MBC patients’ family members had 39.7% ($403 [CI = $293–$555] vs $289 [CI = $253–$330]; p = 0.06) higher personal leave-related indirect costs compared to EBC (Table 7). In addition, family members’ occupation type (oil/gas extraction/mining vs manufacturing non-durable goods, Exp (β) = 2.978; p < 0.001) and geographic region (south vs west, Exp (β) = 2.124; northeast vs west, Exp (β) = 2.064; both p < 0.001) were significant influential factors as well.

Table 7. Regression of indirect costs of MBC vs EBC family members.

Variable	Exp(B)	95% CI	p-value
Intercept	868.726	(267.81, 2818)	<0.001
MBC vs EBC’s family members	1.397	(0.99, 1.98)	0.060
Age	0.987	(0.97, 1.01)	0.214
Region (ref = West)
South	2.124	(1.45, 3.12)	<0.001
Northeast	2.064	(1.35, 3.14)	<0.001
North Central	1.323	(0.85, 2.05)	0.213
Charlson Comorbidity Index	1.772	(0.95, 3.3)	0.071
Occupation type (ref = Manufacturing non-durable Goods)
Transportation/Communications/Utilities	0.517	(0.22, 1.2)	0.124
Services	0.704	(0.43, 1.15)	0.164
Oil/Gas Extraction/Mining	2.978	(1.85, 4.78)	<0.001
Manufacturing Durable Goods	1.033	(0.82, 1.31)	0.787
Health Plan (ref = HMO)
PPO	1.078	(0.71, 1.64)	0.725
POS	1.111	(0.65, 1.9)	0.700
Other	2.606	(1.16, 5.84)	0.020

Discussion

To our knowledge, our study is the first attempt to assess the indirect costs due to paid time off from work and short-term disability incurred by employees with MBC and their family members for taking care of the patients. This analysis revealed that indirect costs due to productivity loss (short-term disability or sick leave) were higher for employees who were diagnosed with MBC than those without breast cancer. Further, indirect costs due to short-term disability were higher for those who had MBC than those of EBC. MBC patients’ family members also incurred higher costs due to absence of work compared to EBC patients’ family members.

Fu et al.’s12 study examined the incremental indirect costs due to absenteeism or STDI and found that breast cancer patients had significant indirect costs due to absenteeism or STDI compared to those without breast cancer diagnosis—the incremental cost was $8068 ($1911 due to absenteeism plus $6157 due to short-term disability). Lidgren et al.’s13 study also reported a high annual indirect cost due to productivity loss in Sweden, which ranged from $17,535–$21,086 in 2005 US$. Consistent with the literature12,13, we found a significant difference in the annual indirect costs between patients with and without MBC within the first year of diagnosis—the incremental cost of MBC patients was $3084 ($608 due to sick leave plus $2476 due to short-term disability) compared to EBC patients and $6709 ($1101 due to sick leave plus $5608 due to short-term disability) compared to the population without breast cancer.

It is worth noting that we calculated the annual indirect costs within the first year of diagnosis of EBC or MBC in our analysis, since it is likely most patients take time off in the first year after diagnosis. Drolet et al.’s22 study found that most breast cancer patients took time off in the first year after diagnosis and absence from work in the year 2 or 3 considerably declined compared to year 1. The use of chemotherapy is a crucial factor increasing the absence duration among breast cancer patients22. The trend of MBC treatment is increasingly complex, with introduction of a growing number of new and expensive treatment options.

Given the fact that MBC is very resource-intensive and its treatment is usually longer lasting and more toxic than that of EBC, new technologies that minimize the toxicity of treatment are likely to simultaneously reduce the indirect costs due to the productivity loss from employers’ perspective. It has been recommended23,24 that not only the direct costs of treatment, but also the indirect costs associated with the treatment should be included when performing economic evaluation in cost-effectiveness analyses from a societal perspective24. Omitting the indirect costs could under-estimate the benefits of new techniques with higher efficacy and/or less adverse events. As mentioned above, absenteeism is likely to be caused by treatment-related symptoms25, thus it is important for future studies to examine the impact of new techniques that will not only reduce direct costs, but also influence indirect costs of productivity loss.

In terms of the magnitude of the indirect costs, Fu et al.12 estimated that the indirect costs of overall breast cancer in the US were $4972 in 2009 dollars due to absentiseem and $7199 due to short-term disability per woman within the first year of diagnosis of breast cancer. Our study found lower indirect costs for both sick leave and short-term disability compared to the Fu et al.12 study: the annual indirect costs per employee were estimated at $1775 due to sick leave/$2383 due to short-term disability among employees with EBC and $3690 due to sick leave/$6166 due to short-term disability for MBC employees, respectively. The lower costs we identified may be partly due to the differences in the studied population: our analysis excluded the patients with tumor other than breast cancer from the analysis cohort to minimize the possibility that the cause of productivity loss is related to tumors other than breast cancer. In addition, our study stratified the breast cancer population into early stage breast cancer and metastatic breast cancer cohorts, while Fu et al.’s12 study examined one breast cancer population in which early stage and later stage breast cancer were not differentiated. Another contributing factor is that we differentiated sick leave data from other types of absenteeism data by excluding other types of absenteeism days (e.g., vacation, disability, FMLA, etc.) from the calculation of sick leave. We chose to only use the sick leave data in order to avoid double counting on disability costs and minimize the possibility of involving the leave days not related with the disease, such as vacation.

A meaningful contribution of the present study is the assessment of the productivity loss to employers associated with MBC, which helps to understand the full economic value of treatments to BC/MBC. More new treatments have been emerging for breast cancer in recent years. However, questions are often raised about the value brought by new treatment options if they are small and increase the costs. To answer such questions, it is crucial to understand the full economic cost of treatments to BC/MBC. Numerous studies26 have evaluated the direct medical costs of treating MBC; however, very few studies looked at the economic burden from the societal perspective and the indirect costs are often neglected.

A second contribution of the study is the assessment of the indirect costs associated with the MBC patients’ family members, who may also face financial burden due to the work loss. Previous studies27 have shed light on the burden of family caregivers for breast cancer patients from the physiological perspective, while studies focusing on the economic burden of caregivers were rare. Limited data have been reported for the burden on caregivers for those with advanced breast cancer26. Grunfeld et al.15 conducted a study assessing the burden of family caregivers of patients with advanced breast cancer in Ontario, Canada. Their study found a significant occupational burden, with 5% of the families quitting their job or declining promotions, as well as significant loss in work hours. Our analyses of indirect costs of patients’ family members using a retrospective cohort study design is original since it is unique in a way that the Marketscan HMP database allows one to track the enrollees from the same family (enrolled within the same health insurance plan). After adjusting for key covariates, MBC patients’ families incurred 39.7% higher indirect costs compared to EBC patients’ families. In addition to MBC disease itself, the families’ insurance plan type, region, and their occupation were significant cost influential factors as well. Another intriguing finding is that families who worked in the industry of oil, gas extraction, or mining were more likely to take time off under FMLA or for personal leave.

Our analysis has several limitations. First, the reason for patients’ taking sick leave is unknown. For example, rather than taking work days off for MBC disease related reasons, patients may have taken sick leave for other disease conditions or for their families. The uncertainty of the inclusiveness of sick leave data could be the driving force behind the small difference in sick leave days/costs between MBC and EBC patients. Similarly, without knowing the reasons of workday absence, a conclusive statement that the workdays taken off by patients’ families were used exclusively for caring for the patients cannot be made. To minimize the bias introduced by this limitation, we only calculated the workday absence due to personal reasons or under FMLA for patients’ families. Secondly, the indirect costs are likely to be under-estimated by the current results: the productivity losses of some caregivers, such as time spent by friends to care for the patients, were not examined in this study. The study relied on the sick leave and short-term disability data as the only measure of productivity loss. Additional sources of productivity loss, such as those associated with time spent in seeking cancer treatment and care, transportation costs and presenteeism, were not assessed in our study. Thirdly, it may be difficult to generalize the study results beyond the large- and medium-sized firms represented in the study sample28. These firms may be more likely to offer more generous medical and disability benefits, and may have a greater capacity to accommodate chronically ill workers on the job, which may result in lower sick leave and short-term disability estimates.

In conclusion, MBC imposes a significant societal and employer burden, as indicated by the higher indirect costs (due to sick leave or short-term disability) of those who had MBC compared to those with EBC or without breast cancer. Understanding the indirect costs associated with MBC is essential in understanding the economic burden of disease and provides the information needed to assess the cost-effectiveness of medical techniques from the societal perspective.

Transparency

Declaration of funding

Financial support for this study was provided by Eisai, Inc. to Pharmerit North America; Eisai also authorized submission of the manuscript for publication.

Declaration of financial/other relationships

Yin Wan, Xin Gao, and Sonam Mehta are employees of Pharmerit, an independent contract research organization with previous and ongoing engagements with Eisai, Inc. as well as other pharmaceutical manufacturers. Lee Schwartzberg is a practicing oncologist and has no relevant financial relationships to disclose. Claudio Faria and Zhixiao Wang are employees of Eisai, Inc. JME Peer Reviewers on this manuscript have no relevant financial relationships to disclose.

Acknowledgments

The author thanks John Cater of Pharmerit for helping with the reviewing, formatting, and submission of this article.