Annual patient and caregiver burden of oncology clinic visits for granulocyte-colony stimulating factor therapy in the US

Abstract

Objective:

Prophylactic treatment with granulocyte-colony stimulating factors (G-CSFs) is indicated for chemotherapy patients with a significant risk of febrile neutropenia. This study estimates the annual economic burden on patients and caregivers of clinic visits for prophylactic G-CSF injections in the US.

Methods:

Annual clinic visits for prophylactic G-CSF injections (all cancers) were estimated from national cancer incidence, chemotherapy treatment and G-CSF utilization data, and G-CSF sales and pricing information. Patient travel times, plus time spent in the clinic, were estimated from patient survey responses collected during a large prospective cohort study (the Prospective Study of the Relationship between Chemotherapy Dose Intensity and Mortality in Early-Stage (I–III) Breast Cancer Patients). Economic models were created to estimate travel costs, patient co-pays and the economic value of time spent by patients and caregivers in G-CSF clinic visits.

Results:

Estimated total clinic visits for prophylactic G-CSF injections in the US were 1.713 million for 2015. Mean (SD) travel time per visit was 62 (50) min; mean (SD) time in the clinic was 41 (68) min. Total annual time for travel to and from the clinic, plus time at the clinic, is estimated at 4.9 million hours, with patient and caregiver time valued at $91.8 million ($228 per patient). The estimated cumulative annual travel distance for G-CSF visits is 60.2 million miles, with a total transportation cost of $28.9 million ($72 per patient). Estimated patient co-pays were $61.1 million, ∼$36 per visit, $152 per patient. The total yearly economic impact on patients and caregivers is $182 million, ∼$450 per patient.

Limitations:

Data to support model parameters were limited. Study estimates are sensitive to the assumptions used.

Conclusions:

The burden of clinic visits for G-CSF therapy is a significant addition to the total economic burden borne by cancer patients and their families.

Keywords::

• Cancer
• Supportive care
• Granulocyte-colony stimulating factors
• Economic burden

Introduction

Cancer patients with solid tumors or non-myeloid malignancies receiving myelosuppressive anti-cancer drug therapy often have suppressed white blood cell counts and are at risk for febrile neutropenia (FN) and infections. FN is a potentially life-threatening condition1 and the costs of treating FN and its clinical sequelae are substantial, both to patients and to the healthcare system. Incidence of inpatient neutropenic complications among cancer patients receiving chemotherapy ranges from 11–19%2–4, depending on the therapy regimen, with the cost of treatment for an inpatient neutropenic event in the US averaging over $22,0005. Treatment with granulocyte colony-stimulating factors (G-CSFs) at least 24 h after administration of myelosuppressive chemotherapy (MC) is indicated for patients whose risk factors are associated with a clinically significant incidence of FN, as observed in clinical trials6. G-CSFs are safe and effective in decreasing the incidence of chemotherapy-induced FN, the major dose-limiting side-effect of chemotherapy. As compared to overall FN-related hospitalizations, real-world incidence of inpatient neutropenic events among patients treated with G-CSFs ranges from only ∼2–5%7–9, and the American Society of Clinical Oncology (ASCO) and the National Comprehensive Cancer Network Clinical Practice Guidelines in Oncology (NCCN Guidelines) each recommend prophylaxis with a G-CSF when FN risk is high (≥ 20%), based on either chemotherapy regimen risk alone or a combination of regimen risk, patient risk factors (e.g., age, type and number of comorbidities, pre-existing neutropenia or infection), and disease characteristics (e.g., tumor type and stage)10–12.

Because G-CSFs rapidly induce the proliferation of myeloid progenitor cells—which may be especially sensitive to myelotoxic agents, and could adversely impact the risk of FN—prescribing information specifies that G-CSFs should not be administered between 14 days before and 24 h after administration of myelosuppressive chemotherapy. Consequently, most doses of G-CSFs given in the US are usually administered during a separate visit after the chemotherapy session. Relatively few doses are self-administered at home by the patient or an informal caregiver such as a spouse or other family member. For patients on short-acting, daily-dosed G-CSFs, multiple additional clinic visits are required during each chemotherapy cycle, the time and costs of which can be burdensome for patients and the informal caregivers who accompany them on these visits. Travel time plus time in the clinic for each injection has been reported to be typically 2 h or more13–15, and travel costs plus out-of-pocket co-pays not covered by insurance can add up quickly16–18. Travel barriers (including long travel times) may be deterrents to proper clinical care in cancer patients19–34, leading to sub-optimal outcomes.

To better understand the total economic impacts of G-CSF therapy after MC, this study estimated the total time and cost burden on patients and informal caregivers of clinic visits for G-CSF injections in the US for 1 year.

Patients and methods

This study used a population modeling approach based on secondary data sources. The project consisted of a literature review for the collection of empirical data to use in generation of model estimates, development of national estimates of oncology office/hospital outpatient department (OPD) visit volume for G-CSF therapy, creation of per-visit cost and utilization estimation models, and generation of patient and caregiver cost estimates based on cost and utilization parameters found in the literature, from government and private websites and from patient questionnaire data collected as part of an ongoing prospective observational study (the Prospective Study of the Relationship between Chemotherapy Dose Intensity and Mortality in Early-Stage (I–III) Breast Cancer Patients, also called the ‘CABS’ study35). Deterministic models were created in Microsoft Excel to estimate the patient and caregiver economic burden of post-MC G-CSF therapy under the current standard of care, as of early 2015. This included treatment with both daily dosed (filgrastim, tbo-filgrastim) as well as long-acting (pegfilgrastim) G-CSF injections. No intervention effects or G-CSF market changes were estimated (e.g., for a change in the mix or volume of G-CSFs used as a result of biosimilars entering the US market). Our estimates attempted to mirror current clinical practice in the use of G-CSFs for primary prophylaxis in the prevention of chemotherapy-induced FN. G-CSF use in other settings (e.g., hospital inpatient, pharmacy) or for other indications (e.g., treatment of severe chronic FN, use in myeloid malignancies and hematologic cancers except Non-Hodgkins Lymphoma) were also excluded from the estimates.

Factors associated with G-CSF treatment decisions that were included in the models were type and stage of cancer, whether or not myelosuppressive chemotherapy is used to treat each case, whether or not G-CSFs are used prophylactically to reduce risk of FN and for how many cycles, what type of G-CSF is prescribed, how many days per cycle G-CSFs are prescribed (for daily G-CSFs), and whether the treatments will be self-administered at home or given in the clinic. Inputs which determined the patient and caregiver costs per office visit included the travel time to the clinic, an average travel speed (to convert travel time to miles), what mode of transportation the patient used to get to the clinic, whether they were accompanied by someone else to the clinic, how much time was spent at the clinic for the G-CSF therapy, and how much of the cost of the visit was not covered by the patient’s insurance (Figure 1).

Figure 1. Treatment volume and cost estimation model overview.

Definition of costs

The models estimated direct and indirect economic costs to patients and to informal caregivers who accompany patients to G-CSF office visits. Direct costs included out-of-pocket costs for transportation and office visit costs not covered by insurance (co-pays). Indirect costs included mileage allowances for use of private autos for travel to the clinic, and opportunity costs of patient and caregiver time spent traveling to and from the clinic visit and of the time in the clinic. Patient time was valued using a foregone wage approach, while caregiver time was valued based on a combination of a foregone wage and a replacement cost approach36. These time valuation methods are consistent with standard practices in cost-effectiveness research37. More specifically, patient time was valued at the national average hourly wage for all occupations, and the economic value of caregiver time was calculated as the average of (a) state minimum wages, (b) average Home Health Aide wages, and (c) the median private pay cost of hiring a home health aide (see Table 1 for sources). For mileage allowances, patient-driven miles were valued at the IRS Medical mileage rate, which is what the patient could theoretically value their mileage at in a Federal tax return. Caregiver miles are valued based on a replacement cost approach using the IRS business mileage rate.

Table 1. Summary of data sources and assumptions used for estimating patient and caregiver costs.

Category	Source(s)	Assumptions/estimates used
National annual clinic visits for G-CSF injection following MC
Cancer Prevalence, by type	SEER Cancer Statistics Review 1975–201238	2012 5-year prevalence estimates, 1st invasive tumor ever
Percentage of cancer patients receiving chemotherapy	American Cancer Society report Cancer Treatment & Survivorship Facts & Figures 2014–201539	Estimates made in seven cancer categories. Average across all non-myeloid cancers was 28%
Chemotherapy cycles per patient	Chemocare.com: How Long is Chemotherapy Given?40	Used four cycles per patient as conservative estimate
G-CSF cycle penetration rates, by cancer type	OSCER database41	Estimates made in seven cancer categories. Average across all non-myeloid cancers was 31.5%
Percentage of G-CSF cycles by daily vs long-acting	OSCER database41	68% long-acting, 32% daily
Average days treated per cycle (Daily G-CSFs)	OSCER database41	2.4 days per cycle
Percentage delivered in office (vs at home)	Daily G-CSFs: IMS G-CSF Report from June 201341. Long-acting G-CSFs: from Phillips et al.42	60% daily G-CSFs 91% long-acting G-CSFs
Patient and caregiver time devoted to G-CSF visits (per visit)
Mean travel time (round-trip)	CABS Survey41	62 min
Mean clinic time	CABS Survey41	41 min
Percentage of trips accompanied	CABS Survey41	66.3%
Economic value of patient time	BLS National average wage all occupations as of May 201443	$22.71 per h
Economic value of caregiver time	Average of state minimum wages44, BLS mean home health aide wage43 and Median private pay cost of hiring a homehealth aide45	$12.80 per h
Travel costs
Average travel distance to treatment (round-trip)	Converted travel time to miles, based on mean travel speed of 34 MPH (from Houts et al.14)	35.13 miles
Mode of transport	CABS Survey41	33.5% self-drive, 64.5% other driver, 1% taxi, 1% public transit
Per mile costs for self-driven, other driver miles	Based on 2015 IRS mileage rates46. Medical rate applied to self-transported miles. Full business rate applied to other driver miles	Self-drive: 23 cents per mile Other drivers: 57.5 cents per mile
Costs of other modes of transport	American Public Transit Association (APTA) 2014 Fact Book47. Summary of US Taxi Fares from taxifarefinder.com48	Cost per mile for Taxis: $2.91 Cost per trip: Taxis: $2.85 + mileage, Public Transit: $2.13 per trip
Insurance cost-sharing/co-pays
G-CSF market share by insurance segment	Payer Mix for 2014 for the Neulasta and Neupogen franchises (Amgen, Inc. data on file41)	Assumed payer mix in office/OPD setting would mirror mix across all settings
Cost-share (co-pay) per visit	Commercial Plans: from Tomic et al.16. Medicare FFS: Based on 20% coinsurance for Part B services, multiplied by Medicare Allowed amounts49. Medicaid: Based on average of state-level office co-pays50	Assumed 83% of Medicare FFS had supplemental insurance51 Medicare Advantage: same as commercial. VA, Military, Uninsured: assumed no cost sharing

MC, myelosuppressive chemotherapy; BLS, Bureau of Labor Statistics; FFS, Fee for Service.

Costs were estimated for all US non-myeloid cancer patients undergoing chemotherapy and traveling to oncology offices or hospital OPD clinics for prophylactic treatment with G-CSFs during a single calendar year. Where necessary, cost estimates were updated to 2015 US dollars using the Medical component of the Consumer Price Index (US Bureau of Labor Statistics52).

Literature search

We conducted a literature search for studies of the utilization of G-CSFs in the treatment and prevention of FN in non-myeloid cancer patients in the US, and the patient and caregiver time and cost burden of office visits for cancer-related therapy. We also reviewed information from government and private websites to augment the literature search. All articles, reports, and data-sets found that had estimates relevant to our models were selected for use in populating the models with either base case and/or alternate assumptions and estimates to be used in sensitivity analysis.

Models, estimates, and data sources

Annual treatment/visit volume

There were no national estimates of the treatment volume of interest available from any public or private source that we could find. Therefore, we developed estimates from four modeling perspectives, each of which used a different set of primary inputs, and averaged the results for our base case estimate. Data sources for each model are described in Table 1. The perspectives of these models were:

1. Epidemiologic perspective. This model started with national estimates of cancer prevalence, estimates of proportions of patients by cancer type who received chemotherapy, G-CSF penetration rates as a percentage of chemotherapy cycles in each cancer type, percentage of G-CSF treatments by setting and by G-CSF type (daily vs long-acting).

2. OSCER sample-based model. This model used data from the Oncology Services Comprehensive Electronic Records (OSCER) database, which includes data on cancer patients from a large US-based national sample of outpatient community and hospital-affiliated oncology practices that utilize a common EMR system for data capture. The OSCER database includes ∼600,000 cancer patients from over 400 clinical sites around the country53–55. A methodology to extrapolate this sample data to national estimates has been developed and validated53, and the database includes many statistics on G-CSF utilization. Data from OSCER provided the national estimates used in the second model.

3. Sales and pricing data perspective. This model used total sales data from Amgen public filings56 and average sales price data from Medicare49 to estimate the annual number of G-CSF injections given in the US. Other sources were used to estimate the proportion given prophylactically in the clinic setting.

4. Insured market segment perspective. The last model started with estimates of the total US insured population by insured market segment and estimated the proportion of each segment that received G-CSF therapy, using data from the literature, from databases, and from Amgen data on file.

Travel and time estimates: The CABS study35

During 2014–2015, data were collected during chemotherapy treatment on patient-reported time and travel burdens associated with G-CSF injections/visits. This information included patient travel times, plus time in the clinic, method of travel to the clinic, whether the patient was accompanied by someone, and other aspects of the patient burden of G-CSF visits. Responses by 598 patients completing Phase I of the data collection were available in early 2015, and were used to estimate these time and travel parameters in our model. CABS study patients were broadly distributed across the US to provide a nationally-representative sample of travel and time estimates. By region, 18% were in the Midwest, 20% Northeast, 25% South Atlantic, 24% South Central, and 13% Western region. Most patients/clinics in the study were in urban areas; rural and suburban patients represented only ∼8% of respondents.

Cost estimates

Once estimates of total national annual visit volume and estimates of per treatment travel and time were created, models were developed to quantify the cost burden per visit to patients and caregivers. Cost estimates were developed for travel costs, time-related costs, and patient cost-sharing for the cost of the office visit not covered by insurance (co-pays). A summary description of data sources and major assumptions used to generate these estimates is shown in Table 1.

Validity and reliability

To ensure validity and reliability of our models and estimates, most estimates and assumptions used in the base case were cross-checked and validated for reasonableness against alternate data sources. The ranges of these alternate estimates were used in sensitivity analysis (see next paragraph). For example, the self-reported travel time and time in the clinic averages from the CABS data, and the percentage of clinic visits where the patient drove him/herself or was accompanied and driven by another, were compared to similar estimates from the literature to ensure reasonableness. As another example, model-derived estimates of G-CSF cycle penetration rates by cancer type were compared to similar estimates from other studies in the published literature. Models were checked for internal consistency (between-model consistency of inputs, and reasonable comparability of outputs).

Sensitivity analysis

We ran a series of one-way sensitivity analyses to determine the key drivers of costs estimated in the model. We ran this for the estimated total costs of travel, plus patient and caregiver time lost due to G-CSF visits, and insurance cost-sharing combined. Ten key variables or parameters in the model were tested. The effects of changes from the baseline assumption for each single variable in the analysis were isolated in the model by holding all other factors constant (at baseline estimates) and then modifying the test variable through a range of reasonable alternate assumptions. We compared model results for total (cumulative) costs under these alternate assumptions to the baseline result. We developed plausible ranges of values from the literature to define the boundaries for the one-way sensitivity analyses. Sources and rationales for our assumptions are given in Table 2 and the accompanying footnotes.

Table 2. Sensitivity analysis: Plausible ranges of values, sources, and rationales.

Variable	Lower boundary	Baseline	Upper boundary
Mean time in clinic visit (h)^a	0.517	0.685	1.20
Mean one-way travel time (h)^b	0.450	0.517	0.667
Transport cost per mile^c	$0.230	$0.457	$0.575
Time opportunity cost per hour^d	$17.95	$18.76	$21.63
% Trips accompanied by caregiver^e	40.0%	66.3%	66.3%
Mean travel speed (MPH)^f	26.7	34.0	43.0
Annual visits^g	1,668,000	1,713,000	1,743,000
Average co-pay per visit^h	$33.86	$35.64	$37.42
% of trips self-driven^i	60.0%	33.5%	33.5%
% of trips by automobile^j	100%	98%	93%

^aLower bound is from Beveridge et al.57. Upper bound is from Houts et al.14.

^bLower bound is from Houts et al.14. Upper bound is from Meehan et al.58.

^cLower bound is the 2015 IRS mileage rate for medical travel. Upper bound is the 2015 IRS mileage rate for business travel46.

^dLower bound represents the weighted average of the Bureau of Labor Statistics (BLS)43 national average wage for all occupations applied to patients’ time, and the national average Home Health Aide wage applied to caregivers time. Upper bound uses the same wage as lower bound for patients’ time and the 2014 median private pay cost of hiring a home health aide45 applied to caregivers time.

^eLower bound is from Haithcox et al.15. The percentage of trips with other drivers was also adjusted to this percentage. Upper bound is the same as baseline—no higher estimates were found in the literature.

^fLower bound is from Beveridge et al.57. Upper bound is from Shreay et al.59.

^gLower bound is from the Sales and Pricing model. Upper bound is from the Epidemiology model.

^hRange is ±5% of the baseline, based on sensitivity testing in the co-pay model.

ⁱLower bound is from Haithcox et al.15. Trips driven by other drivers was adjusted down to get to 60% self-drive. Upper bound is the same as baseline—no higher estimates were found in the literature.

^jLower bound is set to maximum possible value. Upper bound is from Meehan et al.60. Self-drive and other drive percentages from baseline were reduced proportionally and taxi and public transit percentages were increased proportionally to obtain the target mix of transport types.

Results

The estimated total number of clinic visits for prophylactic G-CSF injections in the US following MC for all non-myeloid malignancies was 1.713 million visits for ∼403,000 patients in 2015. The range of total annual visit estimates from the four treatment volume estimation models was 1,668,000 (Sales and Pricing model) to 1,743,000 (Epidemiology model). Two-thirds of CABS patients reported that they were accompanied to the visit by an informal caregiver, which translates to 1.14 million accompanied visits in our national estimate.

Mean (SD) travel time per visit was 62 (50) min; mean (SD) time in the clinic was 41 (68) min. The estimated round-trip travel distance per visit was 35.1 miles. Most trips (98%) were reported to be by automobile, with 1% of trips by taxi and 1% by public transit. Patients drove themselves to the clinic for 33.5% of visits, with a companion driving for 64.5% of visits. Transportation costs were assigned to the caregiver for automobile trips where the patient did not drive, and were valued at the higher IRS business mileage rate, as noted in the methods. Therefore, cumulative transportation costs were higher for caregivers than for patients.

The cumulative annual travel costs for G-CSF clinic visits were $28.9 million ($72 per patient), and the cumulative annual travel distance was estimated at 60.2 million miles. The total national patient and informal caregiver time commitment for 1 year for travel to and from the clinic, plus time at the clinic, was 4.9 million hours (12.1 h per patient), valued at $91.8 million ($228 per patient). Patient co-pays were estimated at $61.1 million ($35.64 per visit). The total economic impact of lost time, transportation costs, and patient cost-sharing was $182 million, or $451 per patient. The average patient and caregiver cost burden per visit was $106, of which half was for the value of time lost for the visit.

Summary results from the models are shown in Tables 3–5.

Table 3. 2015 Estimated patient and caregiver cost burden of G-CSF clinic visits.

	n	n (Visits)	Transport costs	Time value	Cost-sharing	Total costs
Patients	403,000	1,713,000	$6.6 million	$66.8 million	$61.1 million	$134.5 million
Caregivers	268,000	1,135,719	$22.3 million	$25.0 million	–	$47.3 million
Total	403,000	1,713,000	$28.9 million	$91.8 million	$61.1 million	$181.8 million

Table 4. 2015 Patient and caregiver cost burden per patient and per visit.

	Transport costs*	Time value*	Cost-sharing	Total costs*
Per patient	$71.71	$227.79	$151.61	$451.11
Per visit	$16.87	$53.59	$35.64	$106.10

*Includes caregiver costs.

Table 5. 2015 Estimated patient and caregiver travel and time burden of G-CSF visits.

	Clinic visits	Travel miles	Travel time (h)	Clinic time (h)	Total time (h)
Patients	1,713,000	60.2 million	1.8 million	1.1 million	2.9 million
Caregivers	1,135,719	38.8 million	1.2 million	0.8 million	2.0 million
Total	1,713,000	60.2 million	3.0 million	1.9 million	4.9 million

Sensitivity analysis

Figure 2 summarizes the results of the one-way sensitivity analysis. The largest effects on the model cost estimates result from the assumptions of time spent in the oncology clinic and travel time to the clinic. The range of results varied from 5% below baseline to 15% above baseline when we substituted the alternate assumptions for mean time spent in the clinic, and the range of costs varied from 6% below baseline to 13% above baseline when we applied the alternate assumptions for travel time to the clinic. Unit costs for travel per mile and the cost of lost time per hour had the next largest effects. Depending on which IRS mileage rate(s) were used, the results varied from −7% to +4% of baseline costs. This model is sensitive to this variable because almost all of the trips to the clinic are via automobile. The hourly wage applied to caregiver time also moderately affects the overall cost estimates in the model, ranging from −2% from baseline when using the foregone wage approach to +8% when using the replacement cost approach.

Figure 2. One-way sensitivity analysis—key factors influencing costs.

Discussion

Using a deterministic model based on parameters derived from recent primary research and from published literature, we estimated the total annual costs to US patients and informal caregivers of oncology clinic visits for prophylactic G-CSF therapy to be ∼$182 million for 2015, or roughly $450 per patient, across all non-myeloid cancer types. This compares to an estimated range of $1730–$2364 in total annual out-of-pocket costs for actively treated cancer patients, depending on insurance status61–63. The travel burden on patients and informal caregivers in this study was estimated to be ∼60 million miles, and time devoted to going to oncology clinics for G-CSF injections after MC was estimated at 4.9 million hours, or over 12 h per patient treated.

Treatment with G-CSFs significantly reduces the risk and incidence of FN in cancer patients undergoing chemotherapy. In the pivotal, placebo-controlled trial of pegfilgrastim, overall incidence of FN was 17% with placebo vs 1% with pegfilgrastim treatment—a 94% reduction in FN64. However, as evidenced in recent literature, G-CSFs are currently used sub-optimally relative to clinical guidelines for patients at high risk for FN following chemotherapy, and FN hospitalizations continue to negatively impact the delivery of optimal chemotherapy doses65–69. Reasons for such sub-optimal outcomes may include not using G-CSFs prophylactically, as indicated (i.e., not using G-CSFs on the first and every cycle, and/or daily G-CSF use lasting less than the indicated 10–11 days per cycle70), and administering them contrary to the approved prescribing directions (i.e., some patients receive G-CSFs on the same day as the last administration of chemotherapy71).

Patient non-adherence to chemotherapy follow-up visits may also be a factor, as patients engage in cost-coping strategies in light of the high economic costs of adherence to prescribed treatment63^,72. To the extent that the time and economic burden on patients reduces the likelihood of attendance at follow-up visits for G-CSF prophylaxis, efforts should be made to reduce these burdens. Travel barriers and high out-of-pocket costs have been shown to have detrimental effects on cancer patients and on the likelihood of receiving optimal care19–21^,28–31. Conversely, patients being treated with G-CSFs after MC have reported better satisfaction with their care when they have to make fewer follow-up visits per cycle57 and have even reported a willingness-to-pay significantly more in out-of-pocket costs to reduce the number of G-CSF injections per cycle72. One way to reduce the number of injections per cycle is through the use of longer-acting G-CSFs like pegfilgrastim. Follow-up visits can also be reduced or eliminated if patients self-administer G-CSFs at home, but prohibitive pharmacy co-pays and doctor and patient concerns about proper timing and frequency of administration may limit the use of this option. A third alternative is G-CSF administration in the home via a timed-release mechanism administered by a clinician during the chemotherapy visit73.

This study fills a gap in the literature by estimating the total US national economic burden of G-CSF visits on patients and caregivers, and further translates this burden to a per patient and per visit level. However, there are additional patient and caregiver costs that were not captured by this study. Specifically, we did not estimate lost work time, which could be in addition to the time estimated to travel to and from the clinic visit and the time spent in the clinic. Estimates of lost work time that could be directly attributed to the G-CSF visit were not available. Other incidental costs which were not captured include parking fees, tolls, and telephone charges to co-ordinate clinic visits, but these costs are likely to be quite minimal57^,74.

This was an economic analysis using a population-based modeling approach, with inherent limitations. These include limitations of the empirical estimates drawn from the CABS study and from the literature. In particular, precise data on the number of G-CSF office visits across the US are lacking, as well as the current mix of short vs long-acting G-CSF injections. The variation in office visit time and travel distance estimates from the literature was high. Better measurement of total US G-CSF market share by payer category, by drug, by indication, and by setting would improve the precision of cost estimates such as those that we undertook to examine. Our estimates of the opportunity cost of patient and caregiver time do not reflect actual wages of the patients and caregivers, but are generalized assumptions based on national average wages. Finally, the cost of office visit co-pays is subject to many variables including availability of secondary/supplemental insurance for Medicare patients, state-level (Medicaid) or insurer-specific (Commercial) benefit designs, and whether any cost-sharing is borne by the uninsured. In general, we attempted to be conservative in the estimates and assumptions used throughout the process.

Another important element that was of interest in this study but which we could not evaluate was any disparity between travel time and cost for rural vs urban patients. The CABS data (source of our travel time estimates) contained almost no rural patients, making estimates for rural patient travel from this primary data source untenable. Previous research has suggested that transportation barriers for rural cancer patients may lead to sub-optimal care19–21^,28, so an understanding of the magnitude of this issue for these patients would be useful. Although the lack of rural patient data probably adds to the overall conservatism in our model cost estimates, future research on G-CSF therapy should include a better sampling of rural patients.

Conclusions

The patient and informal caregiver time and cost burden of clinic visits for G-CSF therapy following myelosuppressive chemotherapy is a significant addition to the total economic burden borne by cancer patients and their families. Travel barriers and out-of-pocket costs may translate into sub-optimal access to, or adherence to, needed care. Administration practices which reduce the economic and time burden of G-CSF therapy should be thoroughly explored between clinicians and patients. Finally, further research is needed on the impacts of travel distance and costs in the delivery of G-CSF therapy and any consequences of sub-optimal care, especially for rural patients.

Transparency

Declaration of funding

This study was sponsored by Amgen, Inc. The study sponsor played a role in study design, interpretation of data, and writing of this manuscript.

Declaration of financial/other relationships

JMS received consulting fees from Amgen, Inc. related to this study. XL, MR, and ST are, or were, employees of Amgen, Inc. at the time this study was conducted and own stock and/or stock options in Amgen Inc. JME peer reviewers on this manuscript have no relevant financial or other relationships to disclose.

Acknowledgments

The authors wish to thank Wayne Sheldon, Amgen, Inc., who provided data analysis support for this study.