Estimating the Social Value of G-CSF Therapies in the United States

Jacqueline Vanderpuye-Orgle, PhD; Alison Sexton Ward, PhD; Caroline Huber, MPH; Chelsey Kamson, BS; and Anupam B. Jena, MD, PhD

Despite major advances in cancer treatment, neutropenia is a life-threatening complication of chemotherapy.1-3 Severe neutropenia is a below-normal count of neutrophils in the blood (less than 500 per mm), which impairs the body’s ability to combat opportunistic infections.4 Neutropenia can progress to febrile neutropenia (FN) when accompanied by a temperature of 38ËšC (100.4°F) or higher, and can lead to hospitalizations, mortality, and chemotherapy dose reductions or delays, all of which adversely affect patient outcomes4-7 and medical costs.3,8 The risk of FN is determined by the chemotherapy regimen (CR) and patient factors, such as age and comorbidities.9
For patients with solid tumors or nonmyeloid malignancies, granulocyte-colony stimulating factors (G-CSFs) can enable them to undergo and remain on myelosuppressive chemotherapy with lower risk of FN and infection. Despite well-established clinical benefits of G-CSFs,10-14 these therapies have been the subject of recent concern due to large variation in their utilization. Research has indicated that G-CSFs may be overprescribed for patients at a low risk of FN (≤20%) and underprescribed for high-risk (≥20%) patients.15,16
Several studies have evaluated the cost-effectiveness of G-CSFs in particular tumor types, from a payer’s perspective1,17-21; however, these studies may understate the value of G-CSFs from a societal perspective. For example, reductions in lost workdays or disability provide tremendous value to society, but may be excluded from existing studies.11,22 Therefore, a better understanding of the total social value (TSV) of G-CSFs is needed to appropriately determine the cost-effectiveness and value afforded to patients and others affected by chemotherapy-induced neutropenia. In this study, we estimated the TSV of G-CSFs delivered to patients with cancer in the United States in 2014 by combining various components of value derived from the literature.
Targeted Literature Review
To identify evidence on the value of G-CSFs, we conducted a targeted literature review of pivotal studies. Specifically, we limited the search to articles published between 1991 (the year G-CSFs were approved in the United States) and 2014 that examined both adverse outcomes associated with myelosuppressive chemotherapy-induced FN, as well as the impacts of G-CSFs on these outcomes.
The searches yielded 77 potentially relevant articles that were narrowed to 58 for data extraction (Table 1 summarizes the articles by type of analysis). Articles were selected if they investigated patients with nonmyeloid cancers at risk of the clinical definition of chemotherapy-induced FN who were treated with G-CSFs. We focused on studies that compared outcomes for patients who received G-CSFs with patients who did not.
Our literature review provided information on 6 clinical outcomes: FN hospitalization, neutropenia events, length of hospital stay, reductions in relative dose intensity of chemotherapy, overall mortality, and antibiotic use. To associate these outcomes with savings or social value (SV), we mapped them to 4 clinical value components: savings from reductions in FN hospitalizations, reductions in FN mortality, reductions in mortality due to the ability to deliver higher-dose chemotherapy, and reductions in antibiotic use. Then, using cost estimates from the literature, along with clinical improvements, we estimated the SV for all 4 components and, in aggregate, the TSV of G-CSFs.23-26
Using the limited literature on nonclinical outcomes associated with G-CSFs, we collated information on productivity loss and quality of life.27,28 These outcomes were mapped into similar SV components: savings from reduced indirect costs (productivity loss) and improved quality of life. Again, we calculated the value for each component and in total for the nonclinical SV of G-CSFs.23-26
The TSV of G-CSFs was calculated as the sum of all clinical and nonclinical value components. All reported estimates were adjusted to 2014 US dollars using the Medical Care Consumer Price Index, which is a measure of changes in the price level of medical care commodities and services.29 To maintain the integrity of the outputs derived from the literature and to facilitate replication of our study methods, we kept all decimal places of the estimates, recognizing that the level of precision implied by our estimates is less than should be expected of our modeling exercise.
G-CSF Population
In both the United States and the European Union, G-CSFs are indicated for patients with nonmyeloid malignancies receiving myelosuppressive chemotherapy associated with a clinically significant (≥20%) risk of FN. They also may be considered for patients with a 10% to 20% risk and certain individual factors, such as age or stage of cancer.11,22 We derived an estimate of this population using the Surveillance, Epidemiology, and End Results (SEER) program, which estimated the overall cancer incidence in 2014 at 1,665,540 cases.30 To account for patients with myeloid malignancies who fall outside the approved indication, we used an incidence rate of 0.015% (15 individuals with myeloid malignancies per 100,000 population), as reported by the 2014 World Trade Center Health Program.31 Subtracting the myeloid malignancies from the SEER all-cancer incidence suggests that, in 2014, there were approximately 1,617,665 individuals diagnosed with nonmyeloid malignancies in the United States. Given that not all of these individuals would receive a CR associated with a significant risk of FN, we applied a G-CSF usage rate of 19.4%, as reported in Naeim et al (2013), to estimate a 2014 population of patients eligible for GSCFs of 314,442.32
Clinical SV
FN hospitalization. The occurrence of FN often results in immediate hospitalization to treat the associated infection.33 Crawford et al (2008) used medical claims data to estimate that 10.7% of patients with breast, lung, colorectal, and ovarian cancers, as well as lymphoma, develop FN in the first 3 cycles of chemotherapy.34 We found that patients with nonmyeloid malignancies receive 6 cycles of chemotherapy on average. Thus, we used the growth rate in incidence of FN from cycles 1 to 2 and cycles 2 to 3, reported in Crawford et al, to predict the growth rate from cycles 3 to 4, 4 to 5, and 5 to 6 in order to estimate a 6-cycle incidence rate (IFN)  of 22%.34
Further, many trials have demonstrated that, when used prophylactically, G-CSFs significantly reduce the incidence of FN. Cooper et al (2011) and Kuderer et al (2007) both reviewed randomized controlled trials and reported relative risk ratios (RRs) for developing FN (RRFN) for all solid tumor patients receiving primary G-CSFs of 0.51 and 0.44, respectively.35,36 We applied the average of these RRs (0.48) to our calculated incidence measure to estimate that G-CSFs reduce the incidence of FN to 10%. Using cost estimates from Dulisse et al (2013), we used the following equation to estimate that G-CSFs prevented 35,988 FN hospitalizations in 2014, thus creating $768 million in SV derived from avoided FN hospitalizations24:
FN Hospitalizations Value = [{IFN – (IFN  × RRFN)} × GCSF Population] × CostFN = $768,000,000
FN Hospitalizations Value = [{0.22 – (0.22 × 0.48)} × 314,442] × $21,341 = $768,000,000
FN mortality. In addition to hospitalizations, FN is also associated with high mortality. Using claims data, Dulisse et al (2013) found that the inpatient mortality rate (MR) for patients with breast cancer, lung cancer, colorectal cancer, ovarian cancer, non-Hodgkin lymphoma, and Hodgkin lymphoma with FN was 10.6%.24 Similarly, Kuderer et al (2006) and Caggiano et al (2005) reviewed discharge data for all patients with cancer with FN and reported mortality (MRFN) rates of 8.3% and 6.8%, respectively.3,8 We used the average of all 3 rates (8.6%) to estimate that in the absence of G-CSFs, there would have been 5872 FN-related deaths in the United States in 2014.
G-CSFs reduce the chances of developing FN and the probability of death among patients who develop FN. In their systematic review of the G-CSF literature, Kuderer et al (2007) estimated a 0.55 RR of FN-related mortality (RRFNM) with G-CSFs.36 Applying this estimate to the reduced FN population derived above, we calculated that G-CSFs prevented approximately 4171 deaths in 2014.
To calculate the SV generated from the reduction in FNM, we estimated the average value of each life saved. Specifically, we used estimates of average life expectancy from diagnosis for several types of cancers.37 Weighting these values using patient populations provided in Dulisse et al (2013) and Caggiano et al (2005), we estimated that, on average, a patient who dies because of FN has lost 8.81 years of life (LYL).3,24 Valuing each year of life (VLY) in perfect health at $100,00025,26 and adjusting it to account for the average long-term quality-adjusted life-years (QALYs) weight of 0.72 for patients with cancer,38 we used the following equation to estimate a total annual value of reduced FN deaths of $2.65 billion:
FNM Value = [{([IFN – (IFN × RRFN M)] × GCSF Population) × [MRFN × (1 – RRFN M)]} × LYL FNM] × VLY = $2,645,923,449
FNM Value = [{([0.22 – (0.22 × 0.55)] × 314,442) × [0.086 × (1 – 0.55)]} × 8.81] × 100,000 = $2,645,923,449
Mortality Due to Reduced Chemotherapy Relative Dose Intensity
Many patients receiving chemotherapy do not achieve their planned relative dose intensity (RDI) because of treatment toxicity complications. Lower RDI has been associated with adverse clinical outcomes, including reduced life expectancy.39 We identified several studies that estimated an incidence of chemotherapy dose reductions (IDR) greater than 15% among some cancer types, including an estimate of 40% based on a systematic literature review by Kuderer et al (2007) in a robust cancer population.36
Several studies have also found that G-CSFs significantly reduce a patient’s risk of a chemotherapy dose reduction. Although our literature review identified 3 separate estimates of the odds ratio of a dose reduction (ORDR) from 2 different systematic literature reviews, these were all based on very select populations.7,20 Assuming that patients in our population would experience similar reductions in risk, we calculated the average of all 3 reported  estimates (0.63). This estimate suggests that G-CSFs reduced the number of patients requiring chemotherapy dose reductions (IDR) by approximately 45,979 patients in 2014, or nearly 25% of our total G-CSF population.
To value survival gains associated with avoided chemotherapy dose reductions, we used estimates from Havrilesky et al (2014) who found that an RDI greater than 85% was associated with increased survival for patients with breast and lung cancers.40 Using population weights from Naeim et al (2013), we calculated an average survival gain (SDR) of 17.5 months per G-CSF patient, attributable to chemotherapy dose reductions avoided.32 We then used an estimate of the average cost of CRs ($24,392) calculated from data provided in Naeim et al to account for higher treatment costs associated with this increased dosing (assuming a 15% increase).32 Clinical experts suggested these increased costs should only apply to the roughly 50% of patients receiving chemotherapy from multi-use vials. Applying the value of a cancer-adjusted life-year at $72,000, we used the following equation to estimate the value of lowering the incidence of chemotherapy dose reductions (dose-related mortality [DRM value]) to be $4.8 billion, annually:
DRM Value = [{([IDR – (IDR  × ORDR)] × GCSF Population) × SDR} × VLY] – [(CR × 0.15 ) × ([IDR – (IDR × ORDR)] × GCSF Population)] = $4,827,749,599
DRM Value = {([0.4 – (0.4 × 0.63)] × 314,442) × 1.46} × 100,000 – [($24,392 × 0.15) × [0.4 – (0.4 ×  0.63)] × 314,442] = $4,827,749,599
Antibiotic Use
Myelosuppressive chemotherapy is associated with antibiotic use (AU) to fight infection. According to a randomized controlled trial published by Vogel et al (2005) and a claims analysis by Almenar et al (2009), 10% of patients with breast cancer receive antibiotics.41,42 Our literature review did not provide estimates of the incidence of AU (IAU) for other cancer types; therefore, to derive a proxy estimate, we assumed that other patients with nonmyeloid malignancies receive antibiotics at a similar rate. The estimate would be that 31,444 patients received antibiotics in our 2014 population.
G-CSFs have been estimated to reduce the use of antibiotics substantially. We identified 3 studies that reported relative RRs for AU (RRAU) from randomized controlled trials of 0.64 for patients with small cell lung cancer,39 and 0.80 for patients with breast cancer receiving G-CSFs.41,42 Assuming all tumor types respond similarly, the average of these 2 estimates (0.72) suggests that G-CSFs prevented antibiotic use in roughly 8837 patients in 2014, an incidence rate of 7%. Using the equation below and the cost of AU (CostAU) reported in Elder-Lissai (2008), we estimated G-CSFs generated $2.3 million in value due to fewer patients receiving antibiotics23:
AU Value = [{IAU – (IAU × RRAU)} * GCSF Population] × CostAU = $2,297,596
AU Value = [{0.1 – (0.1 × 0.72)} × 314,442] × $260 = $2,297,596
Total Clinical SV

Table 2 reports the estimated total clinical SV of G-CSFs in 2014. The annual value for each component ranged from $2.3 million for AU to $4.8 billion for reduction in mortality related to chemotherapy dose reductions avoided, summing to a total value of $11 billion.
Nonclinical SV
Indirect costs. Our literature review identified 1 article that analyzed the ICs associated with neutropenia events. Calhoun et al (2001) define total ICs (TICs) as the sum of patient work loss, caregiver work loss, and payments for caregivers.27 Using a sample of patients with ovarian cancer, they estimated the cost of each component was, on average, $4038, $1185, and $1170 per patient per event, respectively, for a TIC of $6393 per patient. We multiplied this by the reduction in FN events due to G-CSFs using the following equation to estimate a total value of $230 million:
IC Value = [{IFN – (IFN × RRFN)} × GCSF Population] × TIC = $230,087,690
IC Value = [{0.22 – (0.22 × 0.48)} × 314,442] × 6393 = $230,087,690
Quality of life. Few studies have measured how FN and G-CSF usage affect patients’ health-related quality of life (HRQoL). Our review identified 1 study that used the difference in Short Form Health Survey 36 (SF-36) scores for a group of patients with neutropenia in the last 7 days and a control group. Fortner et al (2005) found the only significant difference in responses between the groups was for bodily pain.28
To associate this difference in SF-36 scores with a dollar value, we used a published algorithm to calculate the equivalent QALY reduction.43 Using a QALY value of $100,000,25,26 we estimated that neutropenia and FN are associated with an $85 reduction in HRQoL per FN event (HRQoLFN). Multiplying this amount by the reduction in FN events due to G-CSFs, we used the following equation to estimate a total annual value of $1.9 million:
HRQoL Value = [{IFN – (IFN × RRFN)} × GCSF Population] × HRQoLFN = $1,930,031
HRQoL Value = [{0.22 – (0.22 × 0.48)} × 314,442] × $85 = $1,930,031

Nonclinical SV
Although literature on the nonclinical value of G-CSFs was limited, we used all available data to create what can be viewed as a conservative estimate of the nonclinical SV from G-CSF use.27,28 Table 2 shows that the indirect costs avoided account for 99% of our estimates, with an annual savings of $232 million.
Total Social Value
Table 2 presents the estimated TSV of G-CSFs used prophylactically as indicated in 2014. The vast majority of the SV stems from the clinical, rather than nonclinical, benefits of G-CSFs.
Value to Society Versus to Manufacturers

An important question in the debate about G-CSFs, and expensive medical technologies more broadly, is how the value created by these therapies is divided between pharmaceutical manufacturers and patients. For manufacturers, value is represented as profits, and for patients, it is “consumer surplus”—an economic concept that reflects the difference between what individuals are willing to pay for a therapy and what they actually pay. To evaluate this question, we estimated the 2014 profits accruing to G-CSF manufacturers in the United States.
The US G-CSF market is dominated by 2 Amgen products: filgrastim, which maintains around 83% market share despite patent expiration in the United States in 2013; and pegfilgrastim, with patent expiration in October 2015.44 Thus, we based our cost and revenue estimates on Amgen’s 2014 annual report scaled to represent the entire market.45,46 Revenues were reported for their G-CSFs; however, the only cost data available were for all products and included all operating costs. We therefore applied the all-product profit margin to filgrastim and pegfilgrastim revenues to estimate 2014 G-CSF profits of $1.306 billion (see Table 3). Compared with our estimate of $8.5 billion in TSV created by G-CSFs, manufacturer profits account for approximately 15% (Figure 1). This low rate of producer surplus is not surprising when considering the monopolistic form of competition that happens across branded drugs and therapies.
Sensitivity Analysis
When possible, the parameters used in the baseline analysis were based on an average (weighted when appropriate) of estimates from multiple studies. However, 2 stronger assumptions were required to translate the mortality benefits associated with G-CSFs into monetary values. For each FN-related death that G-CSFs prevented, we developed an average number of life-years gained using estimates of life expectancy at diagnosis reported in Liu et al (2013).37
On average, the number of life-years gained in the baseline model assumed FN is a random event among patients with cancer receiving myelosuppressive chemotherapy; however, in practice, older and weaker patients may be more susceptible. These patients likely have below-average life expectancies, which bias our baseline estimate toward longer survival. To test the sensitivity of our results, we recalculated the TSV of G-CSFs assuming the average life-years gained was 50% lower (4.4 instead of 8.8 years). This reduction decreased the TSV estimate by $1.8 billion (Figure 2).
Additionally, the 2 largest drivers of SV in our model are the reduction in FN-related mortality (33%) and the value of avoiding chemotherapy RDI reductions (60%). We performed additional sensitivity analyses using confidence intervals (CIs) presented in the literature for these parameters. For example, Kuderer et al (2007) provided a CI of 0.33 to 0.90 for their estimate of the RR of FN-related mortality with G-CSFs (0.55). The lower end of this interval would add $603 million to the TSV estimate, and the upper end of the interval would reduce the TSV by $945 million. Similarly, Kuderer et al (2007) also provide a CI of 0.30 to 0.65 for their estimate of the incidence of dose reductions (0.4). Our sensitivity analysis suggests that the lower end of the CI would reduce the TSV by almost $1.68 billion, but the upper end of the CI would increase the TSV by almost $4.8 billion.
This study had a number of limitations. First, TSV estimates were limited by the lack of available research, particularly on nonclinical burdens imposed by FN. For example, by reducing the likelihood of chemotherapy dose reductions, G-CSFs may provide patients with less anxiety about complications.
Second, previous research on the value of G-CSFs has been constrained by the limited scope of the study population considered. When possible, we based our SV estimates on results from studies with large sample populations that covered all, or most, of the malignancies for which G-CSFs are typically indicated; however, there were limited data for several SV components. For these value components, our estimates were based on more select samples, and we assumed the outcomes would, on average, be similar across other tumor types. Additionally, prior studies focus almost exclusively on patients for whom G-CSFs are indicated (as previously described), making it impossible to generate estimates of SV for patients for whom G-CSFs are not indicated and for whom TSV would be expected to be lower.
Third, our modeling exercise relied on parameter estimates drawn from the literature. Our findings, therefore, reflect uncertainty stemming from the various study designs on which parameter estimates that populate our model were drawn.
This study estimates that 314,442 patients with cancer were potential candidates for prophylactic G-CSFs in the United States in 2014. Based on the parameters and assumptions used in our calculations, G-CSFs generated a total of almost $8.5 billion in SV. More than 97% of this value was attributable to estimated improvements in clinical outcomes. Importantly, reductions in overall mortality from both reduced FN hospitalizations and chemotherapy dose reductions avoided accounted for $10 billion in SV. Since we found few studies that investigated improvements in nonclinical outcomes associated with G-CSFs, our SV estimate of $232 million may be a conservative lower bound for the total nonclinical value.
By considering all possible benefits of G-CSFs, our study results suggest these therapies provide substantial value for society, particularly for patients. The portion of value accruing to manufacturers in the form of profits is approximately 15%. These estimates are similar to previous work that suggests that between 5% and 19% of the value generated by gains in cancer survival have been appropriated by manufacturers in the form of profits.47 Our analysis further demonstrates the need to evaluate the cost-effectiveness of medical innovations in a way that incorporates broader impacts to society, such as nonclinical benefits.  Failure to include these components can significantly underestimate the economic value of medical innovations and the value to patients. 
Print | AJMC Printing...