A comparison of mortality and costs associated with FOLFOX versus FOLFIRI in stage IV colorectal cancer

Abstract

Objective:

To examine cost and mortality among stage IV colorectal cancer (CRC) patients treated with 5-fluorouracil (5FU)/leucovorin/oxaliplatin (FOLFOX) or 5FU/leucovorin/irinotecan (FOLFIRI).

Methods:

Adult CRC patients newly treated with FOLFOX or FOLFIRI were identified from a large database using medical and pharmacy claims for services delivered January 1, 2002 through December 31, 2005. Cancer stage for a subset of patients was abstracted from medical records. Outcomes were annualized costs calculated for 4 years of observation, and deaths as recorded by the National Death Index. Cost was analyzed using generalized linear modeling; mortality was modeled using Cox proportional hazards analysis.

Results:

Unadjusted annualized median and mean costs were $134,401 and $152,213, respectively, for the FOLFOX cohort (n = 41) and $103,150 and $107,994 for the FOLFIRI cohort (n = 86). Death occurred among five (12%) FOLFOX and 42 (53%) FOLFIRI patients. Adjusted analysis revealed no significant difference in cost between cohorts, even after adjusting for reduced irinotecan costs due to generic availability. Incremental costs associated with one additional life saved per year were only $1,236 higher for patients treated with FOLFOX compared with FOLFIRI. Cox analysis revealed a significant survival advantage for FOLFOX over FOLFIRI (HR = 5.2; 95% CI: 1.7–15.8).

Conclusions:

A significant survival benefit was seen for CRC patients receiving FOLFOX versus FOLFIRI; multivariate analysis revealed no significant cost differences. However, the small sample size may have resulted in lack of adequate power to detect a difference between cohorts. There may be factors influencing mortality that were not included in the multivariate modeling.

Key words::

• Colorectal neoplasms
• Chemotherapy
• Drug costs
• FOLFOX regimen
• IFL protocol
• Mortality

Introduction

Colorectal cancer (CRC) is the third most common cancer among both men and women in the United States, with an estimated 106,100 new cases of colon cancer and 40,870 new cases of rectal cancer expected to occur in 20091. CRC ranks behind only lung and prostate cancer in men and lung and breast cancer in women in terms of number of deaths, with an estimated 49,920 deaths (representing 9% of all cancer deaths) expected to occur in 20091. Estimated annual expenditures for CRC in the United States are in the range of $5.3–6.5 billion2.

For more than 40 years, 5-fluorouracil (5FU) has been the chemotherapy of choice for CRC. Recent advances in CRC treatment have led to a wide range of treatment options available to physicians for adjuvant therapy and metastatic disease. As initial therapy for metastatic CRC in patients with good tolerance to intensive therapy, the National Comprehensive Cancer Network guidelines recommend FOLFOX (infusional 5FU, leucovorin, and oxaliplatin), FOLFIRI (infusional 5FU, leucovorin, and irinotecan), or CapeOX (capecitabine and oxaliplatin), with or without biologic therapy (i.e., bevacizumab, cetuximab, or panitumumab). The 5FU/leucovorin combination is reserved for patients who cannot tolerate oxaliplatin or irinotecan3,4.

Few studies have directly compared FOLFOX and FOLFIRI in metastatic CRC, and evaluation of these studies is complicated by the variation in 5FU administration (infusion vs bolus dosing) across these studies. As a result, recommendations regarding which of these regimens should be used as first-line treatment are conflicting. In one recent review, the authors stated that response and survival are similar with oxaliplatin and irinotecan, and therefore, other factors (e.g., adverse effects) should be considered when tailoring therapy to a particular patient5. Conversely, other authors have suggested that FOLFOX holds an advantage in terms of overall survival6. Clearly, more data are needed to clarify this issue.

In the absence of a clearly superior chemotherapy regimen, a wider variety of factors must be considered when choosing which regimen to use. One factor that may be considered is costs – not just the acquisition costs of the therapies included in the regimen, but also the effects of the regimen on downstream costs related to management and supportive care. The results of a recent literature review indicate that FOLFOX and FOLFIRI are cost effective relative to 5FU/leucovorin but also failed to indicate a preference for either regimen2.

More data are needed to better inform the choice of therapy for metastatic CRC. To that end, the purpose of this study was to compare both mortality and economic outcomes among managed-care patients receiving either FOLFOX or FOLFIRI for treatment of metastatic CRC.

Methods

Study design

This was a retrospective analysis of healthcare claims, medical records, and National Death Index (NDI) data. Claims data were evaluated for CRC diagnosis and evidence of FOLFOX or FOLFIRI treatment, as well as healthcare utilization and costs. Medical records abstractions were performed for a subset of patients to collect information relating to cancer stage, tumor grouping, lymph node involvement, presence of surgery and/or radiation, medications and doses received, other treatments, cancer severity, and adverse effects. Medical records were made available for the study on a voluntary basis by treating physicians or chemotherapy centers. Only records for patients who met the study criteria (analyzed in the claims data) and which had complete data on key study elements were retained and analyzed. The mortality status of the patients included in the study was determined by matching the study population to the NDI database.

Data sources

The administrative healthcare claims database included electronic pharmacy and medical claims and enrollment data from a large US managed-care provider affiliated with i3 Innovus. During the study period, the health plan encompassed approximately 26 million members. The population is geographically diverse across the United States, with the majority of patients located in the South and Midwest. The data were accessed using Health Insurance Portability and Accountability Act (HIPAA)-compliant techniques, and no identifiable protected health information (PHI) was extracted except where necessary and allowable under a waiver granted by The Privacy Board of the New England Institutional Review Board. Further details about the database have been described elsewhere7.

Following privacy board approval, PHI and physician-specific information necessary for medical records abstraction and matching to NDI were extracted by study staff. Medical records were abstracted by a professional medical records abstraction firm. Patient name, birth date, and social security number were matched with the death data from NDI to extract date of death for study patients. All data were de-identified prior to study analyses.

Study patients

This study included commercial health plan members with at least one medical claim with a diagnosis code for CRC (ICD-9-CM: 153.xx, 154.0, 154.1, 154.8, 230.3, or 230.4) and evidence of either FOLFOX or FOLFIRI (based upon a claim for 5FU, leucovorin, and either oxaliplatin or irinotecan on the same day; codes for chemotherapy claims are provided in Table 1) during the identification period of January 1, 2002 through December 31, 2005. The service date on the first occurring claim during the identification period was defined as the index date. Patients were excluded if they had evidence of oxaliplatin, irinotecan, 5FU, or capecitabine during the year prior to the index date, if they were <18 years of age as of the year of the index date, or if they were not continuously enrolled with medical and pharmacy coverage for at least 12 months prior to and at least 30 days after the index date. Patients were placed into either the FOLFOX or FOLFIRI cohort based upon the regimen received on the index date. All analyses were based on the index regimen and did not account for changes in regimen.

Table 1.  Codes for chemotherapy medications.

Medications	HICL	CPT/HTPT
Oxaliplatin	016687	J9263, C9205
Irinotecan	010778	J9206
5FU	003907, 009467	J9190
Capecitabine	018385	J8520, J8521
Bevacizumab	025964	J9035, C9214
Cetuximab	025948	J9055, C9215
Panitumumab	034055	C9235

Each patient was observed for at least 1 year prior to the index date as early as January 1, 2001 (baseline period), and for a minimum of 30 days following the index date. For the follow-up period, all patients were observed until the last date of enrollment in the health plan, or until December 31, 2006, whichever occurred earlier. Cost analyses included all patients with up to 4 years of follow-up, while the mortality analysis was limited to patients with no more than 3 years of follow-up to account for availability of death data through 2006 only.

Study measures and outcomes

Demographic and clinical characteristics

Baseline demographic and clinical characteristics were identified from administrative enrollment and claims data. Demographic characteristics focused on patients’ gender, age, and geographic region. Presence of comorbid diabetes, cardiovascular disease, hypertension, and/or liver disease was identified from medical claims using the Agency for Healthcare Research and Quality (AHRQ) categorizations for these comorbid conditions9. In addition, a comorbidity score was calculated based on the presence of diagnosis codes on medical claims in the pre-index period using the Deyo adaptation of the Charlson Comorbidity Index10,11. For this analysis, the score was modified to exclude CRC. Treatments received other than the index regimen were also captured in the healthcare claims database, and were subsequently confirmed by consulting the patient’s medical record. Specifically, it was determined whether patients received a biologic agent (i.e., bevacizumab, cetuximab, or panitumumab), radiation therapy, and/or surgical intervention.

Patient deaths

Data from the NDI (administered by the National Center for Health Statistics) were accessed in order to identify whether patient death had occurred during the study period. For patients who died, the date of death was retained form the NDI files.

Cost measures

Healthcare costs were computed as the combined health plan- and patient-paid amounts during the follow-up period based on actual reimbursed amount. Cost measures were calculated for combined outpatient prescriptions and medical services (including ambulatory, emergency, inpatient, and other medical services), as well as total healthcare costs. Because patients had varying follow-up times, costs were annualized prior to analysis. Costs were adjusted using the annual medical care component of the Consumer Price Index (CPI) to reflect inflation between the earliest year of data and 20068.

Statistical methods

Data extraction and statistical analyses were performed using SAS, version 8.2 (SAS Institute, Cary, NC, USA) and STATA 9 (StataCorp, College Station, TX, USA). All study variables, including baseline and outcome measures, were analyzed descriptively. Results were stratified by treatment cohort. Costs were estimated using a generalized linear model (GLM) with a gamma distribution and log link to assess the incremental cost associated with each cohort. Multivariable analysis of time to death was conducted using a Cox proportional hazards model. The primary predictor in these models was treatment cohort. Covariates included age, gender, geographic region, comorbidity indicators (diabetes, cardiovascular disease, hypertension, and liver disease), comorbidity score, indicator for use of radiation, indicator for use of surgery, and indicator for use of biologic therapy.

A cost-effectiveness analysis (CEA) was performed to compare the relative value of FOLFOX and FOLFIRI, as measured by incremental cost-effectiveness ratio (ICER). The ICER is the ratio of difference in costs to the difference in effects between two treatments, or:

For this analysis, the time horizon was 1 year (i.e., costs and mortality were annualized). The cost measure was annualized total healthcare costs. The effect measure was the number of surviving patients per 100 patients per year.

In order to obtain confidence interval of ICER, a bootstrap method was applied. The basic concept behind bootstrapping is to treat the study sample as if it were the population, the premise being that it is better to draw inferences from the sample at hand rather than to make potentially unrealistic assumptions about the underlying population. To generate a bootstrap distribution for an ICER, the following steps were performed. First, a random sample of the same size as the original sample was generated from the data, with replacement, separately for FOLFOX and FOLFIRI. The ICER was then calculated for this bootstrap resample. This procedure was repeated 1000 times to get 1000 bootstrap estimates of the ICER. These estimates then defined the empirical sampling distribution of ICER.

Recently, irinotecan became available as a generic. Therefore, as a crude sensitivity analysis, acquisition costs for irinotecan were reduced by 20% and by 40%, and costs for the FOLFIRI cohort were recalculated. The reduction values for the sensitivity analyses were estimated based on a blended acquisition cost for generic and name-branded products.

Results

Patient characteristics

Medical records were abstracted for 259 patients with CRC, of whom 131 had received FOLFOX and 128 had received FOLFIRI. Of these, 41 FOLFOX patients and 86 FOLFIRI patients were confirmed to have stage IV CRC and were retained for analysis. Demographic and clinical characteristics of these patients are included in Table 2. A slight majority of patients were male, and mean age was approximately 57 years. Comorbidities were common but did not differ significantly between cohorts. Nearly all patients in each cohort underwent a surgical procedure (88% of FOLFOX and 94% of FOLFIRI patients; p = 0.21), while previous radiation therapy was less common (17% and 21%, respectively; p = 0.61). Bevacizumab was the most frequently used biologic agent during follow-up, with a significantly higher rate of use among FOLFOX patients (68 vs. 29%; p < 0.001).

Table 2.  Demographic and clinical characteristics.

	FOLFOX	FOLFIRI	p-value
Number of subjects	41	86	–
Male, n (%)	24 (58.5%)	44 (51.2%)	0.44
Age, years, mean (SD)	57.9 (10.3)	56.1 (9.1)	0.31
Region, n (%)
West	6 (14.6%)	12 (14.0%)	0.92
Midwest	11 (26.8%)	22 (25.6%)	0.88
Northeast	1 (2.4%)	8 (9.3%)	0.16
South	23 (56.1%)	44 (51.2%)	0.60
Comorbidities, n (%)
Diabetes	13 (31.7%)	26 (30.2%)	0.87
Cardiovascular disease	30 (73.2%)	72 (83.7%)	0.16
Hypertension	28 (68.3%)	49 (57.0%)	0.22
Liver disease	28 (68.3%)	59 (68.6%)	0.97
Charlson Comorbidity Index, mean (SD)	6.3 (1.9)	6.6 (1.6)	0.28
Other treatments received, n (%)
Radiation	7 (17.1%)	18 (20.9%)	0.61
Surgery	36 (87.8%)	81 (94.2%)	0.21
Bevacizumab	28 (68.3%)	25 (29.1%)	<0.001
Cetuximab	8 (19.5%)	10 (11.6%)	0.23
Panitumumab	0 (0%)	0 (0%)	–

FOLFOX, 5-fluorouracil (5FU)/leucovorin/oxaliplatin; FOLFIRI, 5FU/leucovorin/irinotecan.

Healthcare costs

Unadjusted mean and median costs for FOLFOX and FOLFIRI patients are included in Table 3. Mean total annualized healthcare costs were significantly higher for the FOLFOX cohort ($152,213 vs. $107,994; p = 0.04). This difference was driven primarily by the difference in ambulatory medical costs, which include the cost of chemotherapy administration ($99,125 vs. $55,437; p = 0.001) and, in kind, total medical costs ($121,273 vs. $79,396; p = 0.01). Because healthcare costs are often skewed12, a nonparametric comparison of costs was also performed (Mann–Whitney U-test). As was noted for mean costs, median ambulatory medical costs were significantly higher among FOLFOX patients ($102,085 vs. $71,626; p = 0.02). Although both cohorts had median emergency room (ER) costs of $0, a larger proportion of FOLFOX patients had $0 costs (93 vs. 71%); thus, nonparametric testing found ER costs to be significantly higher for FOLFIRI (p = 0.009). Neither total medical nor total healthcare costs were significantly different between cohorts (p = 0.07 and 0.06, respectively).

Table 3.  Unadjusted annualized healthcare costs*.

	FOLFOX	FOLFIRI	p-value†
Pharmacy costs
Mean (SD)	$4381 ($4345)	$3547 ($4190)	0.30
Median	$3085	$2876	0.11
Medical costs
Mean (SD)	$147,832 ($120,830)	$104,447 ($70,860)	0.04
Median	$132,267	$97,767	0.07
Ambulatory costs
Mean (SD)	$117,474 ($95,629)	$75,602 ($51,703)	0.01
Median	$102,085	$71,626	0.02
Emergency costs
Mean (SD)	$263 ($1164)	$1,075 ($4366)	0.11
Median	$0	$0	0.009
Inpatient costs
Mean (SD)	$20,836 ($54,353)	$25,086 ($45,164)	0.64
Median	$1016	$8541	0.10
Total healthcare costs
Mean (SD)	$152,213 ($122,500)	$107,994 ($72,939)	0.04
Median	$134,401	$103,150	0.06

*Costs were analyzed using generalized linear model with a gamma distribution and log link.

†T-test comparisons of means and Mann–Whitney U (rank sum) tests of medians used to derived unadjusted p-values.

FOLFOX, 5-fluorouracil (5FU)/leucovorin/oxaliplatin; FOLFIRI, 5FU/leucovorin/irinotecan; SD, standard deviation.

After adjusting for covariates, multivariate analysis revealed that total healthcare costs were not significantly different between cohorts (cost ratio for FOLFIRI:FOLFOX = 0.79; p = 0.16; Table 4). When irinotecan costs were decreased in a crude sensitivity analysis by 20%, the cost ratio was 0.76 (95% CI: 0.54–1.07; p = 0.11), while reduction of irinotecan costs by 40% resulted in a cost ratio of 0.74 (95% CI: 0.52–1.03; p = 0.08).

Table 4.  Multivariate analysis* of annualized total cost during the follow-up period, FOLFIRI:FOLFOX (all subjects).

	Ratio	SE	95% CI	p-value
FOLFIRI	0.79	0.13	(0.56, 1.10)	0.162
Age	0.97	0.01	(0.95, 0.98)	0.000
Male	0.89	0.13	(0.67, 1.18)	0.414
Midwest region†	0.47	0.14	(0.25, 0.85)	0.013
South region†	0.56	0.16	(0.32, 0.98)	0.044
West region†	0.71	0.23	(0.37, 1.35)	0.294
Diabetes	0.93	0.15	(0.68, 1.27)	0.643
Cardiovascular disease	0.83	0.15	(0.58, 1.19)	0.308
Hypertension	1.14	0.18	(0.84, 1.55)	0.401
Liver disease	1.09	0.16	(0.82, 1.46)	0.550
Charlson score modified for colorectal cancer	1.11	0.07	(0.99, 1.25)	0.079
Biologics use	1.30	0.20	(0.96, 1.77)	0.089
Presence of radiation	0.79	0.15	(0.55, 1.14)	0.212
Presence of surgery	1.05	0.28	(0.62, 1.79)	0.855
n = 127

*Costs were analyzed using generalized linear model with a gamma distribution and log link.

†Northeast region is the reference group.

FOLFOX, 5-fluorouracil (5FU)/leucovorin/oxaliplatin; FOLFIRI, 5FU/leucovorin/irinotecan.

Mortality

Among patients treated with FOLFOX, death occurred in only five patients (12.2%), whereas 41 FOLFIRI patients (52.6%) died during follow-up. This difference in mortality was significant (p < 0.0001). The number of days to death was similar, however; mean (SD) days to death was 463 (291) days in the FOLFOX cohort and 450 (296) days in the FOLFIRI cohort (p = 0.82). Multivariate analysis of mortality demonstrated that use of FOLFIRI was associated with a significantly higher hazard of death versus FOLFOX (hazard ratio [HR]: 4.01; 95% CI: 1.31–12.28; Table 5). Use of biologic therapy was not associated with a statistically significant difference for hazard of death.

Table 5.  Cox proportional hazards analysis of mortality risk during the follow-up period, FOLFIRI:FOLFOX (all subjects).

	Ratio	SE	95% CI	p-value
FOLFIRI	4.01	2.29	(1.31, 12.28)	0.015
Age	1.02	0.02	(0.98, 1.06)	0.323
Male	0.64	0.22	(0.33, 1.25)	0.191
Midwest region*	0.29	0.18	(0.09, 0.99)	0.049
South region*	0.33	0.18	(0.11, 0.98)	0.046
West region*	0.67	0.44	(0.19, 2.39)	0.539
Diabetes	1.37	0.50	(0.67, 2.81)	0.384
Cardiovascular disease	0.64	0.31	(0.25, 1.66)	0.360
Hypertension	0.63	0.21	(0.33, 1.21)	0.168
Liver disease	0.88	0.32	(0.44, 1.78)	0.731
Charlson score modified for colorectal cancer	1.29	0.17	(0.99, 1.67)	0.055
Presence of radiation	0.51	0.25	(0.20, 1.31)	0.162
Presence of surgery	0.32	0.19	(0.10, 1.02)	0.054
Biologics use	0.63	0.22	(0.32, 1.26)	0.191
n = 127

*Northeast region is the reference group.

FOLFOX, 5-fluorouracil (5FU)/leucovorin/oxaliplatin; FOLFIRI, 5FU/leucovorin/irinotecan.

Cost effectiveness

The estimated annual total cost of FOLFOX per 100 surviving patients was $40,835 greater than the annual total cost of FOLFIRI, while the number of surviving patients per 100 patients per year for FOLFOX was 33.78 higher. After running bootstrap estimates, the incremental costs associated with one additional life saved per year were only $1236 higher for patients treated with FOLFOX compared with FOLFIRI. Although quality-of-life data were not available in this analysis, the ICER associated with FOLFOX is well below the commonly accepted threshold of $50,000 per QALY13, which demonstrates that FOLFOX is highly cost-effective as compared with FOLFIRI in this patient population. This threshold is generally used for cost-effectiveness comparisons that incorporate utility data, which is a limitation of making this comparison. However, given the chemotherapy drugs being compared, it is not anticipated that incorporating the utility data would increase the ICER sufficiently to exceed this threshold.

Discussion

The goal of this study was to compare mortality, healthcare costs, and, subsequently, cost effectiveness associated with a FOLFOX chemotherapy regimen versus a FOLFIRI regimen for CRC patients. The study used a combination of claims data, medical records data, and mortality data to create an integrated dataset. Claims data were used to identify patients and assign them to the treatment cohorts using data for services delivered from January 2005 through December 2007. Variables relating to costs and patient characteristics were derived from claims. Medical records were retrieved following privacy board approval and were used to identify cancer stage, chemotherapy cycles, other treatments, and adverse effects. Mortality data were used to determine whether patients survived or died during the study period. A total of 127 patients were included in the analysis, based on having stage IV CRC, as documented in the medical record.

While few comparative studies exist, there are data to suggest that FOLFOX may offer a greater survival benefit compared with FOLFIRI4,5. The data in this regard are conflicting, however. A phase III trial comparing IFL to FOLFOX found that FOLFOX reduced the risk of mortality by 34% over IFL (or a nearly 3 times higher risk of death for IFL), while median survival was increased from 15.0 to 19.5 months14. However, IFL and FOLFIRI may not be equivalent in terms of mortality, as the difference in administration of 5FU in these two regimens (i.e., bolus in IFL vs. infusion in FOLFIRI) may have an impact on efficacy5. Another study that did compare FOLFIRI to FOLFOX found similar response rates for these two regimens in the first-line setting (54 vs. 51%)15. There did appear to be an efficacy advantage for FOLFOX in the second-line setting, however. The results of the present study indicate that there is a survival benefit for stage IV CRC patients within this managed-care population. Hazard of death was more than four times greater among patients receiving FOLFIRI than for patients receiving FOLFOX, vastly exceeding the differences seen in the published clinical trials. It can only be speculated as to the reasons for these pronounced differences. One possibility is that the greater proportion of patients in both arms of the study who underwent surgical resection (88% of FOLFOX and 94% of FOLFIRI patients in the present study vs. 22% of FOLFOX and 9% of FOLFIRI patients in the Tournigand et al. study15) played a major role in patient survival. Another potential reason is that the number of metastases, the use of other chemotherapeutic agents, and patient adherence to treatment regimen (e.g., completion of all cycles) may have differed between the two study populations; these were variables that were not included in our multivariate modeling.

The observed survival benefit among patients receiving the FOLFOX regimen may come at a cost. Within the present study population, patients in the FOLFOX cohort incurred significantly higher unadjusted mean annual healthcare costs than FOLFIRI patients. The difference in these costs was approximately $44,000, which is higher than the incremental cost of $29,523 for FOLFOX over IFL reported by Hillner et al.6, but lower than the $65,170/QALY incremental cost-effectiveness ratio for FOLFOX versus FOLFIRI treatment reported by Tumeh et al.16, who developed a Markov decision model using a hypothetical cohort of patients with metastatic colorectal cancer. However, the costs used in that analysis were based upon utilization rates from the clinical trial and costs from public data sources, while the costs in the present analysis were based upon real-world utilization patterns and actual paid amounts. Of note, after adjusting for other factors, such as demographics, comorbidity, and other treatments received, total healthcare costs were not significantly different between FOLFOX and FOLFIRI. This lack of significance was present even after reducing the cost of irinotecan by as much as 40% to account for the change in acquisition costs for generic product.

Finally, biologic therapy use was not significantly associated with survival in the analyses. Also, a significant association between biologic use and total costs was not observed. The biologic use variable is of particular interest as biologic use was significantly higher in the FOLFOX cohort. Further exploration of the relationship between biologic use and total costs and mortality outcomes was performed to determine whether FOLFOX use was independently associated with lower mortality, and whether costs remained equal after accounting for the impact of biologic use. The subgroups for both the FOLFOX and FOLFIRI cohorts that had biologic use in the follow-up period were identified and compared on cost and mortality outcomes. These results are provided in Tables 6 and 7. Total cost and mortality outcomes did not differ significantly between FOLFOX and FOLFIRI for the subgroups with biologic use. However, this finding may be affected by the small sample size (n = 63) for this analysis.

Table 6.  Multivariate analysis of annualized total cost during the follow-up period, FOLFIRI:FOLFOX (subjects with biologics use).

Follow-up annualized total cost of biologics	Ratio	SE	95% CI	p-value
FOLFIRI	0.67	0.14	(0.44, 1.02)	0.059
Age	0.98	0.02	(0.95, 1.01)	0.168
Male	0.87	0.17	(0.59, 1.26)	0.457
South region*	1.23	0.28	(0.79, 1.93)	0.361
West region*	0.67	0.18	(0.40, 1.12)	0.128
Diabetes	0.80	0.19	(0.50, 1.29)	0.366
Cardiovascular disease	0.77	0.16	(0.51, 1.17)	0.220
Hypertension	1.05	0.29	(0.60, 1.82)	0.870
Liver disease	0.91	0.18	(0.62, 1.34)	0.647
Charlson score modified for colorectal cancer	0.99	0.13	(0.76, 1.28)	0.918
n = 63

*No subjects from the Northeast region. Midwest is the reference group.

FOLFOX, 5-fluorouracil (5FU)/leucovorin/oxaliplatin; FOLFIRI, 5FU/leucovorin/irinotecan.

Table 7.  Cox proportional hazards analysis of mortality risk during the follow-up period, FOLFIRI:FOLFOX (subjects with biologics use).

	Ratio	SE	95% CI	p-value
FOLFIRI	1.36	1.04	(0.30, 6.08)	0.686
Age	1.02	0.04	(0.94, 1.11)	0.609
Male	0.27	0.22	(0.05, 1.32)	0.106
Midwest region*	0.02	0.03	(0.00, 0.54)	0.019
South region*	0.01	0.02	(0.00, 0.39)	0.012
West region*	0.08	0.14	(0.00, 2.37)	0.145
Diabetes	3.82	3.22	(0.73, 19.90)	0.112
Cardiovascular disease	0.33	0.31	(0.05, 2.06)	0.236
Hypertension	0.16	0.14	(0.03, 0.88)	0.035
Liver disease	1.81	1.63	(0.31, 10.54)	0.508
Charlson score modified for colorectal cancer	1.26	0.45	(0.63, 2.52)	0.508
Presence of radiation	0.53	0.42	(0.11, 2.49)	0.419
Presence of surgery	2.19	3.00	(0.15, 32.22)	0.568
n = 63

*Northeast region is the reference group.

FOLFOX, 5-fluorouracil (5FU)/leucovorin/oxaliplatin; FOLFIRI, 5FU/leucovorin/irinotecan.

The minimum hazard ratio of FOLFIRI compared to FOLFOX that can be detected with a power of 80% with a two-sided 0.05 level test is 0.49 for a sample size of 63 and 0.61 for a sample size of 127.

Nevertheless, when determining which chemotherapy regimen to use, patients and providers must weigh the potential for improved survival versus increases in healthcare costs. The ICER analysis demonstrates that a significant case can be made for the selection of FOLFOX, as the additional cost of this regimen is far outweighed by the survival benefit.

The findings of this study must be considered within the limitations of the data and study design. These limitations do not reduce the strength of the study, but they must be taken into account during interpretation of the results. First is the degree to which claims data can accurately capture an individual’s medical use history. These data are subject to possible coding errors and omissions. Furthermore, this study may not be generalizable to the overall population. The study population was identified by selecting patients from a managed-care plan and may not be applicable to a setting outside of the managed-care population.

A limitation specific to this study is the small sample size. The lack of significance in some of the analyses may be due to the lack of adequate power to detect a difference between the study cohorts. Also, not all data were available for all patients, so some important clinical information could not be incorporated into the analyses. For example, patient adherence to their regimens was not assessed during chart review. Another limitation that should be taken into account is that patients were only required to be enrolled in the health plan for a minimum of 30 days of follow-up. Reasons for disenrollment from the health plan cannot be determined from the available data. In some cases, disenrollment may have been due to employer-based decisions to switch plans. And finally, there may have been factors influencing mortality that were not included in the multivariate modeling. These factors may include use of other chemotherapeutic agents and patient adherence to the treatment regimen (including completion of all cycles).

Conclusions

This study demonstrates a significant survival benefit for CRC patients treated with FOLFOX over those treated with FOLFIRI, while no significant cost differences were observed using multivariate analysis. The substantial survival benefit may outweigh other factors when patients and physicians are faced with the decision of which therapy to use. Future studies assessing these outcomes, involving larger patient samples, are needed to guide providers and payers in making decisions regarding the optimal treatment for CRC patients.

Transparency

Declaration of funding

Support for this study was provided by Sanofi-Aventis Pharmaceuticals.

Declaration of financial/other relationships

At the time this study was conducted, M.A.N, S.S., M.K., and C.H. were employed by i3 Innovus, and B.S. was employed by Sanofi-Aventis US. C.H. and M.K. are currently employed by Innovus. All authors were involved in the study design, analysis, and manuscript preparation.

Acknowledgments

The authors would like to acknowledge Herb Heien for his assistance with statistical analysis and Victoria Porter for her assistance with preparation of this manuscript.