Fax: (011) 44 (0)1895 520039
Cost-effectiveness analysis of oxaliplatin compared with 5-fluorouracil/leucovorin in adjuvant treatment of stage III colon cancer in the US†
Version of Record online: 30 JAN 2007
Copyright © 2007 American Cancer Society
Volume 109, Issue 6, pages 1082–1089, 15 March 2007
How to Cite
Aballéa, S., Chancellor, J. V.M., Raikou, M., Drummond, M. F., Weinstein, M. C., Jourdan, S. and Bridgewater, J. (2007), Cost-effectiveness analysis of oxaliplatin compared with 5-fluorouracil/leucovorin in adjuvant treatment of stage III colon cancer in the US. Cancer, 109: 1082–1089. doi: 10.1002/cncr.22512
This study was carried out by i3 Innovus under a grant from Sanofi-Aventis. Sanofi-Aventis provided the raw MOSAIC dataset and reviewed article drafts; i3 Innovus carried out all analyses and interpreted data and results independently.
- Issue online: 7 MAR 2007
- Version of Record online: 30 JAN 2007
- Manuscript Accepted: 6 DEC 2006
- Manuscript Revised: 29 NOV 2006
- Manuscript Received: 5 JUL 2006
- colon cancer;
- cost effectiveness;
- survival analysis
The MOSAIC trial demonstrated that oxaliplatin/5-fluorouracil/leucovorin (FU/LV) (FOLFOX4) as adjuvant treatment of TNM stage II and III colon cancer significantly improves disease-free survival compared with 5-FU/LV alone. For stage III patients the 4-year disease-free survival (DFS) was 69% in the FOLFOX4 arm vs 61% in the LV5FU2 arm, P = .002). The cost-effectiveness of FOLFOX4 in stage III patients was evaluated from a US Medicare perspective.
By using individual patient-level data from the MOSAIC trial (median follow-up: 44.2 months), DFS and overall survival (OS) were estimated up to 4 years from randomization. DFS was extrapolated from 4 to 5 years by fitting a Weibull model and subsequent survival was estimated from life tables. OS beyond 4 years was predicted from the extrapolated DFS estimates and observed survival after recurrence. Costs were calculated from trial data and external estimates of resources to manage recurrence.
Patients on FOLFOX4 were predicted to gain 2.00 (95% confidence interval [CI]: 0.63, 3.37) years of DFS over those on 5-FU/LV. The predicted life expectancy of stage III patients on FOLFOX4 and 5-FU/LV was 17.61 and 16.26 years, respectively. Mean total lifetime disease-related costs were $56,300 with oxaliplatin and $39,300 with 5-FU/LV. Compared with 5-FU/LV, FOLFOX4 was estimated to cost $20,600 per life-year gained and $22,800 per quality-adjusted life-year (QALY) gained, discounting costs and outcomes at 3% per annum.
FOLFOX4 is likely to be cost-effective compared with 5-FU/LV in the adjuvant treatment of stage III colon cancer. The incremental cost-effectiveness ratio compares favorably with other funded interventions in oncology. Cancer 2007 © 2007 American Cancer Society.
For patients with early colon cancer (TNM stage II and III), chemotherapy with 5-fluorouracil plus leucovorin (5-FU/LV) improves survival and has been standard treatment for the past decade. In November 2004, the US Food and Drug Administration (FDA) approved oxaliplatin (Eloxatin, Sanofi-Aventis, Paris, France) in combination with 5-FU/LV as adjuvant treatment of patients with stage III colon cancer based on results of the MOSAIC study, a large, international, randomized Phase III trial involving 2246 patients with stage II or III colon cancer in 146 centers that compared the FOLFOX4 regimen (oxaliplatin/5-FU/LV) with 5-FU/LV alone.1 With a median follow-up of 3 years, the addition of oxaliplatin reduced the risk of recurrence by 23% in patients with stage II and III who had undergone surgery for their primary tumor and 24% in stage III patients. This benefit is confirmed for stage III patients with a median follow-up of 4 years (25% risk reduction).2
Increasingly, healthcare payers require evidence of economic as well as clinical value. Economic evaluations have shown that adjuvant 5-fluorouracil (5-FU) + levamisole in stage III colon cancer is highly cost-effective compared with surgery alone,3–5 whereas a modeled study suggests that the addition of oxaliplatin to 5-FU/LV is incrementally cost-effective compared with 5-FU/LV alone.6 Our objective was to confirm the cost-effectiveness of adjuvant oxaliplatin using stage III patient-level data from the MOSAIC trial and extrapolating these results to a time horizon sufficiently distant to capture all costs and outcomes of relevance.7
MATERIALS AND METHODS
This economic evaluation was a cost-effectiveness analysis (CEA) comparing the FOLFOX4 regimen against 5-FU/LV using individual patient data from the MOSAIC trial. The intention-to-treat (ITT) population consisted of 2246 randomized patients, including 40% of stage II (Dukes B2) and 60% of stage III (Dukes C) patients. Consistent with the FDA-approved label, this study considered the subgroup of 1347 stage III patients, of which 672 were randomized to FOLFOX4 and 675 to 5-FU/LV (see baseline characteristics for the stage III population in Table 1). Patients were scheduled to receive 12 cycles of chemotherapy at 2-week intervals with either of 2 regimens: 1) FOLFOX4: oxaliplatin 85 mg/m2 on Day 1, leucovorin 200 mg/m2 on Days 1 and 2, 5-FU bolus 400 mg/m2 and 5-FU continuous infusion 600 mg/m2 on Days 1 and 2. 2) 5-FU/LV: leucovorin 200 mg/m2 and Days 1 and 2, 5-FU bolus 400 mg/m2 and 5-FU continuous infusion 600 mg/m2 on Days 1 and 2.
|Characteristic||FOLFOX4 N = 672||5-FU/LV N = 675|
|Age <65, y, No. (%)||436 (64.9)||448 (66.4)|
|Sex, No. (%)|
|Men||373 (55.5)||353 (52.3)|
|Women||299 (44.5)||322 (47.7)|
|Karnofsky performance status score, No. (%)|
|<60||2 (0.3)||3 (0.4)|
|60–70||100 (14.9)||89 (13.2)|
|80–100||570 (84.8)||583 (86.4)|
|Depth of invasion, No. (%)|
|T1||6 (0.9)||8 (1.2)|
|T2||51 (7.6)||54 (8.0)|
|T3||486 (72.3)||491 (72.7)|
|T4||129 (19.2)||121 (17.9)|
|Bowel obstruction, No. (%)||130 (19.3)||130 (19.3)|
|Perforation, No. (%)||40 (6.0)||35 (5.2)|
|Histologic appearance, No. (%)|
|Well differentiated||128 (19.0)||103 (15.3)|
|Moderately differentiated||421 (62.6)||433 (64.1)|
|Poorly differentiated||95 (14.1)||106 (15.7)|
|Unknown||28 (4.2)||33 (4.9)|
|No. of nodes involved, No. (%)|
|1–4||499 (74.3)||513 (76.0)|
|>4||170 (25.3)||160 (23.7)|
|Unknown||3 (0.4)||2 (0.3)|
The outcome measures for the CEA were life-years, disease-free life-years, and quality-adjusted life-years (QALYs), estimated to a lifetime horizon. QALYs were derived by weighting overall survival (OS) estimates by utility values representing preferences for health states.7 Lifetime costs of colon cancer-related healthcare resources were estimated from a US Medicare perspective. Outcomes and costs accruing beyond 1 year were discounted to present values at a rate of 3% per annum.7 Incremental cost-effectiveness ratios (ICERs) were calculated based on cost per life-year gained, cost per disease-free life-year gained, and cost per QALY gained. All statistical analyses were conducted with SAS v. 8.2 (SAS Institute, Cary, NC).
The primary endpoint of the trial was disease-free survival (DFS). As the median period of trial follow-up was 44.2 months, OS for a full lifespan was derived by extrapolating the trial data as follows. DFS by group was estimated from 0 to 48 months by the Kaplan-Meier method and extrapolated to 5 years by fitting a parametric model to the tail of the curve for patients alive and free of disease at 36 months, as proposed by Gelber et al.8 After 5 years patients were assumed in the base case to be no longer at risk of recurrence, and all-cause mortality was the same as in the US population adjusted for age and sex. This assumption was based on longer-term follow-up from another large study of adjuvant therapy in stage III CRC.9 OS from 0 to 48 months was estimated from empirical data by the Kaplan-Meier method. To estimate OS beyond 48 months we derived a mathematical relation between DFS and OS, making use of the empirical data on survival time after recurrence and its dependency on the disease-free period before recurrence. In MOSAIC, 61% (252) of stage III patients with recurrence died within the observation period, which was deemed to provide sufficient information to estimate survival after recurrence parametrically. The survival function after recurrence was modeled using a Weibull distribution, adjusted for time of recurrence. Models were estimated separately for each treatment arm and validated by confirming that the predicted OS curve fell within the 95% confidence limits of the respective Kaplan-Meier estimates over the 0 to 48-month period for which the latter were available.
The mean number of disease-free years (DFYs) and life-years (LYs) for patients in each arm was calculated as the areas under the DFS and OS curves, respectively. They were calculated by partitioning the analytic timespan into time intervals and summing the probabilities of being alive at the beginning of each interval multiplied by the duration of the interval. QALYs were calculated from DFS and OS by weighting each survival in each interval by a utility value for each possible health state.
We assigned utility weights, as shown in Table 2, to allow quality adjustment of survival time. Following Ramsey et al.,10 who administered the HUI-3 questionnaire to 173 survivors of colorectal cancer, we assigned a utility value of 0.85 to survivors and applied a reduction of 0.2 to those with recurrence for their remaining lifetime. We also applied utility reductions for episodes of toxicity. From year 4 onwards, when the risk of recurrence has diminished, the utility value for disease-free survivors was considered to be the same as that of the general population. We applied EQ-5D utility norms by age and sex derived from the 2001 Medical Expenditure Panel Survey (MEPS)11 and US population-based preference weights12 to calculate a weighted average utility for each year of follow-up.
|Neutropenia||5 days||Launois et al., 199629|
|Neuropathy (no hospitalization)||Until time of next visit at which grade drops to 0 or 1||Launois et al., 199629|
|Grade 2/3, disease-free||−30%|
|Grade 2/3, after relapse||−23%|
|Nausea, vomiting (no hospitalization)||−19%||5 days||Grunberg et al., 200230|
|Diarrhea (no hospitalization)||−36%||5 days||Cook et al., 199431|
|Hospitalization (except neutropenia)||−50%||7 days||Assumption|
Resource Utilization and Cost Estimation
We estimated costs incurred for resources used during and subsequent to trial follow-up, notably treatment of patients with recurrence. Resource utilization during trial follow-up was derived from individual patient history data. However, observations for many patients were censored, so that subsequent costs and recurrences were unknown. To avoid bias due to censoring we applied the cost history method developed by Lin et al.13
Study chemotherapy costs were calculated using doses actually administered. Drug administration resources included clinic attendances, infusion pumps, and premedication. For each cycle the administration cost was estimated at $547 for the FOLFOX4 arm and $543 for the 5-FU/LV arm, based on 2 attendances, 1 infusion pump, and premedications including dexamethasone, 5HT3 inhibitors, calcium gluconate, and magnesium. Second-line adjuvant chemotherapy costs, for example, among patients who discontinued study drug due to toxicity, were included as were costs of routine follow-up, further procedures, and management of serious adverse events (SAEs) and toxicity recorded in the trial case report forms. Resource utilization for toxicity episodes treated on an outpatient basis was estimated for individual patients based on standard treatment for the management of nausea NCI CTC Grade 2 or higher, leucopenia NCI CTC Grade 3 or 4, and diarrhea NCI CTC Grade 2 or higher. Mean costs of neurotoxicity episodes from a case series in ovarian cancer were used as a proxy. In addition, the descriptions of each SAE episode were reviewed and assigned to a Diagnosis-Related Group (DRG), from which we deduced the corresponding Medicare Average National Payment.
Costs were estimated from the US Medicare perspective using 2003 data. Unit cost data were obtained from the Red Book for pharmaceuticals, from the DRG System for hospitalizations, from the Medicare Physician Fee Schedule for physician services, and from the Medicare Hospital Outpatient Prospective Payment System for outpatient procedures. For drugs administered in the hospital the average wholesale price (AWP) less 15% was applied, reflecting Medicare reimbursement rates for these.
The cost of a recurrence was estimated according to the type of recurrence (local recurrence, liver metastasis, lung metastasis, or disseminated disease) and treatment pathways for each, including the costs of possible surgery and first-line and second-line chemotherapy. Pathways and probabilities were determined using the literature and validated using expert opinion. For recurrences observed during trial follow-up, the type of recurrence was known and costed accordingly; otherwise a weighted average was used.
Expected mean costs and outcomes per patient were calculated for each treatment for the within-trial period (to 48 months), extrapolated to 60 months and then to 50 years. Mean incremental cost-effectiveness ratios (ICERs) of cost per life-year gained, cost per DFY gained, and cost per QALY gained were calculated.
The within-trial DFS data are sample estimates with known variance. We estimated the uncertainty in the resulting cost-effectiveness estimates by bootstrapping.14 Where input data were uncertain, we separately estimated the impact of feasible alternative assumptions in the following sensitivity analyses:
- 1We applied a more conservative assumption for the survival of patients free of disease at 5 years, in which the annual recurrence rate for both arms declined from 1.5% at year 5, based on the Sargent et al.15 pooled analysis of clinical trials.
- 2In a permutation of the above, we assumed the effect of FOLFOX4 persisted beyond the trial by applying the relative risk reduction (RRR) of 24% seen within trial to the post-year-5 recurrence rates.
- 3We applied a set of assumptions leading to a lower cost of managing recurrences: neither bevacizumab nor cetuximab would be used, fewer (90%) instead of all patients unfit for secondary resection would receive further chemotherapy, and that only 70% (instead of all) of patients undergoing secondary resection would receive subsequent chemotherapy.
- 4We used the upper limit of the 95% confidence interval (CI) around the survival curves after recurrence in both arms to assess the impact of a potential extension of life expectancy after recurrence attributable to innovative treatments for metastatic disease.
Within-trial survival estimates
For the outcome ‘death or recurrence’ (on which DFS is based), there were 194 and 245 events in the FOLFOX4 and 5-FU/LV groups of stage III patients, respectively. The DFS probability at 48 months was 69% and 61% for FOLFOX4 and 5-FU/LV, respectively (log-rank test: P = .002), equivalent to a mean gain per patient of 0.21 DFYs.
There were 280 deaths in the stage III subgroup, accounting for 80% of deaths in the overall trial population, of which 132 were in patients on FOLFOX4 and 148 on 5-FU/LV. OS to 48 months was 78.2% and 76.6% for FOLFOX4 and 5-FU/LV, respectively, equivalent to a mean gain per patient on FOLFOX4 of 0.060 life-years. The OS difference did not reach statistical significance (log-rank test: P = .24).
Survival after recurrence
Among 411 cases of recurrence in the stage III subgroup, 252 (61%) died within the trial observation period. The median survival after recurrence was 16.8 (95% CI: 15.3; 19.8) months, and 36-month survival rate after recurrence was 19.5%. Twenty-eight stage III patients died during the trial without having recurrence. Survival time after recurrence did not differ between treatments but was strongly associated with the time of recurrence (log-rank test: P < .0001), those with later recurrences surviving longer.
Extrapolated survival estimates
We extrapolated DFS to 5 years by fitting a Weibull model to trial data for patients free of disease at 36 months. This predicted a gain of 0.084 DFYs per patient at year 5. The estimated gain after 5 years was 1.70 DFYs. The extrapolation of OS predicted a gain of 0.038 life-years per patient at year 5 and 1.25 life-years per patient after year 5.
Hence, the OS estimates over a lifetime were composed of 3 components: OS observed within-trial to 4 years, extrapolated OS for year 5, and extrapolated OS beyond 5 years. Summing these, the estimated mean OS gain per patient on FOLFOX4 was 0.06 + 0.04 + 1.25 = 1.35 years (95% CI: 0.04–2.66). The majority of the gain was due to the predicted impact on survival post 5 years. Applying the utility weights and discount factors to the survival data, the mean gain of 1.35 LYs was equivalent to 0.75 discounted QALYs (95% CI: 0.09–1.41).
Resource Use and Costs
Study chemotherapy, at approximately $29,000 per patient in the FOLFOX4 group and $6,500 in the comparator group, was the largest cost component. Next most costly was resource use associated with recurrence, with respective mean costs of $16,600 and $23,700 per patient. Hence, although oxaliplatin added over $22,500 to the ingredient cost of chemotherapy, this was partially offset by savings of $7,100 because of the reduced numbers of recurrences. Overall, patients on FOLFOX4 and 5-FU/LV incurred mean estimated costs of $56,300 and $39,300, respectively, a difference of $17,000. These costs were predicated on the incidence of recurrences and the costs of recurrences, which were estimated at $67,900 for a local recurrence, $62,400 for a liver metastasis, $61,000 for a lung metastasis, and $56,600 for other types of disseminated disease in the 5-FU/LV arm, and lower by $2,400 for the FOLFOX4 arm, as these patients were less likely to receive oxaliplatin postrecurrence. This gave a weighted average standard cost of a recurrence of $61,200 for patients receiving adjuvant 5-FU/LV as treatment and $58,800 for patients on adjuvant FOLFOX4. In the sensitivity analysis in which lower proportions of patients were assumed to receive chemotherapy and bevacizumab and cetuximab were excluded from treatment options, while holding the survival prognosis constant, the estimated mean cost of recurrence decreased to $43,200 and $45,400 in the FOLFOX4 and 5-FU/LV arms, respectively.
ICERs were calculated as point estimates, as reported in Table 3. Under the base case, the use of FOLFOX4 instead of 5-FU/LV would cost approximately $12,900 per DFY gained, $20,600 per LY gained, and $22,800 per QALY gained.
|Base case||Sensitivity analysis 1: recurrence may occur beyond year 5, no difference between arms||Sensitivity analysis 2: recurrence may occur beyond year 5, difference between arms|
|Cost per DFY gained||$12,873||$14,135||$11,873|
|Cost per LY gained||$20,603||$22,031||$18,144|
|Cost per QALY gained||$22,804||$24,614||$20,393|
ICERs were also calculated for three alternative scenarios, as follows. When post-5-year recurrence rates as reported by Sargent et al.15 instead of zero were assumed, the mean ICERs increased to $14,100 per DFY gained, $22,000 per LY gained, and $24,600 per QALY gained. When the treatment effect of oxaliplatin was assumed to persist beyond 5 years and applied to Sargent et al.15 recurrence rates, the ICERs decreased to $11,900 per DFY gained, $18,100 per LY gained, and $20,400 per QALY gained. Under the lower-cost scenario for postrecurrence chemotherapy, the incremental cost increased to $19,264 and the ICERs increased to $14,600 per DFY gained, $23,300 per LY gained, and $25,800 per QALY gained. When the upper limit of the 95% CI around survival after recurrence was used, life-expectancy after recurrence was extended by approximately 8 months. As a result, the number of life-years gained decreased from 1.35 to 1.20 and the ICERs increased to $22,896 per life-year gained and $24,405 per QALY gained (the incremental cost per DFY gained was not affected).
A probabilistic analysis was conducted using the bootstrap method to estimate the level of confidence in the cost-effectiveness results, given the sampling variability in the MOSAIC trial data. We found that FOLFOX4 has a 91% to 96% probability of being cost-effective, assuming a willingness-to-pay (WTP) of $50,000 to $100,000 per QALY gained.
Adjuvant chemotherapy improves survival by reducing the incidence of recurrence. After demonstration of the benefit of bolus and infused 5-FU,16 various regimens have been studied in the adjuvant setting. The only positive studies have been MOSAIC1, 17 and the NSABP C-07 study,18 both comparing 5-FU to 5-FU and oxaliplatin. Here we analyzed the cost-effectiveness of the FOLFOX4 regimen in stage III patients based on the MOSAIC data.
Key to this analysis is the value of DFS as a predictor of longer-term OS. The MOSAIC trial demonstrated a significant advantage of FOLFOX4 in DFS, but not in OS. In fact, MOSAIC was not powered to detect a difference in OS, but only to detect a difference in DFS at 3 years. However, a meta-analysis of clinical trials on adjuvant colon cancer has shown that DFS at 3 years is a strong predictor of OS at 5 years.19 On the basis of 16 comparative trials, Sargent et al.19 found that the mean difference in 5-year OS is equivalent to 0.85 times the difference in 3-year DFS. This prediction is more optimistic than those we applied to the MOSAIC study, where despite a DFS difference of approximately 7% (FOLFOX 72.2% vs 5-FU/LV 65.3%) at 3 years,1 our projected difference in 5-year OS is 4% (FOLFOX 74.8% vs 5-FU/LV 70.8%). There are several possible explanations. Recurrences may have been diagnosed earlier and treatment of advanced disease improved since those studies used by Sargent et al. to generate their formula. In addition, in MOSAIC, excess deaths not attributable to colon cancer recurrences occurred during the early period of follow-up in the FOLFOX4 group. However, the difference in OS is predicted to ‘catch up’ partially with the DFS benefit by 5 years. This is supported by the 4-year data: de Gramont et al.2 reported a hazard ratio of 0.86 (95% CI: 0.69–1.08) for 4-year OS, which is comparable to the 0.85 hazard ratio predicted by the model when it is applied over that period. As expected, the OS absolute difference has increased in sequential analyses of stage III patients.17
The extrapolation beyond 4 years relies on 2 key assumptions: that the survival function after recurrence estimated from the trial is representative of survival after recurrence in reality and that the survival of nonrecurring patients is similar to the survival of the general population, controlling for age and gender. On the basis of data from the SEER registry, 88.4% of patients alive 5 years after diagnosis of stage III colon cancer were still alive at 10 years. In our model, approximately 80% of all patients who were alive at 5 years and 89% of those who were free of recurrence at 5 years were still alive at 10 years.20 This suggests that our assumption that nonrecurring patients have similar survival as the general population is realistic. The former assumption, ie, that survival after recurrence in the trial is representative of reality, is perhaps more questionable, given the evolution of therapy in advanced disease. An increase in survival after recurrence would reduce the life-expectancy gain associated with FOLFOX4. However, in a scenario where mean survival after recurrence was increased by 8 months, our analysis predicted that the impact on cost-effectiveness would be rather small. As a possible point of comparison, it can be noted that the addition of bevacizumab to the IFL regimen as first-line treatment of metastatic colon cancer was shown to increase median survival by approximately 5 months.21 Similarly, the median survival of patients with previously untreated metastatic colon cancer was greater by 4.5 months in patients treated with FOLFOX than in those treated with IFL.22
The estimated CIs around the numbers of DFYs, LYs, and QALYs gained over the trial period indicate that these gains are not statistically significant. Comparing the lower limit of the difference in DFYs to zero does not provide an appropriate test for the superiority of FOLFOX in terms of DFS, as the power of this test is low. Previous publications based on MOSAIC have established that the difference in DFS between FOLFOX4 and 5-FU/LV is highly significant.1, 17 As the risk of recurrence beyond the trial time-horizon is small, we concluded that it is reasonable to project disease-free survival over the long term. The justification for extrapolating gains in OS over a lifetime in the absence of significant observed difference in OS within trial may be less apparent. However, as explained above, the projected difference in OS after 4 years is obtained as a direct consequence of the difference in DFS (not OS) within trial.
We assumed that a proportion of patients with recurrence would receive biological therapies, although in the absence of patient-level data from MOSAIC on the treatment of advanced disease, we applied a standardized treatment pathway and duration for both arms of the study. As noted by Schrag,23 the Mayo clinic regimen costs $63, FOLFOX costs $11,889, and FOLFOX combined with bevacizumab costs $21,033 per 8-week course, at 95% of the May 2004 average wholesale price. However, patients are likely to remain on regimens that include bevacizumab for longer, partly as a result of increased survival.21 Accounting for differential costs and survival in advanced disease was beyond the scope of this study, but we explored a scenario in which biologicals were not used, holding survival constant. This resulted in a modest increase in the ICER for FOLFOX4.
A difficulty encountered in performing economic evaluation for the US as a whole is that the unit costs of healthcare vary substantially across regions. The approach taken here was to use average national costs, in line with recommendations of the US Panel on Cost-Effectiveness in Health and Medicine.7 If similar analyses were performed locally, the cost-effectiveness results would show some variation, primarily because of the resources used subsequent to recurrences and the unit costs of these. Another issue with regard to the estimation of costs is that the dataset contained little information about treatment of toxicities. Importantly, all adverse events requiring hospitalization were reported, with a description of the cause. Thus, we were able to infer associated costs on an individual basis by determining the DRG corresponding to each cause of admission. However, in the case of less serious toxicity, where hospital admission (or prolongation of hospital stay) is not required, the costs had to be predicted on the basis of the assumed standard treatment. Whereas practice may vary between physicians, the impact of variation on the standard assumptions was very small. In the base case analysis, costs of nonserious toxicity episodes (including neutropenia, nausea, diarrhea, and neuropathy) accounted for less than 5% of the total difference in costs. For example, we assumed that all patients with grade 2 nausea received a 5-HT3 antagonist. If we assumed instead that this treatment was only used in 50% of episodes of grade 2 nausea and others were not treated, ICERs would vary by less than 0.15%.
Cost-effectiveness analysis using QALYs offers the opportunity to consider trade-offs between quantity and quality of survival, and we estimated the impact of toxicity on QALYs. Although patients on FOLFOX4 had a greater incidence of toxicity than those on 5-FU/LV, the negative impact on QALYs was greatly outweighed by the QALYs gained from reduction in recurrences and improved survival with FOLFOX4.
By using a threshold value of $50,000 per QALY, the estimated probability that FOLFOX4 is cost-effective compared with 5-FU/LV is 91%. Although $50,000 has widely been quoted as the minimum value of a QALY in the US, based on the ‘dialysis standard,’ others have argued that this is unrealistically low.24–26 We conclude that FOLFOX4 is cost-effective in the adjuvant chemotherapy of stage III colon cancer, compared with 2-drug regimens based on 5-FU/LV with a probability of at least 91%. Confirmation of these findings awaits real-time follow-up, but in the meantime the adoption of FOLFOX4 appears to represent an efficient use of Medicare resources. These conclusions are consistent with those drawn from previous studies of the cost-effectiveness of adjuvant chemotherapy in colorectal cancer.4, 27, 28
In the MOSAIC study the difference in 4-year DFS for those patients with high-risk stage II disease was 5.4%.17 There is therefore a strong clinical case for individual patients with high-risk stage II disease to receive FOLFOX, although we have not examined the cost-effectiveness of this scenario.
We conclude that the FOLFOX4 regimen is likely to be cost-effective in adjuvant treatment of patients with stage III colon cancer based on analysis of data from the MOSAIC trial, with median follow-up of 44.2 months and extrapolation of outcomes and costs over a lifetime.
We thank Dr. Carlos Becerra (University of Texas Southwestern Medical Center) and Dr. Bert O'Neil (University of North Carolina, Chapel Hill) for invaluable assistance in reviewing the pathways for management postrecurrence and toxicity episodes and the costing of these. Professor Alistair McGuire of the London School of Economics provided advice on the cost analysis of censored data and extrapolation methods. Dr. Margaret Hutka, a former employee of Innovus Research, performed the DRG coding of individual records of adverse events. Dr. Remi Brouard of Sanofi-Aventis conceived the project and provided the raw dataset from MOSAIC to i3 Innovus Research for analysis.
- 2Oxaliplatin/5FU/LV in the adjuvant treatment of stage II and stage III colon cancer: efficacy results with a median follow-up of 4 years. J Clin Oncol. 2005; 23(16S, part I of II): 246S. Abstract 3501., , , et al.
- 5Adjuvant chemotheraphy in patients with resectable stage III colon cancer: lifetime cost-effectiveness and cost-utility analysis. Cancer J. 1998; 11: 39–47., , , , , .
- 7Cost-Effectiveness in Health and Medicine. New York: Oxford University Press; 1996., , , .
- 15Adjuvant Colon Cancer Endpoints (ACCENT) Group. Endpoints for Colon Adjuvant Clinical Trials (CACT): recommendations based on individual patient data (IPD) from 20898 patients (pts) and 18 randomized trials. J Clin Oncol (Suppl). 2005; 23(16S Part I of II): 249s. Abstract 3512.,
- 17Oxaliplatin/5FU/LV in the adjuvant treatment of stage II and stage III colon cancer: efficacy results with a median follow-up of 4 years. J Clin Oncol. 2005; 23(16S, Part I of II): 246S. Abstract 3501., , , et al.
- 18A phase III trial comparing FULV to FULV + oxaliplatin in stage II or III carcinoma of the colon: results of NSABP Protocol C-07. J Clin Oncol (Meeting Abstracts). 2005; 23: LBA3500., , , et al.
- 19http://www.fda.gov/cder/drug/cancer_endpoints/sargent/sld001.htm. Accessed on 15-6-2004, , . 3 Year DFS vs. 5 year OS as an endpoint for adjuvant colon cancer studies: data from randomized trials. Available at URL:
- 20Surveillance Epidemiology and End Results (SEER) Program. SEER*Stat Database: Incidence-SEER 17 Regs Public-Use, Nov 2005 Sub (1973–2003 varying). National Cancer Institute, DCCPS, Surveillance Research Program, Cancer Statistics Branch. Available from URL: www.seer.cancer.gov Accessed on 10-10-2006.