Feasibility of Economic Analysis of Radiation Therapy Oncology Group (RTOG) 91-11 Using Medicare Data

Int. J. Radiation Oncology Biol. Phys., Vol. 79, No. 2, pp. 436–442, 2011 Copyright Ó 2011 Elsevier Inc. Printed in the USA. All rights reserved 0360-3016/$–see front matter

doi:10.1016/j.ijrobp.2009.11.059

CLINICAL INVESTIGATION

Head and Neck

FEASIBILITY OF ECONOMIC ANALYSIS OF RADIATION THERAPY ONCOLOGY GROUP (RTOG) 91-11 USING MEDICARE DATA ANDRE KONSKI, M.D., M.B.A., M.A.,* MYTHREYI BHARGAVAN, PH.D.,y JEAN OWEN, PH.D.,z REBECCA PAULUS, B.S.,x JAY COOPER, M.D.,k ARLENE FORASTIERE, M.D.,x K. KIAN ANG, M.D., PH.D.,{ AND DEBORAH WATKINS-BRUNER, R.N., PH.D.** From the *Department of Radiation Oncology, Wayne State University School of Medicine, Detroit, MI; yAmerican College of Radiology, Reston, VA; zRadiation Therapy Oncology Group, Philadelphia, PA; xDepartment of Radiation Oncology, Maimonides Medical Center, Brooklyn, NY; kDepartment of Oncology, Johns Hopkins University School of Medicine, Baltimore, MD; {Department of Radiation Oncology, MD Anderson Cancer Center, Houston, TX; and **Department of Nursing, University of Pennsylvania, Philadelphia, PA Purpose: The specific aim of this analysis was to evaluate the feasibility of performing a cost-effectiveness analysis using Medicare data from patients treated on a randomized Phase III clinical trial. Methods and Materials: Cost data included Medicare Part A and Part B costs from all providers—inpatient, outpatient, skilled nursing facility, home health, hospice, and physicians—and were obtained from the Centers for Medicare & Medicaid Services for patients eligible for Medicare, treated on Radiation Therapy Oncology Group (RTOG) 9111 between 1992 and 1996. The 47-month expected discounted (annual discount rate of 3%) cost for each arm of the trial was calculated in 1996 dollars, with Kaplan-Meier sampling average estimates of survival probabilities for each month and mean monthly costs. Overall and disease-free survival was also discounted 3%/year. The analysis was performed from a payer’s perspective. Incremental cost-effectiveness ratios were calculated comparing the chemotherapy arms to the radiation alone arm. Results: Of the 547 patients entered, Medicare cost data and clinical outcomes were available for 66 patients. Reasons for exclusion included no RTOG follow-up, Medicare HMO enrollment, no Medicare claims since trial entry, and trial entry after 1996. Differences existed between groups in tumor characteristics, toxicity, and survival, all which could affect resource utilization. Conclusions: Although we were able to test the methodology of economic analysis alongside a clinical trial using Medicare data, the results may be difficult to translate to the entire trial population because of non-random missing data. Methods to improve Medicare data capture and matching to clinical trial samples are required. Ó 2011 Elsevier Inc. Cost-effectiveness, Laryngeal preservation, Chemoradiation, Medicare data capture.

INTRODUCTION New cancer therapies or techniques are investigated prospectively in clinical trials, making this an environment ideal for performing economic analysis. Because of the specific prescriptions, little variation should exist between treatments and the resulting cost differences should reflect an actual difference between outcome or toxicity differences and not be a result of the variations in the treatment given. Randomized clinical trials, however, are not designed or powered to perform economic analyses. Collection of clinical data can be challenging without the additional difficulties of trying to collect economic data from many different sources. In addi-

tion, payment rates differ between payers. Medicare, being available to the majority of patients 65 and older, is a data source that may be used to evaluate the cost of care. Medicare has relatively uniform coverage policies across the country and its use avoids problems with geographic variability of insurance. With the majority of cancers occurring in those older than age 50, it may be feasible to assess economic information by using Medicare as the primary source of economic information. Radiation Therapy Oncology Group (RTOG) clinical trial 91-11 investigated the effect of induction chemotherapy and radiation therapy vs. concomitant chemotherapy and radiation therapy vs. radiation therapy alone on survival with

Reprint requests to: Andre Konski, M.D., M.B.A., M.A., F.A.C.R., Department of Radiation Oncology, Wayne State University School of Medicine, 540 E. Canfield, Scott Hall Detroit, MI 48201. Tel: (313) 966-2274; Fax: (313) 966-9400; E-mail: akonski@ med.wayne.edu

Supported by the Pennsylvania Commonwealth Universal Research Enhancement (C.U.R.E.) Program ME-02-149 grant. Conflict of interest: none. Received April 3, 2009, and in revised form Oct 10, 2009. Accepted for publication Nov 10, 2009. 436

Cost-effectiveness of Laryngeal Sparing Therapy d A. KONSKI et al.

preservation of laryngeal function in patients with previously untreated, Stage III and IV squamous cell carcinoma of the glottic and supraglottic larynx. Although no difference was noted in overall survival, both chemotherapy containing arms had improved disease-free survival and the concurrent chemotherapy and radiation therapy arm had a greater proportion of patients with an intact larynx (1). Both combined modality arms, however, had a higher rate of toxicity; the mucosal toxicity of the concurrent radiotherapy and cisplatin treatment regimen being nearly twice as frequent as the mucosal toxicity of the other two treatment regimens. The specific aim of this study was to evaluate the feasibility of performing an economic analysis, using Medicare data, alongside a randomized Phase III clinical trial. A secondary aim was to determine the cost-effectiveness of combination treatment relative to radiation alone. METHODS AND MATERIALS The eligibility for RTOG 91-11 included patients 18 years or older with previously untreated Stage III or IV squamous-cell carcinoma of the glottic or supraglottic larynx that would require a total laryngectomy should surgery be performed were eligible for this study. Patients were considered eligible if they were considered curable with surgery and postoperative radiation. Pretreatment eligibility, pretreatment staging studies, and treatment details have been previously reported (1). Statistical comparisons for RTOG 91-11 of association of categorical covariates were carried out using the chi-square test. Overall survival was measured from the date of randomization to the date of death or last follow-up. Disease-free survival was measured from the date of randomization to the first occurrence of a locoregional failure, distant metastasis, a new primary, or death without progression. The Kaplan-Meier method was used to estimate overall and disease-free rates, and the log–rank test was used to compare treatment arms (2, 3). A list of patients with a social security number who were $65 years of age or listed Medicare as their primary insurance carrier was submitted to the Centers for Medicare and Medicaid Services to obtain cost data to link to clinical outcome. The Centers for Medicare and Medicaid Services approved the application for no more than 5 years of data. Cost of treatment was calculated from Medicare Part A and Part B reimbursement from all providers—inpatient, outpatient, skilled nursing facility, home health, hospice, and physicians and other Part B providers for patients treated on RTOG 91-11 between 1992 and 1996. Forty-seven month expected discounted costs for each arm of the trial indexed to 1996 dollars was calculated with the Lin method, which uses average monthly survival estimates and monthly costs for as many months for which data are available (4–6). The 47-month interval was chosen because that was the longest spell for which data were available for analysis on any patient. (Months are denoted sequentially after randomization, and average costs and survival probabilities for each month after randomization are used to estimate expected costs.) By using data on all patients for whom information is available in any month, the Lin method permits estimation of costs without having to restrict the analysis to a set calendar period. Costs and effects were discounted back to the time of entry onto the trial, using an annual discount rate of 3%; this corrects for the higher value that most people assign to immediate benefits and costs than to benefits and costs in the distant future. (The 3% is a standard

437

discount rate used in health-related cost-benefit analyses.) Also, to adjust for inflation, costs were indexed to 1996 dollars using the Consumer Price Index (Bureau of Labor Statistics. Consumer Price Index—All Urban Consumers [CPI-U], U.S. city average, all items (1982-1984 = 100), ftp://ftp.bls.gov/pub/special.requests/cpi/cpiai. txt. Accessed November 18, 2005. US Department of Labor). The analysis was performed from a payer’s perspective (i.e., Medicare). Incremental cost-effective ratios were calculated comparing the chemotherapy containing schedules to radiotherapy alone. (An incremental cost-effectiveness ratio [ICER] is defined as the ratio of the difference between the average effectiveness of the test and control arms to the difference between the mean costs of the test and control arms. Mean costs are used rather than median costs, despite the presence of outliers, because this is the most useful measure for decision making regarding resource allocation for treatment of patients. Although the mean is misleading as a representation of the typical patient, it is a useful measure for policy makers who need estimates of total costs of opting for a treatment method.) In addition, the cost to prevent a laryngectomy was calculated comparing patients undergoing induction chemotherapy and radiation to patients undergoing radiation alone and patients undergoing concurrent chemoradiotherapy to radiation alone. The cost of care for patients undergoing radiation was subtracted from the cost of care for patients have induction chemotherapy to obtain an incremental cost. This was then divided by the number of laryngectomies prevented by undergoing induction chemotherapy and radiation compared with radiation alone. Likewise, the cost/laryngectomy prevented for patients undergoing concurrent chemoradiotherapy was calculated in a similar fashion with the cost of care for patients undergoing concurrent chemoradiotherapy substituted in place of induction chemotherapy and radiation and number of laryngectomies prevented by concurrent chemoradiotherapy used in place of the number of laryngectomies prevented by induction chemotherapy and radiation. Sensitivity analysis was performed evaluating the effect on treatment cost of using only International Classification of Diseases (ICD)-9 codes associated with head-and-neck cancer and the effect of age at the time of diagnosis, <65 compared with $65, on expected Medicare cost per patient. ICD-9 code 161 was used to identify claims for the relevant cancer. Nonparametric bootstrap analysis was used to generate 95% confidence intervals on the cost-effectiveness plane and to plot cost-effective acceptability curves (7–10). (The bootstrap analysis involved drawing 1,000 samples with replacement from the data and calculating the ICER for each sample. These estimates are all plotted on a plane, and the middle 95% of estimates are identified as being within the 95% confidence interval.) This study was reviewed and approved by the Institutional Review Boards of the American College of Radiology and Fox Chase Cancer Center. The manuscript was also reviewed by the Centers for Medicare and Medicaid Services in accordance with our Data Use Agreement to ensure patients participating in the study could not be identified.

RESULTS Patient randomization, inclusion and exclusion criteria used in this economic analysis are depicted in the Consolidated Standards of Reporting Trials diagram in Fig. 1. Table 1 lists the stage of the patients in the three groups: in cost study (ICS) (i.e., patients with Medicare insurance and matched claims), no Medicare data (NMD) (i.e., patients of $65 but without Medicare data), and non-Medicare age (NMA) (i.e., patients <65 years of age). Patients in the ICS

438

I. J. Radiation Oncology d Biology d Physics

Volume 79, Number 2, 2011

Table 1. Patient stage by group

ICS (n = 66) RT only Induction chemotherapy and radiation Concurrent chemoradiation NMD (n = 104) RT only Induction chemotherapy and radiation Concurrent chemoradiation NMA (n = 345) RT only Induction chemotherapy and radiation Concurrent chemoradiation

Stage III

Stage IV

18 (78%) 18 (78%) 16 (80%)

5 (22%) 5 (22%) 4 (20%)

22 (16%) 20 (61%) 21 (64%)

16 (42%) 13 (39%) 12 (36%)

71 (65%) 73 (62%) 75 (64%)

39 (35%) 44 (38%) 43 (36%)

Abbreviations: ICS = in cost study; NMD = no Medicare data; NMA = non-Medicare age.

Fig. 1. Diagram documenting the inclusion criteria for patients included in the cost analysis. Only 12% of randomized patients had sufficient cost data to be included in this analysis.

subgroup had statistically significant lower stage tumors compared with patients in the NMD and NMA subgroups (p = 0.014 and 0.017, respectively). Toxicity differed between the three groups. Although there were no differences in Grade 3+ toxicity among the three groups, the ICS and NMA patients receiving concurrent chemoradiation had a higher percentage of Grade 3+ toxicity compared with the NMD group. Patients in the ICS and NMD groups receiving radiation alone had higher acute toxicity compared with patients in the NMA group. An even distribution of late toxicity was seen between groups in patients receiving concurrent chemoradiation, but differences did exist between groups in patients receiving induction chemotherapy and radiation and radiation alone. Patients in the NMD and NMA groups receiving induction chemotherapy and radiation had greater late toxicity compared with the ICS group, whereas patients receiving radiation alone in the ICS subgroup had higher late toxicity compared with the NMD and NMA subgroups.

Patients in the ICS subgroup had consistently poorer median survival compared with patients in the NMA and NMD subgroups. There was, however, no significant difference in overall survival between treatment arms within the ICS subgroup. There was no significant difference between disease-free survival between arms, but patients in the NMD group receiving concurrent chemoradiation had higher disease-free survival, 4.3 years, compared with patients in the NMA group, 3.1 years, and ICS group, 2.3 years. Table 2 shows the expected mean Medicare cost, discounted survival (both overall and disease-free), and ICER. Medicare costs are the main cost measure for this study. In comparison, the mean protocol treatment costs as determined by clinical data were $11,662 for radiation alone, $17,439 for induction chemotherapy and radiation, and $12,433 for radiation and concurrent chemotherapy. Figure 2 is the incremental cost-effectiveness scatterplot comparing induction chemotherapy and radiation to radiation alone. The 95% confidence ellipse, however, crosses both the cost and effectiveness axes. Figure 3 is the incremental cost-effectiveness scatterplot comparing concurrent chemoradiation with radiation alone. Once again, the 95% confidence ellipse crosses both the cost and effect axes, but the majority of points lie below the $50,000/life year line. Five patients in the ICS subgroup having radiation alone experienced a primary relapse, with 4 undergoing laryngectomy. Three patients receiving induction chemotherapy and radiation experienced a primary relapse and 2 underwent laryngectomy, whereas 2 patients having concurrent chemoradiation had a primary relapse, but none of the patients had a laryngectomy. This results in a cost savings of $4,170/laryngectomy prevented comparing induction chemotherapy and radiation with radiation alone and $128.25/laryngectomy prevented for concurrent chemoradiation compared radiation alone. Sensitivity analysis revealed costs were less when only claims for a cancer diagnosis were used but the costeffectiveness conclusion did not change. The cost decreased to $31,318 for radiotherapy only, for concurrent chemoradiation to $25,470, and to $31,300 for induction chemoradiation.

Cost-effectiveness of Laryngeal Sparing Therapy d A. KONSKI et al.

439

Table 2. Expected mean cost, survival, and ICER of patients treated in RTOG 9111 Treatment Radiation alone Induction chemotherapy and radiation Concurrent chemoradiation

Expected mean 47-month cost (95% CI)

Discounted survival (y)

Discounted disease-free survival (y)

Survival ICER

DFS ICER

$57,357 ($38,870–78,6240) $49,018 ($37,583–49,338)

3.95 4.84

2.43 3.61

–$9,336

–$7,031

$57,870 ($40,309–70,468)

4.68

2.68

$697

$2,048

Abbreviations: ICER = incremental cost-effective ratios; DFS = disease-free survival; RTOG = Radiation Therapy Oncology Group.

Conversely, all of the costs increased when total costs to all payers (Medicare, beneficiary, and other primary payer were considered, but once again the overall cost-effectiveness results did not change.) Costs increased to $63,250 for radiotherapy only, $64,653 for concurrent chemoradiation, and $56,458 for induction chemotherapy and radiation. Costs in all treatment groups were higher in patients $65 compared with patients <65, but there were only 16 patients <65 in the analysis, so statistical comparison is not possible between the two groups. A sensitivity analysis was also performed evaluating the effect of the 3 high-cost patients in the radiation alone arm. The 47-month expected Medicare cost per patient of the radiotherapy-only arm decreases to $41,874 when the 3 highcost patients are removed, with an increase to 4.22 years for the overall survival and 2.66 years of the disease-free survival. The incremental cost-effectiveness ratios increase to $11,640/life year and $34,939/life year for induction chemoradiation and concurrent chemoradiation, respectively. Both are within the range of accepted cost-effectiveness.

DISCUSSION The methodology needed to perform pharmacoeconomic analyses alongside clinical trials has been outlined, but is rarely undertaken (11–16). Hillner has proposed factors supporting the inclusion of an economic analysis alongside a clinical trial, and RTOG 91-11 fulfills a number of the outlined criteria (17). This economic analysis was performed after the conclusion of the study and was performed using Medicare data. But once performed, can the results of economic analysis be translated to the general population? Were the characteristics of the population selected in the economic analysis representative of the entire trial population or the general population on the whole? Some have debated that randomized clinical trials may not be the best venue for health economic evaluations of treatment regardless of economic data source (18). Entry to a trial is controlled and patients monitored more carefully; therefore, patients treated on clinical trials are usually not representative of the general population as a whole and the

RT + Induction vs. RT-Only $s per year of survival 150,000 Acceptable ICER = 50,000

100,000

95% CI Ellipse (LCL=-22095300, UCL=2001273)

Cost

50,000

0 -2

0

-1

1

2

-50,000

3

4 Mean ICER = $-9336

-100,000

-150,000

Survival Cost within bootstrapped 95% CI of ICER

ICER, ( Survival = 0.89, Cost = -8338)

Cost outside bootstrapped 95% CI of ICER

Acceptable ICER = 50,000

Fig. 2. Scatterplot showing the 95% confidence ellipse and confidence intervals comparing patients treated with induction chemotherapy and radiation to patients receiving radiotherapy alone. The 95% confidence ellipse crosses the X and Y axes because of little difference between the two treatments resulting in no statistical difference in cost-effectiveness between the two treatments.

I. J. Radiation Oncology d Biology d Physics

440

Volume 79, Number 2, 2011

RT + Concomitant vs. RT Only $s per year of survival 150,000 Acceptable ICER = 50,000

100,000

50,000

Cost

Mean ICER = $697 0 -2

0

-1

1

2

-50,000

3 95% CI Ellipse (LCL=-11519202, UCL=3968193)

4

-100,000

-150,000

Survival Cost within bootstrapped 95% CI of ICER Cost outside bootstrapped 95% CI of ICER

ICER, ( Survival = 0.74, Cost = 514)

Fig. 3. Scatterplot showing the 95% confidence ellipse and confidence intervals comparing patients treated with concurrent chemoradiotherapy to radiotherapy alone. Once again, as seen in Fig. 2, the 95% confidence ellipse crosses the X and Y axes because of little difference between the two treatments resulting in no statistical difference in cost-effectiveness between the two treatments.

results may not be translatable to the general population (13, 14, 16). Clinical trials have high internal validity and are relatively bias-free for the assessment of the most effective of two or more treatments, their external validity and the estimates of treatment effect, impact on quality of life, and economic costs may all be questioned (19). The number of patients included in economic analyses who were treated in clinical trials is variable, pointing to the inherent difficulties in performing this type of analysis regardless of source of economic data. Smith et al. measured costs in all patients participating in a clinical trial evaluating the cost-effectiveness of two forms of bone marrow transplant in patients with Hodgkin’s and non-Hodgkin’s lymphoma (20). Bloomfield et al. used a detailed retrospective chart review of resources used in 70% of patients to perform an economic analysis of chemotherapy with mitoxantrone plus prednisone for patients with symptomatic hormone-resistant prostate cancer (21). Costs were collected prospective in all patients enrolled on a Phase III clinical trial comparing two different chemotherapy regimens in the treatment of non–small-cell lung cancer (22). Conversely, at the low end Bennett et al. used detailed financial data from only 7 of 88 patients who survived induction chemotherapy on a Phase III Eastern Cooperative Oncology Group trial to inform an economic model evaluating the economic benefits of a yeast-derived granulocyte-macrophage colony-stimulating factor (12). There was no mention how these 7 patients were selected or how they compared with the other trial participants (12). In a later analysis of the same Eastern Cooperative Oncology Group trial, Bennett et al. collected detailed economic analysis from 24 patients from seven major contributing institutions (23). The

conclusion of the analysis, whether using economic data from 7 or 24 patients, did not differ between the reports. Only 12% of the total RTOG 91-11 trial patients were able to be included in the economic analysis using Medicare data. Non-random missing data overall and the small number of patients included in the economic analysis from each arm affect the results. The costs for the initial treatment were higher than expected for patients in the radiotherapy alone group. Three patients experienced significantly higher costs, with 1 patient having $99,000 in costs in month 0 and another with $51,000 by month 5. The small number of patients in each treatment arm allow a small number of outliers to affect the overall cost calculation. This is further seen in our sensitivity analysis when we eliminate the 3 high-cost, radiotherapy-only patients. Including only patients with Medicare data limits the extrapolation of the results of this study to the general population (24). The characteristic of the ICS group differed from both the NMA and NMD groups. Differences were also noted in toxicities experienced between the groups that could have affected resource utilization. And finally, the ICS subgroup had worse survival compared with the other subgroups, which also would affect the results. On the positive side, using administrative claims data, such as Medicare, reduces the potential complexity and additional cost of adding an economic analysis to a clinical trial compared with attempting to collect the data prospectively (13, 15, 16). The cost-effectiveness from the addition of chemotherapy to radiation was a secondary end point in our study. The mean incremental cost-effectiveness ratios were favorable, but are they considered statistically significant? The 95%

Cost-effectiveness of Laryngeal Sparing Therapy d A. KONSKI et al.

confidence ellipses of the incremental cost-effectiveness scatterplots of overall survival comparing induction chemotherapy and radiation and concurrent chemoradiation cross the y axis indicating no statistical difference in survival between the treatments, which was also shown in the original report. Because of the small difference in cost between treatments and small number of patients in each group, the 95% confidence ellipses also cross the x axis. This illustrates a common problem in evaluating cost-effectiveness alongside clinical trials. Clinical trials are powered to evaluate differences in outcome and would require significantly more patients to prove statistical difference in cost-effectiveness (14, 25). Given the clinical and economic results of RTOG 91-11, a total of 165 patients in each arm comparing induction chemotherapy and radiation to radiation alone and 243 patients in each arm comparing concurrent chemoradiation to radiation alone would be required to show statistical significance if $500/quality adjusted life years was the minimum meaningful acceptable ICER and $50,000/QALY was the maximal acceptable ICER (26). Therefore, this study was underpowered from an economic analysis perspective, given the above calculation, to detect a significant difference in cost-effectiveness since we only had 20–23 patients in each arm. We did find, however, the dollar amounts acceptable per laryngectomy prevented. There are, however, no other reported studies using this metric to compare against. Using administrative claims data, such as Medicare, reduces the potential complexity and additional cost of adding an economic analysis to a clinical trial compared with attempting to collect the data prospectively (13, 15, 16). Therefore, it may be worth exploring possibilities of obtaining more claims data to facilitate cost comparisons. Among the Medicare eligible patients considered for this study (NMA), claims data were unavailable for the following reasons: they were in HMOs, and we could not get any claims for them; they had other sources of care, such as Veterans Administration or Canadian health care; and they were not in the trial for the years for which we had Medicare data. This resulted in data that had many observations that were missing, and these observations were not missing at random. The only factor that can be readily addressed is the third one, and the solution would be to obtain more years of Medicare data. These data pulls are expensive. Medicare data will restrict our inferences to patients older than age 65; therefore, these data are useful for studies of clinical trials where a substantial number of patients are age 65 or

441

older. Even if the findings are not generalizable across all patients, sufficient sample sizes may permit meaningful and useful inferences about patients age 65 and older. With more observations, the differences between treatment arms will not be driven by outliers. The pattern of missing data may still not be random because characteristics of patients with HMOs or other payer benefits may still be different from characteristics of patients with fee-for-service Medicare, but the underlying characteristics for patients for whom we have Medicare claims may not differ as starkly from characteristics of patients with missing data as they do now. The larger underlying finding is that administrative claims data are a plausible but limited means of comparing economic outcomes. We use Medicare because that is the only feasible payer from whom we can get all claims for a subset of the population. If there were a multicenter trial with one center in Canada, it might be feasible to use claims data and our methods to obtain meaningful representative results for the Canadian patients of all ages. CONCLUSION Secondary analyses of clinical trials, such as this one, will be important in the future as more expensive targeted therapies are introduced into clinical practice. The use of Medicare data to inform the cost component of an economic analysis although technically achievable may result in outcomes that are not translatable to the general population as a whole because of small patient numbers and non-random missing data. Differences existed between the groups in this study that could have influenced the outcome of the analysis. The small numbers in each group also affected the ability to determine if a statistical difference existed between groups. Therefore, economic analysis using Medicare data would only be useful if the study and general populations were similar or included sufficient patients in the Medicare eligible age group (age 65 and older) to determine statistical difference between the groups. Analyses using other claims data where a payer covers most of the patients in a locality may be a promising option. Another approach would be to prospectively plan to collect Medicare data for studies in which the over-65 population is particularly important even if the results cannot be generalized to the rest of the patient population. Data could be obtained each year for all patients in all such RTOG studies providing full data for all patients covered by Medicare (non-HMO) throughout the study period.

REFERENCES 1. Forastiere AA, Goepfert H, Maor M, et al. Concurrent chemotherapy and radiotherapy for organ preservation in advanced laryngeal cancer. N Engl J Med 2003;349:2091–2098. 2. Kaplan E, Meier P. Nonparametric estimation from incomplete observations. J Am Stat Assoc 1958;53:447–457. 3. Cox DR. Regression models and life tables. J Royal Stat Soc 1972;34:187–220. 4. Lin DY, Feuer EJ, Etzioni R, et al. Estimating medical costs from incomplete follow-up data. Biometrics 1997;53:419–434.

5. Brown ML, Riley GF, Schussler N, et al. Estimating health care costs related to cancer treatment from SEER-Medicare data. Med Care 2002;40:IV-104–17. 6. Etzioni R, Riley GF, Ramsey SD, et al. Measuring costs: Administrative claims data, clinical trials, and beyond. Med Care 2002;40:III63–72. 7. Warren JL, Brown ML, Fay MP, et al. Costs of treatment for elderly women with early-stage breast cancer in fee-for-service settings. J Clin Oncol 2002;20:307–316.

442

I. J. Radiation Oncology d Biology d Physics

8. Manning WG, Fryback DG, Weinstein MC. Reflecting uncertainty in cost-effectiveness analysis. In: Gold MR, Siegel JE, Russell LB, Weinstein MC, editors. Cost-effectiveness in health and medicine. New York: Oxford University Press; 1996. p. 247–275. 9. Fenwick E, Byford S. A guide to cost-effectiveness acceptability curves. Br J Psychiatry 2005;187:106–108. 10. Fenwick E, O’Brien BJ, Briggs A. Cost-effectiveness acceptability curves—facts, fallacies and frequently asked questions. Health Econ 2004;13:405–415. 11. Van Enckevort P, TenVergert EM, Kingma J, Koeter GH, Rutten FFH. Factors to consider when designing phase III clinical trials involving economic evaluations. Percept Motor Skills 1999;89:1059–1072. 12. Bennett CL, Golub R, Waters TM, et al. Economic analyses of phase III cooperative cancer group clinical trials: Are they feasible? Cancer Invest 1997;15:227–236. 13. Bennett CL, Armitage JL, Buchner D, et al. Economic analysis in phase III clinical cancer trials. Cancer Invest 1994;12:336–342. 14. O’Sullivan AK, Thompson D, Drummond MF. Collection of health-economic data alongside clinical trials: Is there a future for piggyback evaluations? Value Health 2005;8:67–79. 15. Mauskopf J, Schulman K, Bell L, et al. A strategy for collecting pharmacoeconomic data during phase II/III clinical trials. Pharmacoeconomics 1996;9:264–277. 16. Bennett CL, Westerman IL. Economic analysis during phase III clinical trials: Who, what, when, where, and why? Oncology (Williston Park) 1995;9:169–175. 17. Hillner BE. Potential evaluation of the incremental cost-effectiveness of paclitaxel in advanced non-small-cell lung cancer (Eastern Cooperative Oncology Group 5592). J Natl Cancer Inst Monogr 1995;19:65–67. 18. Fayers PM, Hand DJ. Generalisation from phase III clinical trials: Survival, quality of life, and health economics. Lancet 1997;350:1025–1027.

Volume 79, Number 2, 2011

19. Rittenhouse B, O’Brien BJ. Threats to the validity of pharmacoeconomics analyses based on clinical trial data. In: Spiker B, editor. Quality of life and pharmacoeconomics in clinical trials. Philadelphia: Lippincott-Raven; 1996. p. 1215–1223. 20. Smith TJ, Hillner BE, Schmitz N, et al. Economic analysis of a randomized clinical trial to compare filgrastim-mobilized peripheral-blood progenitor-cell transplantation and autologous bone marrow transplantation in patients with Hodgkin’s and non-Hodgkin’s lymphoma. J Clin Oncol 1997;15: 5–10. 21. Bloomfield DJ, Krahn MD, Neogi T, et al. Economic evaluation of chemotherapy with mitoxantrone plus prednisone for symptomatic hormone-resistant prostate cancer: Based on a Canadian randomized trial with palliative end points. J Clin Oncol 1998; 16:2272–2279. 22. Sacristan JA, Kennedy-Martin T, Rosell R, et al. Economic evaluation in a randomized phase III clinical trial comparing gemcitabine/cisplatin and etoposide/cisplatin in non-small cell lung cancer. Lung Cancer 2000;28:97–107. 23. Bennett CL, Stinson TJ, Tallman MS, et al. Economic analysis of a randomized placebo-controlled phase III study of granulocyte macrophage colony stimulating factor in adult patients (>55 to 70 years of age) with acute myelogenous leukemia. Eastern Cooperative Oncology Group (E1490). Ann Oncol 1999;10:177–182. 24. Drummond M. Introducing economic and quality of life measurements into clinical studies. Ann Med 2001;33: 344–349. 25. Drummond MF, Davies L. Economic analysis alongside clinical trials. Revisiting the methodological issues. Int J Technol Assess Health Care 1991;7:561–573. 26. Willan AR, O’Brien BJ. Sample size and power issues in estimating incremental cost-effectiveness ratios from clinical trials data. Health Econ 1999;8:203–211.