Counting the Costs of Chemotherapy in a National Cancer Institute of Canada Randomized Trial in Nonsmall-Cell Lung Cancer By Liisa Jaakkimainen, Pamela J. Goodwin, Joseph Pater, Padraig Warde, Nevin Murray, and Edna Rapp An economic evaluation was undertaken of a previously reported National Cancer Institute of Canada (NCIC) trial of chemotherapy in advanced nonsmallcell lung cancer (NSCLC). That trial had demonstrated a survival benefit associated with the use of either vindesine and cisplatin (VP) or cyclophosphamide, doxorubicin, and cisplatin (CAP) in relation to best supportive care (BSC). The economic technique used in this evaluation was cost-effectiveness analysis (CEA). All costs were determined from the viewpoint of two provincial health care plans. When compared with BSC, the survival benefit of 8 weeks in favor of patients receiving CAP chemotherapy was associated with an economic saving of $949.49 (in 1984 Canadian dollars). This translated into a savings of $6,171.69 per year of life gained. The mean survival benefit of 12.8 weeks that was obtained with VP
THE ECONOMIC implications of new treat-
ment strategies in medicine have rarely been evaluated. This is particularly true in the area of cancer therapy where investigators have usually concentrated on establishing therapeutic efficacy (or lack thereof) of new therapies and have uncommonly examined the costs of these treatments. As it has become increasingly apparent that the demands for the provision of health care may exceed available resources, economic factors have become more important in decision making regarding optimal allocation of health care resources. Full economic evaluations, which look at both costs and consequences of medical care, are urgently needed so that a more complete assessment of new therapeutic interventions and diagnostic technologies can be undertaken. The economic implications of new treatments are particularly important in common diseases such as nonsmall-cell lung cancer (NSCLC). The vast majority of patients with NSCLC either present with, or develop, advanced disease unsuitable for potentially curative therapy (either surgery or radiation therapy). The role of cytotoxic chemotherapy in these patients was assessed in a randomized controlled clinical trial (BR-5) by the National Cancer Institute of Canada (NCIC).
chemotherapy compared with BSC was associated with an increased cost of $3,637.60 per patient, or $14,777.75 per year of life gained. The economic evaluation demonstrated that the majority of costs on each of the three treatment arms was related to hospitalization and not to the use of chemotherapy agents. These results compare favorably with estimates of cost-effectiveness (CE) of commonly used treatments for other diseases and demonstrate that a policy of supportive care is associated with costs that may exceed those of active treatment. It is concluded that economic factors should not adversely affect decisions regarding the use of chemotherapy in advanced NSCLC. J Clin Oncol 8: 1301-1309. @ 1990 by American Soci-
ety of ClinicalOncology.
The trial compared two drug regimens, vindesine and cisplatin (VP) or cyclophosphamide, Adriamycin (doxorubicin; Adria Laboratories, Columbus, OH) and cisplatin (CAP) with a policy of best supportive care (BSC). The therapeutic results of this trial have been published' and they show a small, statistically significant improvement in median survival in patients treated with VP or CAP.
From the Departmentof Preventive Medicine and Biostatistics, University of Toronto, Toronto, Ontario;the Department of Medicine and the Research Institute Division of ClinicalEpidemiology, Mount Sinai Hospital,University of Toronto, Toronto, Ontario;the Department of Community Health and Epidemiology, Queen's University, National CancerInstitute of CanadaClinical Trials Group,Kingston, Ontario;the Department of Radiation Oncology, The Princess MargaretHospital, Toronto, Ontario;the Department of Medicine, Cancer Control Agency of British Columbia, Vancouver, British Columbia; and the Department of Medicine, Tom Baker Cancer Centre, Calgary, Alberta, Canada. Submitted January29, 1990; accepted March 27, 1990. Dr Goodwin is a CareerScientist of the OntarioMinistry ofHealth. Address reprint requests to Pamela J. Goodwin, MD, Mount Sinai Hospital, 600 University Ave, Suite 1224, Toronto, Ontario,CanadaM5G IX5. © 1990 by American Society of Clinical Oncology. 0732-183X/90/0808-0014$3.00/0
Journalof Clinical Oncology, Vol 8, No 8 (August), 1990: pp 1301-1309
Downloaded from ascopubs.org by 80.82.77.83 on February 24, 2017 from 080.082.077.083 Copyright © 2017 American Society of Clinical Oncology. All rights reserved.
1301
1302
JAAKKIMAINEN ET AL
A retrospective economic evaluation of this trial has now been completed, and the results are reported here. Our purpose was to assess the cost-effectiveness (CE) of these two chemotherapeutic regimens, VP and CAP, in comparison to a policy of BSC in patients with metastatic or inoperable NSCLC. Although the results of this economic evaluation provide important information regarding the costs of various treatment strategies in patients with NSCLC, readers are cautioned that economic factors should not be used alone, and other factors, such as the toxicity of treatments, must also be taken into account when treatment decisions are made. METHODS
GeneralMethod of Economic Evaluation A retrospective economic evaluation of the NCIC BR-5 randomized trial in advanced NSCLC' was undertaken. This trial compared the effectiveness of the two chemotherapeutic regimens described above to a policy of BSC. A minority of centers in the trial randomized patients between the two chemotherapy arms only; patients from these centers were not included in the economic evaluation because of concern regarding their comparability to patients on the three-arm portion of the trial. The study sample for this economic evaluation therefore included all eligible patients in the three-arm portion of the BR-5 study. Measures of the effectiveness of each treatment were taken directly from the trial data. The economic technique used was that of CE 23 analysis (CEA). ' The costs considered included costs of all hospital-based services determined from the viewpoint of two provincial health care plans. Other costs incurred by the provincial health care plans (such as care provided in the home) and costs to patients (such as out-of-pocket expenses and lost wages), to private health insurance plans (such as oral drugs taken at home), or to other groups within society were not included in the evaluation. All costs were determined in 1984 Canadian dollars, one of the years during which patients were entered into the BR-5 study. These costs were estimated for all the patients in the three-arm trial from the time they were randomized until death. Discounting of health costs and health effectiveness to adjust for differential timing4 was not undertaken because the mean and median survivals on all treatment arms were less than 1 year. Sensitivity analyses3 were conducted to assess the effect on the results of the CEA of variability in the estimates of cost and therapeutic efficacy that were used in this evaluation.
BR-5 PatientPopulationand Study Design Patients with nonresectable NSCLC received either CAP or VP or BSC. No chemotherapy was administered to patients on the BSC arm. Palliative radiation could be given to any patient in the trial, and patients on any arm could receive any other form of supportive care deemed appropriate by the treating physicians. Between February 1983 and
January 1986, 137 eligible patients were enrolled at centers who participated in the three-arm portion of the study.' There were 50 eligible patients randomized to the BSC arm, 43 to the CAP arm, and 44 to the VP arm. Two provinces, British Columbia (BC) and Ontario, contributed 62.1% of the patients on the three-arm study. The economic evaluation was conducted in these two provinces.
Treatment Outcomes The measures of health effectiveness were obtained directly from the therapeutic results of the BR-5 trial. The primary treatment outcome was mean survival (in weeks), which corresponded closely to median survival. The incremental therapeutic effectiveness was determined by subtracting the mean survival time on one treatment arm from the mean survival time on the comparison arm.
Determinationof Costs It was not feasible to retrospectively determine the individual costs for each patient in the trial because individual patient costs are not routinely identified in the Canadian health care system. Therefore, costs of the resources used by patients in this study were determined at two centers only: The Princess Margaret Hospital (PMH) in Toronto, Ontario and the A. Maxwell Evans Clinic (AMEC) in Vancouver, BC. Both of these centers randomized patients to the threearm portion of the BR-5 trial. Costs at each institution were determined separately and it was assumed that these costs were representative of costs at other institutions across the country. Resource use was obtained from all patients in the three-arm portion of the study except when stated otherwise. Average costs per patient were calculated by multiplying the unit cost of each resource by the average use of that resource per patient. The costs incurred in the treatment of patients were divided into four categories: chemotherapy, hospitalization, clinic visits, and radiotherapy. The majority of costs were determined in 1984 dollars. When it was necessary to determine costs in another year, costs were converted to 1984 dollars using the Health and Personal Care Index of the Consumer Price Index.5' 6 Chemotherapy costs. Chemotherapy costs included drug costs and other costs related to chemotherapy administration. Drug costs represented the actual 1984 drug prices paid by the institution. Other costs included those that were constant for each chemotherapy treatment regardless of whether chemotherapy was administered on an inpatient or outpatient basis (laboratory and radiology costs, physician and resident costs, and the cost for pharmacy preparation time of the chemotherapeutic drugs). Costs such as nursing costs were included in clinic costs or hospitalization costs depending on the location of treatment. Costs were grouped in this way to facilitate sensitivity analyses that examined the impact on CE of changing the location (inpatient v outpatient) of chemotherapy administration. The number of courses of CAP chemotherapy and VP chemotherapy were obtained directly from trial data. Hospitalization costs. The hospitalization costs were determined using the hotel approximation method as described by Hull et al.7 This method distinguishes "hotel"
Downloaded from ascopubs.org by 80.82.77.83 on February 24, 2017 from 080.082.077.083 Copyright © 2017 American Society of Clinical Oncology. All rights reserved.
1303
COST OF CHEMOTHERAPY IN NSCLC costs from "medical care" costs. Hotel costs are those costs that are constant for all patients (for example heating, lighting, security, and housekeeping). Costs associated with the specific types of patients included in the study, such as laboratory investigations and medical procedures, are included in the medical care costs. A direct allocation method was used to allocate hotel costs to inpatient services using an allocation basis that was appropriate for each cost center (for example, square footage for plant operations). 8 This method ignores the interaction of overhead departments and results in costs that are slightly lower than other methods, which include the costs of these interactions.' The hospitalization hotel and medical care costs for the PMH were obtained from a previous economic evaluation of a clinical trial in small-cell lung cancer.9 In that economic evaluation, the hospitalization medical care costs were determined in 1984 Canadian dollars for an oncology ward at the PMH that treated patients with solid tumors similar to those on the BR-5 study. The hospitalization medical care costs at the AMEC included nursing and ward staff, ward supplies, pharmacy (excluding chemotherapeutic drugs), radiology, other procedures and physician and resident costs. They were derived from patients admitted to a ward that treated patients with solid tumors. Because the AMEC moved to a different geographic location during 1984, some costs were determined in 1984 dollars and some in 1985 dollars. All costs were ultimately converted to 1984 dollars. The hospital charts of the 32 patients on the BR-5 study treated at the AMEC were reviewed to identify procedures and radiologic tests conducted on these patients. Chemotherapy admissions were distinguished from nonchemotherapy admissions, and the total number of procedures and radiologic tests performed on each patient during either type of admission was recorded. The BC health plan fee schedule was used to determine costs for each procedure or laboratory test. A per day cost was then obtained by dividing these total costs by the number of inpatient days involved. The hospitalization rates for all 29 Ontario patients were obtained from the Ontario Cancer Registry, which contains admission records for all cancer patients in that province. The accuracy of the Ontario Cancer Registry has previously been shown to be high.' The hospitalization rates for the 32 patients treated at the AMEC were obtained from a review of AMEC charts and compared with the data recorded in study charts at the NCIC clinical trials group central offices. The fact that their data were not obtained from a prospectively maintained registry potentially reduces the reliability of this data. Clinic costs. The clinic costs were also determined using the hotel approximation method." The clinic medical care costs for chemotherapy clinic visits and nonchemotherapy (assessment) clinic visits included nursing, antiemetics, equipment, physicians, residents, laboratory, and radiology costs. Care was taken to avoid double counting of costs included as chemotherapy costs. The number of each type of chemotherapy clinic visit for each patient was obtained from trial data. Nonchemotherapy clinic visits were assumed to occur monthly (excluding the months patients received chemotherapy) on all arms.
Radiotherapy costs. The hotel approximation methods was used to determine the radiotherapy costs. The radiotherapy medical care costs varied according to the type of radiotherapy visit (planning, review, or treatment visit). Medical care costs for a planning visit included the costs of staff physicians and residents, radiographers, a treatment simulator, and a planning clinic clerk; and those for a review visit included the cost of staff physicians, residents, and radiographers. Finally, medical care costs for a treatment visit included the costs of radiographers and treatment machines (cobalt or high energy linear accelerator). Because detailed fractionation data were not available for all patients, a radiation oncologist at the PMH and another at the AMEC reviewed the radiation therapy given to all patients in the trial. They estimated the number of each type of visit associated with each course of therapy that would have occurred if the patient had been treated at their institution.
Analysis of Costs Total costs. For each cost category, an average cost per patient was calculated for the BSC, CAP, and VP treatment arms at each institution. The estimated average chemotherapy, clinic, and radiotherapy costs per patient were calculated from costs determined at either the PMH or the AMEC and resource use by all patients on the three-arm study. The estimated average hospitalization cost per patient was based on costs determined at either the PMH or the AMEC, and the hospitalization rates were obtained for all patients treated in Ontario or all patients treated at the AMEC. Incrementalcosts. Incremental costs were determined by subtracting the average total cost per patient calculated in one arm of the study from the average total cost per patient calculated in the comparison arm. Combined costs. To simplify sensitivity analyses, the costs determined at the PMH and the AMEC were combined. The combined chemotherapy, clinic, and radiotherapy costs per patient were determined by averaging the costs per patient calculated at each institution. The hospitalization rates and per diem costs were averaged separately and then combined to provide an average hospitalization cost per patient.
CE Ratio The CE ratios represent the incremental health costs (IC) divided by the incremental health effectiveness.l This ratio is calculated using the following formula: CE = IC (chemotherapy) + IC (clinic) + IC (radiotherapy) + IC (hospital)/ incremental mean survival. CE ratios were determined for VP chemotherapy relative to BSC, CAP chemotherapy relative to BSC, and VP chemotherapy relative to CAP chemotherapy.
Sensitivity Analyses Sensitivity analyses were performed to determine the impact on CE of varying all assumptions and estimations that were made during identification of costs or resource use over a range of possible alternatives. The less the CE ratios changed in a sensitivity analysis (ie, the greater the robustness of the results), the greater the confidence in the results. The effect
Downloaded from ascopubs.org by 80.82.77.83 on February 24, 2017 from 080.082.077.083 Copyright © 2017 American Society of Clinical Oncology. All rights reserved.
1304
JAAKKIMAINEN ET AL
on the final CE ratios of varying each important cost center and resource by ±+25% was determined in this series of analyses. An additional sensitivity analysis was conducted to determine the impact on CE of giving CAP chemotherapy on an outpatient versus inpatient basis. In this analysis, the proportion of CAP chemotherapy given on an inpatient or outpatient basis was varied from 0% to 100%. RESULTS
BR-5 Treatment Outcomes As previously reported, 1 the median survival on the BSC, CAP, and VP arms was 17 weeks (mean 24.2 weeks), 24.7 weeks (mean 32.2 weeks), and 32.6 weeks (mean 37.0 weeks), respectively. The difference between VP and BSC was statistically significant (Ps .01) as was the difference between CAP and BSC (P < .05). Mean survival was the outcome measure used in this economic evaluation. The incremental mean survival time for VP compared with CAP and BSC were 4.8 and 12.8 weeks, respectively. Incremental mean survival for CAP compared with BSC was 8.0 weeks. Costs The average unit costs for selected items at the two institutions are provided in Table 1. Hospitalization rates and the average total hospitalization cost per patient in each arm of the trial are shown in Table 2. The chemotherapy-related admission cost per day was estimated to be $282.04 at the PMH and $324.50 at the AMEC, and the cost per day for nonchemotherapy-related admissions was estimated to be $291.72 and $322.77, respectively. These costs per day were averaged between the institutions and are shown in Table 2. Hospitalization costs were highest for the BSC arm followed by the VP and CAP arms. At the AMEC, hospitalization costs were highest in the VP arm followed by the BSC and CAP arms. The CAP arm had the lowest hospitalization cost at both institutions. The Ontario and AMEC VP and CAP hospitalization rates (the number of days per patient) demonstrated variability for reasons other than chemotherapy. The Ontario and AMEC hospitalization rates on the BSC arm were similar. The possible reasons for these discrepancies are explored in the Discussion section. The average costs per patient at the two centers combined for each cost category are shown in Table 3. For all treatment arms, the
Table 1. Selected Unit Costs (1984 Canadian Dollars)
Item
Drugs Cyclophosphamide (1,000 mg vial) Adriamycint (50 mg vial) Cisplatin (50 mg vial) Vindesine (5 mg vial) Hospital (per day) Hotel Medical Care Nonchemaotherapy Chemotherapy Clinic (per visit) Hotel Medical Care CAP Vindesine Assessment Radiotherapy (per visit) Hotel Medical Care Staff Planning Review Treatment Therac 20§ Clinac 35§ Theratron II§ Philips 250 kV§
Average Unit Cost*
$ 5.25 $111.50 $163.01 $153.04 $ 86.18 $165.18 $166.91 $ 53.11 $ 75.96 $ 13.29 $116.92t $ 16.29
$263.45 $ 36.16 $ 34.44 $ 21.30 $ 9.21 $112.8111
*These costs were obtained by averaging the unit costs estimated at the PMH and the AMEC.
"tDoxorubicin; Adria Laboratories,
Columbus, OH. lIncludes laboratory and radiologic investigation. These costs were included in the chemotherapy cost category for the chemotherapy arms. §Therac 20, Atomic Energy Canada Ltd, Ottawa, Canada; Clinac 35, Varian Associates, Palo Alto, CA; Theratron II, Atomic Energy Canada, Ltd, Phillips Medical System, Toronto, Canada. IUnit costs are higher for this machine because fewer patients were treated per year.
largest single cost was inpatient hospitalization. The combined costs indicate that the BSC arm had the highest hospitalization cost per patient. The BSC arm also had the highest radiotherapy costs and the lowest clinic and chemotherapy costs. CEA Using the combined costs, the incremental costs for each cost category and the CE ratios were calculated for each chemotherapy arm in comparison to BSC and for the two chemotherapy arms compared with each other. The results are presented in Table 4. When comparing either chemotherapy arm with BSC, there was a nega-
Downloaded from ascopubs.org by 80.82.77.83 on February 24, 2017 from 080.082.077.083 Copyright © 2017 American Society of Clinical Oncology. All rights reserved.
1305
COST OF CHEMOTHERAPY IN NSCLC Table 2. Hospitalization Rates and Costs at the PMH and the AMEC No. of Days per Patient (mean) Classification CAP chemotherapy arm Chemotherapy administration Nonchemotherapy administration Total VP chemotherapy arm Chemotherapy administration Nonchemotherapy administration Total BSC arm Total
Average No. of
Average Unit Cost per Patient*
Total Cost per Patient
3.5 11.2 14.7
$303.27 $307.25
$1,046.28 $3,425.84 $4,472.12
4.9 19.4 24.3
4.8 17.1 21.8
$303.27 $307.25
$1,440.53 $5,253.98 $6,694.51
22.6
23.6
$307.25
$7,235.74
PMH*
AMECt
Days per Patient
2.3 14.3 16.7
4.6 8.0 12.6
4.6 14.8 19.3 24.5
*Hospitalization data obtained for each patient from the date of randomization to date of death (before December 31, 1985). Two patients died after December 31, 1985 and were not included in the analysis since complete hospitalization data could not be obtained. The data are for patients (N = 29) treated in Ontario. tHospitalization data obtained for each patient from the date of randomization to date of death. The data are for patients (N = 32) treated at the AMEC. *These costs were obtained by averaging the unit costs for a chemotherapy and nonchemotherapy hospitalization visit estimated at the PMH and the AMEC.
tive incremental hospitalization cost in keeping with decreased hospitalization rates and costs when chemotherapy was used. The same was true with respect to the radiotherapy costs. When CAP was compared with BSC there was a negative incremental total cost per patient indicating a net economic saving of $949.39 per patient. The use of VP compared with BSC and with CAP had net economic costs of $3,637.60 per patient and $4,587.09 per patient, respectively. The CE of VP compared with BSC and with CAP was $14,777.75 and $49,693.48 per year of life gained, respectively. The CE of CAP compared with BSC was -$6,171.69 per year of life gained, that is, a saving was identified when CAP was used.
only to variability in the nonchemotherapy hospitalization rates and costs. Table 5 shows the effect of varying the location of administration of CAP (inpatient v outpatient). While the magnitude of the CAP to BSC CE ratio changes when the location of chemotherapy administration is varied, the direction (positive or negative) does not change, and the use of CAP remains more CE than BSC regardless of the location of chemotherapy administration (inpatient v outpatient). These sensitivity analyses indicate that the results of this economic evaluation are robust to considerable variation in the estimates of resource use and costs used in this evaluation. Therefore, inaccuracies in these estimates were unlikely to significantly alter the basic results of this evaluation.
Sensitivity Analyses The variation of estimates of resource use or costs by + 25% did not change the direction of the CE ratios in any of the sensitivity analyses. The final magnitude of the CE ratios was sensitive
DISCUSSION
The BR-5 trial provided a unique opportunity to examine the patterns of resource use and the Table 4. Measures of Economic Evaluation
(1984 Candian Dollars) Table 3. Average Costs Per Patient (1984 Canadian Dollars) Cost Category
BSC (N = 50)
CAP (N = 43)
VP (N = 44)
Chemotherapy Hospitalization Clinic Radiotherapy Total
0 $7,235.74 $ 992.92 $ 366.19 $8,594.85
$1,941.44 $4,472.12 $1,042.57 $ 189.23 $7,645.36
$ 3,973.48 $ 6,694.51 $ 1,352.47 $ 211.99 $12,232.45
Incremental Cost
Chemotherapy Hospitalization Clinic Radiotherapy Total CE (per life year gained)
VP to BSC
+$ 3,973.48 - $ 541.23 +$ 359.55 -$ 154.20 + $ 3,637.60 $14,777.75
CAP to BSC
+$1,941.44 - $2,763.62 +$ 49.65 -$ 176.96 -$ 949.49 -$6,171.69
Downloaded from ascopubs.org by 80.82.77.83 on February 24, 2017 from 080.082.077.083 Copyright © 2017 American Society of Clinical Oncology. All rights reserved.
VP to CAP
+$ 2,032.04 +$ 2,222.39 +$ 309.90 +$ 22.76 +$ 4,587.09 $49,693.48
1306
JAAKKIMAINEN ET AL
Table 5. Sensitivity Analysis of Varying the Proportion of Outpatient/Inpatient CAP Treatments Outpatient/Inpatient CAP Treatments
CE Ratio of CAP to BSC (per year of life gained)
0 0.25 0.50 0.75 1.0
-$ 7,423 -$ 8,309 -$ 9,194 -$10,079 -$10,965
costs associated with the use of palliative chemotherapy in patients with advanced NSCLC. The results of the economic evaluation are based on patients entered onto the three-arm part of the NCIC BR-5 study in which patients were randomized to receive either VP or CAP chemotherapy or to a policy of BSC without chemotherapy. Not only was a survival advantage of 8 weeks seen in association with the use of CAP chemotherapy, but the average cost of medical resources used by patients in the CAP chemotherapy arm was less than the average cost of medical resources used by patients in the BSC arm. Therefore, the use of CAP chemotherapy prolonged survival at a decreased cost to the hospitalbased services of the provincial health care plans. The VP chemotherapy arm resulted in both increased costs and increased survival when compared with BSC. It was, therefore, less economically favorable than CAP, but the costs associated with its use are probably still acceptable when compared with the benefits. The drug costs for the CAP chemotherapy regimen made up about 17% of the total cost of treating patients in that arm of the study, while those for VP were approximately 22% of the costs. Thus, chemotherapy drug costs represent a small proportion of the total treatment costs for patients with malignant disease, an observation that we have made previously. 9 Even though drug costs represent a small proportion of the total cost, reduction of costs through the introduction of generic cisplatin improved the CE of CAP relative to BSC. When 1989 drug costs (which include the costs of generic products) are incorporated into the economic evaluation, the use of CAP relative to BSC results in a saving of $7,625.54 per year of life gained, and the use of VP relative to BSC costs $10,088.16 per year of life gained. This represents a further improvement over the previous estimate of CE. The major costs on all three arms were related to
inpatient hospital care indicating that savings are most likely to occur if hospitalization rates can be decreased, regardless of whether chemotherapy is administered. The cost per fraction of radiation was estimated to be between $65.00 and $80.00 (in 1984 Canadian dollars) at the two cancer centers. The radiotherapy costs determined in this study were based on the specific resources used by the patients in the BR-5 trial. The treatments administered to these patients were palliative, and it is unlikely the costs would be the same for patients receiving curative radiotherapy requiring detailed planning and set-up techniques, or the use of custom made immobilization devices. Because of this and because the methodology used to determine radiotherapy costs in this study differs from that used in other studies determining radiotherapy costs, 11-14 direct comparison to radiotherapy costs reported in the other studies is not advisable. The analysis of toxicity of the two chemotherapy regimens was based on the 178 patients who received chemotherapy. The difference in toxicity between the treatment groups was significant only for neurologic toxicity (P: .001).' Although this toxicity may have had an important effect on quality of life, it was unlikely to result in additional costs beyond those reflected in the hospitalization and outpatient costs already considered. Therefore, additional costs were not included for toxicity alone. Only costs relevant to the hospital-based services of the provincial health care plans were determined in this study. If a more comprehensive viewpoint had been taken, for example, that of society in general, other costs (such as food, travel, accommodation, or even psychosocial distress) would have to be considered. The effects of inclusion of some of these costs can be estimated. For example, transportation and accommodation costs for patients receiving treatment away from their home towns would have been related to the number of clinic visits on each arm of the study. Since the BSC and VP chemotherapy arms had the higher hospitalization rates and more frequent clinic visits than the CAP arm, the costs of these items would probably be greatest on those arms. The predicted effects on the results of the economic evaluation would be to lower the cost of CAP relative to BSC, increase the VP cost relative to CAP, and to have little effect on the
Downloaded from ascopubs.org by 80.82.77.83 on February 24, 2017 from 080.082.077.083 Copyright © 2017 American Society of Clinical Oncology. All rights reserved.
1307
COST OF CHEMOTHERAPY IN NSCLC
cost of VP relative to BSC. The inclusion of other costs (such as food and clothing and lost wages) would probably be fairly equal in each arm and would not have a major effect on the estimates of CE. The CE ratios identified in this economic evaluation would likely be similar in countries with similar health care systems. The magnitude of the ratios might vary in different countries; however, their direction probably would not. In the United States, for example, the magnitude of the CE ratio would be influenced by several factors. Simple conversion of Canadian dollars to American dollars would lower the magnitude of the CE ratios; however, the higher overall American health care costs would likely counterbalance this effect. For example, if the radiotherapy charges of $500 to $1,000 per fraction that are commonly encountered in the United States were used in the current evaluation, the savings associated with the use of CAP compared with BSC would increase from $6,171 per year of life gained to between $12,000 and $20,000 per year of life gained, and the cost of VP relative to BSC would fall from $14,700 per year of life gained to between $5,800 and $10,600 per year of life gained. Because both hospitals costed in this study were tertiary care university-affiliated institutions, their costs may not accurately represent those at a nonteaching general hospital. Many of the nonchemotherapy-related hospital admissions were at primary care institutions. It is difficult to accurately predict the difference in costs between the two types of institutions. If nonchemotherapy hospitalization costs were actually lower at these nonteaching institutions, the magnitude of the CE ratio of CAP to BSC would decrease slightly. There were major differences in the hospitalization practices for the administration of CAP chemotherapy at the two hospitals evaluated. Ontario patients received 65.2% of their CAP treatments as outpatients, whereas no patients at the AMEC received CAP chemotherapy on an outpatient basis. This was due to a policy decision that CAP be given on an inpatient basis at AMEC and not because of medical differences between the two groups of patients. A sensitivity analysis demonstrated that the CE of CAP relative to BSC became increasingly favorable as an increasing proportion of CAP treatments was
given on an outpatient basis. Similar observations have been made by others.9'5' 6 This observation emphasizes the importance of reducing hospitalization rates, particularly those related to chemotherapy administration, to improve the CE of a given therapy. The Ontario Cancer Registry, which provided the hospitalization data for all the patients treated in Ontario, is a central registry that is prospectively maintained. Its accuracy is high 9 and the data on hospitalization rates for Ontario patients used in this economic evaluation are considered accurate. The hospitalization data collected for patients treated at the AMEC were obtained by a review of the AMEC charts and the charts prepared by the clinical trials nurse at that institution and not from a central registry. While the chemotherapy hospitalization data thus obtained are probably accurate, the nonchemotherapy data are subject to error, and the error may have occurred to a different extent in each arm of the study. The importance of collecting accurate data, particularly of a high-cost resource such as hospitalization, is demonstrated by the effect on the estimated CE ratios of using the hospitalization rates determined in Ontario (an accurate data source) versus those at AMEC. The cost saving associated with the use of CAP chemotherapy compared with a policy of BSC was almost twice as great at PMH than at AMEC, an effect that was due mainly to differences in estimated hospitalization rates for chemotherapy administration. The prospective collection of complete data relating to hospitalization rates in future clinical trials would facilitate economic evaluations of those trials. The purpose of conducting economic evaluations is to compare costs and benefits of different programs so that decisions concerning the allocation of scarce resources can be made. Comparison of programs included in one evaluation is reasonable. However, comparison of programs across different economic evaluations should be done cautiously as differing methodologies used in the evaluations may render their results less than fully comparable. With this caveat in mind, the results of previously published economic evaluations, which used a health outcome of life years gained, may be compared with the current evaluation. The use of CAP chemotherapy compared with BSC has a CE ratio that is comparable to reported estimates
Downloaded from ascopubs.org by 80.82.77.83 on February 24, 2017 from 080.082.077.083 Copyright © 2017 American Society of Clinical Oncology. All rights reserved.
1308
JAAKKIMAINEN ET AL
of CE for treatments such as routine betaadrenergic-antagonist therapy for medium-risk and high-risk hypertensive patients,17 routine digital rectal examination to screen for rectal carcinoma,is or alternating chemotherapy for small-cell lung cancer. 9 The use of VP chemotherapy in comparison to BSC was somewhat more costly, with estimates of CE comparable to those reported for coronary artery bypass graft surgery,"9 home dialysis or cadaveric renal transplantation for renal failure, 20 therapeutic leukocyte transfusion in neutropenic patients,21 or sigmoidoscopy as a screening procedure for rectal carcinoma in asymptomatic patients aged 50 to 80 years. 22 An important observation made in this study was that a policy of supportive care for patients with advanced NSCLC was associated with substantial costs. The BSC patients in this trial had higher hospitalization rates, and they received more palliative radiotherapy treatments than those who received chemotherapy, leading to increased costs. This indicates that the use of a chemotherapy treatment regimen that prolongs survival, even when given with palliative intent, may significantly reduce health care costs. Treatment policies for NSCLC patients should address the economic impact of providing palliative care for these patients, whether this includes chemotherapy or not. The common occurrence of NSCLC and the frequency with which it is unresectable underscore the importance of consideration of the costs of palliative treatment of patients with this malignancy. Centers that have a policy of supportive care without the use of chemotherapy may incur significantly increased costs to the hospital-based health care system than those using chemotherapy. This is particularly true if chemotherapy can be given on an outpatient basis. If supportive care is administered at centers other than the cancer treatment center, these costs may be shifted to other hospitals, but they are still incurred by the health care system as a whole and should not be ignored. It would be tempting to argue that the differ-
ences in hospitalization rates among the three arms may reflect differences in the quality of life and that chemotherapy improved quality of life because it resulted in hospitalization rates that were lower than when chemotherapy was not given (particularly on the CAP arm). It must be emphasized that no formal quality of life indicators were available in the randomized trial, and this evaluation cannot be considered a quality of life assessment of patients on the BR-5 study. If treatment decisions or policy development are to be based on quality of life factors, it is important that these be measured prospectively using appropriate techniques. The incorporation of quality of life measurements in future clinical trials in this area is needed to address this issue. The results of this economic evaluation are intended to provide information for policy makers who must decide how scarce health care resources are to be allocated. Economic factors should not be used alone when these policy decisions are made but should be considered along with other factors, including treatment efficacy and feasibility, the CE of competing programs and societal preferences. Treatment decisions regarding individual patients are clearly influenced by treatment policies, but they are made primarily after consideration of each individual patient's situation and treatment preferences after discussion between physician and patient. They should not be made solely on the basis of economic factors. The conclusions of this economic evaluation simply indicate that economic factors should not adversely affect decisions regarding the use of chemotherapy in patients with advanced NSCLC. Similarly, these conclusions should not be used to support the use of chemotherapy in situations where it has not been demonstrated to be effective. ACKNOWLEDGMENT We want to express our appreciation to Anika Kind at The Princess Margaret Hospital and Dr Ed Kostashuk, Ruth Page, and Don Morrison at the A. Maxwell Evans Clinic for their assistance with this study.
REFERENCES 1. Rapp E, Pater JL, Willan A, et al: Chemotherapy can prolong survival in patients with advanced non-small cell lung cancer-Report of a Canadian multicenter randomized trial. J Clin Oncol 6:633-641, 1988 2. Weinstein MC: Cost-effectiveness analysis for clinical
procedures in oncology. Bull Cancer (Paris) 67:491-500, 1980 3. Drummond MF, Stoddart GL, Torrance GW: Methods for the Economic Evaluation of Health Care Programmes. New York, NY, Oxford University Press, 1987
Downloaded from ascopubs.org by 80.82.77.83 on February 24, 2017 from 080.082.077.083 Copyright © 2017 American Society of Clinical Oncology. All rights reserved.
1309
COST OF CHEMOTHERAPY IN NSCLC 4. Weinstein MC, Stason WB: Foundations of costeffectiveness analysis for health and medical practices. New Engl J Med 296:716-721, 1977 5. Statistics Canada: Consumer price index: Health and personal care. Ottawa, Canada 14:36, 1989 6. Statistics Canada: Consumer price index: Health and personal care. Ottawa, Canada 10:35, 1985 7. Hull RD, Hirsh J, Sackett DL, et al: Cost-effectiveness of primary and secondary prevention of fatal pulmonary embolism in high-risk surgical patients. Can Med Assoc J 127:990-995, 1982 8. Horngren CT: Cost accounting. A managerial emphasis (ed 5). Englewood Cliffs, NJ, Prentice-Hall, 1982, pp 43-78, 363-402, 438-474, 475-530 9. Goodwin P, Feld R, Evans WK, et al: Cost effectiveness of cancer chemotherapy: An economic evaluation of a randomized trial in small-cell lung cancer. J Clin Oncol 6:1537-1547, 1988 10. Shepard DS, Thompson MS: First principles of costeffectiveness analysis in health. Pub Health Rep 96:535-543, 1979 11. Atherton L: The cost of radiotherapy treatments on a linear accelerator. Br J Radiol 57:106-107, 1984 12. Ilbery PLT: Cost of health care services in the treatment of cancer. Aust Fam Phys 7:1062-1074, 1978 13. Greene DR: The cost of radiotherapy treatments on a linear accelerator. Br J Rad 56:189-191, 1983
14. Wodinsky HB, Jenkin RDT: The cost of radiation treatment at an Ontario regional cancer centre. Can Med Assoc J 137:906-909, 1987 15. Friedlander ML, Tattersall MHN: Counting the costs of cancer therapy. Eur J Cancer Clin Oncol 18:1237-1241, 1982 16. Calman JC, McVie JG, Soukop M, et al: Cost of outpatient chemotherapy. Br Med J 1:493-494, 1978 17. Goldman L, Sia B, Cook EF, et al: Cost and effectiveness of routine therapy with long-term beta-adrenergic antagonists after acute myocardial infarction. N Engl J Med 319:152-157, 1988 18. Love RR, Fryback DG, Kimbrough SR: A costeffectiveness analysis of screening for carcinoma of the prostate by digital examination. Med Decis Making 5:263278, 1985 19. Pliskin JS, Stason WB, Weinstein MC, et al: Coronary artery bypass graft surgery: Clinical decision making and cost-effectiveness analysis. Med Decis Making 1:10-28, 1981 20. Roberts SD, Maxwell DR, Gross TL: Cost-effective care of end-stage renal disease: A billion dollar question. Ann Intern Med 93:499-500, 1980 21. Rosenshein MS, Farewell VT, Price TH, et al: The cost-effectiveness of therapeutic and prophylactic leukocyte transfusion. N Engl J Med 302:1508-1062, 1980 22. Weinstein MC: Estrogen use in postmenopausal women--Costs, risks, and benefits. N Engl J Med 303:308316, 1980
Downloaded from ascopubs.org by 80.82.77.83 on February 24, 2017 from 080.082.077.083 Copyright © 2017 American Society of Clinical Oncology. All rights reserved.
THE ECONOMIC implications of new treat-
ment strategies in medicine have rarely been evaluated. This is particularly true in the area of cancer therapy where investigators have usually concentrated on establishing therapeutic efficacy (or lack thereof) of new therapies and have uncommonly examined the costs of these treatments. As it has become increasingly apparent that the demands for the provision of health care may exceed available resources, economic factors have become more important in decision making regarding optimal allocation of health care resources. Full economic evaluations, which look at both costs and consequences of medical care, are urgently needed so that a more complete assessment of new therapeutic interventions and diagnostic technologies can be undertaken. The economic implications of new treatments are particularly important in common diseases such as nonsmall-cell lung cancer (NSCLC). The vast majority of patients with NSCLC either present with, or develop, advanced disease unsuitable for potentially curative therapy (either surgery or radiation therapy). The role of cytotoxic chemotherapy in these patients was assessed in a randomized controlled clinical trial (BR-5) by the National Cancer Institute of Canada (NCIC).
chemotherapy compared with BSC was associated with an increased cost of $3,637.60 per patient, or $14,777.75 per year of life gained. The economic evaluation demonstrated that the majority of costs on each of the three treatment arms was related to hospitalization and not to the use of chemotherapy agents. These results compare favorably with estimates of cost-effectiveness (CE) of commonly used treatments for other diseases and demonstrate that a policy of supportive care is associated with costs that may exceed those of active treatment. It is concluded that economic factors should not adversely affect decisions regarding the use of chemotherapy in advanced NSCLC. J Clin Oncol 8: 1301-1309. @ 1990 by American Soci-
ety of ClinicalOncology.
The trial compared two drug regimens, vindesine and cisplatin (VP) or cyclophosphamide, Adriamycin (doxorubicin; Adria Laboratories, Columbus, OH) and cisplatin (CAP) with a policy of best supportive care (BSC). The therapeutic results of this trial have been published' and they show a small, statistically significant improvement in median survival in patients treated with VP or CAP.
From the Departmentof Preventive Medicine and Biostatistics, University of Toronto, Toronto, Ontario;the Department of Medicine and the Research Institute Division of ClinicalEpidemiology, Mount Sinai Hospital,University of Toronto, Toronto, Ontario;the Department of Community Health and Epidemiology, Queen's University, National CancerInstitute of CanadaClinical Trials Group,Kingston, Ontario;the Department of Radiation Oncology, The Princess MargaretHospital, Toronto, Ontario;the Department of Medicine, Cancer Control Agency of British Columbia, Vancouver, British Columbia; and the Department of Medicine, Tom Baker Cancer Centre, Calgary, Alberta, Canada. Submitted January29, 1990; accepted March 27, 1990. Dr Goodwin is a CareerScientist of the OntarioMinistry ofHealth. Address reprint requests to Pamela J. Goodwin, MD, Mount Sinai Hospital, 600 University Ave, Suite 1224, Toronto, Ontario,CanadaM5G IX5. © 1990 by American Society of Clinical Oncology. 0732-183X/90/0808-0014$3.00/0
Journalof Clinical Oncology, Vol 8, No 8 (August), 1990: pp 1301-1309
Downloaded from ascopubs.org by 80.82.77.83 on February 24, 2017 from 080.082.077.083 Copyright © 2017 American Society of Clinical Oncology. All rights reserved.
1301
1302
JAAKKIMAINEN ET AL
A retrospective economic evaluation of this trial has now been completed, and the results are reported here. Our purpose was to assess the cost-effectiveness (CE) of these two chemotherapeutic regimens, VP and CAP, in comparison to a policy of BSC in patients with metastatic or inoperable NSCLC. Although the results of this economic evaluation provide important information regarding the costs of various treatment strategies in patients with NSCLC, readers are cautioned that economic factors should not be used alone, and other factors, such as the toxicity of treatments, must also be taken into account when treatment decisions are made. METHODS
GeneralMethod of Economic Evaluation A retrospective economic evaluation of the NCIC BR-5 randomized trial in advanced NSCLC' was undertaken. This trial compared the effectiveness of the two chemotherapeutic regimens described above to a policy of BSC. A minority of centers in the trial randomized patients between the two chemotherapy arms only; patients from these centers were not included in the economic evaluation because of concern regarding their comparability to patients on the three-arm portion of the trial. The study sample for this economic evaluation therefore included all eligible patients in the three-arm portion of the BR-5 study. Measures of the effectiveness of each treatment were taken directly from the trial data. The economic technique used was that of CE 23 analysis (CEA). ' The costs considered included costs of all hospital-based services determined from the viewpoint of two provincial health care plans. Other costs incurred by the provincial health care plans (such as care provided in the home) and costs to patients (such as out-of-pocket expenses and lost wages), to private health insurance plans (such as oral drugs taken at home), or to other groups within society were not included in the evaluation. All costs were determined in 1984 Canadian dollars, one of the years during which patients were entered into the BR-5 study. These costs were estimated for all the patients in the three-arm trial from the time they were randomized until death. Discounting of health costs and health effectiveness to adjust for differential timing4 was not undertaken because the mean and median survivals on all treatment arms were less than 1 year. Sensitivity analyses3 were conducted to assess the effect on the results of the CEA of variability in the estimates of cost and therapeutic efficacy that were used in this evaluation.
BR-5 PatientPopulationand Study Design Patients with nonresectable NSCLC received either CAP or VP or BSC. No chemotherapy was administered to patients on the BSC arm. Palliative radiation could be given to any patient in the trial, and patients on any arm could receive any other form of supportive care deemed appropriate by the treating physicians. Between February 1983 and
January 1986, 137 eligible patients were enrolled at centers who participated in the three-arm portion of the study.' There were 50 eligible patients randomized to the BSC arm, 43 to the CAP arm, and 44 to the VP arm. Two provinces, British Columbia (BC) and Ontario, contributed 62.1% of the patients on the three-arm study. The economic evaluation was conducted in these two provinces.
Treatment Outcomes The measures of health effectiveness were obtained directly from the therapeutic results of the BR-5 trial. The primary treatment outcome was mean survival (in weeks), which corresponded closely to median survival. The incremental therapeutic effectiveness was determined by subtracting the mean survival time on one treatment arm from the mean survival time on the comparison arm.
Determinationof Costs It was not feasible to retrospectively determine the individual costs for each patient in the trial because individual patient costs are not routinely identified in the Canadian health care system. Therefore, costs of the resources used by patients in this study were determined at two centers only: The Princess Margaret Hospital (PMH) in Toronto, Ontario and the A. Maxwell Evans Clinic (AMEC) in Vancouver, BC. Both of these centers randomized patients to the threearm portion of the BR-5 trial. Costs at each institution were determined separately and it was assumed that these costs were representative of costs at other institutions across the country. Resource use was obtained from all patients in the three-arm portion of the study except when stated otherwise. Average costs per patient were calculated by multiplying the unit cost of each resource by the average use of that resource per patient. The costs incurred in the treatment of patients were divided into four categories: chemotherapy, hospitalization, clinic visits, and radiotherapy. The majority of costs were determined in 1984 dollars. When it was necessary to determine costs in another year, costs were converted to 1984 dollars using the Health and Personal Care Index of the Consumer Price Index.5' 6 Chemotherapy costs. Chemotherapy costs included drug costs and other costs related to chemotherapy administration. Drug costs represented the actual 1984 drug prices paid by the institution. Other costs included those that were constant for each chemotherapy treatment regardless of whether chemotherapy was administered on an inpatient or outpatient basis (laboratory and radiology costs, physician and resident costs, and the cost for pharmacy preparation time of the chemotherapeutic drugs). Costs such as nursing costs were included in clinic costs or hospitalization costs depending on the location of treatment. Costs were grouped in this way to facilitate sensitivity analyses that examined the impact on CE of changing the location (inpatient v outpatient) of chemotherapy administration. The number of courses of CAP chemotherapy and VP chemotherapy were obtained directly from trial data. Hospitalization costs. The hospitalization costs were determined using the hotel approximation method as described by Hull et al.7 This method distinguishes "hotel"
Downloaded from ascopubs.org by 80.82.77.83 on February 24, 2017 from 080.082.077.083 Copyright © 2017 American Society of Clinical Oncology. All rights reserved.
1303
COST OF CHEMOTHERAPY IN NSCLC costs from "medical care" costs. Hotel costs are those costs that are constant for all patients (for example heating, lighting, security, and housekeeping). Costs associated with the specific types of patients included in the study, such as laboratory investigations and medical procedures, are included in the medical care costs. A direct allocation method was used to allocate hotel costs to inpatient services using an allocation basis that was appropriate for each cost center (for example, square footage for plant operations). 8 This method ignores the interaction of overhead departments and results in costs that are slightly lower than other methods, which include the costs of these interactions.' The hospitalization hotel and medical care costs for the PMH were obtained from a previous economic evaluation of a clinical trial in small-cell lung cancer.9 In that economic evaluation, the hospitalization medical care costs were determined in 1984 Canadian dollars for an oncology ward at the PMH that treated patients with solid tumors similar to those on the BR-5 study. The hospitalization medical care costs at the AMEC included nursing and ward staff, ward supplies, pharmacy (excluding chemotherapeutic drugs), radiology, other procedures and physician and resident costs. They were derived from patients admitted to a ward that treated patients with solid tumors. Because the AMEC moved to a different geographic location during 1984, some costs were determined in 1984 dollars and some in 1985 dollars. All costs were ultimately converted to 1984 dollars. The hospital charts of the 32 patients on the BR-5 study treated at the AMEC were reviewed to identify procedures and radiologic tests conducted on these patients. Chemotherapy admissions were distinguished from nonchemotherapy admissions, and the total number of procedures and radiologic tests performed on each patient during either type of admission was recorded. The BC health plan fee schedule was used to determine costs for each procedure or laboratory test. A per day cost was then obtained by dividing these total costs by the number of inpatient days involved. The hospitalization rates for all 29 Ontario patients were obtained from the Ontario Cancer Registry, which contains admission records for all cancer patients in that province. The accuracy of the Ontario Cancer Registry has previously been shown to be high.' The hospitalization rates for the 32 patients treated at the AMEC were obtained from a review of AMEC charts and compared with the data recorded in study charts at the NCIC clinical trials group central offices. The fact that their data were not obtained from a prospectively maintained registry potentially reduces the reliability of this data. Clinic costs. The clinic costs were also determined using the hotel approximation method." The clinic medical care costs for chemotherapy clinic visits and nonchemotherapy (assessment) clinic visits included nursing, antiemetics, equipment, physicians, residents, laboratory, and radiology costs. Care was taken to avoid double counting of costs included as chemotherapy costs. The number of each type of chemotherapy clinic visit for each patient was obtained from trial data. Nonchemotherapy clinic visits were assumed to occur monthly (excluding the months patients received chemotherapy) on all arms.
Radiotherapy costs. The hotel approximation methods was used to determine the radiotherapy costs. The radiotherapy medical care costs varied according to the type of radiotherapy visit (planning, review, or treatment visit). Medical care costs for a planning visit included the costs of staff physicians and residents, radiographers, a treatment simulator, and a planning clinic clerk; and those for a review visit included the cost of staff physicians, residents, and radiographers. Finally, medical care costs for a treatment visit included the costs of radiographers and treatment machines (cobalt or high energy linear accelerator). Because detailed fractionation data were not available for all patients, a radiation oncologist at the PMH and another at the AMEC reviewed the radiation therapy given to all patients in the trial. They estimated the number of each type of visit associated with each course of therapy that would have occurred if the patient had been treated at their institution.
Analysis of Costs Total costs. For each cost category, an average cost per patient was calculated for the BSC, CAP, and VP treatment arms at each institution. The estimated average chemotherapy, clinic, and radiotherapy costs per patient were calculated from costs determined at either the PMH or the AMEC and resource use by all patients on the three-arm study. The estimated average hospitalization cost per patient was based on costs determined at either the PMH or the AMEC, and the hospitalization rates were obtained for all patients treated in Ontario or all patients treated at the AMEC. Incrementalcosts. Incremental costs were determined by subtracting the average total cost per patient calculated in one arm of the study from the average total cost per patient calculated in the comparison arm. Combined costs. To simplify sensitivity analyses, the costs determined at the PMH and the AMEC were combined. The combined chemotherapy, clinic, and radiotherapy costs per patient were determined by averaging the costs per patient calculated at each institution. The hospitalization rates and per diem costs were averaged separately and then combined to provide an average hospitalization cost per patient.
CE Ratio The CE ratios represent the incremental health costs (IC) divided by the incremental health effectiveness.l This ratio is calculated using the following formula: CE = IC (chemotherapy) + IC (clinic) + IC (radiotherapy) + IC (hospital)/ incremental mean survival. CE ratios were determined for VP chemotherapy relative to BSC, CAP chemotherapy relative to BSC, and VP chemotherapy relative to CAP chemotherapy.
Sensitivity Analyses Sensitivity analyses were performed to determine the impact on CE of varying all assumptions and estimations that were made during identification of costs or resource use over a range of possible alternatives. The less the CE ratios changed in a sensitivity analysis (ie, the greater the robustness of the results), the greater the confidence in the results. The effect
Downloaded from ascopubs.org by 80.82.77.83 on February 24, 2017 from 080.082.077.083 Copyright © 2017 American Society of Clinical Oncology. All rights reserved.
1304
JAAKKIMAINEN ET AL
on the final CE ratios of varying each important cost center and resource by ±+25% was determined in this series of analyses. An additional sensitivity analysis was conducted to determine the impact on CE of giving CAP chemotherapy on an outpatient versus inpatient basis. In this analysis, the proportion of CAP chemotherapy given on an inpatient or outpatient basis was varied from 0% to 100%. RESULTS
BR-5 Treatment Outcomes As previously reported, 1 the median survival on the BSC, CAP, and VP arms was 17 weeks (mean 24.2 weeks), 24.7 weeks (mean 32.2 weeks), and 32.6 weeks (mean 37.0 weeks), respectively. The difference between VP and BSC was statistically significant (Ps .01) as was the difference between CAP and BSC (P < .05). Mean survival was the outcome measure used in this economic evaluation. The incremental mean survival time for VP compared with CAP and BSC were 4.8 and 12.8 weeks, respectively. Incremental mean survival for CAP compared with BSC was 8.0 weeks. Costs The average unit costs for selected items at the two institutions are provided in Table 1. Hospitalization rates and the average total hospitalization cost per patient in each arm of the trial are shown in Table 2. The chemotherapy-related admission cost per day was estimated to be $282.04 at the PMH and $324.50 at the AMEC, and the cost per day for nonchemotherapy-related admissions was estimated to be $291.72 and $322.77, respectively. These costs per day were averaged between the institutions and are shown in Table 2. Hospitalization costs were highest for the BSC arm followed by the VP and CAP arms. At the AMEC, hospitalization costs were highest in the VP arm followed by the BSC and CAP arms. The CAP arm had the lowest hospitalization cost at both institutions. The Ontario and AMEC VP and CAP hospitalization rates (the number of days per patient) demonstrated variability for reasons other than chemotherapy. The Ontario and AMEC hospitalization rates on the BSC arm were similar. The possible reasons for these discrepancies are explored in the Discussion section. The average costs per patient at the two centers combined for each cost category are shown in Table 3. For all treatment arms, the
Table 1. Selected Unit Costs (1984 Canadian Dollars)
Item
Drugs Cyclophosphamide (1,000 mg vial) Adriamycint (50 mg vial) Cisplatin (50 mg vial) Vindesine (5 mg vial) Hospital (per day) Hotel Medical Care Nonchemaotherapy Chemotherapy Clinic (per visit) Hotel Medical Care CAP Vindesine Assessment Radiotherapy (per visit) Hotel Medical Care Staff Planning Review Treatment Therac 20§ Clinac 35§ Theratron II§ Philips 250 kV§
Average Unit Cost*
$ 5.25 $111.50 $163.01 $153.04 $ 86.18 $165.18 $166.91 $ 53.11 $ 75.96 $ 13.29 $116.92t $ 16.29
$263.45 $ 36.16 $ 34.44 $ 21.30 $ 9.21 $112.8111
*These costs were obtained by averaging the unit costs estimated at the PMH and the AMEC.
"tDoxorubicin; Adria Laboratories,
Columbus, OH. lIncludes laboratory and radiologic investigation. These costs were included in the chemotherapy cost category for the chemotherapy arms. §Therac 20, Atomic Energy Canada Ltd, Ottawa, Canada; Clinac 35, Varian Associates, Palo Alto, CA; Theratron II, Atomic Energy Canada, Ltd, Phillips Medical System, Toronto, Canada. IUnit costs are higher for this machine because fewer patients were treated per year.
largest single cost was inpatient hospitalization. The combined costs indicate that the BSC arm had the highest hospitalization cost per patient. The BSC arm also had the highest radiotherapy costs and the lowest clinic and chemotherapy costs. CEA Using the combined costs, the incremental costs for each cost category and the CE ratios were calculated for each chemotherapy arm in comparison to BSC and for the two chemotherapy arms compared with each other. The results are presented in Table 4. When comparing either chemotherapy arm with BSC, there was a nega-
Downloaded from ascopubs.org by 80.82.77.83 on February 24, 2017 from 080.082.077.083 Copyright © 2017 American Society of Clinical Oncology. All rights reserved.
1305
COST OF CHEMOTHERAPY IN NSCLC Table 2. Hospitalization Rates and Costs at the PMH and the AMEC No. of Days per Patient (mean) Classification CAP chemotherapy arm Chemotherapy administration Nonchemotherapy administration Total VP chemotherapy arm Chemotherapy administration Nonchemotherapy administration Total BSC arm Total
Average No. of
Average Unit Cost per Patient*
Total Cost per Patient
3.5 11.2 14.7
$303.27 $307.25
$1,046.28 $3,425.84 $4,472.12
4.9 19.4 24.3
4.8 17.1 21.8
$303.27 $307.25
$1,440.53 $5,253.98 $6,694.51
22.6
23.6
$307.25
$7,235.74
PMH*
AMECt
Days per Patient
2.3 14.3 16.7
4.6 8.0 12.6
4.6 14.8 19.3 24.5
*Hospitalization data obtained for each patient from the date of randomization to date of death (before December 31, 1985). Two patients died after December 31, 1985 and were not included in the analysis since complete hospitalization data could not be obtained. The data are for patients (N = 29) treated in Ontario. tHospitalization data obtained for each patient from the date of randomization to date of death. The data are for patients (N = 32) treated at the AMEC. *These costs were obtained by averaging the unit costs for a chemotherapy and nonchemotherapy hospitalization visit estimated at the PMH and the AMEC.
tive incremental hospitalization cost in keeping with decreased hospitalization rates and costs when chemotherapy was used. The same was true with respect to the radiotherapy costs. When CAP was compared with BSC there was a negative incremental total cost per patient indicating a net economic saving of $949.39 per patient. The use of VP compared with BSC and with CAP had net economic costs of $3,637.60 per patient and $4,587.09 per patient, respectively. The CE of VP compared with BSC and with CAP was $14,777.75 and $49,693.48 per year of life gained, respectively. The CE of CAP compared with BSC was -$6,171.69 per year of life gained, that is, a saving was identified when CAP was used.
only to variability in the nonchemotherapy hospitalization rates and costs. Table 5 shows the effect of varying the location of administration of CAP (inpatient v outpatient). While the magnitude of the CAP to BSC CE ratio changes when the location of chemotherapy administration is varied, the direction (positive or negative) does not change, and the use of CAP remains more CE than BSC regardless of the location of chemotherapy administration (inpatient v outpatient). These sensitivity analyses indicate that the results of this economic evaluation are robust to considerable variation in the estimates of resource use and costs used in this evaluation. Therefore, inaccuracies in these estimates were unlikely to significantly alter the basic results of this evaluation.
Sensitivity Analyses The variation of estimates of resource use or costs by + 25% did not change the direction of the CE ratios in any of the sensitivity analyses. The final magnitude of the CE ratios was sensitive
DISCUSSION
The BR-5 trial provided a unique opportunity to examine the patterns of resource use and the Table 4. Measures of Economic Evaluation
(1984 Candian Dollars) Table 3. Average Costs Per Patient (1984 Canadian Dollars) Cost Category
BSC (N = 50)
CAP (N = 43)
VP (N = 44)
Chemotherapy Hospitalization Clinic Radiotherapy Total
0 $7,235.74 $ 992.92 $ 366.19 $8,594.85
$1,941.44 $4,472.12 $1,042.57 $ 189.23 $7,645.36
$ 3,973.48 $ 6,694.51 $ 1,352.47 $ 211.99 $12,232.45
Incremental Cost
Chemotherapy Hospitalization Clinic Radiotherapy Total CE (per life year gained)
VP to BSC
+$ 3,973.48 - $ 541.23 +$ 359.55 -$ 154.20 + $ 3,637.60 $14,777.75
CAP to BSC
+$1,941.44 - $2,763.62 +$ 49.65 -$ 176.96 -$ 949.49 -$6,171.69
Downloaded from ascopubs.org by 80.82.77.83 on February 24, 2017 from 080.082.077.083 Copyright © 2017 American Society of Clinical Oncology. All rights reserved.
VP to CAP
+$ 2,032.04 +$ 2,222.39 +$ 309.90 +$ 22.76 +$ 4,587.09 $49,693.48
1306
JAAKKIMAINEN ET AL
Table 5. Sensitivity Analysis of Varying the Proportion of Outpatient/Inpatient CAP Treatments Outpatient/Inpatient CAP Treatments
CE Ratio of CAP to BSC (per year of life gained)
0 0.25 0.50 0.75 1.0
-$ 7,423 -$ 8,309 -$ 9,194 -$10,079 -$10,965
costs associated with the use of palliative chemotherapy in patients with advanced NSCLC. The results of the economic evaluation are based on patients entered onto the three-arm part of the NCIC BR-5 study in which patients were randomized to receive either VP or CAP chemotherapy or to a policy of BSC without chemotherapy. Not only was a survival advantage of 8 weeks seen in association with the use of CAP chemotherapy, but the average cost of medical resources used by patients in the CAP chemotherapy arm was less than the average cost of medical resources used by patients in the BSC arm. Therefore, the use of CAP chemotherapy prolonged survival at a decreased cost to the hospitalbased services of the provincial health care plans. The VP chemotherapy arm resulted in both increased costs and increased survival when compared with BSC. It was, therefore, less economically favorable than CAP, but the costs associated with its use are probably still acceptable when compared with the benefits. The drug costs for the CAP chemotherapy regimen made up about 17% of the total cost of treating patients in that arm of the study, while those for VP were approximately 22% of the costs. Thus, chemotherapy drug costs represent a small proportion of the total treatment costs for patients with malignant disease, an observation that we have made previously. 9 Even though drug costs represent a small proportion of the total cost, reduction of costs through the introduction of generic cisplatin improved the CE of CAP relative to BSC. When 1989 drug costs (which include the costs of generic products) are incorporated into the economic evaluation, the use of CAP relative to BSC results in a saving of $7,625.54 per year of life gained, and the use of VP relative to BSC costs $10,088.16 per year of life gained. This represents a further improvement over the previous estimate of CE. The major costs on all three arms were related to
inpatient hospital care indicating that savings are most likely to occur if hospitalization rates can be decreased, regardless of whether chemotherapy is administered. The cost per fraction of radiation was estimated to be between $65.00 and $80.00 (in 1984 Canadian dollars) at the two cancer centers. The radiotherapy costs determined in this study were based on the specific resources used by the patients in the BR-5 trial. The treatments administered to these patients were palliative, and it is unlikely the costs would be the same for patients receiving curative radiotherapy requiring detailed planning and set-up techniques, or the use of custom made immobilization devices. Because of this and because the methodology used to determine radiotherapy costs in this study differs from that used in other studies determining radiotherapy costs, 11-14 direct comparison to radiotherapy costs reported in the other studies is not advisable. The analysis of toxicity of the two chemotherapy regimens was based on the 178 patients who received chemotherapy. The difference in toxicity between the treatment groups was significant only for neurologic toxicity (P: .001).' Although this toxicity may have had an important effect on quality of life, it was unlikely to result in additional costs beyond those reflected in the hospitalization and outpatient costs already considered. Therefore, additional costs were not included for toxicity alone. Only costs relevant to the hospital-based services of the provincial health care plans were determined in this study. If a more comprehensive viewpoint had been taken, for example, that of society in general, other costs (such as food, travel, accommodation, or even psychosocial distress) would have to be considered. The effects of inclusion of some of these costs can be estimated. For example, transportation and accommodation costs for patients receiving treatment away from their home towns would have been related to the number of clinic visits on each arm of the study. Since the BSC and VP chemotherapy arms had the higher hospitalization rates and more frequent clinic visits than the CAP arm, the costs of these items would probably be greatest on those arms. The predicted effects on the results of the economic evaluation would be to lower the cost of CAP relative to BSC, increase the VP cost relative to CAP, and to have little effect on the
Downloaded from ascopubs.org by 80.82.77.83 on February 24, 2017 from 080.082.077.083 Copyright © 2017 American Society of Clinical Oncology. All rights reserved.
1307
COST OF CHEMOTHERAPY IN NSCLC
cost of VP relative to BSC. The inclusion of other costs (such as food and clothing and lost wages) would probably be fairly equal in each arm and would not have a major effect on the estimates of CE. The CE ratios identified in this economic evaluation would likely be similar in countries with similar health care systems. The magnitude of the ratios might vary in different countries; however, their direction probably would not. In the United States, for example, the magnitude of the CE ratio would be influenced by several factors. Simple conversion of Canadian dollars to American dollars would lower the magnitude of the CE ratios; however, the higher overall American health care costs would likely counterbalance this effect. For example, if the radiotherapy charges of $500 to $1,000 per fraction that are commonly encountered in the United States were used in the current evaluation, the savings associated with the use of CAP compared with BSC would increase from $6,171 per year of life gained to between $12,000 and $20,000 per year of life gained, and the cost of VP relative to BSC would fall from $14,700 per year of life gained to between $5,800 and $10,600 per year of life gained. Because both hospitals costed in this study were tertiary care university-affiliated institutions, their costs may not accurately represent those at a nonteaching general hospital. Many of the nonchemotherapy-related hospital admissions were at primary care institutions. It is difficult to accurately predict the difference in costs between the two types of institutions. If nonchemotherapy hospitalization costs were actually lower at these nonteaching institutions, the magnitude of the CE ratio of CAP to BSC would decrease slightly. There were major differences in the hospitalization practices for the administration of CAP chemotherapy at the two hospitals evaluated. Ontario patients received 65.2% of their CAP treatments as outpatients, whereas no patients at the AMEC received CAP chemotherapy on an outpatient basis. This was due to a policy decision that CAP be given on an inpatient basis at AMEC and not because of medical differences between the two groups of patients. A sensitivity analysis demonstrated that the CE of CAP relative to BSC became increasingly favorable as an increasing proportion of CAP treatments was
given on an outpatient basis. Similar observations have been made by others.9'5' 6 This observation emphasizes the importance of reducing hospitalization rates, particularly those related to chemotherapy administration, to improve the CE of a given therapy. The Ontario Cancer Registry, which provided the hospitalization data for all the patients treated in Ontario, is a central registry that is prospectively maintained. Its accuracy is high 9 and the data on hospitalization rates for Ontario patients used in this economic evaluation are considered accurate. The hospitalization data collected for patients treated at the AMEC were obtained by a review of the AMEC charts and the charts prepared by the clinical trials nurse at that institution and not from a central registry. While the chemotherapy hospitalization data thus obtained are probably accurate, the nonchemotherapy data are subject to error, and the error may have occurred to a different extent in each arm of the study. The importance of collecting accurate data, particularly of a high-cost resource such as hospitalization, is demonstrated by the effect on the estimated CE ratios of using the hospitalization rates determined in Ontario (an accurate data source) versus those at AMEC. The cost saving associated with the use of CAP chemotherapy compared with a policy of BSC was almost twice as great at PMH than at AMEC, an effect that was due mainly to differences in estimated hospitalization rates for chemotherapy administration. The prospective collection of complete data relating to hospitalization rates in future clinical trials would facilitate economic evaluations of those trials. The purpose of conducting economic evaluations is to compare costs and benefits of different programs so that decisions concerning the allocation of scarce resources can be made. Comparison of programs included in one evaluation is reasonable. However, comparison of programs across different economic evaluations should be done cautiously as differing methodologies used in the evaluations may render their results less than fully comparable. With this caveat in mind, the results of previously published economic evaluations, which used a health outcome of life years gained, may be compared with the current evaluation. The use of CAP chemotherapy compared with BSC has a CE ratio that is comparable to reported estimates
Downloaded from ascopubs.org by 80.82.77.83 on February 24, 2017 from 080.082.077.083 Copyright © 2017 American Society of Clinical Oncology. All rights reserved.
1308
JAAKKIMAINEN ET AL
of CE for treatments such as routine betaadrenergic-antagonist therapy for medium-risk and high-risk hypertensive patients,17 routine digital rectal examination to screen for rectal carcinoma,is or alternating chemotherapy for small-cell lung cancer. 9 The use of VP chemotherapy in comparison to BSC was somewhat more costly, with estimates of CE comparable to those reported for coronary artery bypass graft surgery,"9 home dialysis or cadaveric renal transplantation for renal failure, 20 therapeutic leukocyte transfusion in neutropenic patients,21 or sigmoidoscopy as a screening procedure for rectal carcinoma in asymptomatic patients aged 50 to 80 years. 22 An important observation made in this study was that a policy of supportive care for patients with advanced NSCLC was associated with substantial costs. The BSC patients in this trial had higher hospitalization rates, and they received more palliative radiotherapy treatments than those who received chemotherapy, leading to increased costs. This indicates that the use of a chemotherapy treatment regimen that prolongs survival, even when given with palliative intent, may significantly reduce health care costs. Treatment policies for NSCLC patients should address the economic impact of providing palliative care for these patients, whether this includes chemotherapy or not. The common occurrence of NSCLC and the frequency with which it is unresectable underscore the importance of consideration of the costs of palliative treatment of patients with this malignancy. Centers that have a policy of supportive care without the use of chemotherapy may incur significantly increased costs to the hospital-based health care system than those using chemotherapy. This is particularly true if chemotherapy can be given on an outpatient basis. If supportive care is administered at centers other than the cancer treatment center, these costs may be shifted to other hospitals, but they are still incurred by the health care system as a whole and should not be ignored. It would be tempting to argue that the differ-
ences in hospitalization rates among the three arms may reflect differences in the quality of life and that chemotherapy improved quality of life because it resulted in hospitalization rates that were lower than when chemotherapy was not given (particularly on the CAP arm). It must be emphasized that no formal quality of life indicators were available in the randomized trial, and this evaluation cannot be considered a quality of life assessment of patients on the BR-5 study. If treatment decisions or policy development are to be based on quality of life factors, it is important that these be measured prospectively using appropriate techniques. The incorporation of quality of life measurements in future clinical trials in this area is needed to address this issue. The results of this economic evaluation are intended to provide information for policy makers who must decide how scarce health care resources are to be allocated. Economic factors should not be used alone when these policy decisions are made but should be considered along with other factors, including treatment efficacy and feasibility, the CE of competing programs and societal preferences. Treatment decisions regarding individual patients are clearly influenced by treatment policies, but they are made primarily after consideration of each individual patient's situation and treatment preferences after discussion between physician and patient. They should not be made solely on the basis of economic factors. The conclusions of this economic evaluation simply indicate that economic factors should not adversely affect decisions regarding the use of chemotherapy in patients with advanced NSCLC. Similarly, these conclusions should not be used to support the use of chemotherapy in situations where it has not been demonstrated to be effective. ACKNOWLEDGMENT We want to express our appreciation to Anika Kind at The Princess Margaret Hospital and Dr Ed Kostashuk, Ruth Page, and Don Morrison at the A. Maxwell Evans Clinic for their assistance with this study.
REFERENCES 1. Rapp E, Pater JL, Willan A, et al: Chemotherapy can prolong survival in patients with advanced non-small cell lung cancer-Report of a Canadian multicenter randomized trial. J Clin Oncol 6:633-641, 1988 2. Weinstein MC: Cost-effectiveness analysis for clinical
procedures in oncology. Bull Cancer (Paris) 67:491-500, 1980 3. Drummond MF, Stoddart GL, Torrance GW: Methods for the Economic Evaluation of Health Care Programmes. New York, NY, Oxford University Press, 1987
Downloaded from ascopubs.org by 80.82.77.83 on February 24, 2017 from 080.082.077.083 Copyright © 2017 American Society of Clinical Oncology. All rights reserved.
1309
COST OF CHEMOTHERAPY IN NSCLC 4. Weinstein MC, Stason WB: Foundations of costeffectiveness analysis for health and medical practices. New Engl J Med 296:716-721, 1977 5. Statistics Canada: Consumer price index: Health and personal care. Ottawa, Canada 14:36, 1989 6. Statistics Canada: Consumer price index: Health and personal care. Ottawa, Canada 10:35, 1985 7. Hull RD, Hirsh J, Sackett DL, et al: Cost-effectiveness of primary and secondary prevention of fatal pulmonary embolism in high-risk surgical patients. Can Med Assoc J 127:990-995, 1982 8. Horngren CT: Cost accounting. A managerial emphasis (ed 5). Englewood Cliffs, NJ, Prentice-Hall, 1982, pp 43-78, 363-402, 438-474, 475-530 9. Goodwin P, Feld R, Evans WK, et al: Cost effectiveness of cancer chemotherapy: An economic evaluation of a randomized trial in small-cell lung cancer. J Clin Oncol 6:1537-1547, 1988 10. Shepard DS, Thompson MS: First principles of costeffectiveness analysis in health. Pub Health Rep 96:535-543, 1979 11. Atherton L: The cost of radiotherapy treatments on a linear accelerator. Br J Radiol 57:106-107, 1984 12. Ilbery PLT: Cost of health care services in the treatment of cancer. Aust Fam Phys 7:1062-1074, 1978 13. Greene DR: The cost of radiotherapy treatments on a linear accelerator. Br J Rad 56:189-191, 1983
14. Wodinsky HB, Jenkin RDT: The cost of radiation treatment at an Ontario regional cancer centre. Can Med Assoc J 137:906-909, 1987 15. Friedlander ML, Tattersall MHN: Counting the costs of cancer therapy. Eur J Cancer Clin Oncol 18:1237-1241, 1982 16. Calman JC, McVie JG, Soukop M, et al: Cost of outpatient chemotherapy. Br Med J 1:493-494, 1978 17. Goldman L, Sia B, Cook EF, et al: Cost and effectiveness of routine therapy with long-term beta-adrenergic antagonists after acute myocardial infarction. N Engl J Med 319:152-157, 1988 18. Love RR, Fryback DG, Kimbrough SR: A costeffectiveness analysis of screening for carcinoma of the prostate by digital examination. Med Decis Making 5:263278, 1985 19. Pliskin JS, Stason WB, Weinstein MC, et al: Coronary artery bypass graft surgery: Clinical decision making and cost-effectiveness analysis. Med Decis Making 1:10-28, 1981 20. Roberts SD, Maxwell DR, Gross TL: Cost-effective care of end-stage renal disease: A billion dollar question. Ann Intern Med 93:499-500, 1980 21. Rosenshein MS, Farewell VT, Price TH, et al: The cost-effectiveness of therapeutic and prophylactic leukocyte transfusion. N Engl J Med 302:1508-1062, 1980 22. Weinstein MC: Estrogen use in postmenopausal women--Costs, risks, and benefits. N Engl J Med 303:308316, 1980
Downloaded from ascopubs.org by 80.82.77.83 on February 24, 2017 from 080.082.077.083 Copyright © 2017 American Society of Clinical Oncology. All rights reserved.
