Special Article
Efficacy and Cost Effectiveness of Adjuvant Chemotherapy in Women with Node-Negative Breast Cancer — A Decision-Analysis Model
N Engl J Med 1991; 324:160-168January 17, 1991
- Abstract
- Abstract
Background.
In 1988 the National Cancer Institute issued a Clinical Alert that has been widely interpreted as recommending that all women with node-negative breast cancer receive adjuvant chemotherapy. Acceptance of this recommendation is controversial, since many women who would not have a recurrence would be treated.
Methods.
Using a decision-analysis model, we studied the cost effectiveness of chemotherapy in cohorts of 45-year-old and 60-year-old women with node-negative breast cancer by calculating life expectancy as adjusted for quality of life. The analysis evaluated different scenarios of the benefit of therapy: improved disease-free survival for five years, with a lesser effect on overall survival (base line); a lifelong benefit from chemotherapy; and a benefit in disease-free survival with no change in overall survival by year 10. The base-line analysis assumed a 30 percent reduction in the relative risk of recurrence for five years after treatment.
Results.
For the 45-year-old woman, the base-line analysis found an average lifetime benefit from chemotherapy of 5.1 quality-months at a cost of $15,400 per quality-year. The 60-year-old women gained 4.0 quality-months at a cost of $18,800 per quality-year. Under the more and less optimistic scenarios, the benefit of chemotherapy varied from 1.4 to 14.0 quality-months for both groups.
Conclusions.
Chemotherapy substantially increases the quality-adjusted life expectancy of an average woman at a cost comparable to that of other widely accepted therapies. This benefit decreases markedly if the changes in long-term survival are less than in disease-free survival. Given its uncertain duration, the benefit may be too small for many women to choose chemotherapy. Selective use of chemotherapy to maximize the benefit to individual patients may be possible with refinements in risk stratification and explicit assessment of the patients' risk preferences. (N Engl J Med 1991; 324:160–8.)
Media in This Article
Figure 1
Abbreviated Version of the Markov Model Used in the Study.Figure 2
Alternative Models of the Duration of the Survival Benefit of Treatment in Patients with Cancer.Article Activity
- Article
IN 1990 breast cancer affected more than 150,000 women in the United States; it will eventually be the cause of death of 25 percent of them.1 The 1985 Consensus Development Conference on Breast Cancer2 and the 1988 Clinical Alert3 affected both the approximately 75,000 women with node-negative breast cancer and the health care professionals who care for them. The Clinical Alert stated that "chemotherapy can have a meaningful impact on the natural history of node-negative breast cancer patients," which was widely interpreted as a recommendation for adjuvant therapy in these women,4 although controversy has continued over the treatment of such patients who are not in clinical trials.5 , 6
The effectiveness of adjuvant chemotherapy in prolonging disease-free survival in women with node-negative breast cancer is measurable but moderate.7 8 9 10 The effectiveness of adjuvant therapy in prolonging overall survival is less clear.11 , 12 Chemotherapy causes a short-term decrease in the quality of life of most patients, and this must be weighed against a possible chance of prolonged life for the few who will benefit. The chance of dying because of adjuvant therapy is low, but most treated patients have side effects, and the estimated cost of chemotherapy is high — $338 million for the entire group.6
The purpose of this study was to explore the effectiveness of adjuvant therapy in women with node-negative breast cancer at different levels of risk of recurrence. We created a model that incorporated the variables of the risk of recurrence, efficacy of adjuvant therapy, duration of benefit from adjuvant therapy, and quality of life. We used decision analysis to evaluate the clinical model, perform sensitivity analyses, and define thresholds that would assist clinicians in choosing therapy.13
Methods
Basic Model
We stated the study problem as follows: a woman has undergone surgery for Stage I or lia breast cancer that is node negative and estrogen-receptor negative. Should she receive chemotherapy? The decision to accept therapy is based on the patient's and her physician's perceptions of benefit, which depend on the patient's prognosis according to clinical variables, the risk and benefit of chemotherapy, and the patient's preference about the prospect of treatment toxicity.
We developed a decision-analysis model involving two groups: 45-year-old premenopausal women and 60-year-old postmenopausal women. Decision analysis uses a decision tree to consider all the available options and their possible outcomes systematically in solving a problem. The methodologie background has recently been reviewed.13 14 15
These groups represent two typical kinds of patients: premenopausal women who commonly are advised to receive chemotherapy, and postmenopausal, older women who have the highest incidence of breast cancer. Our model used a Markov process to calculate the cumulative value of outcomes in each cohort with and without adjuvant therapy. A Markov process is a modeling technique used for conditions in which the prognosis is described by a series of chance events (adverse outcomes), and the value of these outcomes depends on whether and when they occur.16 A Markov process has been useful for modeling diseases in which the same event — the occurrence of osteoporotic fractures17 or thromboemboli,16 , 18 for example — can happen repeatedly over time.
Figure 1Figure 1
Abbreviated Version of the Markov Model Used in the Study. shows an abbreviated version of the Markov model of prognosis that we used. The figure shows only four of the nine health states used in the model. "Health states" are part of the Markov vocabulary and describe defined categories of health or disease that apply to a person for a finite period of time. time-dependent probabilities determined how the patients in our study moved from state to state. All the patients began in the well state. They were tracked through the defined health states at annual intervals. We counted them each year by taking a computer "snapshot" of each woman, and we assigned credits according to the number of women in each health state. We adjusted the credits to reflect the quality of life in each health state (Table 1Table 1
Health States in the Study Model and Their Incremental Values.*). For example, a patient who had her first recurrence of breast cancer during the year contributed only 70 percent of a year to the cohort's net credits. The model was run until the entire cohort died or reached the age of 90. We calculated the average number of years of life, as adjusted for quality, for each cohort. A quality-adjusted year of life was a numerical description that combined expected survival and expected quality of life; it was based on the value people place on life in a given state of health. The difference between the groups was thus a measure of the benefit of treatment in terms of quality-adjusted life expectancy.For each of the two age groups, one cohort received chemotherapy and one did not. Similar Markov processes were used for the four cohorts; only the probability of toxic effects of chemotherapy in the first year and the probability of a first recurrence of breast cancer differed (Table 1). All the patients were assumed to be without metastatic disease at the outset. In each subsequent year a woman could remain free of recurrence (well) or have her first recurrence of breast cancer. If a first recurrence occurred, the patient might respond to salvage therapy (thus entering the post-first-recurrence state) or die. Once a patient had a first recurrence, returning to the well state was impossible. During the year of a first recurrence, the patient could die or survive (first recurrence). If the patient survived, during the next year the patient could continue to respond to salvage therapy given during the first recurrence (thus entering the post-first-recurrence state) or could have a second recurrence. In the year of the second recurrence, the patient could die or survive (second recurrence). We considered a maximum of three recurrences. Almost all patients died within four years. The patients who died in the year of a recurrence contribute nothing to their cohort's net credits.
During the first year, the cohorts that received chemotherapy might have complications associated with treatment — minor toxicity, major toxicity, or death. Minor toxicity was defined as severe nausea and vomiting or weakness sufficient to require a reduction in the activities of daily living but not hospitalization, and major toxicity as complications sufficient to require hospitalization. These health states represented short-term morbidity; with no recurrence of breast cancer, the patient returned to being well in the model's next year.
Modeling the Duration of Benefit from Chemotherapy
A fundamental question is, What is the long-term prognosis for patients treated with chemotherapy? The majority of studies on adjuvant chemotherapy for node-negative cancer have found increases in disease-free survival but no effect on overall survival at the end of the study period. Many oncologists believe that these results are due to insufficient follow-up and that if patients were followed for a sufficient period, a commensurate increase in overall survival would be evident. This belief is based on the assumption that some patients are permanently cured.19 It is possible, however, that adjuvant therapy delays but does not prevent the eventual recurrence of breast cancer. A tumor might respond to chemotherapy to the extent that it becomes clinically undetectable, but eventually the surviving cells start to grow again, producing a clinical recurrence. Since tumors may be kinetically as well as biochemically heterogeneous, both these outcomes may occur. If so, the initial benefit in disease-free survival would eventually decrease, but a benefit in overall survival would remain.
We repeated our analyses with each of the following assumptions: chemotherapy increases disease-free survival but has less effect on overall long-term survival, chemotherapy has a lifelong benefit in altering both disease-free and overall survival, and chemotherapy increases disease-free survival but does not change overall survival after 10 years. Survival curves for four hypothetical cohorts of patients with breast cancer with an annual probability of recurrence of 10 percent and a relative efficacy of initial treatment of 50 percent are shown in Figure 2Figure 2
Alternative Models of the Duration of the Survival Benefit of Treatment in Patients with Cancer.. Relative efficacy or reduction in relative risk is the annual decrease in the risk of cancer. The absolute or total reduction in risk would be lower; for the first year it would be 5 percent (10 percent in the untreated group as compared with 5 percent if all patients were treated).20 With the assumption that the benefit of therapy is limited to the first five years, the risk of recurrence was reduced by 50 percent in the first five years but unchanged in years 6 to 10. The model of maximal benefit assumed that treatment is effective indefinitely. To model an initial benefit in disease-free survival but no change in long-term survival required an increased risk of recurrence in later years. We therefore increased the annual probability by 20 percent for years 6 to 10. With this assumption, the treated groups had a greater average life expectancy due to the benefits of therapy in the first five years, but the number of survivors at year 10 was unchanged. Figure 2 shows that the net benefit of treatment, which is represented by the area between any of the treatment curves and the no-treatment curve, varied markedly between assumptions.Which approach should be used? Long-term trials in patients with node-positive disease have consistently shown greater benefits of chemotherapy in increasing disease-free survival and smaller benefits in improving long-term overall survival.21 Since trials7 8 9 10 in patients with node-negative disease have shown benefits only in disease-free survival, we assumed no benefit in long-term overall survival for years beyond the years of follow-up in published reports. We therefore limited the survival benefit of chemotherapy to the first five years after treatment, and we considered the alternative approaches in our sensitivity analyses. A sensitivity analysis is a test of the stability of the results over a range of clinically relevant estimates of probability, structural assumptions, or value judgments. Our other assumptions included the following: that chemotherapy alters only the probability of a first recurrence of breast cancer, making the probability of second and third recurrences the same for both cohorts; that all patients receiving chemotherapy complete the treatment, receive full therapeutic doses, and have no long-term disability (unless there is a fatal complication); that the chemotherapy consists of six cycles of treatment with cyclophosphamide, methotrexate, and fluorouracil2; that no patient has a second primary breast cancer; and that this combination of chemotherapy does not increase the risk of secondary tumors.22
Summary of the Data
Natural History of Node-Negative Breast Cancer
The risk of recurrence of breast cancer depends on the biologic characteristics of the tumor and is independent of the patient's age (Table 2Table 2
Probability and Cost of Recurrence, Complications of Chemotherapy, and Death in Women with Node-Negative Breast Cancer.).4 , 11 For the typical patient, the annual probability of a first recurrence is 2 to 7 percent.23 , 24 , 30 Our estimated probability of recurrence of 4 percent per year corresponded to a rate of diseasefree survival of 80.4 percent or a rate of recurrence of 19.6 percent after five years. In the sensitivity analysis, we varied the probability of recurrence from 1 to 10 percent per year.10 , 23 , 25 After the first recurrence of cancer, the rate of response to salvage chemotherapy or hormonal therapy varies with the location of the recurrence but increases long-term survival very little.26 Most patients receive a palliative benefit, but median survival is about two years.27 28 29 Our probability estimates for second and third recurrences were based on a review of the literature and on expert estimates.27 , 28 , 30 The salvage chemotherapy in our model was cyclophosphamide, doxorubicin, and fluorouracil.27
Effect and Toxicity of Chemotherapy
We modeled the efficacy of chemotherapy in reducing the risk of a first recurrence of breast cancer by decreasing the probability of recurrence by 30 percent per year for the first five years, on the basis of recently reported trials.7 8 9 10 We limited the duration of benefit because of the experience in patients with node-positive disease, in whom initial disease-free survival exceeded long-term survival.21 The probability of a first recurrence was therefore 2.8 percent per year in the treated group and 4 percent per year in the untreated group for each of the first five years after therapy.20 Thereafter, the probability of a first recurrence was the same (4 percent per year). The net effect of chemotherapy after 10 years was thus an absolute gain of about 12 percent in survival. As noted earlier, the toxic effects of chemotherapy could occur only during the year in which it was received. Minor and major toxicity reduced a patient's quality of life and required unscheduled visits to a physician, but only major toxicity required hospitalization. For rates of death from other causes, we used annual age-specific mortality rates for white women in the United States.31
Quality of Life
Explicit consideration was given to changes in health or quality of life associated with recurrences of breast cancer and complications of adjuvant therapy. The appropriate way to determine a patient's preferences is controversial, and different methods provide inconsistent results.32 , 33 Despite these limitations, there are gross differences in the level of well-being during the stages of treatment and the natural history of breast cancer. We assigned an incremental value for quality of life to each health state. The values shown in Table 1 were determined by a survey of our oncology staff. The adjustments in quality of life with adjuvant therapy apply only to the year in which chemotherapy was given.
Cost of Treatment
The financial cost of therapy varied with the patient's state of health (Table 2). We based costs on 1989 charges at our medical center for physician, laboratory, office, and hospital services and on estimates from Medicare data29; our figures were similar to other published estimates.6 Indirect costs due to loss of earnings and the cost of treatment for other medical conditions were excluded. The cost of chemotherapy was based on six months of outpatient therapy. There were additional costs associated with minor toxicity, among them more office visits, laboratory studies, and drug therapy. The cost of major toxicity was based on a five-day hospitalization plus additional follow-up care. Costs for both cohorts included the annual expenses of scheduled visits and mammograms. Annual costs of treatment assumed outpatient care for nonfatal first recurrences and seven days of hospitalization for nonfatal second or third recurrences. Costs for terminal care varied with the number of recurrences and were greater if death occurred in the year of first recurrence. These aggregate costs were used to determine the cost effectiveness of chemotherapy, defined as the cost per quality-adjusted year of life.15 , 34 We used a discount rate of 5 percent15 to account for the fact that health benefits gained and dollars spent in the future are worth less than health benefits and dollars in the present. We discounted both costs and benefits and varied the rate from 0 to 8 percent in the sensitivity analyses. Using a discount rate of 0 percent and excluding adjustments for quality, we calculated the life expectancy of the average member of each cohort. All calculations were performed with the decision-analysis software program Smltree (Smltree 2.9.1989 [J.P. Hollenberg]).
Results
Base-Line Analysis
Table 3Table 3
Base-Line Analyses in Two Cohorts Receiving Chemotherapy, According to Three Assumptions of Benefit.* shows the results in the cohorts of 45- and 60-year-old women receiving chemotherapy for each of the three assumptions of the duration of efficacy of treatment. For example, a 45-year-old woman's average quality-adjusted life expectancy was 11.03 quality-years if given chemotherapy and 10.60 quality-years if not treated; the benefit was therefore 0.43 quality-years or 5.1 quality-months. The incremental cost per quality-adjusted year was 115,400. The average life expectancy after chemotherapy was 19.5 years, a gain of 11.0 months over life expectancy without treatment. (For comparison, the average life expectancy of a 45-year-old woman without breast cancer is 35.8 years.) For a 60-year-old woman given chemotherapy, the quality-adjusted life expectancy was 9.49 quality-years, a benefit of 4.0 quality-months at an incremental cost of $18,800 per quality-adjusted year. The treated women had an average life expectancy of 14.7 years, an increase of 7.7 months. This compares with an average life expectancy of 22.5 years for 60-year-old women without breast cancer.Sensitivity Analyses
Sensitivity analyses for plausible ranges of rates of first recurrence of breast cancer and the relative efficacy of chemotherapy in the 45-year-old cohorts are shown in Figure 3Figure 3
Sensitivity Analyses in the 45-Year-Old Cohorts.. The benefit curve intersects the horizontal axis at the threshold point, the finite point at which two strategies have equal benefits or values for any or all variables.12 By definition, the cost-effectiveness ratio at a threshold point is infinite. An incremental cost-effectiveness ratio was determined at probabilities of recurrence or efficacies of chemotherapy greater than the threshold levels.As Figure 3A shows, if the annual probability of recurrence is above 0.4 percent, the use of chemotherapy will be beneficial. The net benefit in quality-years gradually increases in a nonlinear curve as the probability of recurrence increases. Although therapy remains effective at higher risks of recurrence, the increases in the size of the benefit decline. This is due to the limited life expectancy in which to overcome the initial quality-of-life penalty that chemotherapy imposes. Cost effectiveness changes minimally for probabilities of recurrence higher than 5 percent per year, and it plateaus at $9,000 per quality-year. For the 60-year-old cohorts the results were similar, but the absolute benefit for any probability of recurrence was lower and the costs were higher.
Figure 3B illustrates the results of changes in the relative efficacy of chemotherapy. The minimal efficacy that will produce a benefit or threshold point is 4.2 percent. The cost-effectiveness curve is very steep at low estimates of efficacy. If a conservative relative efficacy of 15 percent is used, the benefit decreases to about two quality-months at an incremental cost of $36,800 per quality-year.12 With the best results of recent trials, a 45 percent relative efficacy,10 the benefit is eight quality-months and the incremental cost drops to $9,900 per quality year.
Figure 4Figure 4
Structural Sensitivity Analyses. shows the anticipated benefits of chemotherapy in the 45-year-old cohorts, given different assumptions about the duration of benefit. If chemotherapy is curative in a subset of patients, then that group will have a lifelong benefit in disease-free survival and overall survival. We modeled this optimistic view by allowing the base-line 30 percent reduction in the risk of recurrence to apply for the patient's lifetime. As Figure 4 shows, the threshold for a benefit of therapy is a 2 percent relative reduction in the rate of recurrence; at the base-line relative efficacy of 30 percent the benefit is 14.0 quality-months. An alternative point of view is that chemotherapy is not curative and all women with micrometastatic disease will eventually relapse. We modeled this less optimistic view by increasing the rate of recurrence in years 6 to 10 so that the survival curves for treated and untreated women meet after 10 years and show no further survival benefit. Patients treated with adjuvant therapy still have a slightly increased life expectancy, since the rate of recurrence increases only after year 5. With this less optimistic assumption, chemotherapy is beneficial only if the relative efficacy exceeds 10 percent; at the base-line relative efficacy of 30 percent the benefit is about 1.7 quality-months. The reason for this small benefit is that the penalty in reduced quality of life associated with chemotherapy is immediate, whereas the benefit in improved disease-free survival occurs later and is discounted.Since there is disagreement about the correct estimates for several of the key variables used in our analyses, we performed sensitivity analyses using best-case and worse-case combinations of recurrence and treatment efficacy (Table 4Table 4
Sensitivity Analyses of the Cost and Quality-Adjusted Benefit of Chemotherapy.*). In addition, risk preferences may vary greatly among women. We therefore assessed various combinations of discount rates and quality-of-life decrements associated with chemotherapy (Table 4).Patients' Questions
Our decision-analysis model allows doctors and patients to explore relevant clinical issues even when data from clinical trials are not complete. We posed these in the form of questions that women might ask their doctors. For the answers to each of the following questions we first assumed our base-line estimates, in which the risk of recurrence was 4 percent per year and chemotherapy reduced the risk by 30 percent for the first five years and improved survival by 12 percent at year 10.
I am willing to accept the initial toxicity if you can add one or two years to my life with chemotherapy. Can you do that? Probably. The question offers another way of quantifying the decrease in quality of life due to chemotherapy. Women who have undergone chemotherapy with cyclophosphamide, methotrexate, and fluorouracil have said that maybe one and certainly two added years would be sufficient benefit to warrant undergoing therapy again.35 Whether the efficacy of chemotherapy meets these thresholds depends on the duration of the benefit. Given our base-line assumption that the benefit lasts for five years after treatment, chemotherapy must be at least 33 percent effective for 45-year-old women and 47 percent effective for 60-year-old women. With the optimistic assumption that the benefit is lifelong, the minimal efficacy for 45-year-olds drops to 11 percent for one year and 20 percent for two years. Current reports of chemotherapy for women with node-negative breast cancer show relative benefits in disease-free survival of 15 to 45 percent after three to five years7 , 9 , 10; the long-term survival benefit is unknown but is probably smaller.11 , 12 , 21 If early disease-free survival does not translate into a later survival advantage, chemotherapy may not provide the magnitude of benefit most women require in choosing it.
If quality-of-life issues are important to me, during both chemotherapy and a possible recurrence, and if I accept that the present has greater value than the future (the principle of discounting), can chemotherapy add one year to my life? Probably not. For a 45-year-old woman, chemotherapy must be at least 63 percent effective to gain an additional quality-adjusted year. As discussed above, the reported benefits of adjuvant therapy are substantially below these thresholds. The duration of benefit has a dramatic effect on the thresholds. Given the optimistic view of a lifelong benefit, the thresholds for a gain of one quality-adjusted year drop to 27 percent relative efficacy for 45-year-olds, which is less than our initial estimate, and 38 percent for 60-year-olds, which has been achieved in some studies.
Is chemotherapy effective enough to increase by 10 percent my chance of being disease-free five years from now? Yes, but only for women at high risk. Adjuvant chemotherapy must be more than 55 percent effective to increase life expectancy by 10 percent after five years. Simes et al. found that most women would accept the toxicity of adjuvant chemotherapy if the benefit were a 10 percent greater chance of being alive and disease-free after five years.35 This absolute gain of 10 percent is an unrealistic goal for a woman at the standard risk of recurrence of 4 percent per year, since the chance of recurrence or death in five years is about 20 percent. For a woman at high risk, however, a 10 percent absolute gain may be achievable. If risk stratification could identify women at high risk of recurrence (10 percent per year, or about 50 percent after five years), adjuvant therapy would still have to be at least 30 percent effective in 45-year-olds and 37 percent effective in 60-year-olds to achieve a 10 percent absolute gain.
If my breast cancer is small (≤1 cm) and has good prognostic factors, will chemotherapy be of any benefit? Yes, but very little. Risk stratification based on tumor size and other prognostic factors has become increasingly feasible and advocated.36 Tumors no larger than 1 cm have a particularly good prognosis — about 10 percent recurrence after 10 years.25 We therefore assumed an annual risk of recurrence of 1 percent. Chemotherapy provides a benefit of 38 quality-days for 45-year-old women and 27 quality-days for 60-year-old women, at a cost of $65,000 to $91,000 per quality-year. Chemotherapy is still beneficial because of the very low annual chance of death from other causes. As a group, 45-year-old women have an expected chance of death of only 2 per 1000 in the next year. Even a low-risk tumor, if untreated, thus increases the chance of dying three- to fivefold.
If chemotherapy does not cure me, but just delays recurrence, should I still undergo it? Probably, especially if the risk of recurrence is average or high. This worst-case scenario explores the ramifications if early disease-free survival does not translate into a long-term survival benefit. There may be an overall benefit if the disease-free years outweigh the early toxicity. However, each woman's risk preferences have a great effect on this balance. To make informed choices, women who are willing to risk the early toxicity must know how effective chemotherapy is likely to be in their case. Given base-line adjustments in quality at standard risks of recurrence for both 45- and 60-year-old white women, chemotherapy must be at least 10 percent effective in delaying the recurrence of disease to offer any benefit at all. As Table 4 shows, different risk preferences and rates of discounting markedly affect the decision.
Discussion
Our analysis shows that there is a definite benefit of chemotherapy in women with node-negative, estrogen-receptor—negative disease, if we assume that all women are treated and that the average risk of recurrence is 4 percent per year. For a premenopausal woman of 45, life expectancy would be increased by 11 months; adjusted for quality of life and discounted, the benefit would be about 5 quality-months. For a postmenopausal woman of 60, the benefit would be only about 8 months; adjusted, the benefit would be only 4 quality-months. The financial cost of gaining an additional quality-year of life would be approximately $15,400 in the premenopausal group and $18,800 in the postmenopausal group.
What are the implications of these additional days and dollars to women with breast cancer, their physicians, insurance companies, and policy makers? Women with breast cancer generally choose adjuvant therapy over the uncertainty of observation. One preliminary report found that nearly 90 percent of the women who had already undergone adjuvant chemotherapy with cyclophosphamide, methotrexate, and fluorouracil indicated that either of two conditions would be sufficient enticement to choose adjuvant therapy again35: an additional two years of life expectancy or a 10 percent increase in the chance of being alive after five years. Women who were married or had dependent children were more likely to choose chemotherapy. Unfortunately, women faced with the decision about adjuvant therapy may not receive full information about risks and benefits. In one carefully constructed study of the decision about adjuvant therapy in node-positive breast cancer (in which the benefit in survival is established), patients markedly overestimated their chance of cure: 60 percent of the patients overestimated their chance of cure by 20 percent or more, and nearly 50 percent overestimated their chance of cure even when given numerical values.37 This suggests that women will overestimate their chance of benefit and choose chemotherapy even when the benefit is known to be small. What will be the choice when the benefit, in terms of long-term survival, is unknown? We suspect (although data are not available to answer the question) that most women, and particularly younger women with dependents, will overestimate their benefit and underestimate toxicity38; most will choose chemotherapy if the toxicity is acceptable.39 Our analysis suggests that for women at a 4 percent risk of annual recurrence, the efficacy of chemotherapy necessary to gain one quality-adjusted year is 62 percent, far greater than the currently attainable level.
What is the proper role of the physician in counseling women with node-negative breast cancer about adjuvant therapy? Physicians usually supply the patient with a clear recommendation, and the patient usually follows it.37 Our perception is that oncologists are recommending chemotherapy for women with node-negative breast cancer even without a demonstrated survival benefit. Oncologists are prone to accept new forms of treatment as standard before their value has been well documented.39 The National Cancer Institute's Clinical Alert and the fear of litigation if chemotherapy is withheld appear to allow oncologists only two choices: give chemotherapy even if the benefit is small, or enroll the patient in a clinical trial. Since only a small fraction of eligible patients enter clinical trials and since physicians are reluctant to discuss frankly adverse outcomes (such as lack of benefit or risk of morbidity and mortality), we suspect that most patients will be offered chemotherapy — and most will accept.
Can physicians define a subgroup of women with breast cancer for whom chemotherapy is clearly of sizable benefit, so that only those women will bear the toxicity and cost? The issue is complex, since there is no perfect marker for a low risk of recurrence and no evidence that women at high risk have a better chance of responding to chemotherapy. Patients with a low risk of recurrence may be identifiable by tumor size, morphologic features, or DNA techniques. Tumors with the best prognosis (those less than 2 cm in diameter) may have only a 5 percent chance of causing death in the first five years.23 Expert morphologic and histologie analysis can identify a subgroup with an excellent prognosis, but this has not yet become standard practice.40 DNA analysis, carried out by flow cytometry and reported as ploidy and S-phase analysis, may allow the identification of a low-risk subset.41 Unfortunately, such analysis can be technically difficult, lacks a defined standard, and is not available at many institutions.
Is it possible for physicians to define a subset of women with breast cancer in whom chemotherapy will be most effective? Ideally, this would be the subset at highest risk of recurrence and death, probably those with large (>3 cm), estrogen-receptor-negative, aneuploid, or high-S-phase tumors. The minimal data that are available to answer this critical question suggest that patients whose tumors indicate a poor prognosis are no more intrinsically curable by chemotherapy than those whose tumors indicate a good prognosis. The single study that showed a correlation between the thymidine-labeling index and the response to chemotherapy showed no correlation between the thymidine-labeling index and survival.42
If chemotherapy is relatively ineffective in the group at highest risk of recurrence and death, then the benefit to that group is likely to be less than predicted. If we assume in our model that adjuvant therapy is equally effective in cancers with a high risk of recurrence, then therapy is most cost effective in patients at high risk. If chemotherapy is not curative and merely postpones recurrence with no change in long-term survival, then the benefit of therapy decreases markedly. Observation will produce more benefit than treatment unless the relative efficacy of therapy exceeds 10 percent.
When compared with the choices of women who have undergone adjuvant therapy with cyclophosphamide, methotrexate, and fluorouracil, our results suggest that the level of benefit is very close to the typical woman's threshold for choosing chemotherapy. If the optimistic perspective is supported by the results of long-term follow-up of subjects in reported trials, then most women should receive treatment. If these studies show no change in long-term survival, then the benefit appears to be insufficient. However, most women and their oncologists will choose treatment when faced with uncertainty and the overestimation of personal benefit as opposed to risk.37
What are the policy implications of our analysis of cost effectiveness for those who pay for cancer care? For the individual patient, the decision has been clear-cut: treat to prolong life despite the cost. In an era of limited health care resources, this approach is increasingly untenable, however. Those who pay for the care of patients with breast cancer might point to the high cost of gaining an additional quality-adjusted year of life in an attempt to discourage the standard use of chemotherapy in women with node-negative disease. The conundrum for cost-conscious physicians is that the expense of this therapy is within the range of current medical interventions. For both age groups of women, an incremental cost of less than $20,000 per quality-year would be generally acceptable. The true cost effectiveness may actually be lower, since we based the costs of chemotherapy on charge estimates. Other widely accepted medical interventions with costs per quality-adjusted year ranging between $10,000 and $25,000 include hemodialysis for end-stage renal disease, treatment of three-vessel coronary disease, bone marrow transplantation for acute non-lymphocytic leukemia, and treatment of hypertension.43 44 45 46 But if chemotherapy only delays recurrence, then the incremental cost increases markedly. Is it acceptable to spend more than $50,000 for one quality-year? Alternatively, is it acceptable to spend $6,000 for chemotherapy to gain less than two months of life? How much is society willing to spend for a small potential gain, since the effect of adjuvant therapy on survival is unknown? Should the potential benefit be limited to younger women, for whom it is more cost effective, even though a woman alive at the age of 60 may be expected to live another 22 years?
It is important to keep in mind that these decisions affect women who are faced with a frightening, disfiguring, and often fatal disease. It is also an expensive disease with increasingly costly options for treatment. Decisions about treatment that look at cost have always been a part — even if unspoken — of medical practice. Costs and risk preferences should be part of the discussion of the efficacy and toxicity of treatment in these and all other decisions made at the bedside.47
Source Information
From the Division of General Internal Medicine, Department of Medicine (B.E.H.), and the John N. Dalton Oncology Clinics, Massey Cancer Center (T.J.S.), Medical College of Virginia, Virginia Commonwealth University, Richmond. Address reprint requests to Dr. Hillner at the Division of General Internal Medicine, Box 170, MCV Station, Richmond, VA 23298.
- References
References
1
Silverberg E, Boring CC, Squires TS. Cancer statistics, 1990. CA 1990; 40:9–26.
2
Adjuvant chemotherapy for breast cancer . JAMA 1985; 254:3461–3.
CrossRef | Web of Science3
National Cancer Institute. Clinical alert. Bethesda, Md.: National Cancer Institute, May 16–18, 1988.
4
Ingle
Full Text | Web of Science | Medline5
DeVita
Full Text | Web of Science | Medline6
McGuire
Full Text | Web of Science | Medline7
Fisher
Full Text | Web of Science | Medline8
Early Breast Cancer Trialists' Collaborative Group
Full Text | Web of Science | Medline9
The Ludwig Breast Cancer Study Group
Full Text | Web of Science | Medline10
Mansour
Full Text | Web of Science | Medline11
Henderson
Full Text | Web of Science | Medline12
Ginsburg
Web of Science13
Pauker
Full Text | Web of Science | Medline14
Sox HC Jr, Blatt MA, Higgins MC, Marton KI. Medical decision making. Boston: Butterworths, 1988.
15
Weinstein MC, Feinberg HV, Elstein AS, McNeil B. Clinical decision analysis. Philadelphia: W.B. Saunders, 1980.
16
Beck
CrossRef | Web of Science | Medline17
Hillner
CrossRef | Web of Science | Medline18
Tsevat
CrossRef | Web of Science | Medline19
Norton
Web of Science | Medline20
Fletcher RH, Fletcher SW, Wagner EH. Clinical epidemiology: the essentials. 2nd ed. Baltimore: Williams & Wilkins, 1988.
21
Tannock
Web of Science22
Henderson
Web of Science | Medline23
Carter
CrossRef | Web of Science | Medline24
Fisher
Web of Science | Medline25
Rosen
Web of Science | Medline26
Henderson IC, Harris JR, Kinne DW, Hellman S. Cancer of the breast. In: DeVita VT Jr, Hellman S, Rosenberg SA, eds. Cancer: principles and practice of oncology. 3rd ed. Philadelphia: J.B. Lippincott, 1989.
27
Bonadonna
Web of Science | Medline28
Henderson
Web of Science | Medline29
Baker MS, Kessler LG, Smucker RC. Site-specific treatment costs for cancer: an analysis of the Medicare continuous history sample file. In: Scheffler RM, Andrew NC, eds. Cancer care and costs: DRGs and beyond. Ann Arbor, Mich.: Health Administration Press Perspectives, 1989:127–38.
30
Bonadonna
Web of Science | Medline31
Bureau of the Census. Statistical abstract of the United States, 1989. Washington, DC: Government Printing Office, 1989:73.
32
Nelson
CrossRef | Web of Science | Medline33
Mulley
CrossRef | Web of Science | Medline34
Detsky
CrossRef | Web of Science | Medline35
Simes 36
McGuire
CrossRef | Web of Science | Medline37
Siminoff
Web of Science | Medline38
Weinstein
CrossRef | Web of Science | Medline39
Goldsmith
Medline40
Fisher ER. Prognostic and therapeutic significance of pathological features of breast cancer. In: NCI Monographs. No. 1. Washington, DC: Government Printing Office, 1986:29–34. (NIH publication no. 86–2860.)
41
Clark
Full Text | Web of Science | Medline42
Sulkes
Web of Science | Medline43
Stange
Full Text | Web of Science | Medline44
Weinstein
Web of Science45
Welch
Full Text | Web of Science | Medline46
Edelson
CrossRef | Web of Science | Medline47
Durbin
Web of Science | Medline
- Citing Articles (59)
Citing Articles
1
Masahide Kondo, Shu-Ling Hoshi, Hiroshi Ishiguro, Masakazu Toi. (2012) Economic evaluation of the 70-gene prognosis-signature (MammaPrint®) in hormone receptor-positive, lymph node-negative, human epidermal growth factor receptor type 2-negative early stage breast cancer in Japan. Breast Cancer Research and Treatment
CrossRef2
H.E. Campbell, D. Epstein, D. Bloomfield, S. Griffin, A. Manca, J. Yarnold, J. Bliss, L. Johnson, H. Earl, C. Poole, L. Hiller, J. Dunn, P. Hopwood, P. Barrett-Lee, P. Ellis, D. Cameron, A.L. Harris, A.M. Gray, M.J. Sculpher. (2011) The cost-effectiveness of adjuvant chemotherapy for early breast cancer: A comparison of no chemotherapy and first, second, and third generation regimens for patients with differing prognoses. European Journal of Cancer 47:17, 2517-2530
CrossRef3
Masahide Kondo, Shu-Ling Hoshi, Takeharu Yamanaka, Hiroshi Ishiguro, Masakazu Toi. (2011) Economic evaluation of the 21-gene signature (Oncotype DX®) in lymph node-negative/positive, hormone receptor-positive early-stage breast cancer based on Japanese validation study (JBCRG-TR03). Breast Cancer Research and Treatment 127:3, 739-749
CrossRef4
David R Nerenz, Yung-wen Liu, Keoki L Williams, Kaan Tunceli, Huiwen Zeng. (2011) A simulation model approach to analysis of the business case for eliminating health care disparities. BMC Medical Research Methodology 11:1, 31
CrossRef5
Wen Chen, Zefei Jiang, Zhimin Shao, Qiang Sun, Kunwei Shen. (2009) An Economic Evaluation of Adjuvant Trastuzumab Therapy in HER2-Positive Early Breast Cancer. Value in Health 12, S82-S84
CrossRef6
Per E. Lønning, Jürgen Geisler. (2007) Aromatase inhibitors—Socioeconomical issues. The Journal of Steroid Biochemistry and Molecular Biology 106:1-5, 55-61
CrossRef7
Hirohisa Imai, Katsumasa Kuroi, Shozo Ohsumi, Michikazu Ono, Kojiro Shimozuma. (2007) Economic evaluation of the prevention and treatment of breast cancer-present status and open issues. Breast Cancer 14:1, 81-87
CrossRef8
J Cassidy, J-Y Douillard, C Twelves, J J McKendrick, W Scheithauer, I Bustová, P G Johnston, K Lesniewski-Kmak, S Jelic, G Fountzilas, F Coxon, E Díaz-Rubio, T S Maughan, A Malzyner, O Bertetto, A Beham, A Figer, P Dufour, K K Patel, W Cowell, L P Garrison. (2006) Pharmacoeconomic analysis of adjuvant oral capecitabine vs intravenous 5-FU/LV in Dukes' C colon cancer: the X-ACT trial. British Journal of Cancer 94:8, 1122-1129
CrossRef9
Reinhold Kilian, Franz Porzsolt. (2005) When to recommend and to pay for first-line adjuvant breast cancer treatment? A structured review of the literature. The Breast 14:6, 636-642
CrossRef10
Bruce E. Hillner, Deborah Schrag, Daniel J. Sargent, Charles S. Fuchs, Richard M. Goldberg. (2005) Cost-effectiveness projections of oxaliplatin and infusional fluorouracil versus irinotecan and bolus fluorouracil in first-line therapy for metastatic colorectal carcinoma. Cancer 104:9, 1871-1884
CrossRef11
Arash Naeim, Emmett B. Keeler. (2005) Is Adjuvant Therapy for Older Patients with Node (+) Early Breast Cancercost-effective?*. Breast Cancer Research and Treatment 94:2, 95-103
CrossRef12
Nina Oestreicher, Scott D. Ramsey, Hannah M. Linden, Jeannine S. McCune, Laura J. van???t Veer, Wylie Burke, David L. Veenstra. (2005) Gene expression profiling and breast cancer care: What are the potential benefits and policy implications?. Genetics in Medicine 7:6, 380-389
CrossRef13
Per E. Lønning, Jürgen Geisler. (2005) Aromatase inhibitors—Socio-economical issues. The Journal of Steroid Biochemistry and Molecular Biology 95:1-5, 137-142
CrossRef14
Arash Naeim, Emmett B. Keeler. (2005) Is adjuvant therapy for older patients with node (−) early breast cancer cost-effective?. Critical Reviews in Oncology/Hematology 53:1, 81-89
CrossRef15
Michael W. Kattan, Dilip Giri, Katherine S. Panageas, Amanda Hummer, Milicent Cranor, Kimberly J. Van Zee, Clifford A. Hudis, Larry Norton, Patrick I. Borgen, Lee K. Tan. (2004) A tool for predicting breast carcinoma mortality in women who do not receive adjuvant therapy. Cancer 101:11, 2509-2515
CrossRef16
Bruce E. Hillner. (2004) Benefit and projected cost-effectiveness of anastrozole versus tamoxifen as initial adjuvant therapy for patients with early-stage estrogen receptor-positive breast cancer. Cancer 101:6, 1311-1322
CrossRef17
Richard R. Love, Susan M. Love, Adriano V. Laudico. (2004) Breast Cancer from a Public Health Perspective. The Breast Journal 10:2, 136-140
CrossRef18
T. D. Szucs, P. Wyss, K. J. Dedes. (2003) Cost-effectiveness studies in ovarian cancer. International Journal of Gynecological Cancer 13:s2, 212-219
CrossRef19
M. Jefford, M. R. Stockler, M. H. N. Tattersall. (2003) Outcomes research: what is it and why does it matter?. Internal Medicine Journal 33:3, 110-118
CrossRef20
Jennifer L. Malin, Emmett Keeler, Cynthia Wang, Robert Brook. (2002) Using Cost-Effectiveness Analysis to Define a Breast Cancer Benefits Package for the Uninsured. Breast Cancer Research and Treatment 74:2, 143-153
CrossRef21
BRET TABACK, ARMANDO E. GIULIANO, NORA M. HANSEN, DAVE S. B. HOON. (2001) Microsatellite Alterations Detected in the Serum of Early Stage Breast Cancer Patients. Annals of the New York Academy of Sciences 945:1, 22-30
CrossRef22
Mary Cianfrocca, Lori J. Goldstein. (2001) Operable breast cancer. Current Treatment Options in Oncology 2:2, 157-167
CrossRef23
Pippa M. Simpson, John A. Spratt, John S. Spratt. (2001) Statistical methods in cancer research. Journal of Surgical Oncology 76:3, 201-223
CrossRef24
Stephen E. Karp. (2000) Clinical management of BRCA1- and BRCA2-associated breast cancer. Seminars in Surgical Oncology 18:4, 296-304
CrossRef25
J Brown. (1999) Using modelling in a phased approach to the economic evaluation of adjuvant therapy for early breast cancer. Critical Reviews in Oncology/Hematology 32:2, 95-103
CrossRef26
Richard K. Orr, Nananda F. Col, Karen M. Kuntz. (1999) A cost-effectiveness analysis of axillary node dissection in postmenopausal women with estrogen receptor–positive breast cancer and clinically negative axillary nodes. Surgery 126:3, 568-576
CrossRef27
Scott B. Cantor, Theodore G. Ganiats. (1999) Incremental Cost-Effectiveness Analysis. Journal of Clinical Epidemiology 52:6, 517-522
CrossRef28
LAURA M. McCONNELL, MARY KANE GOLDSTEIN. (1999) The Application of Medical Decision Analysis to Genetic Testing: An Introduction. Genetic Testing 3:1, 65-70
CrossRef29
Wright, Janice C., Weinstein, Milton C., . (1998) Gains in Life Expectancy from Medical Interventions — Standardizing Data on Outcomes. New England Journal of Medicine 339:6, 380-386
Full Text30
Thomas D. Szucs, Georg Guggenberger, Karin Berger, Winfried März, Jürgen R. Schäfer. (1998) Pharmakoökonomische Bewertung von Pravastatin in der Sekundärprävention der koronaren Herzkrankheit bei Patienten mit durchschnittlichen Cholesterinwerten. Herz 23:5, 319-329
CrossRef31
Wolfgang Hoffmann. (1998) Wirtschaftlichkeitsanalyse der adjuvanten Interferon-alpha-2b-Therapie bei Hochrisiko-Melanompatienten aufgrund von Ergebnissen der Studie ECOG 1684. Strahlentherapie und Onkologie 174:5, 288-289
CrossRef32
Vic Velanovich. (1998) Axillary lymph node dissection for breast cancer: A decision analysis of T1 lesions. Annals of Surgical Oncology 5:2, 131-139
CrossRef33
N DALYSCHVEITZER. (1997) L'
valuation du co
CrossRef34
Ana Ortega, George Dranitsaris, Jeremy Sturgeon, Heather Sutherland, Amit Oza. (1997) Cost–Utility Analysis of Paclitaxel in Combination with Cisplatin for Patients with Advanced Ovarian Cancer. Gynecologic Oncology 66:3, 454-463
CrossRef35
Kenneth A. Freedberg, Calvin J. Cohen, Thomas W. Barber. (1997) Prophylaxis for Disseminated Mycobacterium avium Complex(MAC) Infection in Patients With AIDS. Journal of Acquired Immune Deficiency Syndromes and Human Retrovirology 15:4, 275-282
CrossRef36
R. James Brenner, Edward A. Sickles. (1997) Surveillance mammography and stereotactic core breast biopsy for probably benign lesions: A cost comparison analysis. Academic Radiology 4:6, 419-425
CrossRef37
Analee E. Beisecker, Mary R. Cook, Janet Ashworth, Jeanne Hayes, Marci Brecheisen, Linda Helmig, Siobhan Hyland, Darcy Selenke. (1997) Side Effects of Adjuvant Chemotherapy: Perceptions of Node-Negative Breast Cancer Patients. Psycho-Oncology 6:2, 85-93
CrossRef38
Karin Berger, Gerald Klose, Thomas D. Szucs. (1997) Die Wirtschaftlichkeit von Arzneimitteltherapien am Beispiel von Pravastatin. Medizinische Klinik 92:6, 363-369
CrossRef39
Schrag, Deborah, Kuntz, Karen M., Garber, Judy E., Weeks, Jane C., . (1997) Decision Analysis — Effects of Prophylactic Mastectomy and Oophorectomy on Life Expectancy among Women with BRCA1 or BRCA2 Mutations. New England Journal of Medicine 336:20, 1465-1471
Full Text40
Bruce E. Hillner. (1997) Economic Issues in Breast Cancer Care. The Breast Journal 3:s1, 32-35
CrossRef41
Thomas J. Garrett, Linda T. Vahdat, David W. Kinne. (1997) Systemic adjuvant therapy of breast cancer. Journal of Surgical Oncology 64:2, 167-172
CrossRef42
D Meltzer. (1997) Accounting for future costs in medical cost-effectiveness analysis. Journal of Health Economics 16:1, 33-64
CrossRef43
Bruce E. Hillner, Harry D. Bear, Laurie L. Fajardo. (1996) Estimating the cost-effectiveness of stereotaxic biopsy for nonpalpable breast abnormalities: A decision analysis model. Academic Radiology 3:4, 351-360
CrossRef44
Dwight J. Rouse, John Owen, Robert L. Goldenberg, Suzanne P. Cliver. (1995) Determinants of the optimal time in gestation to initiate antenatal fetal testing: A decision-analytic approach. American Journal of Obstetrics and Gynecology 173:5, 1357-1363
CrossRef45
IAN D. McRURY, JAMES G. WHAYNE, DAVID E. HAINES. (1995) Temperature Measurement as a determinant of Tissue Heating During Radiofrequency Catheter ablation: An Examination of Electrode Thermistor Positioning for Measurement Accuracy. Journal of Cardiovascular Electrophysiology 6:4, 268-278
CrossRef46
RANDI E. ISAACS. (1995) Advances in the Treatment of Breast Cancer: Balancing Technology and Economics. Journal of Women's Health 4:1, 15-26
CrossRef47
Joel Tsevat, Jane C. Weeks, Edward Guadagnoli, Anna N. A. Tosteson, Carol M. Mangione, Joseph S. Pliskin, Milton C. Weinstein, Paul D. Cleary. (1994) Using health-related quality-of-life information. Journal of General Internal Medicine 9:10, 576-582
CrossRef48
Lawrence Tim Goodnough, Judy E. Grishaber, John D. Birkmeyer, Terri G. Monk, William J. Catalona. (1994) Efficacy and cost effectiveness of autologous blood predeposit in patients undergoing radical prostatectomy procedures. Urology 44:2, 226-231
CrossRef49
James Studnicki, David V. Schapira, Douglas D. Bradham, Robert A. Clark, Anne Jarrett. (1993) Response to the National Cancer Institute alert: The effect of practice guidelines on two hospitals in the same medical community. Cancer 72:10, 2986-2992
CrossRef50
H. D. Banta. (1993) Quality of life 1. Pharmacy World & Science 15:2, 45-49
CrossRef51
Gerald W. Chodak. (1993) Screening and early detection of prostate cancer. Cancer 71:S3, 981-983
CrossRef52
Bruce E. Hillner, Thomas J. Smith, Christopher E. Desch. (1993) Assessing the cost effectiveness of adjuvant therapies in early breast cancer using a decision analysis model. Breast Cancer Research and Treatment 25:2, 97-105
CrossRef53
Sandra R. Wolman, Robert J. Pauley, Anwar N. Mohamed, Peter J. Dawson, Daniel W. Visscher, Fazlul H. Sarkar. (1992) Genetic markers as prognostic indicators in breast cancer. Cancer 70:S4, 1765-1774
CrossRef54
Rodrigo Arriagada, Lars Erik Rutqvist, Lambert Skoog, Hemming Johansson, Andrew Kramar. (1992) Prognostic factors and natural history in lymph node-negative breast cancer patients. Breast Cancer Research and Treatment 21:2, 101-109
CrossRef55
Bruce E. Hillner, Thomas J. Smith. (1992) Should women with node-negative breast cancer receive adjuvant chemotherapy? — Insights from a decision analysis model. Breast Cancer Research and Treatment 23:1-2, 17-27
CrossRef56
George W. Sledge, Worta McCaskill-Stevens. (1992) Adjuvant Chemotherapy for Node-Negative Breast Cancer. Cancer Investigation 10:6, 595-602
CrossRef57
Martin Schumacher, Manfred Olschewski, Gabi Schulgen. (1991) Assessment of quality of life in clinical trials. Statistics in Medicine 10:12, 1915-1930
CrossRef58
Ulrich C Dubach, Bernard Rosner, Til Stürmer. (1991) Epidemiologic study of analgesic abuse: Mortality study in 7275 working women (1968–1987). Kidney International 40:4, 728-733
CrossRef59
Michael A. Bookman, Lori J. Goldstein, Richard M. Scher. (1991) Medical management of early-stage breast cancer. Current Problems in Cancer 15:4, 161-232
CrossRef
