- Email: [email protected]
Cost-Effectiveness of Treatment of Early Stage Endometrial Cancer
Gynecologic Oncology 74, 208 –216 (1999) Article ID gyno.1999.5427, available online at http://www.idealibrary.com on
Cost-Effectiveness of Treatment of Early Stage Endometrial Cancer Heidi Ashih, M.S.,* Tara Gustilo-Ashby, M.D.,† Evan R. Myers, M.D., M.P.H.,† Jeffrey Andrews, M.D.,† Daniel L. Clarke-Pearson, M.D.,† Donald Berry, Ph.D.,* ,‡ and Andrew Berchuck, M.D.† ,1 *Institute of Statistics and Decision Sciences, †Department of Obstetrics and Gynecology, and ‡Cancer Center Biostatistics, Duke University, Durham, North Carolina 27710 Received October 13, 1998
patients found to have occult metastatic disease, or who are considered at “high risk” of developing recurrent disease, often receive adjuvant therapy, usually in the form of external radiation. Prospective randomized trials have not been performed, but available data from case series suggest that adjuvant radiation may effect a modest increase in survival, perhaps by about 5% overall [6 – 8]. Historically, adverse prognostic factors such as poor histologic grade and deep myometrial invasion have dictated the use of adjuvant therapy in about 30 – 40% of cases [9]. More recently, surgical staging including lymph node sampling has been performed to identify more accurately the 15–20% of truly “high-risk” individuals who are most likely to benefit from adjuvant radiation [4, 10 –12]. However, lymph node sampling has been associated with higher perioperative morbidity and expense in some series [13]. We used decision analytic techniques to compare the effectiveness, as measured in years of life gained, and cost-effectiveness, as measured by cost per life-year gained, of surgery and radiation for early stage endometrial cancer.
Objective. The purpose of this study was to determine the average life-years gained and cost per life-year gained in treatment of early endometrial cancer. Methods. We performed a decision analysis using statistical models for survival after treatment for Stage I endometrial cancer. Estimates for survival probabilities without treatment, with surgery alone, and with surgery and radiation were derived from the literature. Charges and costs of treatment were estimated based on data from our institution. We calculated the average number of life-years gained and the cost per life-year gained of various treatment options based on these estimates. Sensitivity analyses were performed to determine the effect of uncertainty about parameter estimates on the results derived from our model. Results. Based on the assumptions of our model, most of the life-years gained in treatment of early endometrial cancer are attributable to hysterectomy, with a very low associated cost. For the “average” woman with endometrial cancer, about 10 life-years are gained from hysterectomy at a cost of $1000 per life-year gained, whereas adjuvant radiation yields on average 1 year of life gained at $4000 per life-year gained. Both life-years gained and cost are dramatically affected by age at diagnosis and to a lesser extent by histologic grade and comorbid medical conditions. Conclusions. This analysis suggests that the use of hysterectomy and adjuvant radiation in treatment of early endometrial cancer is a worthwhile use of health care resources. More sophisticated models may help determine the cost-effectiveness of various treatment strategies in specific subgroups of patients. © 1999 Academic Press
METHODS
INTRODUCTION
Our methods conformed to the guidelines of the U.S. Public Health Service Panel on Cost-Effectiveness in Health and Medicine as detailed below [14]. We modeled the life expectancy of women treated with surgery alone or with adjuvant radiation under varying assumptions about cost, competing mortality risk, and effectiveness of treatment. All calculations were performed using JMP software (SAS Institute, Cary, NC). Perspective. We took a medical center-based perspective. We did not include nonmedical costs or the costs associated with treatment of diseases in patients who survive endometrial cancer. We also did not include the costs of treatment of recurrent or terminal endometrial cancer. Time frame and analytic horizon. Our time frame for treatment was the initial hospitalization and treatment. Our analytic horizon was from the time of treatment to age 85 years. Measures of outcome. We calculated the average number of life-years gained and the cost per life-year gained associated
Endometrial cancer is the most common gynecologic malignancy in the United States, with approximately 35,000 new cases annually [1, 2]. The median age at diagnosis is in the early 60s, but cases occur throughout the peri- and postmenopausal years. Most patients (80 –90%) present with cancers that grossly appear to be confined to the uterus, and relatively high cure rates (70 –90%) are achieved with hysterectomy [1, 3–5]. Although surgical therapy is curative in most early stage cases, 1 To whom correspondence and reprint requests should be addressed at the Duke University Medical Center, DUMC 3079, Durham, NC 27710. Fax: (919) 684-8719. E-mail: [email protected].
0090-8258/99 $30.00 Copyright © 1999 by Academic Press All rights of reproduction in any form reserved.
208
209
ENDOMETRIAL CANCER COST-EFFECTIVENESS
TABLE 1 Estimates of 5-Year Survival after Treatment of Early Stage Endometrial Cancer (Sources: Refs. 3, 6 – 8, 12, 15–24) Percentage of women cured of cancer
Grade Well-differentiated
Moderately differentiated
Poorly differentiated
Age
Hysterectomy only (%)
Hysterectomy 1 radiation (%)
,60 years 60–70 years .70 years
88 85 82
90 87 84
,60 years 60–70 years .70 years ,60 years 60–70 years .70 years
75 70 65 60 50 40
80 75 70 70 60 50
with the different treatment strategies. We know of no data for estimating quality of life using standard measures for endometrial cancer patients. We did not adjust life expectancy for quality of life. We calculated the incremental cost per life-year gained for adjuvant radiation compared to surgery alone. Discount rate. We did not adjust for time preferences for economic or health outcomes. Since we considered only the costs of the initial hospitalization, there was no need to adjust economic measures. Using a moderate-sized annual discount rate such as 3% for life-years saved [14] would result in slightly higher cost-effectiveness ratios, but would not change our qualitative conclusions. Efficacy of treatment. Prospective randomized studies of treatment for Stage I adenocarcinoma of the uterus have not been performed. Existing population and hospital-based reports in the literature were reviewed to develop a set of assumptions regarding the curative potential of surgery and radiation (Table 1) [3, 6 – 8, 12, 15–24]. Overall, we assumed that hysterectomy yields a 75% cure rate and that adding radiation increases the cure rate to 80% (defining cure as disease-free survival). Because it is apparent from the literature
that age and histologic grade are important determinants of outcome, we estimated cure rates for various combinations of ages and grades as well. For this analysis we assumed that surgical staging including lymph node sampling provides prognostic information that guides the use of adjuvant therapy, but that it does not affect cure rate. We also assumed that if no treatment were performed survival would range from 1 to 4 years depending on the age at diagnosis and grade of the cancer. We assumed no excess mortality resulting from treatment and that any extra morbidity associated with lymph node sampling or radiation would not have an impact on survival. Costs of treatment. Data on actual allocated hospital costs for treatment of early stage endometrial cancer were obtained from patients treated at Duke University Medical Center in 1995. Professional fees were discounted 60% to represent the approximate level of reimbursement typically obtained from private insurers. The total cost of abdominal hysterectomy at our institution averaged about $9000. This included hospital care costs of about $3000 for an average hospital stay of 5– 6 days, $4000 for operating room costs, and expected collection of about $2000 for physician services. This is consistent with other data sources. The Health Care Financing Administration’s Medicare Provider Analysis Review files of costs by diagnosis-related group (DRG) for 1996 showed a mean charge of $8489 for DRG 355, “Non-ovarian or adnexal malignancy without comorbidity or complications,” and $14,022 for DRG 354, “Non-ovarian or adnexal malignancy with comorbidity or complication.” Mean Medicare reimbursements for the same DRGs were $3537 and $6776, respectively (http://www.hcfa.gov/stats/medpar/ss96d&s.txt). 1994 data from the Agency for Health Care Policy and Research’s Healthcare Cost and Utilization project showed average hospital charges of $11,641 (http://www.ahcpr.gov/data/94dcchpr.htm#11). If surgical staging including pelvic and aortic lymph node sampling also was performed at Duke, the incremental cost was estimated to be $1000. For administration of 4500 CGy pelvic radiation, actual cost data were not available, but charges average about $5000. This is consistent with the range of costs, $3466 to $9153, of various radiation regimens reported by Konski et al. [25]. In Table 2, the average cost for
TABLE 2 Estimates of Average Cost of Treatment of Early Stage Endometrial Cancer (Source: Duke University Medical Center)
Strategy Hysterectomy followed by adjuvant radiation of 40% of cases Hysterectomy and surgical staging followed by radiation of 20% of cases
Intervention Hysterectomy Radiation Total Hysterectomy Surgical staging Radiation Total
Proportion of patients treated (%)
Cost
Average cost
100 40
$9000 $5000
100 100 20
$9000 $1000 $5000
$9,000 $2,000 $11,000 $9,000 $1,000 $1,000 $11,000
210
ASHIH ET AL.
treatment of early endometrial cancer is presented using two strategies typically employed in treating endometrial cancer. In one strategy, all patients undergo hysterectomy and about 40% of patients also receive radiation on the basis of prognostic factors such as histologic grade and depth of myometrial invasion. Alternatively, all patients undergo hysterectomy and surgical staging and 20% of patients with nodal metastases or other evidence of metastatic disease undergo radiation. The average cost of both strategies is similar at our institution ($11,000). More importantly, the average incremental costs above those for hysterectomy alone are identical for the two strategies ($2000). Statistical analysis. The life tables for 1992 (National Center for Health Statistics, 1996) provide the number of survivors at any given age (up to 85 years) of 100,000 people born alive. To calculate the expected additional life-years of a woman who is healthy and alive at age a, we estimate the survival probability for each year beyond age a. This is the conditional probability of surviving a particular year given being alive at the start of the year (which is also 1 2 hazard). The probability of surviving to age (a 1 b) is the product of the b single-year survival probabilities. Summing these probabilities for b 5 1, 2, etc., gives the expected additional life-years out to age 85. Hazards for each succeeding year of age for a woman alive at a given age are calculated based on survival numbers from life tables (National Center for Health Statistics, 1996: http:// www.cdc.gov/nchswww/data/lifetb92.pdf):
TABLE 3 Relationship between Expected Additional Life-Years (x1) Gained at Various Ages and Relative Risk of Death Age
Relative risk of death 1.0
N aliveage~i! 2 N aliveage~i11! . N aliveage~i!
(1)
Because conditions such as obesity, hypertension, and diabetes that alter life expectancy are common in women with endometrial cancer [26 –29], we also calculated estimates of life-years saved and cost per life-year saved in which adjustments were made for these competing risks. The literature provides some information regarding the effect of different medical conditions on a patient’s relative risk of death over time [30 –33]. For example, women who are hypertensive have a 150% greater chance of dying than normal women (relative risk, r 5 2.5) [30], while diabetic women have about a 50% greater chance of dying than normal women (r 5 1.5) [31, 32]. Chance of survival until age i given that the patient was alive at age a is then calculated using the hazards and relative risk r: P~survive until age ~i!u alive at age ~a!! 5
5P
1, i5a i21 . ~1 2 hazardage~ j! !, i . a j5a
(2)
Since these probabilities were calculated in single-year incre-
2.0
2.5
3.0
31.8 27.4 23.2 19.3 15.7 12.4 9.5 6.9 4.3
30.1 25.8 21.7 17.9 14.5 11.4 8.7 6.3 3.9
Additional life-years gained 40 45 50 55 60 65 70 75 80
38.7 34.0 29.4 24.9 20.7 16.6 12.7 8.9 5.2
35.9 31.4 26.9 22.7 18.7 14.9 11.5 8.2 4.9
33.7 29.2 24.9 20.8 17.1 13.6 10.4 7.5 4.6
ments for age i, summing these survival probabilities is an estimate of x1, the expected number of additional life-years to age 85, given that the woman is alive at age a and has relative risk r: P~survive until age ~i!u alive at age ~a! with 1.25 hzd cnst! 5
hazardage~i! 5
1.5
5P
1, i5a i21 . ~1 2 1.25 3 hazardage~ j! !, i . a j5a
(3)
This calculation is repeated across different ages a (from age 40 to 80, in 5-year increments) and different relative risks r (1.0, 1.5, 2.0, 2.5, 3.0) to produce the expected number of life-years gained at each age and level of relative risk (Table 3). The expected additional life-years of a patient if no treatment was sought was calculated as a linear relationship on age, depending on grade and varied at age 40 from 4 years for a well-differentiated cancer to 2 years for a poorly differentiated cancer. At age 80, survival with no treatment was estimated at 1 year for a well-differentiated cancer and 0.5 year for a poorly differentiated cancer. The expected additional life-years of a patient who received only hysterectomy, given the patient’s grade, age, and relative risk, are based on the percentages in Table 1, x1 values in Table 3, and linearly interpolated values for no treatment. If a woman is cured of cancer, she is expected to live the same additional number of years as a healthy woman, x1; however, if the woman is not cured of cancer even though she had treatment, she is expected to live the same number of additional years as if she had not received treatment. For example, for a patient with a well-differentiated cancer the following calculations would be performed to determine her life expectancy:
211
ENDOMETRIAL CANCER COST-EFFECTIVENESS
FIG. 1. Life expectancy following no treatment, hysterectomy alone (TAH), or hysterectomy and adjuvant radiation (RT) by age and risk of death from other cause. Risk 5 1.0, normal health for age; Risk 5 3.0, poor health for age (relative risk of death from other causes three times normal).
1. For age ,60 years, Life expectancy 5 0.88 3 (x1) 1 (0.12 3 Life expectancy with no treatment). 2. For age between 60 and 70 years, Life expectancy 5 0.85 3 (x1) 1 (0.15 3 Life expectancy with no treatment). 3. For age .70 years, Life expectancy 5 0.82 3 (x1) 1 (0.18 3 Life expectancy with no treatment). Note that her relative risk is already incorporated into her value of x1. A similar calculation is performed for expected additional number of life-years if she receives hysterectomy and adjuvant radiation, the only distinction being the cure rate used from Table 1. The difference in life-years saved between hysterectomy alone and hysterectomy/adjuvant radiation yields the additional expected life-years due to adjuvant radiation. Sensitivity analysis. Because of the uncertainty of many of our estimates, we tested the impact of varying key parameters on the results of our model. We examined the effect of best case (10% cure rate for adjuvant therapy at an additional cost of $1000) and worst case (2% cure rate for adjuvant therapy at an additional cost of $5000) scenarios on life-years gained and cost/life-years gained.
RESULTS Life-years gained. The average number of life-years gained at various ages due to treatment is based on life tables (National Center for Health Statistics, 1996: http://www.cdc.gov/nchswww/data/lifetb92.pdf) and expected cure rates for well, moderately, and poorly differentiated cancers. In Figure 1, the average life expectancy for women with well, moderate, and poorly differentiated endometrial cancers with no treatment, hysterectomy, and hysterectomy 1 radiation of high-risk cases is presented for both normal healthy women (relative risk r 5 1) and women with severe medical problems that place them at a threefold increased hazard (relative risk r 5 3) of death from other causes. Calculations were performed for every 5-year interval between the ages of 40 and 80, and relative risks r at every 0.5 interval between 1.0 and 3.0. Since results did not vary strikingly across the different hazards, only results for the two relative risks (r 5 1 and r 5 3) are presented. Life-years gained by a given strategy at a given age are represented by the distance between the curves representing alternate strategies. Thus, the life-years gained by hysterectomy are represented by the distance between the No therapy
212
ASHIH ET AL.
curve and the hysterectomy (TAH) curve, while the distance between the TAH and the RT curves represents the additional life-years saved by adjuvant radiation. Table 4 presents data on life-years gained at age 40 and 80 dependent on grade and hazard of death. In both Fig. 1 and Table 4, it is clear that most of the life-years gained in treatment of early endometrial cancer are attributable to hysterectomy. For a 60-year-old woman with grade 2 endometrial cancer and hazard 2.0, about 10 life-years are gained from hysterectomy compared to about 1 year from radiation. Cost per life-year gained. The cost per life-year gained with treatment of various grades of endometrial cancer at ages 40 to 80 were calculated from our cost estimates and the average number of life-years gained at different ages and across different hazards of death from other causes. Figure 2 depicts the life-years gained and cost per life gained for adjuvant radiation for different grades of endometrial cancer and risks of death from other causes. Because the gains in life expectancy decrease for older patients and patients with comorbidity, the cost per life-year gained increases. Table 4 shows the costs, life-years gained, and cost per life-year gained from hysterectomy alone and for hysterectomy plus adjuvant radiation for the three grades at ages 40 and 80. Although the cost per life-year gained from hysterectomy is affected by age at diagnosis, grade, and comorbidity, it is very low under all circumstances. For an “average” woman with endometrial cancer (e.g., age 60, moderately differentiated cancer, hazard 2.0) the cost per life-year saved is only about $1000. Table 4 also demonstrates that the incremental cost per life-year gained for adjuvant therapy is also relatively low, although higher than that for hysterectomy alone. The effect of age at diagnosis, grade, and comorbidity on cost-effectiveness is apparent, with cost per life-years gained being lowest in healthy young women with poorly differentiated cancers. For the “average” woman with endometrial cancer mentioned above, the cost per life-year saved is about $4000. Sensitivity analysis. Figure 3 shows the results of our best and worst case scenarios with respect to curative potential and additional cost of adjuvant radiation. The incremental costeffectiveness is clearly quite sensitive to assumptions about the efficacy and costs of such treatment. The cost per additional year of life gained for patients older than 70 years rises rapidly at low estimates of incremental efficacy (2%).
TABLE 4 Life-Years Gained and Cost per Life-Year Gained for Hysterectomy Alone and Hysterectomy Plus Adjuvant Radiation, for Various Grades and Relative Risks of Death
Life- Cost per Incremental Age Average years life-year Incremental life-years (years) cost gained gained cost gained
Incremental cost per life-year gained
Hysterectomy: Grade 1, relative risk of noncancer death 1.0 40 80
$9000 $9000
34.5 4.5
$261 $2000
— —
— —
— —
Adjuvant radiation: Grade 1, relative risk of noncancer death 1.0 40 80
$2000 $2000
0.7 0.1
$2,881 $23,714
Hysterectomy: Grade 1, relative risk of noncancer death 3.0 40 80
$9000 $9000
27.0 3.5
$333 $2571
— —
— —
— —
Adjuvant radiation: Grade 1, relative risk of noncancer death 3.0 40 80
$2000 $2000
0.5 0.1
$3,825 $33,448
Hysterectomy: Grade 2, relative risk of noncancer death 1.0 40 80
$9000 $9000
29.8 3.7
$302 $2432
— —
— —
— —
Adjuvant radiation: Grade 2, relative risk of noncancer death 1.0 40 80
$2000 $2000
1.8 0.2
$1,120 $8,995
Hysterectomy: Grade 2, relative risk of noncancer death 3.0 40 80
$9000 $9000
23.4 2.9
$385 $3103
— —
— —
— —
Adjuvant radiation: Grade 2, relative risk of noncancer death 3.0 40 80
$2000 $2000
1.4 0.2
$1,474 $12,347
Hysterectomy: Grade 3, relative risk of noncancer death 1.0 40 80
$9000 $9000
24.0 2.4
$375 $3750
— —
— —
— —
Adjuvant radiation: Grade 3, relative risk of noncancer death 1.0 40 80
$2000 $2000
3.7 0.5
$545 $4,240
Hysterectomy: Grade 3, relative risk of noncancer death 3.0 40 80
$9000 $9000
18.9 1.9
$476 $4737
— —
— —
— —
DISCUSSION Adjuvant radiation: Grade 3, relative risk of noncancer death 3.0
This analysis demonstrates that for a relatively high assumed efficacy and low cost, hysterectomy for treatment of early endometrial cancer represents an economically sound investment of health care resources. The $1000 cost per life-year gained for the “average” 60-year-old woman in our model is small in comparison with other medical interventions. For example, the use of tissue plasminogen activator instead of
40 80
$2000 $2000
0.5 0.3
$4,240 $5,731
streptokinase after acute myocardial infarction is estimated to cost $30,000 per life-year saved [34]. Even if the cure rate with hysterectomy is actually 20% lower and the cost 50% higher,
ENDOMETRIAL CANCER COST-EFFECTIVENESS
213
FIG. 2. Incremental life-years gained and incremental cost per life-year gained attributable to adjuvant radiation therapy compared to hysterectomy alone by age, grade of endometrial cancer, and risk of death from other causes. Risk 5 1.0, normal health for age; Risk 5 3.0, poor health for age (relative risk of death from other causes three times normal).
the cost per life-year gained would remain very low. Although there is no consensus in our society regarding the acceptable upper limit of cost per life-year gained, many decision analyses have assumed that interventions costing less than $50,000 per life-year, and even up to $100,000 per life-year, are worth implementing [35, 36]. One of the most striking aspects of this analysis for most clinicians will likely be the extent to which life-years gained from hysterectomy dwarfs that attributable to adjuvant radiation. The 5% overall increment in survival attributed to adjuvant radiation of high-risk cases has a much smaller effect on life-years gained and comes at a significantly higher cost. It is also instructive to observe how the cost per life-year gained varies significantly with age and to a lesser extent with grade and comorbidity. Because the elderly have a shorter life expectancy, fewer life-years can be gained while the cost of treatment is constant. As a result, the cost per life-year gained rises dramatically in older women. On the other hand, although
the life-years gained from adjuvant radiation are few and average only about 1 life-year for a 60-year-old woman with a moderately differentiated cancer and hazard of 2.0, the associated $4000 cost per life-year gained is acceptable compared to many other interventions. However, it is possible that for some patients—those over 70,with multiple medical problems, for example—the cost-effectiveness ratio may be above the $50,000 range. This analysis represents an approximation of the cost-effectiveness of treatment of early endometrial cancer. Although we accounted for factors that affect outcome and cost most profoundly, some variables were omitted. Factors not included in our model are either difficult to quantitate because of the paucity of useful studies or believed to have a smaller impact on costs and outcomes. Although many of the model assumptions are open to challenge, we believe that they are reasonably representative of practice patterns and costs in the United States. The next several paragraphs address some of the po-
214
ASHIH ET AL.
FIG. 3. Sensitivity analysis of adjuvant radiation cure rate: incremental life-years gained and incremental cost per life-year gained attributable to adjuvant radiation compared to hysterectomy alone if incremental cure rate is 2 or 10% (compared to 5% for base-case model).
tential shortcomings of our model that might be addressed in future attempts to model the economics of treatment of early endometrial cancer. First, the assumptions on which our model was based were derived from our own experience and a review of the existing literature. Comparability between studies is limited by variations in age and racial distribution, staging techniques, stage distribution, type of treatment, length of follow-up, and other factors that affect outcome. For example, the proportion of well-differentiated cases is usually substantially higher in population-based registries compared to case series from large centers, where more poorly differentiated cases may be preferentially referred for treatment [4, 37]. In addition, it has been demonstrated convincingly that the prognosis of African-
American women with early stage endometrial cancer is strikingly worse than that of white women [38 – 41]. In this model, we assumed that surgical staging had no therapeutic benefit, but rather allowed individualization of adjuvant radiation and decreased the fraction of patients receiving adjuvant therapy from 40 to 20%. Likewise, we assumed that the cure rate did not suffer as a result of decreasing the fraction of cases receiving radiation in the surgically staged group. Using these assumptions, the average cost of treatment was the same regardless of whether hysterectomy was followed by radiation of 40% of cases or hysterectomy and surgical staging were followed by radiation of 20% of cases. Although the merits of these two approaches to management of early endometrial cancer have long been debated, it is worth recognizing that the average costs are similar. This model does not account for more subtle potential differences between these two approaches, however. For example, surgical complications may be more likely to occur more frequently in surgically staged patients [13], whereas the incidence of radiation-associated complications will be higher when more women receive this treatment. A more detailed model would be better able to define the benefits, risks, and costs of various treatment options for select subgroups of patients. One of the key assumptions in our model was that selective use of adjuvant radiation in high-risk cases increases overall survival in early stage disease. It appears from the literature that this benefit is most pronounced in poorly differentiated cases, for which the survival advantage may approach 10% [6, 8, 11, 42], but since poorly differentiated cases comprise only a minority of early endometrial cancers, the overall increase in survival is only about 5%. It is possible that we overestimated the curative potential of adjuvant radiation leading to an overly high estimate of life-years gained and an underestimation of the cost per life-year gained. On the other hand, it is conceivable that the increase in cure rate may somewhat exceed 5%. Our sensitivity analysis shows that these factors do alter costeffectiveness estimates. In addition, any benefit from prolongation of survival in individuals who are not cured was not accounted for in this model. Cost is another potentially variable feature of the model. It is often difficult to define the true costs of medical care as opposed to charges. We obtained cost data where available and when cost data were not available discounted professional fees by about 60% to reflect current conditions. Our estimate for the average cost of hysterectomy was $9000, but this can vary significantly between individuals, particularly when significant complications which prolong hospital stay occur. The model presented in this paper assumed a constant cost of treatment for all women regardless of factors such as age and comorbidity that affect the average cost of treatment. This represents another area in which the model can be improved to increase its utility. Age and comorbidity increase the risk of complications and costs of hysterectomy (E. R. Myers, unpublished data). If the costs of hysterectomy are higher for those patients with
ENDOMETRIAL CANCER COST-EFFECTIVENESS
lower expected benefits, then our current model will underestimate the cost-effectiveness ratios. Variations in costs between hospitals might also lead to changes in the cost per life-year gained for a given hospital. In addition to uncertainty regarding the curative potential of adjuvant radiation and the omission of some factors that affect life-years gained and cost-effectiveness, another shortcoming of this analysis was its focus on life-years gained rather than the quality of those life-years. In our clinical experience, the quality of life of most women who survive endometrial cancer is excellent. However, it may be lower in some patients, especially in women with other chronic diseases. Again, since these patients would be expected to have both higher costs of treatment and lower quality-adjusted life-expectancies, our model would underestimate the cost per life-year (or qualityadjusted life-year) gained. Despite these limitations, we believe that this model serves as a reasonable starting point from which to begin to evaluate the cost-effectiveness of treatment of early endometrial cancer. The complexity of this model could be increased in the future by adding additional factors that affect survival and cost. We believe that, in addition to justifying the use of health care resources, this type of data could be useful in guiding and individualizing treatment of women with endometrial cancer. We are currently developing more sophisticated models that will take into account many of these other factors. REFERENCES 1. Creasman WT: Endometrial cancer: incidence, prognostic factors, diagnosis, and treatment. Semin Oncol 24:S141–150, 1997 2. Ries L, Kosary C, Hankey B, Miller B, Edwards B: SEER Cancer Statistics Review, 1975–1995. Bethesda, MD, National Cancer Institute, 1998 3. Larson D, Broste S, Krawisz B: Surgery without radiotherapy for primary treatment of endometrial cancer. Obstet Gynecol 91:355–359, 1998 4. Corn BW, Dunton CJ, Carlson JA, Xie Y, Valicenti RK: National trends in the surgical staging of corpus cancer: a pattern-of-practice survey. Obstet Gynecol 90:628 – 631, 1997 5. Leijon T, Rosenberg P, Boeryd B: Total abdominal hysterectomy and bilateral salpingo-oophorectomy: a sufficient treatment for patients with low risk endometrial carcinoma. Int J Gynecol Cancer 7:376 –380, 1997 6. Boz G, De Paoli A, Innocente R, Talamini R, Scarabelli C, Scozzari G, Trovo MG: Postoperative radiotherapy and surgery in stage I endometrial carcinoma: a 10-year experience. Tumori 84:52–56, 1998 7. Algan O, Tabesh T, Hanlon A, Hogan WM, Boente M, Lanciano RM: Improved outcome in patients treated with postoperative radiation therapy for pathologic stage I/II endometrial cancer. Int J Radiat Oncol Biol Phys 35:925–933, 1996 8. Garcia-Domenech RV, Inesta JM, Asins E, Aznar I, Llixiona J: Prognostic factors in endometrial carcinoma: risk groups and adjuvant radiotherapy. Eur J Gynaecol Oncol 18:164 –170, 1997
215
and responsible challenge—a personal perspective. Gynecol Oncol 66: 179 –189, 1997 11. Greven KM, Corn BW, Case D, Purser P, Lanciano RM: Which prognostic factors influence the outcome of patients with surgically staged endometrial cancer treated with adjuvant radiation? Int J Radiat Oncol Biol Phys 39:413– 418, 1997 12. Maneschi M, Maneschi F, Geraci P, Licata MD, Serruto A, Attanasio P, Curcio P: Surgical pathological staging of endometrial carcinoma and results of treatment. Eur J Gynaecol Oncol 13:30 –35, 1992 13. Cliby W, Clarke-Pearson D, Dodge R, Kohler M, Rodriguez G, Olt G, Soper J, Berchuck A, Soisson A: Acute morbidity and mortality associated with selective pelvic and para-aortic lymphadenectomy in the surgical staging of endometrial adenocarcinoma. J Gynecol Tech 1:19 –26, 1995 14. Gold MR, Siegel JE, Russell LB, Weinstein MC (eds): Cost-Effectiveness in Health and Medicine. New York, Oxford Univ. Press, 1996 15. Akkad AA, Habiba MA, Ismail N, Abrams K, al-Azzawi F: Abnormal uterine bleeding on hormone replacement: the importance of intrauterine structural abnormalities. Obstet Gynecol 86:330 –334, 1995 16. Ackerman I, Malone S, Thomas G, Franssen E, Balogh J, Dembo A: Endometrial carcinoma—relative effectiveness of adjuvant irradiation vs therapy reserved for relapse. Gynecol Oncol 60:177–183, 1996 17. Barhum M, Stein M, Ronsenblatt E, Dale J, Kuten A: Pathological stage I endometrial carcinoma: the role for adjuvant radiotherapy. Tumori 79:405– 409, 1993 18. Foote AJ, Proietto A: Stage 1 endometrial cancer: treatment modalities and factors influencing recurrence. Aust N Z J Obstet Gynaecol 34:448 – 452, 1994 19. Grigsby PW, Perez CA, Kuten A, Simpson JR, Garcia DM, Camel HM, Kao MS, Galakatos AE: Clinical stage I endometrial cancer: results of adjuvant irradiation and patterns of failure. Int J Radiat Oncol Biol Phys 21:379 –385, 1991 20. Kim Y, Niloff J: Endometrial carcinoma: analysis of recurrence in patients treated with a strategy minimizing lymph node sampling and radiation therapy. Obstet Gynecol 82:175–180, 1993 21. Nazeer T, Ballouk F, Malfetano JH, Figge H, Ambros RA: Multivariate survival analysis of clinicopathologic features in surgical stage I endometrioid carcinoma including analysis of HER-2/neu expression. Am J Obstet Gynecol 173:1829 –1834, 1995 22. Rose PG, Cha SD, Tak WK, Fitzgerald T, Reale F, Hunter RE: Radiation therapy for surgically proven para-aortic node metastasis in endometrial carcinoma. Int J Radiat Oncol Biol Phys 24:229 –233, 1992 23. Stryker JA, Podczaski E, Kaminski P, Velkley DE: Adjuvant external beam therapy for pathologic stage I and occult stage II endometrial carcinoma. Cancer 67:2872–2879, 1991 24. Sainz de la Cuesta R, Goff BA, Nikrui N, Flynn DF, Fuller AF Jr: Postoperative management of patients with stage Ib endometrial carcinoma. Eur J Gynaecol Oncol 17:338 –341, 1996 25. Konski AA, Bracy PM, Jurs SG, Weiss SJ, Zeidner SR: Cost minimization analysis of various treatment options for surgical stage I endometrial carcinoma. Int J Radiat Oncol Biol Phys 37:367–373, 1997 26. Adami HO, McLaughlin J, Ekbom A, Berne C, Silverman D, Hacker D, Persson I: Cancer risk in patients with diabetes mellitus. Cancer Causes Control 2:307–314, 1991 27. Le Marchand L, Wilkens LR, Mi MP: Early-age body size, adult weight gain and endometrial cancer risk. Int J Cancer 48:807– 811, 1991
9. Morrow C, Bundy B, Kurman R, Creasman W, Heller P, Homesley H, Graham J: Relationship between surgical–pathological risk factors and outcomes in clinical stage I and II carcinoma of the endometrium: a Gynecologic Oncology Group study. Gynecol Oncol 40:55– 65, 1991
28. Maatela J, Aromaa A, Salmi T, Pohja M, Vuento M, Gronroos M: The risk of endometrial cancer in diabetic and hypertensive patients: a nationwide record-linkage study in Finland. Ann Chir Gynaecol Suppl 208:20 –24, 1994
10. Boronow R: Surgical staging of endometrial cancer: evolution, evaluation,
29. Wideroff L, Gridley G, Mellemkjaer L, Chow WH, Linet M, Keehn S,
216
ASHIH ET AL.
Borch-Johnsen K, Olsen JH: Cancer incidence in a population-based cohort of patients hospitalized with diabetes mellitus in Denmark. J Natl Cancer Inst 89:1360 –1365, 1997 30. Patterns of extra mortality for the common impairments. Stat Bull Metropol Life Insur Co 50:6 –10, 1969
37.
31. Life expectancy among diabetics in the United States. Stat Bull Metropol Life Insur Co 60:6 – 8, 1979
38.
32. Life expectancy among diabetics—II. Stat Bull Metropol Life Insur Co 60:11–13, 1979 33. Andersen B, Ostberg J: Survival rates in surgery of the aged: assessment of long-term prognosis according to coexisting diseases. Gerontol Clin 14:354 –360, 1972 34. Mark DB, Hlatky MA, Califf RM, Naylor CD, Lee KL, Armstrong PW, Barbash G, White H, Simoons ML, Nelson CL: Cost effectiveness of thrombolytic therapy with tissue plasminogen activator as compared with streptokinase for acute myocardial infarction. N Engl J Med 332:1418 – 1424, 1995 35. Mark D. Economic Analysis Methods and Endpoints, in Califf R, Mark D, Wagner G (eds): Acute Coronary Care in the Thrombolytic Era. St. Louis, Mosby, 1995, pp 167–182 36. Laupacis A, Feeny D, Detsky AS, Tugwell PX: How attractive does a new
39.
40. 41.
42.
technology have to be to warrant adoption and utilization? Tentative guidelines for using clinical and economic evaluations. Can Med Assoc J 146:473– 481, 1992 Partridge E, Shingleton H, Menck H: The National Cancer Data Base report on endometrial cancer. J Surg Oncol 61:111–123, 1996 Hill HA, Eley JW, Harlan LC, Greenberg RS, Barrett RJ, Chen VW: Racial differences in endometrial cancer survival: the black/white cancer survival study. Obstet Gynecol 88:919 –926, 1996 Liu JR, Conaway M, Rodriguez GC, Soper JT, Clarke-Pearson DL, Berchuck A: Relationship between race and interval to treatment in endometrial cancer. Obstet Gynecol 86:486 – 490, 1995 Plaxe SC, Saltzstein SL: Impact of ethnicity on the incidence of high-risk endometrial carcinoma. Gynecol Oncol 65:8 –12, 1997 Barrett RJ, Harlan LC, Wesley MN, Hill HA, Chen VW, Clayton LA, Kotz HL, Eley JW, Robboy SJ, Edwards BK: Endometrial cancer: stage at diagnosis and associated factors in black and white patients. Am J Obstet Gynecol 173:414 – 423, 1995 Grigsby PW, Perez CA, Kuten A, Simpson JR, Garcia DM, Camel HM, Kao MS, Galakatos AE: Clinical stage I endometrial cancer: prognostic factors for local control and distant metastasis and implications of the new FIGO surgical staging system. Int J Radiat Oncol Biol Phys 22:905–911, 1992
Cost-Effectiveness of Treatment of Early Stage Endometrial Cancer Heidi Ashih, M.S.,* Tara Gustilo-Ashby, M.D.,† Evan R. Myers, M.D., M.P.H.,† Jeffrey Andrews, M.D.,† Daniel L. Clarke-Pearson, M.D.,† Donald Berry, Ph.D.,* ,‡ and Andrew Berchuck, M.D.† ,1 *Institute of Statistics and Decision Sciences, †Department of Obstetrics and Gynecology, and ‡Cancer Center Biostatistics, Duke University, Durham, North Carolina 27710 Received October 13, 1998
patients found to have occult metastatic disease, or who are considered at “high risk” of developing recurrent disease, often receive adjuvant therapy, usually in the form of external radiation. Prospective randomized trials have not been performed, but available data from case series suggest that adjuvant radiation may effect a modest increase in survival, perhaps by about 5% overall [6 – 8]. Historically, adverse prognostic factors such as poor histologic grade and deep myometrial invasion have dictated the use of adjuvant therapy in about 30 – 40% of cases [9]. More recently, surgical staging including lymph node sampling has been performed to identify more accurately the 15–20% of truly “high-risk” individuals who are most likely to benefit from adjuvant radiation [4, 10 –12]. However, lymph node sampling has been associated with higher perioperative morbidity and expense in some series [13]. We used decision analytic techniques to compare the effectiveness, as measured in years of life gained, and cost-effectiveness, as measured by cost per life-year gained, of surgery and radiation for early stage endometrial cancer.
Objective. The purpose of this study was to determine the average life-years gained and cost per life-year gained in treatment of early endometrial cancer. Methods. We performed a decision analysis using statistical models for survival after treatment for Stage I endometrial cancer. Estimates for survival probabilities without treatment, with surgery alone, and with surgery and radiation were derived from the literature. Charges and costs of treatment were estimated based on data from our institution. We calculated the average number of life-years gained and the cost per life-year gained of various treatment options based on these estimates. Sensitivity analyses were performed to determine the effect of uncertainty about parameter estimates on the results derived from our model. Results. Based on the assumptions of our model, most of the life-years gained in treatment of early endometrial cancer are attributable to hysterectomy, with a very low associated cost. For the “average” woman with endometrial cancer, about 10 life-years are gained from hysterectomy at a cost of $1000 per life-year gained, whereas adjuvant radiation yields on average 1 year of life gained at $4000 per life-year gained. Both life-years gained and cost are dramatically affected by age at diagnosis and to a lesser extent by histologic grade and comorbid medical conditions. Conclusions. This analysis suggests that the use of hysterectomy and adjuvant radiation in treatment of early endometrial cancer is a worthwhile use of health care resources. More sophisticated models may help determine the cost-effectiveness of various treatment strategies in specific subgroups of patients. © 1999 Academic Press
METHODS
INTRODUCTION
Our methods conformed to the guidelines of the U.S. Public Health Service Panel on Cost-Effectiveness in Health and Medicine as detailed below [14]. We modeled the life expectancy of women treated with surgery alone or with adjuvant radiation under varying assumptions about cost, competing mortality risk, and effectiveness of treatment. All calculations were performed using JMP software (SAS Institute, Cary, NC). Perspective. We took a medical center-based perspective. We did not include nonmedical costs or the costs associated with treatment of diseases in patients who survive endometrial cancer. We also did not include the costs of treatment of recurrent or terminal endometrial cancer. Time frame and analytic horizon. Our time frame for treatment was the initial hospitalization and treatment. Our analytic horizon was from the time of treatment to age 85 years. Measures of outcome. We calculated the average number of life-years gained and the cost per life-year gained associated
Endometrial cancer is the most common gynecologic malignancy in the United States, with approximately 35,000 new cases annually [1, 2]. The median age at diagnosis is in the early 60s, but cases occur throughout the peri- and postmenopausal years. Most patients (80 –90%) present with cancers that grossly appear to be confined to the uterus, and relatively high cure rates (70 –90%) are achieved with hysterectomy [1, 3–5]. Although surgical therapy is curative in most early stage cases, 1 To whom correspondence and reprint requests should be addressed at the Duke University Medical Center, DUMC 3079, Durham, NC 27710. Fax: (919) 684-8719. E-mail: [email protected].
0090-8258/99 $30.00 Copyright © 1999 by Academic Press All rights of reproduction in any form reserved.
208
209
ENDOMETRIAL CANCER COST-EFFECTIVENESS
TABLE 1 Estimates of 5-Year Survival after Treatment of Early Stage Endometrial Cancer (Sources: Refs. 3, 6 – 8, 12, 15–24) Percentage of women cured of cancer
Grade Well-differentiated
Moderately differentiated
Poorly differentiated
Age
Hysterectomy only (%)
Hysterectomy 1 radiation (%)
,60 years 60–70 years .70 years
88 85 82
90 87 84
,60 years 60–70 years .70 years ,60 years 60–70 years .70 years
75 70 65 60 50 40
80 75 70 70 60 50
with the different treatment strategies. We know of no data for estimating quality of life using standard measures for endometrial cancer patients. We did not adjust life expectancy for quality of life. We calculated the incremental cost per life-year gained for adjuvant radiation compared to surgery alone. Discount rate. We did not adjust for time preferences for economic or health outcomes. Since we considered only the costs of the initial hospitalization, there was no need to adjust economic measures. Using a moderate-sized annual discount rate such as 3% for life-years saved [14] would result in slightly higher cost-effectiveness ratios, but would not change our qualitative conclusions. Efficacy of treatment. Prospective randomized studies of treatment for Stage I adenocarcinoma of the uterus have not been performed. Existing population and hospital-based reports in the literature were reviewed to develop a set of assumptions regarding the curative potential of surgery and radiation (Table 1) [3, 6 – 8, 12, 15–24]. Overall, we assumed that hysterectomy yields a 75% cure rate and that adding radiation increases the cure rate to 80% (defining cure as disease-free survival). Because it is apparent from the literature
that age and histologic grade are important determinants of outcome, we estimated cure rates for various combinations of ages and grades as well. For this analysis we assumed that surgical staging including lymph node sampling provides prognostic information that guides the use of adjuvant therapy, but that it does not affect cure rate. We also assumed that if no treatment were performed survival would range from 1 to 4 years depending on the age at diagnosis and grade of the cancer. We assumed no excess mortality resulting from treatment and that any extra morbidity associated with lymph node sampling or radiation would not have an impact on survival. Costs of treatment. Data on actual allocated hospital costs for treatment of early stage endometrial cancer were obtained from patients treated at Duke University Medical Center in 1995. Professional fees were discounted 60% to represent the approximate level of reimbursement typically obtained from private insurers. The total cost of abdominal hysterectomy at our institution averaged about $9000. This included hospital care costs of about $3000 for an average hospital stay of 5– 6 days, $4000 for operating room costs, and expected collection of about $2000 for physician services. This is consistent with other data sources. The Health Care Financing Administration’s Medicare Provider Analysis Review files of costs by diagnosis-related group (DRG) for 1996 showed a mean charge of $8489 for DRG 355, “Non-ovarian or adnexal malignancy without comorbidity or complications,” and $14,022 for DRG 354, “Non-ovarian or adnexal malignancy with comorbidity or complication.” Mean Medicare reimbursements for the same DRGs were $3537 and $6776, respectively (http://www.hcfa.gov/stats/medpar/ss96d&s.txt). 1994 data from the Agency for Health Care Policy and Research’s Healthcare Cost and Utilization project showed average hospital charges of $11,641 (http://www.ahcpr.gov/data/94dcchpr.htm#11). If surgical staging including pelvic and aortic lymph node sampling also was performed at Duke, the incremental cost was estimated to be $1000. For administration of 4500 CGy pelvic radiation, actual cost data were not available, but charges average about $5000. This is consistent with the range of costs, $3466 to $9153, of various radiation regimens reported by Konski et al. [25]. In Table 2, the average cost for
TABLE 2 Estimates of Average Cost of Treatment of Early Stage Endometrial Cancer (Source: Duke University Medical Center)
Strategy Hysterectomy followed by adjuvant radiation of 40% of cases Hysterectomy and surgical staging followed by radiation of 20% of cases
Intervention Hysterectomy Radiation Total Hysterectomy Surgical staging Radiation Total
Proportion of patients treated (%)
Cost
Average cost
100 40
$9000 $5000
100 100 20
$9000 $1000 $5000
$9,000 $2,000 $11,000 $9,000 $1,000 $1,000 $11,000
210
ASHIH ET AL.
treatment of early endometrial cancer is presented using two strategies typically employed in treating endometrial cancer. In one strategy, all patients undergo hysterectomy and about 40% of patients also receive radiation on the basis of prognostic factors such as histologic grade and depth of myometrial invasion. Alternatively, all patients undergo hysterectomy and surgical staging and 20% of patients with nodal metastases or other evidence of metastatic disease undergo radiation. The average cost of both strategies is similar at our institution ($11,000). More importantly, the average incremental costs above those for hysterectomy alone are identical for the two strategies ($2000). Statistical analysis. The life tables for 1992 (National Center for Health Statistics, 1996) provide the number of survivors at any given age (up to 85 years) of 100,000 people born alive. To calculate the expected additional life-years of a woman who is healthy and alive at age a, we estimate the survival probability for each year beyond age a. This is the conditional probability of surviving a particular year given being alive at the start of the year (which is also 1 2 hazard). The probability of surviving to age (a 1 b) is the product of the b single-year survival probabilities. Summing these probabilities for b 5 1, 2, etc., gives the expected additional life-years out to age 85. Hazards for each succeeding year of age for a woman alive at a given age are calculated based on survival numbers from life tables (National Center for Health Statistics, 1996: http:// www.cdc.gov/nchswww/data/lifetb92.pdf):
TABLE 3 Relationship between Expected Additional Life-Years (x1) Gained at Various Ages and Relative Risk of Death Age
Relative risk of death 1.0
N aliveage~i! 2 N aliveage~i11! . N aliveage~i!
(1)
Because conditions such as obesity, hypertension, and diabetes that alter life expectancy are common in women with endometrial cancer [26 –29], we also calculated estimates of life-years saved and cost per life-year saved in which adjustments were made for these competing risks. The literature provides some information regarding the effect of different medical conditions on a patient’s relative risk of death over time [30 –33]. For example, women who are hypertensive have a 150% greater chance of dying than normal women (relative risk, r 5 2.5) [30], while diabetic women have about a 50% greater chance of dying than normal women (r 5 1.5) [31, 32]. Chance of survival until age i given that the patient was alive at age a is then calculated using the hazards and relative risk r: P~survive until age ~i!u alive at age ~a!! 5
5P
1, i5a i21 . ~1 2 hazardage~ j! !, i . a j5a
(2)
Since these probabilities were calculated in single-year incre-
2.0
2.5
3.0
31.8 27.4 23.2 19.3 15.7 12.4 9.5 6.9 4.3
30.1 25.8 21.7 17.9 14.5 11.4 8.7 6.3 3.9
Additional life-years gained 40 45 50 55 60 65 70 75 80
38.7 34.0 29.4 24.9 20.7 16.6 12.7 8.9 5.2
35.9 31.4 26.9 22.7 18.7 14.9 11.5 8.2 4.9
33.7 29.2 24.9 20.8 17.1 13.6 10.4 7.5 4.6
ments for age i, summing these survival probabilities is an estimate of x1, the expected number of additional life-years to age 85, given that the woman is alive at age a and has relative risk r: P~survive until age ~i!u alive at age ~a! with 1.25 hzd cnst! 5
hazardage~i! 5
1.5
5P
1, i5a i21 . ~1 2 1.25 3 hazardage~ j! !, i . a j5a
(3)
This calculation is repeated across different ages a (from age 40 to 80, in 5-year increments) and different relative risks r (1.0, 1.5, 2.0, 2.5, 3.0) to produce the expected number of life-years gained at each age and level of relative risk (Table 3). The expected additional life-years of a patient if no treatment was sought was calculated as a linear relationship on age, depending on grade and varied at age 40 from 4 years for a well-differentiated cancer to 2 years for a poorly differentiated cancer. At age 80, survival with no treatment was estimated at 1 year for a well-differentiated cancer and 0.5 year for a poorly differentiated cancer. The expected additional life-years of a patient who received only hysterectomy, given the patient’s grade, age, and relative risk, are based on the percentages in Table 1, x1 values in Table 3, and linearly interpolated values for no treatment. If a woman is cured of cancer, she is expected to live the same additional number of years as a healthy woman, x1; however, if the woman is not cured of cancer even though she had treatment, she is expected to live the same number of additional years as if she had not received treatment. For example, for a patient with a well-differentiated cancer the following calculations would be performed to determine her life expectancy:
211
ENDOMETRIAL CANCER COST-EFFECTIVENESS
FIG. 1. Life expectancy following no treatment, hysterectomy alone (TAH), or hysterectomy and adjuvant radiation (RT) by age and risk of death from other cause. Risk 5 1.0, normal health for age; Risk 5 3.0, poor health for age (relative risk of death from other causes three times normal).
1. For age ,60 years, Life expectancy 5 0.88 3 (x1) 1 (0.12 3 Life expectancy with no treatment). 2. For age between 60 and 70 years, Life expectancy 5 0.85 3 (x1) 1 (0.15 3 Life expectancy with no treatment). 3. For age .70 years, Life expectancy 5 0.82 3 (x1) 1 (0.18 3 Life expectancy with no treatment). Note that her relative risk is already incorporated into her value of x1. A similar calculation is performed for expected additional number of life-years if she receives hysterectomy and adjuvant radiation, the only distinction being the cure rate used from Table 1. The difference in life-years saved between hysterectomy alone and hysterectomy/adjuvant radiation yields the additional expected life-years due to adjuvant radiation. Sensitivity analysis. Because of the uncertainty of many of our estimates, we tested the impact of varying key parameters on the results of our model. We examined the effect of best case (10% cure rate for adjuvant therapy at an additional cost of $1000) and worst case (2% cure rate for adjuvant therapy at an additional cost of $5000) scenarios on life-years gained and cost/life-years gained.
RESULTS Life-years gained. The average number of life-years gained at various ages due to treatment is based on life tables (National Center for Health Statistics, 1996: http://www.cdc.gov/nchswww/data/lifetb92.pdf) and expected cure rates for well, moderately, and poorly differentiated cancers. In Figure 1, the average life expectancy for women with well, moderate, and poorly differentiated endometrial cancers with no treatment, hysterectomy, and hysterectomy 1 radiation of high-risk cases is presented for both normal healthy women (relative risk r 5 1) and women with severe medical problems that place them at a threefold increased hazard (relative risk r 5 3) of death from other causes. Calculations were performed for every 5-year interval between the ages of 40 and 80, and relative risks r at every 0.5 interval between 1.0 and 3.0. Since results did not vary strikingly across the different hazards, only results for the two relative risks (r 5 1 and r 5 3) are presented. Life-years gained by a given strategy at a given age are represented by the distance between the curves representing alternate strategies. Thus, the life-years gained by hysterectomy are represented by the distance between the No therapy
212
ASHIH ET AL.
curve and the hysterectomy (TAH) curve, while the distance between the TAH and the RT curves represents the additional life-years saved by adjuvant radiation. Table 4 presents data on life-years gained at age 40 and 80 dependent on grade and hazard of death. In both Fig. 1 and Table 4, it is clear that most of the life-years gained in treatment of early endometrial cancer are attributable to hysterectomy. For a 60-year-old woman with grade 2 endometrial cancer and hazard 2.0, about 10 life-years are gained from hysterectomy compared to about 1 year from radiation. Cost per life-year gained. The cost per life-year gained with treatment of various grades of endometrial cancer at ages 40 to 80 were calculated from our cost estimates and the average number of life-years gained at different ages and across different hazards of death from other causes. Figure 2 depicts the life-years gained and cost per life gained for adjuvant radiation for different grades of endometrial cancer and risks of death from other causes. Because the gains in life expectancy decrease for older patients and patients with comorbidity, the cost per life-year gained increases. Table 4 shows the costs, life-years gained, and cost per life-year gained from hysterectomy alone and for hysterectomy plus adjuvant radiation for the three grades at ages 40 and 80. Although the cost per life-year gained from hysterectomy is affected by age at diagnosis, grade, and comorbidity, it is very low under all circumstances. For an “average” woman with endometrial cancer (e.g., age 60, moderately differentiated cancer, hazard 2.0) the cost per life-year saved is only about $1000. Table 4 also demonstrates that the incremental cost per life-year gained for adjuvant therapy is also relatively low, although higher than that for hysterectomy alone. The effect of age at diagnosis, grade, and comorbidity on cost-effectiveness is apparent, with cost per life-years gained being lowest in healthy young women with poorly differentiated cancers. For the “average” woman with endometrial cancer mentioned above, the cost per life-year saved is about $4000. Sensitivity analysis. Figure 3 shows the results of our best and worst case scenarios with respect to curative potential and additional cost of adjuvant radiation. The incremental costeffectiveness is clearly quite sensitive to assumptions about the efficacy and costs of such treatment. The cost per additional year of life gained for patients older than 70 years rises rapidly at low estimates of incremental efficacy (2%).
TABLE 4 Life-Years Gained and Cost per Life-Year Gained for Hysterectomy Alone and Hysterectomy Plus Adjuvant Radiation, for Various Grades and Relative Risks of Death
Life- Cost per Incremental Age Average years life-year Incremental life-years (years) cost gained gained cost gained
Incremental cost per life-year gained
Hysterectomy: Grade 1, relative risk of noncancer death 1.0 40 80
$9000 $9000
34.5 4.5
$261 $2000
— —
— —
— —
Adjuvant radiation: Grade 1, relative risk of noncancer death 1.0 40 80
$2000 $2000
0.7 0.1
$2,881 $23,714
Hysterectomy: Grade 1, relative risk of noncancer death 3.0 40 80
$9000 $9000
27.0 3.5
$333 $2571
— —
— —
— —
Adjuvant radiation: Grade 1, relative risk of noncancer death 3.0 40 80
$2000 $2000
0.5 0.1
$3,825 $33,448
Hysterectomy: Grade 2, relative risk of noncancer death 1.0 40 80
$9000 $9000
29.8 3.7
$302 $2432
— —
— —
— —
Adjuvant radiation: Grade 2, relative risk of noncancer death 1.0 40 80
$2000 $2000
1.8 0.2
$1,120 $8,995
Hysterectomy: Grade 2, relative risk of noncancer death 3.0 40 80
$9000 $9000
23.4 2.9
$385 $3103
— —
— —
— —
Adjuvant radiation: Grade 2, relative risk of noncancer death 3.0 40 80
$2000 $2000
1.4 0.2
$1,474 $12,347
Hysterectomy: Grade 3, relative risk of noncancer death 1.0 40 80
$9000 $9000
24.0 2.4
$375 $3750
— —
— —
— —
Adjuvant radiation: Grade 3, relative risk of noncancer death 1.0 40 80
$2000 $2000
3.7 0.5
$545 $4,240
Hysterectomy: Grade 3, relative risk of noncancer death 3.0 40 80
$9000 $9000
18.9 1.9
$476 $4737
— —
— —
— —
DISCUSSION Adjuvant radiation: Grade 3, relative risk of noncancer death 3.0
This analysis demonstrates that for a relatively high assumed efficacy and low cost, hysterectomy for treatment of early endometrial cancer represents an economically sound investment of health care resources. The $1000 cost per life-year gained for the “average” 60-year-old woman in our model is small in comparison with other medical interventions. For example, the use of tissue plasminogen activator instead of
40 80
$2000 $2000
0.5 0.3
$4,240 $5,731
streptokinase after acute myocardial infarction is estimated to cost $30,000 per life-year saved [34]. Even if the cure rate with hysterectomy is actually 20% lower and the cost 50% higher,
ENDOMETRIAL CANCER COST-EFFECTIVENESS
213
FIG. 2. Incremental life-years gained and incremental cost per life-year gained attributable to adjuvant radiation therapy compared to hysterectomy alone by age, grade of endometrial cancer, and risk of death from other causes. Risk 5 1.0, normal health for age; Risk 5 3.0, poor health for age (relative risk of death from other causes three times normal).
the cost per life-year gained would remain very low. Although there is no consensus in our society regarding the acceptable upper limit of cost per life-year gained, many decision analyses have assumed that interventions costing less than $50,000 per life-year, and even up to $100,000 per life-year, are worth implementing [35, 36]. One of the most striking aspects of this analysis for most clinicians will likely be the extent to which life-years gained from hysterectomy dwarfs that attributable to adjuvant radiation. The 5% overall increment in survival attributed to adjuvant radiation of high-risk cases has a much smaller effect on life-years gained and comes at a significantly higher cost. It is also instructive to observe how the cost per life-year gained varies significantly with age and to a lesser extent with grade and comorbidity. Because the elderly have a shorter life expectancy, fewer life-years can be gained while the cost of treatment is constant. As a result, the cost per life-year gained rises dramatically in older women. On the other hand, although
the life-years gained from adjuvant radiation are few and average only about 1 life-year for a 60-year-old woman with a moderately differentiated cancer and hazard of 2.0, the associated $4000 cost per life-year gained is acceptable compared to many other interventions. However, it is possible that for some patients—those over 70,with multiple medical problems, for example—the cost-effectiveness ratio may be above the $50,000 range. This analysis represents an approximation of the cost-effectiveness of treatment of early endometrial cancer. Although we accounted for factors that affect outcome and cost most profoundly, some variables were omitted. Factors not included in our model are either difficult to quantitate because of the paucity of useful studies or believed to have a smaller impact on costs and outcomes. Although many of the model assumptions are open to challenge, we believe that they are reasonably representative of practice patterns and costs in the United States. The next several paragraphs address some of the po-
214
ASHIH ET AL.
FIG. 3. Sensitivity analysis of adjuvant radiation cure rate: incremental life-years gained and incremental cost per life-year gained attributable to adjuvant radiation compared to hysterectomy alone if incremental cure rate is 2 or 10% (compared to 5% for base-case model).
tential shortcomings of our model that might be addressed in future attempts to model the economics of treatment of early endometrial cancer. First, the assumptions on which our model was based were derived from our own experience and a review of the existing literature. Comparability between studies is limited by variations in age and racial distribution, staging techniques, stage distribution, type of treatment, length of follow-up, and other factors that affect outcome. For example, the proportion of well-differentiated cases is usually substantially higher in population-based registries compared to case series from large centers, where more poorly differentiated cases may be preferentially referred for treatment [4, 37]. In addition, it has been demonstrated convincingly that the prognosis of African-
American women with early stage endometrial cancer is strikingly worse than that of white women [38 – 41]. In this model, we assumed that surgical staging had no therapeutic benefit, but rather allowed individualization of adjuvant radiation and decreased the fraction of patients receiving adjuvant therapy from 40 to 20%. Likewise, we assumed that the cure rate did not suffer as a result of decreasing the fraction of cases receiving radiation in the surgically staged group. Using these assumptions, the average cost of treatment was the same regardless of whether hysterectomy was followed by radiation of 40% of cases or hysterectomy and surgical staging were followed by radiation of 20% of cases. Although the merits of these two approaches to management of early endometrial cancer have long been debated, it is worth recognizing that the average costs are similar. This model does not account for more subtle potential differences between these two approaches, however. For example, surgical complications may be more likely to occur more frequently in surgically staged patients [13], whereas the incidence of radiation-associated complications will be higher when more women receive this treatment. A more detailed model would be better able to define the benefits, risks, and costs of various treatment options for select subgroups of patients. One of the key assumptions in our model was that selective use of adjuvant radiation in high-risk cases increases overall survival in early stage disease. It appears from the literature that this benefit is most pronounced in poorly differentiated cases, for which the survival advantage may approach 10% [6, 8, 11, 42], but since poorly differentiated cases comprise only a minority of early endometrial cancers, the overall increase in survival is only about 5%. It is possible that we overestimated the curative potential of adjuvant radiation leading to an overly high estimate of life-years gained and an underestimation of the cost per life-year gained. On the other hand, it is conceivable that the increase in cure rate may somewhat exceed 5%. Our sensitivity analysis shows that these factors do alter costeffectiveness estimates. In addition, any benefit from prolongation of survival in individuals who are not cured was not accounted for in this model. Cost is another potentially variable feature of the model. It is often difficult to define the true costs of medical care as opposed to charges. We obtained cost data where available and when cost data were not available discounted professional fees by about 60% to reflect current conditions. Our estimate for the average cost of hysterectomy was $9000, but this can vary significantly between individuals, particularly when significant complications which prolong hospital stay occur. The model presented in this paper assumed a constant cost of treatment for all women regardless of factors such as age and comorbidity that affect the average cost of treatment. This represents another area in which the model can be improved to increase its utility. Age and comorbidity increase the risk of complications and costs of hysterectomy (E. R. Myers, unpublished data). If the costs of hysterectomy are higher for those patients with
ENDOMETRIAL CANCER COST-EFFECTIVENESS
lower expected benefits, then our current model will underestimate the cost-effectiveness ratios. Variations in costs between hospitals might also lead to changes in the cost per life-year gained for a given hospital. In addition to uncertainty regarding the curative potential of adjuvant radiation and the omission of some factors that affect life-years gained and cost-effectiveness, another shortcoming of this analysis was its focus on life-years gained rather than the quality of those life-years. In our clinical experience, the quality of life of most women who survive endometrial cancer is excellent. However, it may be lower in some patients, especially in women with other chronic diseases. Again, since these patients would be expected to have both higher costs of treatment and lower quality-adjusted life-expectancies, our model would underestimate the cost per life-year (or qualityadjusted life-year) gained. Despite these limitations, we believe that this model serves as a reasonable starting point from which to begin to evaluate the cost-effectiveness of treatment of early endometrial cancer. The complexity of this model could be increased in the future by adding additional factors that affect survival and cost. We believe that, in addition to justifying the use of health care resources, this type of data could be useful in guiding and individualizing treatment of women with endometrial cancer. We are currently developing more sophisticated models that will take into account many of these other factors. REFERENCES 1. Creasman WT: Endometrial cancer: incidence, prognostic factors, diagnosis, and treatment. Semin Oncol 24:S141–150, 1997 2. Ries L, Kosary C, Hankey B, Miller B, Edwards B: SEER Cancer Statistics Review, 1975–1995. Bethesda, MD, National Cancer Institute, 1998 3. Larson D, Broste S, Krawisz B: Surgery without radiotherapy for primary treatment of endometrial cancer. Obstet Gynecol 91:355–359, 1998 4. Corn BW, Dunton CJ, Carlson JA, Xie Y, Valicenti RK: National trends in the surgical staging of corpus cancer: a pattern-of-practice survey. Obstet Gynecol 90:628 – 631, 1997 5. Leijon T, Rosenberg P, Boeryd B: Total abdominal hysterectomy and bilateral salpingo-oophorectomy: a sufficient treatment for patients with low risk endometrial carcinoma. Int J Gynecol Cancer 7:376 –380, 1997 6. Boz G, De Paoli A, Innocente R, Talamini R, Scarabelli C, Scozzari G, Trovo MG: Postoperative radiotherapy and surgery in stage I endometrial carcinoma: a 10-year experience. Tumori 84:52–56, 1998 7. Algan O, Tabesh T, Hanlon A, Hogan WM, Boente M, Lanciano RM: Improved outcome in patients treated with postoperative radiation therapy for pathologic stage I/II endometrial cancer. Int J Radiat Oncol Biol Phys 35:925–933, 1996 8. Garcia-Domenech RV, Inesta JM, Asins E, Aznar I, Llixiona J: Prognostic factors in endometrial carcinoma: risk groups and adjuvant radiotherapy. Eur J Gynaecol Oncol 18:164 –170, 1997
215
and responsible challenge—a personal perspective. Gynecol Oncol 66: 179 –189, 1997 11. Greven KM, Corn BW, Case D, Purser P, Lanciano RM: Which prognostic factors influence the outcome of patients with surgically staged endometrial cancer treated with adjuvant radiation? Int J Radiat Oncol Biol Phys 39:413– 418, 1997 12. Maneschi M, Maneschi F, Geraci P, Licata MD, Serruto A, Attanasio P, Curcio P: Surgical pathological staging of endometrial carcinoma and results of treatment. Eur J Gynaecol Oncol 13:30 –35, 1992 13. Cliby W, Clarke-Pearson D, Dodge R, Kohler M, Rodriguez G, Olt G, Soper J, Berchuck A, Soisson A: Acute morbidity and mortality associated with selective pelvic and para-aortic lymphadenectomy in the surgical staging of endometrial adenocarcinoma. J Gynecol Tech 1:19 –26, 1995 14. Gold MR, Siegel JE, Russell LB, Weinstein MC (eds): Cost-Effectiveness in Health and Medicine. New York, Oxford Univ. Press, 1996 15. Akkad AA, Habiba MA, Ismail N, Abrams K, al-Azzawi F: Abnormal uterine bleeding on hormone replacement: the importance of intrauterine structural abnormalities. Obstet Gynecol 86:330 –334, 1995 16. Ackerman I, Malone S, Thomas G, Franssen E, Balogh J, Dembo A: Endometrial carcinoma—relative effectiveness of adjuvant irradiation vs therapy reserved for relapse. Gynecol Oncol 60:177–183, 1996 17. Barhum M, Stein M, Ronsenblatt E, Dale J, Kuten A: Pathological stage I endometrial carcinoma: the role for adjuvant radiotherapy. Tumori 79:405– 409, 1993 18. Foote AJ, Proietto A: Stage 1 endometrial cancer: treatment modalities and factors influencing recurrence. Aust N Z J Obstet Gynaecol 34:448 – 452, 1994 19. Grigsby PW, Perez CA, Kuten A, Simpson JR, Garcia DM, Camel HM, Kao MS, Galakatos AE: Clinical stage I endometrial cancer: results of adjuvant irradiation and patterns of failure. Int J Radiat Oncol Biol Phys 21:379 –385, 1991 20. Kim Y, Niloff J: Endometrial carcinoma: analysis of recurrence in patients treated with a strategy minimizing lymph node sampling and radiation therapy. Obstet Gynecol 82:175–180, 1993 21. Nazeer T, Ballouk F, Malfetano JH, Figge H, Ambros RA: Multivariate survival analysis of clinicopathologic features in surgical stage I endometrioid carcinoma including analysis of HER-2/neu expression. Am J Obstet Gynecol 173:1829 –1834, 1995 22. Rose PG, Cha SD, Tak WK, Fitzgerald T, Reale F, Hunter RE: Radiation therapy for surgically proven para-aortic node metastasis in endometrial carcinoma. Int J Radiat Oncol Biol Phys 24:229 –233, 1992 23. Stryker JA, Podczaski E, Kaminski P, Velkley DE: Adjuvant external beam therapy for pathologic stage I and occult stage II endometrial carcinoma. Cancer 67:2872–2879, 1991 24. Sainz de la Cuesta R, Goff BA, Nikrui N, Flynn DF, Fuller AF Jr: Postoperative management of patients with stage Ib endometrial carcinoma. Eur J Gynaecol Oncol 17:338 –341, 1996 25. Konski AA, Bracy PM, Jurs SG, Weiss SJ, Zeidner SR: Cost minimization analysis of various treatment options for surgical stage I endometrial carcinoma. Int J Radiat Oncol Biol Phys 37:367–373, 1997 26. Adami HO, McLaughlin J, Ekbom A, Berne C, Silverman D, Hacker D, Persson I: Cancer risk in patients with diabetes mellitus. Cancer Causes Control 2:307–314, 1991 27. Le Marchand L, Wilkens LR, Mi MP: Early-age body size, adult weight gain and endometrial cancer risk. Int J Cancer 48:807– 811, 1991
9. Morrow C, Bundy B, Kurman R, Creasman W, Heller P, Homesley H, Graham J: Relationship between surgical–pathological risk factors and outcomes in clinical stage I and II carcinoma of the endometrium: a Gynecologic Oncology Group study. Gynecol Oncol 40:55– 65, 1991
28. Maatela J, Aromaa A, Salmi T, Pohja M, Vuento M, Gronroos M: The risk of endometrial cancer in diabetic and hypertensive patients: a nationwide record-linkage study in Finland. Ann Chir Gynaecol Suppl 208:20 –24, 1994
10. Boronow R: Surgical staging of endometrial cancer: evolution, evaluation,
29. Wideroff L, Gridley G, Mellemkjaer L, Chow WH, Linet M, Keehn S,
216
ASHIH ET AL.
Borch-Johnsen K, Olsen JH: Cancer incidence in a population-based cohort of patients hospitalized with diabetes mellitus in Denmark. J Natl Cancer Inst 89:1360 –1365, 1997 30. Patterns of extra mortality for the common impairments. Stat Bull Metropol Life Insur Co 50:6 –10, 1969
37.
31. Life expectancy among diabetics in the United States. Stat Bull Metropol Life Insur Co 60:6 – 8, 1979
38.
32. Life expectancy among diabetics—II. Stat Bull Metropol Life Insur Co 60:11–13, 1979 33. Andersen B, Ostberg J: Survival rates in surgery of the aged: assessment of long-term prognosis according to coexisting diseases. Gerontol Clin 14:354 –360, 1972 34. Mark DB, Hlatky MA, Califf RM, Naylor CD, Lee KL, Armstrong PW, Barbash G, White H, Simoons ML, Nelson CL: Cost effectiveness of thrombolytic therapy with tissue plasminogen activator as compared with streptokinase for acute myocardial infarction. N Engl J Med 332:1418 – 1424, 1995 35. Mark D. Economic Analysis Methods and Endpoints, in Califf R, Mark D, Wagner G (eds): Acute Coronary Care in the Thrombolytic Era. St. Louis, Mosby, 1995, pp 167–182 36. Laupacis A, Feeny D, Detsky AS, Tugwell PX: How attractive does a new
39.
40. 41.
42.
technology have to be to warrant adoption and utilization? Tentative guidelines for using clinical and economic evaluations. Can Med Assoc J 146:473– 481, 1992 Partridge E, Shingleton H, Menck H: The National Cancer Data Base report on endometrial cancer. J Surg Oncol 61:111–123, 1996 Hill HA, Eley JW, Harlan LC, Greenberg RS, Barrett RJ, Chen VW: Racial differences in endometrial cancer survival: the black/white cancer survival study. Obstet Gynecol 88:919 –926, 1996 Liu JR, Conaway M, Rodriguez GC, Soper JT, Clarke-Pearson DL, Berchuck A: Relationship between race and interval to treatment in endometrial cancer. Obstet Gynecol 86:486 – 490, 1995 Plaxe SC, Saltzstein SL: Impact of ethnicity on the incidence of high-risk endometrial carcinoma. Gynecol Oncol 65:8 –12, 1997 Barrett RJ, Harlan LC, Wesley MN, Hill HA, Chen VW, Clayton LA, Kotz HL, Eley JW, Robboy SJ, Edwards BK: Endometrial cancer: stage at diagnosis and associated factors in black and white patients. Am J Obstet Gynecol 173:414 – 423, 1995 Grigsby PW, Perez CA, Kuten A, Simpson JR, Garcia DM, Camel HM, Kao MS, Galakatos AE: Clinical stage I endometrial cancer: prognostic factors for local control and distant metastasis and implications of the new FIGO surgical staging system. Int J Radiat Oncol Biol Phys 22:905–911, 1992