SEARCH

SEARCH BY CITATION

Keywords:

  • prostate cancer;
  • treatment;
  • costs;
  • patterns of care;
  • health services

Abstract

  1. Top of page
  2. Abstract
  3. MATERIALS AND METHODS
  4. RESULTS
  5. DISCUSSION
  6. Acknowledgements
  7. CONFLICT OF INTEREST DISCLOSURES
  8. REFERENCES

BACKGROUND.

Data regarding costs of prostate cancer treatment are scarce. This study investigates how initial treatment choice affects short-term and long-term costs.

METHODS.

This retrospective, longitudinal cohort study followed prostate-cancer cases diagnosed in 2000 for 5 years using the Surveillance, Epidemiology, and End Results (SEER)-Medicare database. Men age ≥66 years, in Medicare fee for service, diagnosed with clinically localized prostate cancer in 2000 while residing in a SEER region, were matched to noncancer controls using age, sex, race, region, comorbidity, and survival. On the basis of treatment received during the first 9 months postdiagnosis, patients were assigned to watchful waiting, radiation, hormonal therapy, hormonal + radiation, and surgery (may have received other treatments). Incremental costs for prostate cancer were the difference in costs for prostate cancer cases versus matched controls. Costs were divided into initial treatment (months −1 to 12), long-term (each 12 months thereafter), and total (months −1 to 60). Sensitivity analyses excluded the last 12 months of life.

RESULTS.

A total of 13,769 prostate-cancer cases were matched to 13,769 noncancer controls. Watchful waiting had the lowest initial treatment ($4270) and 5-year total costs ($9130). Initial treatment costs were highest for hormonal + radiation ($17,474) and surgery ($15,197). At $26,896, 5-year total costs were highest for hormonal therapy only followed by hormonal + radiation ($25,097) and surgery ($19,214). After excluding the last 12 months of life, total costs were highest for hormonal + radiation ($23,488) and hormonal therapy ($23,199).

CONCLUSIONS.

Patterns of costs vary widely based on initial treatment. These data can inform patients and clinicians considering treatment options and policy makers interested in patterns of costs. Cancer 2010. © 2010 American Cancer Society.

Prostate cancer is the most frequently diagnosed cancer in men, and because >90% of prostate cancer is clinically localized at diagnosis, the relative 5-year survival rate approaches 100%, the relative 10-year survival rate is 93%, and the relative 15-year survival rate is 79%.1 Over 60% of prostate cancers are diagnosed in men older than 65 years, making it an important issue in the Medicare population.2 Because of the high survival rates, the long-term follow-up care for these men is critical.

With many treatment options, a paucity of comparative effectiveness data, and patients across a wide range of ages, treatment for clinically localized prostate cancer remains controversial.3 Common treatments include surgery, radiation therapy (external beam and brachytherapy), hormonal therapy, or combinations of the above. Because of the indolent nature of the disease, some men are appropriate for watchful waiting and do not require any treatment. However, few randomized controlled trials across modalities have been conducted to inform treatment choice. Patient characteristics such as age, comorbidities, and life expectancy may influence treatment decisions. Some patients, particularly younger ones, may be appropriate for >1 treatment option. Because the treatment options have different side effect profiles, patient preferences are particularly important in prostate cancer treatment decisions.4

The different treatment options also may have important economic consequences. In addition to initial treatment cost differences, prostate cancer therapies can have important short-term and long-term sequelae requiring various amounts of additional treatment.5 Some treatments rely on outpatient care, whereas others require inpatient services. Because of differences in insurance coverage, the setting in which care is delivered may have important economic implications. Yet few studies have investigated how treatment choice affects both the short-term and long-term costs. This study examines how initial treatment choice affects both the short-term and long-term costs for clinically localized prostate cancer.

MATERIALS AND METHODS

  1. Top of page
  2. Abstract
  3. MATERIALS AND METHODS
  4. RESULTS
  5. DISCUSSION
  6. Acknowledgements
  7. CONFLICT OF INTEREST DISCLOSURES
  8. REFERENCES

Study Design and Objectives

This was a retrospective, longitudinal cohort study. This study aimed to compare the incremental costs of prostate cancer based on initial treatment received. To calculate incremental costs, we followed men diagnosed with prostate cancer in the year 2000 from 1 month before diagnosis6 to 5 years postdiagnosis (or until death). We compared the costs of their care to controls without cancer who were matched 1 to 1 to cases, using age, sex, race, geographic region, comorbidity, and survival.

Data Source

We used the Surveillance, Epidemiology, and End Results (SEER)-Medicare linked database, which combines clinical data from the SEER registries with Medicare administrative claims.7 We used the SEER registry to obtain data on the cancer diagnosis and used the Medicare claims to follow the subjects' health services utilization longitudinally. In the year 2000, the SEER-17 registries covered approximately 26% of the US population8; 16 registries participate in the SEER-Medicare linkage. Data from a 5% random sample of Medicare patients in the SEER regions were used to match controls to cancer cases.

Study Population

We included prostate cancer cases meeting these eligibility criteria: 1) diagnosis of clinically localized prostate cancer in 2000 while residing in a SEER region, 2) first or only cancer in the SEER registry, 3) survived at least the initial treatment period (ie, 9 months postdiagnosis), 4) age 66 years or older, and 5) enrolled continuously in the Medicare fee-for-service program (Parts A and B) from 1 year before diagnosis (to calculate baseline comorbidity) through 5 years postdiagnosis or until death. Patients enrolled in Medicare managed care at anytime during the study period were excluded. Controls were matched to cases on an individual basis using age (exact), sex (all male), race (white, black, other), comorbidity index (0, 1, 2+), geographic region, and duration of survival (based on Zeliadt et al9).

Variables

By using the SEER data, we assessed age, race, geographic region, urban/rural location, survival, tumor grade, and socioeconomic status. Socioeconomic status was separated into quintiles and used as a categorical variable. By using the Medicare claims, we constructed a comorbidity index as a categorical variable (0, 1, 2+).10-12 Treatment groups were defined based on the treatments received during the first 9 months after diagnosis, similarly to Burkhardt et al.6 By using the Medicare claims, we assigned patients to the following treatment groups: watchful waiting (if they received no treatment in the first 9 months), radiation therapy only (including external beam and/or brachytherapy), hormonal therapy only, hormonal + radiation combination, and surgery (may have also received other treatments). It is important to note that men may have gone on to receive other treatments (eg, a man assigned to watchful waiting may have received surgery later), but the treatment group assignment was based solely on treatments during the first 9 months from diagnosis.

Costs were assessed using all Medicare paid amounts from all standard analytic files, including inpatient, outpatient, physician/supplier, hospice, home health, and durable medical equipment. Costs were categorized as inpatient, outpatient, and other and combined into overall costs. We calculated costs in the short-term (months −1 through 12 from diagnosis), the long-term (each 12-month period thereafter), and total 5-year costs (months −1 through 60 from diagnosis). We adjusted for the combination of inflation and different Medicare reimbursements by geographic region using methods similar to those of Brown et al13 and Zeliadt et al,9 with data provided by Information Management Services, Inc. (IMC).

Statistical Analysis

We calculated the sociodemographic characteristics for prostate cancer cases and controls, and compared the characteristics across the 5 treatment groups. We used analysis of variance (ANOVA) to compare continuous variables and chi-square tests for categorical variables. We also conducted multinomial logistic analyses using only prostate cancer cases to determine which factors (age, race, comorbidity index, geographic region, urban/rural location, socioeconomic status, grade) are associated with initial treatment choice.

To calculate the incremental costs of prostate cancer, we adjusted for geographic region,9, 13 inflated all costs to 2007 dollars, subtracted the costs of controls from the costs of their matched prostate cancer cases, and calculated the mean incremental costs for each treatment group. The increment was calculated for overall costs as well as for costs grouped by setting (inpatient, outpatient, other). Dollars were discounted 3% for each year beyond the first year.14 For comparisons in each 12-month period, we included only patients who had survived to that time. Differences in incremental costs over time by treatment group were tested with linear mixed effects models that included fixed effects for time and a random effect for each matched pair. Differential changes in incremental costs over time between treatment groups were tested using the same approach. P values for differences are from model-based likelihood ratio tests.

Because patterns of costs are quite different in the period before death,15 we conducted a sensitivity analysis excluding costs for the last 12 months of life. Finally, to minimize the influence of selection bias in treatment groups, a multinomial logistic regression model with treatment group as the outcome and including age, race, comorbidity, SEER region, urban/rural location, socioeconomic status, and grade as predictors was used to generate a propensity score for each matched pair. The inverses of these scores were used as weights to test for differences in the initial treatment and overall 5-year costs across treatment groups.16, 17

Because of the large sample size, we focus on the actual differences between groups rather than P values. When P values are reported, we used a threshold of P<.0001 for statistical significance. The Johns Hopkins School of Medicine Institutional Review Board deemed this study exempt.

RESULTS

  1. Top of page
  2. Abstract
  3. MATERIALS AND METHODS
  4. RESULTS
  5. DISCUSSION
  6. Acknowledgements
  7. CONFLICT OF INTEREST DISCLOSURES
  8. REFERENCES

To obtain our analytic cohort, we started with 37,448 men diagnosed with prostate cancer (site 54) in the year 2000. After limiting to men with clinically localized disease and a valid month of diagnosis, 33,261 men remained. Reasons for exclusion were age younger than 66 years (n = 7652), surviving <9 months (n = 614), not being eligible for Medicare Part A and B during the study period (n = 2734), enrollment in managed care (n = 7632), and no matched control (n = 860). Thus, the final cohort included 13,769 prostate cancer cases matched to 13,769 controls. The most common treatment was hormonal + radiation combination (n = 3992; 29%), followed by watchful waiting (n = 2805; 20%), radiation only (n = 2582; 19%), surgery (n = 2200; 16%), and hormonal therapy only (n = 2190; 16%). As evidence that men in the watchful waiting group were being monitored closely, cases had more physician visits (ranging from an average of 9.9 visits in Year 1 to a low of 7.6 visits in Year 2) than their matched controls (ranging from 7.0 visits in Year 5 to 6.4 visits in Year 2), with visits to urologists predominantly accounting for the difference between groups. Within the surgery group, 118 (5%) men also received radiation therapy, 319 (15%) men also received hormonal therapy, and 67 (3%) men also received both hormonal and radiation treatment in the 9 months postdiagnosis. The average age ranged from 70.5 years for surgery to 78.9 years for hormonal therapy only; between 64% and 77% of men had no comorbidities, and between 85% and 90% of the men in each treatment group were white (Table 1). The percentage of men who survived at least 4 years ranged from 95% in the surgery group to 71% in the hormonal therapy only group.

Table 1. Characteristics of Prostate Cancer Cases and Matched Controls
CharacteristicWatchful Waiting, Survivors and Controls, n=2805aRadiotherapy, Survivors and Controls, n=2582aHormonal Therapy, Survivors and Controls, n=2190aHormonal + Radiotherapy, Survivors and Controls, n=3992aSurgery, Survivors and Controls, n=2200aTotal, n=13,769aANOVA/ Chi-Square, P
  • ANOVA indicates analysis of variance; SD, standard deviation; NA, not applicable; SEER, Surveillance, Epidemiology, and End Results.

  • a

    Sample sizes are equal for prostate cancer cases and control cohorts. Because of matching, descriptive statistics are the same for cases and controls.

  • b

    Applies to prostate cancer cases only.

Age, mean y (SD)76.5 (6.1)73.6 (4.4)78.9 (6.0)74.1 (4.4)70.5 (3.5)NA<.0001
Race, No. (%)
 White2386 (85)2318 (90)1892 (86)3510 (88)1919 (87)12,025 (87)<.0001
 Black289 (10)164 (6)179 (8)264 (7)140 (6)1036 (8) 
 Other130 (5)100 (4)119 (5)218 (5)141 (6)708 (5) 
Comorbidity index, No. (%)
 01895 (68)1807 (70)1392 (64)2707 (68)1696 (77)9497 (69)<.0001
 1513 (18)512 (20)461 (21)810 (20)331 (15)2627 (19) 
 2+397 (14)263 (10)337 (15)475 (12)173 (8)1645 (12) 
Time until death/censoring, No. (%)
 273 days-2 years217 (8)69 (3)278 (13)130 (3)30 (1)724 (5)<.0001
 >2 years-3 years157 (6)70 (3)177 (8)124 (3)35 (2)563 (4) 
 >3 years-4 years171 (6)84 (3)189 (9)128 (3)34 (2)606 (4) 
 >4 years2260 (80)2359 (91)1546 (71)3610 (91)2101 (95)11,876 (86) 
Prostate cancer grade, No. (%)b
 Well differentiated308 (11)112 (4)84 (4)126 (3)44 (2)674 (5)<.0001
 Moderately differentiated1879 (67)2112 (82)1232 (56)2635 (66)1635 (74)9493 (69) 
 Poorly differentiated/ undifferentiated394 (14)285 (11)790 (36)1121 (28)499 (23)3089 (22) 
 Unknown224 (8)73 (3)84 (4)110 (3)22 (1)513 (4) 
SEER region
 San Francisco98 (3)70 (3)48 (2)106 (3)77 (4)399 (3)<.0001
 Connecticut169 (6)191 (7)112 (5)311 (8)98 (4)881 (6) 
 Detroit305 (11)391 (15)201 (9)318 (8)209 (10)1424 (10) 
 Hawaii27 (1)39 (2)33 (2)73 (2)14 (1)186 (1) 
 Iowa237 (8)185 (7)219 (10)224 (6)156 (7)1021 (7) 
 New Mexico90 (3)57 (2)32 (1)64 (2)56 (3)299 (2) 
 Seattle169 (6)154 (6)85 (4)170 (4)149 (7)727 (5) 
 Utah104 (4)73 (3)75 (3)113 (3)119 (5)484 (4) 
 Atlanta and Rural Georgia58 (2)133 (5)31 (1)81 (2)52 (2)355 (3) 
 San Jose64 (2)41 (2)56 (3)61 (2)43 (2)265 (2) 
 Los Angeles222 (8)131 (5)159 (7)187 (5)244 (11)943 (7) 
 Greater California493 (18)436 (17)396 (18)592 (15)454 (21)2371 (17) 
 Kentucky236 (8)183 (7)185 (8)316 (8)147 (7)1067 (8) 
 Louisiana164 (6)113 (4)243 (11)239 (6)179 (8)938 (7) 
 New Jersey369 (13)385 (15)315 (14)1137 (28)203 (9)2409 (17) 

As demonstrated by the ANOVA/chi-square tests (Table 1) and the multinomial logistic regression (Table 2), patient characteristics had an important association with treatment receipt. Compared with men in watchful waiting, men on only hormonal therapy were older (relative risk [RR] per year, 1.07; 95% confidence interval [CI], 1.05-1.08), and men in the radiation only (RR per year, 0.89; 95% CI, 0.88-0.90), hormonal + radiation (RR per year, 0.90; 95% CI, 0.90-0.91), and surgery (RR per year, 0.73; 95% CI, 0.72-0.75) groups were younger (all P < .0001). Compared with white men, black men were less likely to receive each of the treatments except hormonal therapy only versus watchful waiting with RRs ranging from 0.34 to 0.46 (all P < .0001). Compared with men in the lowest socioeconomic status quintile, men in the highest quintile were more likely to receive radiation only (RR, 1.52; 95% CI, 1.26-1.84) or surgery (RR, 1.77; 95% CI, 1.43-2.19) versus watchful waiting (both P < .0001). Compared with men with well-differentiated disease, men with moderately differentiated or poorly differentiated/undifferentiated disease were more likely to receive each of the treatments versus watchful waiting (all P < .0001).

Table 2. Relative Risk Ratios From Multinomial Logistic Regression Models for Associations Between Covariates and Treatment Group Assignmenta
CharacteristicRelative Risk Ratio (95% CI)
Watchful WaitingRadiotherapyHormonalHormonal + RadiotherapySurgery
  • 95% CI indicates 95% confidence interval; Ref, reference category.

  • a

    Also adjusted for Surveillance, Epidemiology, and End Results region.

  • b

    P < .0001.

AgeRef0.89 (0.88-0.90)b1.07 (1.05-1.08)b0.90 (0.90-0.91)b0.73 (0.72-0.75)b
Race
 WhiteRefRefRefRefRef
 BlackRef0.39 (0.31-0.50)b0.75 (0.60-0.95)0.46 (0.37-0.56)b0.34 (0.27-0.44)b
 OtherRef0.85 (0.63-1.15)1.15 (0.86-1.55)1.24 (0.96-1.61)1.52 (1.13-2.04)
Comorbidity index
 0RefRefRefRefRef
 1Ref1.11 (0.96-1.28)1.18 (1.02-1.37)1.11 (0.97-1.27)0.84 (0.71-0.99)
 2+Ref0.80 (0.67-0.96)1.07 (0.90-1.27)0.90 (0.77-1.05)0.67 (0.54-0.83)
Urban/rural location
 RuralRefRefRefRefRef
 UrbanRef1.48 (1.18-1.85)1.21 (0.97-1.49)1.26 (1.03-1.54)1.54 (1.20-1.96)
Socioeconomic status
 1st quintileRefRefRefRefRef
 2nd quintileRef1.26 (1.05-1.50)1.00 (0.85-1.19)1.17 (1.00-1.37)1.31 (1.06-1.62)
 3rd quintileRef1.09 (0.91-1.31)0.83 (0.69-0.99)1.18 (1.01-1.39)1.30 (1.05-1.60)
 4th quintileRef1.22 (1.02-1.47)0.84 (0.70-1.02)1.12 (0.95-1.33)1.44 (1.16-1.77)
 5th quintileRef1.52 (1.26-1.84)b0.80 (0.65-0.98)1.30 (1.09-1.55)1.77 (1.43-2.19)b
Prostate cancer grade
 Well differentiatedRefRefRefRefRef
 Moderately differentiatedRef3.09 (2.45-3.91)b2.44 (1.89-3.16)b3.60 (2.88-4.50)b7.51 (5.31-10.61)b
 Poorly differentiated/undifferentiatedRef2.34 (1.78-3.08)b6.72 (5.10-8.86)b8.43 (6.58-10.79)b13.38 (9.26-19.35)b
 UnknownRef0.91 (0.64-1.29)1.50 (1.06-2.12)1.21 (0.88-1.66)1.09 (0.64-1.86)

Table 3 and Figure 1 present the patterns of incremental costs of prostate cancer over 5 years for each treatment group, adjusted for inflation and geographic region. Overall initial treatment costs are highest for hormonal + radiation ($17,474), followed by surgery ($15,197), with watchful waiting lowest ($4270). After the first year, costs for all treatment groups fall sharply and remain relatively steady, except for watchful waiting, which drops to $733 in Year 2 then rises to $1836 in Year 3 and $2090 in Year 4, with an average total cost of $9130 across the 5 years. Total costs were highest for hormonal therapy only ($26,896), which had relatively low initial treatment costs, but the highest costs in Years 2 through 5. Hormonal + radiation had the second highest total costs ($25,097) across the 5 years, although 70% of these costs were experienced during Year 1. Similarly, for radiation only and surgery, 78% and 79% of total costs were experienced in Year 1, respectively.

thumbnail image

Figure 1. Overall incremental costs of prostate cancer by treatment group is shown for Years 1 through 5.

Download figure to PowerPoint

Table 3. Mean (SD) Incremental Costs of Prostate Cancer by Treatment Group (Overall, Inpatient, Outpatient, Other): Years 1-5a
GroupYear 1 (Months −1 to 12)Year 2 (Months 13 to 24)Year 3 (Months 25 to 36)Year 4 (Months 37 to 48)Year 5 (Months 49 to 60)Totalb (Months −1 to 60)
  • SD indicates standard deviation; RT, radiotherapy.

  • a

    Adjusted for inflation and geographic region. All changes over time by treatment group were statistically significant (P < .0001), and an interaction term to test for differential changes in cost differences over time across treatment groups was also statistically significant (P < .0001) for all groups.

  • b

    Total costs are not the sum of Year 1 through 5 costs because costs were only calculated for years during which the prostate cancer patient survived.

Overall
 Watchful waiting4270 (25,304)733 (24,420)1836 (26,195)2090 (25,905)1278 (27,594)9130 (65,604)
 RT only12,120 (21,138)1237 (22,492)1089 (23,531)750 (23,515)775 (25,931)15,589 (58,381)
 Hormonal only11,034 (28,821)6561 (28,073)4166 (26,723)4571 (29,521)4895 (27,570)26,896 (67,200)
 Hormonal + RT17,474 (21,189)2854 (27,364)3166 (28,585)1367 (24,769)1030 (25,301)25,097 (67,715)
 Surgery15,197 (19,453)1192 (19,074)1385 (20,634)820 (17,665)997 (25,371)19,214 (53,955)
Inpatient
 Watchful waiting2288 (19,841)−257 (19,003)819 (20,771)913 (20,200)358 (21,617)3738 (47,134)
 RT only64 (16,254)152 (18,151)84 (19,099)−62 (18,076)100 (20,674)314 (42,886)
 Hormonal only3060 (23,246)1322 (22,761)177 (21,083)489 (23,425)1633 (21,244)5802 (49,125)
 Hormonal + RT876 (15,960)813 (22,748)1391 (23,582)42 (19,069)0 (19,434)2961 (50,074)
 Surgery8765 (15,090)−66 (14,186)75 (15,995)−249 (12,454)255 (19,483)8766 (36,516)
Outpatient
 Watchful waiting1705 (6705)901 (7430)926 (7884)962 (7840)778 (8581)4679 (23,168)
 RT only12,061 (8577)1113 (6840)1049 (7390)895 (7833)697 (8065)15,436 (23,229)
 Hormonal only7807 (7871)5110 (7679)3852 (8123)3871 (9110)3029 (9395)20,395 (24,987)
 Hormonal + RT16,656 (8764)2047 (7298)1713 (7771)1318 (8500)986 (8632)22,103 (24,972)
 Surgery6152 (6476)1295 (6264)1315 (7766)1121 (7304)823 (9196)10,322 (22,814)
Other
 Watchful waiting277 (3575)89 (3511)92 (3419)216 (3840)142 (3741)712 (11,520)
 RT only−5 (1474)−28 (2083)−43 (2185)−82 (2607)−23 (2779)−161 (6808)
 Hormonal only167 (3667)129 (3477)138 (3715)211 (3906)232 (3681)698 (10,751)
 Hormonal + RT−57 (2481)−7 (2662)62 (2741)7 (2768)43 (2914)33 (8432)
 Surgery281 (2949)−36 (4258)−5 (2392)−53 (2563)−80 (2790)126 (9878)

Because 23% of men in the surgery group also received either radiation therapy, hormonal therapy, or both, we conducted a sensitivity analysis to estimate the initial and total costs across the 5 years for the 1696 men (77%) who received surgery only. Initial treatment costs for men who received surgery only were $13,730 (vs $15,197), and total costs across the 5 years were $16,327 (vs $19,214).

In terms of costs by setting of care, initial inpatient costs were highest for surgery ($8765), followed by $3060 for hormonal therapy only and $2288 for watchful waiting. Initial outpatient costs were highest for hormonal + radiation ($16,656), followed by radiation only ($12,061). Across all 5 years and all treatment groups, other costs tended to be low.

Table 4 presents the sensitivity analysis, excluding the last 12 months of life for subjects who died. There was little change in the patterns of initial costs; hormonal + radiation was still most expensive ($17,384), followed by $15,083 for surgery, with the lowest costs for watchful waiting ($3981). However, the sensitivity analysis did affect the rankings of total costs. Instead of hormonal therapy being the most expensive, it was slightly less expensive ($23,199) versus hormonal + radiation ($23,488).

Table 4. Mean (SD) Overall Incremental Costs of Prostate Cancer by Treatment Group, Excluding Last 12 Months of Life: Years 1 to 5a
Group, Overall CostsYear 1 (Months −1 to 12)Year 2 (Months 13 to 24)Year 3 (Months 25 to 36)Year 4 (Months 37 to 48)Year 5 (Months 49 to 60)Totalb (Months −1 to 60)
  • SD indicates standard deviation; RT, radiotherapy.

  • a

    Adjusted for inflation and geographic region. All changes over time by treatment group were statistically significant (P<.0001), and an interaction term to test for differential changes in cost differences over time across treatment groups was also statistically significant (P<.0001) for all groups.

  • b

    Total costs are not the sum of Year 1 through 5 costs because costs were only calculated for years during which the prostate cancer patient survived.

Watchful waiting3981 (22,541)1029 (20,141)2033 (23,024)2189 (19,787)1095 (21,683)8845 (54,134)
RT only12,194 (17,978)890 (20,382)869 (21,446)657 (18,908)825 (20,760)14,930 (50,365)
Hormonal only10,410 (24,237)6142 (23,391)4287 (20,703)4995 (22,414)2676 (21,094)23,199 (53,514)
Hormonal + RT17,384 (20,032)2419 (23,975)2701 (26,586)1268 (19,519)762 (19,745)23,488 (58,696)
Surgery15,083 (18,221)1291 (15,808)1191 (18,058)860 (15,678)591 (20,357)18,628 (46,946)

Finally, we analyzed initial treatment (months −1 to 12) and 5-year total (months −1 to 60) costs using a propensity score to minimize the impact of selection bias associated with treatment receipt (Table 5). In most cases, use of the propensity score adjusted the means only slightly, likely because of the large sample size. The propensity score had the greatest impact on the initial treatment costs for the watchful waiting ($4270 vs $3936) and hormonal therapy ($11,034 vs $11,710) groups, a decrease of 8% and an increase of 6%, respectively. For total cost estimates, use of the propensity score was associated with a 7% decrease in watchful waiting ($9130 vs $8535) and a 7% increase in radiation therapy ($15,589 vs $16,653).

Table 5. Propensity Score Adjusted Mean Incremental Initial Treatment and 5-Year Total Costs by Treatment Group
GroupInitial Treatment (Months −1 to 12)Five-Year Total (Months −1 to 60)
  1. RT indicates radiotherapy.

Watchful waiting3936 (3078-4794)8535 (6223-10,847)
RT12,319 (11,419-13,219)16,653 (14,228-19,078)
Hormonal therapy11,710 (10,857-12,563)26,643 (24,345-28,941)
Hormonal + RT17,021 (16,121-17,921)24,767 (22,344-27,190)
Surgery15,556 (14,835-16,277)19,481 (17,538-21,424)

DISCUSSION

  1. Top of page
  2. Abstract
  3. MATERIALS AND METHODS
  4. RESULTS
  5. DISCUSSION
  6. Acknowledgements
  7. CONFLICT OF INTEREST DISCLOSURES
  8. REFERENCES

Despite the high prevalence of clinically localized prostate cancer and the controversies associated with determining initial treatment, few studies have examined both the short-term and long-term cost implications of the various treatment options. In this study, we examined the characteristics of men in each of the treatment groups, as well as the short-term and long-term costs associated with treatment choice. As expected, tumor grade and age had an important association with treatment group assignment. Race (and to a lesser extent socioeconomic status) were also associated with treatment group, indicative of potential disparities in care.

The costs of the treatment choices varied widely, not only during the initial treatment period but across all 5 years. Whereas hormonal therapy was the second least expensive for initial treatment, it had the highest 5-year total costs. This cost pattern may result from the relatively high mortality rate in this group, although after excluding the last 12 months of life, hormonal therapy still had the second highest 5-year total costs. To attempt to address the selection bias associated with treatment, we used a propensity score for initial and total costs, but this adjustment had little effect on our findings, likely because of our large sample size, suggesting that our estimates are robust.

Treatment costs tended to decrease from Year 1 and remain stable for most treatment groups, except costs for the watchful waiting group increased after an initial decrease. The different treatment options also varied in terms of setting of care, which has important implications for insurance coverage. For example, in fee-for-service Medicare, Medicare pays the total amount for 60 inpatient days after beneficiaries reach their $1100 deductible for each “spell of illness.” For outpatient services, the annual deductible is only $155, but Medicare generally covers only 80%, with beneficiaries paying 20% coinsurance.18

The study most similar to this study was conducted by Wilson et al19 using the CaPSURE database and self-reported health services utilization questionnaires. They examined costs for the first 6 months of treatment and for 5.5 years of follow-up for men undergoing radical prostatectomy, brachytherapy, cryotherapy, external beam radiation, androgen deprivation (hormonal) therapy, and watchful waiting. Treatment costs during the first 6 months ranged from $2586 for watchful waiting to $24,204 for external beam radiation. After initial treatment, annual costs were highest for androgen deprivation therapy ($12,590) and lowest for watchful waiting ($5843). Because of differences in study populations (the Wilson study was not limited to the Medicare population), treatment groups, cost analysis approach, and follow-up periods, the data are not directly comparable to the findings presented here. Another CaPSURE study only examined first-year costs.20

Other studies have also looked at the costs of localized prostate cancer using the SEER-Medicare linked database. Burkhardt et al6 looked at initial treatment costs of radiation therapy and radical prostatectomy, but compared only the short-term costs of 2 treatments. Other studies also examined initial treatment costs only.21, 22 Zeliadt et al9 used SEER-Medicare, but their study was more focused on how prostate-specific antigen testing affected initial treatment costs. By using a different database, Krupski et al examined costs of care specifically related to androgen deprivation therapy.23 The prostate cancer cost studies included in the review by Yabroff et al24 either focused on short-term costs only or did not compare costs by initial treatment.

Thus, in contrast to most of the previous research in this area, our study includes a variety of treatment groups and both short-term and long-term follow-up. Although the Wilson et al19 study was similar, it relied on self-reported health services utilization data, which may be incomplete. An advantage of the Wilson et al19 study, however, was that it captured noncovered health services (eg, outpatient pharmaceuticals), whereas we were limited to claims for Medicare-covered services.

Other limitations of our analysis also warrant mention. First, because we used the SEER-Medicare database, our sample does not include younger men or men in Medicare managed care. Second, the study may be subject to some misclassification bias, because we relied on the claims data to assign men to treatment groups. Missing or inaccurate codes may have resulted in some men being assigned to the wrong treatment group. Also, some men may have taken >9 months to decide on their treatment choice. Although we made an effort to make the treatment groups as homogenous as possible, heterogeneity in treatments received remains, with for example various lengths of hormonal treatment. Third, we have classified men who received no active treatment in the first 9 months postdiagnosis as watchful waiting. Active surveillance is increasingly used in men with clinically insignificant cancer and a low risk for progression. However, we did find that men in the watchful waiting group had a greater number of physician visits overall, and urologist visits in particular. Fourth, 23% of men in the surgery treatment group also received other treatments; however, a sensitivity analysis that included men who underwent surgery only provides a basis for comparison. Fifth, our cost estimates include only direct medical costs and do not include other direct costs or any indirect costs. Sixth, our cost estimates relied on the incremental costs of men with prostate cancer compared with matched men without a cancer diagnosis. This approach has been used previously,13 and has the advantage of providing a more comprehensive estimate of the costs of prostate cancer, but it does not assess which health services were specifically related to prostate cancer. Seventh, we only followed men for 5 years postdiagnosis. Although this length of follow-up provides useful information on the cost implications beyond initial treatment, important costs may be incurred >5 years out, and these costs were not included in our study. Eighth, prostate cancer treatment is subject to important selection bias. Although we used a propensity score to compare the costs of care for like groups of men, such adjustment may not completely account for differences between men in the various treatment groups. Ninth and finally, this study was not a cost-effectiveness or comparative effectiveness analysis. Cost is only 1 consideration in treatment selection, and quality of life and other outcomes are also important.

Despite these limitations, the findings of this study are informative for a variety of purposes. Because most prostate cancer patients will survive their disease, it is important to consider both the short-term and long-term cost implications of the various treatment options. These data demonstrate that patterns of care vary widely by treatment group. We also assessed the robustness of our cost estimates in sensitivity analyses that excluded the last 12 months of life and that used a propensity score. Finally, we examined the settings of care accounting for the costs by treatment group. Because insurance coverage varies by setting of care, this information is useful. These data can inform patients and clinicians considering treatment options and policy makers who may be interested in patterns of costs.

Acknowledgements

  1. Top of page
  2. Abstract
  3. MATERIALS AND METHODS
  4. RESULTS
  5. DISCUSSION
  6. Acknowledgements
  7. CONFLICT OF INTEREST DISCLOSURES
  8. REFERENCES

We thank Hsin-Chieh (Jessica) Yeh, PhD, Core Faculty of the Johns Hopkins General Internal Medicine Methods Core, for assistance in preparing the data for analysis.

CONFLICT OF INTEREST DISCLOSURES

  1. Top of page
  2. Abstract
  3. MATERIALS AND METHODS
  4. RESULTS
  5. DISCUSSION
  6. Acknowledgements
  7. CONFLICT OF INTEREST DISCLOSURES
  8. REFERENCES

Supported by the Ho Ching Yang Memorial Faculty Fellowship Award from the Johns Hopkins Bloomberg School of Public Health. Drs. Snyder and Carducci are also supported by a Mentored Research Scholar Grant from the American Cancer Society (MRSG-08-011-01-CPPB).

REFERENCES

  1. Top of page
  2. Abstract
  3. MATERIALS AND METHODS
  4. RESULTS
  5. DISCUSSION
  6. Acknowledgements
  7. CONFLICT OF INTEREST DISCLOSURES
  8. REFERENCES