The current study used the linked Surveillance, Epidemiology and End Results (SEER)-Medicare database. The interpretation and reporting of these data are the sole responsibility of the authors. The authors acknowledge the efforts of the Applied Research Program, National Cancer Institute (Bethesda, MD); the Office of Research, Development and Information, Centers for Medicare and Medicaid Services (CMS; Baltimore, MD); Information Management Services, Inc. (Silver Springs, MD); and the SEER program tumor registries in the creation of the SEER-Medicare database.
The role of androgen deprivation therapy in prostate carcinoma is controversial in earlier stages of disease. The authors examined the time trends and patterns of use for androgen deprivation in the form of gonadotropin-releasing hormone (GnRH) agonists or orchiectomy, in population-based tumor registries.
Data were obtained from the linked Surveillance, Epidemiology and End Results-Medicare database. A total of 100,274 men with prostate carcinoma diagnosed from 1991 through 1999 were selected. The main outcome was the proportion of men who received ≥ 1 dose of a GnRH agonist in the first 6 months of diagnosis. This was plotted by year and stratified for age, grade, stage as well as primary versus adjuvant usage. Multiple logistic regression was used to examine predictors of GnRH agonist use in the subset of patients with localized cancer.
There was a consistent increase in GnRH agonist use by year for all ages, stages, and grades. Even in men ≥ 80 years with localized stage and low to moderate grade tumors, primary GnRH agonist use increased over the study period, from 3.7% in 1991 to 30.9% in 1999 (P < 0.001). The multivariable analysis showed that significant variability in GnRH agonist use existed among SEER geographic regions.
Prostate carcinoma is the most common nonskin malignancy in men, with an expected 220,900 incident cases and 28,900 deaths in 2003.1 Whereas the salutary effect of androgen deprivation in advanced-stage prostate carcinoma has been recognized for more than one-half a century,2 there is considerable uncertainty about its use in earlier stages of the disease.3–5
For patients presenting with localized disease (American Joint Committee on Cancer [AJCC] Stages I and II), there is no published clinical trial evidence supporting survival benefit for the use of androgen deprivation therapy alone.6 In 1995, the American Urologic Association presented guidelines on the management of localized prostate carcinoma,7, which briefly mentioned androgen deprivation therapy, mainly to recommend a need for further research. Recently, in 2004, practice guidelines for the treatment of prostate carcinoma were published by the National Comprehensive Cancer Network.8 Based on clinical trial evidence, androgen deprivation therapy together with radiotherapy is recommended for patients with localized stage and high-risk disease, as defined by a poorly differentiated tumor and/or a prostate-specific antigen (PSA) level > 20 ng/mL.9, 10 Otherwise, for localized stage without high-risk disease, androgen deprivation is not listed as a treatment option.
The main forms of androgen deprivation therapy are orchiectomy or administration of gonadotropin-releasing hormone (GnRH) agonists. The GnRH agonists are costly, with Medicare reimbursements totaling > $1 billion dollars, ranking as the second highest Medicare Part B drug expenditure in 2001.11 A number of studies have documented their adverse effects on the quality of life,12, 13 as well as complications such as osteoporotic fractures, depression, and anemia.14
In summary, there is lack of clear evidence pertaining to the most appropriate use of androgen deprivation therapy in localized prostate carcinoma. With the added concerns of substantial economic costs and adverse effects, there is a need to document patterns, trends, and determinants of use of GnRH agonists for prostate carcinoma. Two recent publications have demonstrated a striking increase in the use of androgen deprivation therapy during the 1990s, especially in patients with low-risk disease.15, 16 These were based on the Cancer of the Prostate Strategic Urologic Research Endeavor (CaPSURE) database, established by following patients from 35 selected community and academic urology practices across the United States.17 As such, they may not be fully representative of trends in the larger population. Furthermore, these reports did not differentiate between the various forms of androgen deprivation therapy to provide details on GnRH agonist versus orchiectomy use. We, therefore, used the Surveillance, Epidemiology and End Results (SEER)-Medicare database as a population-based resource to examine usage of GnRH agonist therapy for prostate carcinoma over the period 1991–1999.
MATERIALS AND METHODS
The study protocol was approved by the local institutional review board.
The data used in the current study were derived from the linked SEER-Medicare database.18, 19 The SEER program consists of a group of population-based tumor registries in selected geographic areas, from 11 states, covering approximately 14% of the U.S. population. Medicare is a federal program that covers health services for 97% of persons ≥ 65 years. It provides data in the form of claims submitted by providers for reimbursement, which include information on diagnoses and the service, testing, or procedure performed.
All men ≥ 65 years listed in the SEER-Medicare database who were first diagnosed with prostate carcinoma in the years 1991–1999 were selected, for a total of 143,308 subjects. To ensure complete information, patients not enrolled in both Medicare Part A and Part B for the first 6 months after their cancer diagnosis (n = 13,775), were members of a Health Maintenance Organization (HMO; n = 27,642), or were diagnosed by autopsy or on a death certificate (n = 1617) were excluded. This left 100,274 patients for the primary analyses.
The following Healthcare Common Procedure Coding System (HCPCS) drug administration codes were used to identify usage of GnRH agonists: J9217, J9202, J1950, J9218, and J9219. Other hormonal therapy used included orchiectomy, identified by Current Procedural Terminology codes 54520, 54521, 54522, 54530, and 54535, and International Classification of Diseases-9-Clinical Modification procedure code 62.4.20, 21 Information on the use of orally administered agents such as nonsteroidal anti-androgens was not available as these therapies are not covered by Medicare. Radical prostatectomy and radiotherapy were identified by SEER treatment codes or Medicare claims, using methodology previously described.22, 23
Demographic data were collected on age, SEER geographic region, race, and socioeconomic class variables. Race was divided into non-Hispanic white, black, white Hispanic, and others according to the SEER race code classification. Socioeconomic status (SES) variables were not available at the individual level, so a surrogate value from the 1990 U.S. census was based on census tract residence. SES variables used included education and income.
Cancer characteristics were available from SEER data files. The information in these files was abstracted from various sources by individuals trained by SEER, for each cancer case reported in a SEER tumor registry. Sources include, but are not limited to, patient medical records, reports of radiologic and laboratory studies, and pathology reports from surgical or biopsy specimens. Cancer grade was divided as well differentiated (corresponding to Gleason Grades 2–4), moderately differentiated (Grades 5–7), and poorly differentiated or undifferentiated (Grades 8–10). The SEER staging used in the current study was based on earlier versions of the AJCC system, which included Stages 0–IV.24, 25 For the purposes of the current study, the stages were reclassified to reflect most closely the current AJCC staging system.26 Briefly, Stage 0 was reclassified as Stage I, Stages I and II were reclassified as Stage II, and Stages III and IV were unchanged. One concern was that for > 40% of the patients, the AJCC stage was listed as unknown. It appears that these represent cases managed very conservatively in which extensive staging procedures were not carried out. In support of this, cases with unknown stage were more likely to be diagnosed clinically, without histology, or by PSA alone and were less likely to be managed by radical prostatectomy compared with cases with a known stage. In addition, some of the cases have histology data available but not enough information for staging. In support of this, more cases with unknown stage had prostate needle biopsy alone, without radical prostatectomy, to allow more detailed staging information. Although an alternative, “historic” SEER staging system was also available, this was not used because its definitions changed midway through the study period and it would have made comparisons with other literature on prostate carcinoma difficult.
Comorbidity was assessed by using Klabunde's adaptation of the Charlson comorbidity index.27, 28 This analysis required Medicare claims in the 12 months before the diagnosis of prostate carcinoma and was therefore limited to subjects ≥ 66 years. Higher scores refer to more extensive comorbidity.
The main outcome of interest was usage of GnRH agonist therapy, defined as the percentage of patients receiving ≥ 1 dose within the first 6 months of their cancer diagnosis. Primary usage of GnRH agonist therapy was defined as treatment with those agents in the absence of other definitive therapy such as radiotherapy or radical prostatectomy within the first 6 months of cancer diagnosis. Adjuvant therapy was defined as administration of both GnRH agonist therapy and radiotherapy or surgery within the first 6 months of diagnosis of prostate carcinoma.
The number of doses of GnRH agonist administered was calculated based on each instance of a relevant HCPCS J-code. To differentiate between once a month and once every 3 or 4-month regimens, the dosage for a given injection was determined from the unit designation (1, 3, or 4) of the “carrier miles/time/units/serv count” field or the “revenue center unit count” field of the Medicare claims. All doses were converted to a once-a-month regimen, i.e., an injection given to cover a 3-month period was counted as 3 doses.
The proportion of men who received GnRH agonists in the first 6 months of diagnosis was calculated together with 95% confidence intervals (95% CIs). This was plotted by year with stratification by patient and cancer variables, as well as by type of usage. All time trends were analyzed for significance using the Cochran–Armitage trend test.
Multivariable logistic regression was performed for the subset of patients presenting with AJCC Stage I–II prostate carcinoma, with any use of GnRH agonist therapy in the first 6 months of diagnosis as the dependent variable and all potentially relevant patient and cancer characteristics entered as independent variables. Odds ratios with 95% CIs were calculated for each category of the variables. Significance testing was performed based on the Wald chi-square statistic.
All tests of statistical significance were 2-sided, with P < 0.05 being considered significant. All analyses were performed using SAS Version 8.2 (SAS Institute, Cary, NC).
Patient and Cancer Characteristics
Table 1 presents patient and cancer characteristics of the 100,274 men comprising the study population. Their mean age was 74 years, and 42% of the patients were ≥ 75 years. The majority were non-Hispanic white, with approximately 10% blacks and 4% white Hispanics. Approximately one-half of the cancers were moderately differentiated, with the rest roughly split between well differentiated and poorly differentiated tumors. Thirty-six percent were localized cancers (Stages I–II), whereas 21% were locally advanced or metastatic (Stages III–IV) and 43% were listed as unknown in the SEER-Medicare database.
Table 1. Patient and Cancer Characteristics
No. of patients (n = 100,274)
AJCC: American Joint Committee on Cancer.
Percentage of adults with < 12 years of education.
Percentage of tract living below the poverty line.
Only for patients diagnosed in 1992 or later and with Medicare Part A/B and without Health Maintenance Organization coverage in the 12 months before diagnosis.
Number of Doses of Gonadotropin-Releasing Hormone Agonist Administered
Table 2 presents the number of doses of GnRH agonist administered during the 24-month period after diagnosis in patients initiating therapy within 6 months of diagnosis. Overall, the mean and median number of doses were 11.7 and 9, respectively. There was a slight downward trend in the number of doses with increasing calendar year, but this was no longer present when stratified by primary versus adjuvant usage (data not shown). There was also a shift with time from once-a-month dosing to once every 3-or-4-month dosing regimens. In 1991, 99% of doses were administered once a month, versus only 46% in 1999.
Table 2. Number of Doses of GnRH Agonist Received in a 24-Month Period after Diagnosis among Patients who Initiated Therapy within 6 Months of Diagnosis
GnRH: gonadotropin-releasing hormone; SD: standard deviation.
Only includes subjects who survived to 24 months after diagnosis and had Medicare Part A/B coverage, without Health Maintenance Organization coverage during that period.
All doses were converted to a once-a-month regimen, i.e., an injection given to cover a 3-month period was counted as 3 doses.
I or II
III or IV
Well or moderately differentiated
Type of usage
Adjuvant with radical prostatectomy
Adjuvant with radiotherapy
Year of diagnosis
Time Trends in Use of Androgen Deprivation
Figure 1 shows a plot of the percentage of patients receiving androgen deprivation therapy by year of diagnosis. GnRH agonist usage increased throughout the time period, from 11.5% (95% CI, 11.0–12.0%) in 1991 to 41.1% (95% CI, 40.1–42.1%) in 1999. Over the same period, orchiectomy use decreased from 12.0% (95% CI, 11.5–12.5%) in 1991 to 1.8% (95% CI,1.5–2.1%) in 1999. Overall, the percentage of patients who received either GnRH agonists or orchiectomy increased from 22.9% (95% CI, 22.2–23.6%) in 1993 to 42.7% (95% CI, 41.7–43.8%) in 1999. All time trends were statistically significant (P < 0.001).
Time Trends in Use of Gonadotropin-Releasing Hormone Agonists By Age, Stage, and Grade
Figures 2–4 show plots of GnRH agonist use over time stratified by age, grade, and AJCC stage. There are increases across all stratification groups and categories. Older patients have the highest rates of use (Fig. 2). Patients with higher grades of cancer had greater rates of GnRH agonist use (Fig. 3). However, use in well differentiated tumors also increased substantially, from 6.1% (95% CI, 5.2–6.9%) in 1991 to 24.8% (95% CI, 21.2–28.4%) in 1999 (P for time trend < 0.001). The highest rate of use was in patients with Stage IV disease, with 55.7% (95%CI, 51.9–59.6%) receiving GnRH agonists in 1999 (Fig. 4).
Time Trends in Primary Versus Adjuvant Use of Gonadotropin-Releasing Hormone Agonists
Figure 5 shows a plot of GnRH agonist use as primary therapy, as an adjuvant therapy with radical prostatectomy, and as an adjuvant therapy with radiotherapy. Use of GnRH agonists as an adjuvant therapy with radical prostatectomy varied little over the study period, from 2.3% (95% CI, 2.0–2.5%) in 1991 to 2.0% (95% CI, 1.7–2.2%) in 1999 (P for time trend = 0.06). Adjuvant therapy with radiotherapy increased from 2.6% (95% CI, 2.4–2.9%) in 1991 to 21.5% (95% CI, 20.6–22.3%) in 1999 (P < 0.001). Use of GnRH agonists as primary therapy increased from 6.9% (95% CI, 6.5–7.3%) in 1991 to 17.9% (95% CI, 17.2–18.7%) in 1999 (P < 0.001).
Time Trends in Use of Gonadotropin-Releasing Hormone Agonists for Earlier-Stage Prostate Carcinoma
Figure 6 shows plots of GnRH agonist use for different indications in earlier-stage prostate carcinomas. These include primary therapy in patients with AJCC Stage III cancers, adjuvant therapy combined with radiotherapy in AJCC Stage III cancers, and primary therapy in men ≥ 80 years with localized (AJCC Stage I–II), well to moderately differentiated tumors. In all three settings, GnRH agonist use increased over the study period. In the group of older men with well to moderately differentiated, localized cancers, use increased from 3.7% (95% CI, 2.1–5.3%) of such patients in 1991 to 30.9% (95% CI, 27.1–34.8%) in 1999 (P < 0.001).
Predictors of Gonadotropin-Releasing Hormone Agonist Use in Localized Prostate Carcinoma
Table 3 presents the results of multivariable logistic regression models examining factors determining use of GnRH agonist therapy in localized (Stage I–II) prostate carcinoma. Two models are shown. The first model includes patient and cancer characteristics, whereas the second model, in addition, includes use of radical prostatectomy and radiotherapy. In the first model, the likelihood of use increased with age. Blacks were significantly less likely to receive GnRH agonists than non-Hispanic whites. There were substantial variations in usage across the SEER regions. For example, men in Los Angeles with prostate carcinoma were threefold more likely to receive GnRH agonists than were men in New Mexico. Comorbidity was a modest, marginally significant (P = 0.049) predictor of GnRH agonist use. In the second model, the magnitude of the effect of age was diminished, but remained significant. Comorbidity was no longer a significant predictor. The significance and magnitude of effect of the other variables were unaffected.
Table 3. Multivariable Analysis of GnRH Agonist use in Localized (Stage I and II) Cancer
Multivariable logistic regression with use of GnRH agonists within 6 months of diagnosis as the dependent variable, and the following independent variables entered into the model: age, race, Surveillance, Epidemiology and End Results region, grade, year of diagnosis, income, education, and comorbidity. Results for education, income, and comorbidity are not shown.
As for Model 1, but also included use of radiotherapy and use of radical prostatectomy.
Year of diagnosis
The current study demonstrates a substantial increase in GnRH agonist therapy use for prostate carcinoma over the period 1991–1999, which occurred across all cancer stages and grades. We have further shown that GnRH agonist therapy is replacing orchiectomy as the primary modality for androgen deprivation in prostate carcinoma.
Use of GnRH agonists has been shown to be beneficial in terms of palliation and reduction of cancer-related morbidity for patients with advanced-stage disease.6, 29 Is the rapid growth of GnRH agonist therapy in earlier stages of prostate carcinoma consistent with available evidence of efficacy? The striking increase in the use of GnRH agonists combined with radiotherapy for AJCC Stage III cancers (from 2.4% in 1991 to 20.3% in 1999) is encouraging. There is strong evidence of an overall survival benefit with this regimen based on clinical trials published in the 1990s.30–32 Use of GnRH agonists as primary therapy for patients with AJCC Stage III disease has not increased as steadily (Fig. 6). Nevertheless, there has been a doubling of use between 1997 and 1999, perhaps related to publication of a clinical trial in 1997 which showed a reduction of cancer-related morbidity in this setting.5 There has been minimal use of GnRH agonists as adjuvant therapy with radical prostatectomy through the 1990s (Fig. 5), consistent with findings of no improvement in clinically meaningful outcomes such as overall or progression-free survival.33, 34 For localized prostate carcinoma (AJCC Stages I and II), which has become the most common form of presentation since the advent of PSA testing,35 there is no evidence from clinical trials of a survival benefit for primary use of GnRH agonists. Even under theoretic considerations, it is virtually impossible to show survival benefit from any intervention in such patients, particularly in the elderly, due to the slow natural progression and a competing risk of death from causes other than prostate carcinoma.36, 37 Despite this, our study showed a marked increase in primary GnRH agonist use for the subset of men ≥ 80 years with localized, low to moderate grade tumors, increasing from 3.7% in 1991 to 30.9% of such patients in 1999.
In addition to issues of efficacy, it is important to note recent concerns regarding toxicities related to GnRH agonist use.38 A number of studies document a variety of systemic changes caused by GnRH agonists in addition to the expected adverse effects on sexual function.14 These include a decrease in lean body mass and bone density, and an increase in fat mass and arterial stiffness.39, 40 Furthermore, our own analysis of SEER-Medicare data showed that androgen deprivation for prostate carcinoma is associated with risk of subsequent fractures, with the risk of fracture increasing with the number of GnRH agonist doses administered in the year after diagnosis.41
In light of its uncertain efficacy for localized prostate carcinoma, what is driving the increase in GnRH agonist use in this setting? The reasons probably come from both sides of the patient–physician interaction. The move away from orchiectomy is likely motivated, in large part, by patient desires.42 However, in an observational study comparing quality of life in patients with medical versus surgical androgen deprivation, those receiving GnRH agonists were more likely to report their health as fair or poor, and had more physical discomfort and worry related to their prostate carcinoma.12
There may be extrapolation from studies showing benefit for GnRH agonist use in locally advanced cancer.5, 30–32 Androgen deprivation therapy is also likely being increasingly used in place of observation alone. This is supported by CaPSURE data, which shows a decrease in the rate of watchful waiting for low-risk disease, from 19.4% in 1993–1995 to 8.3% in 1999–2001, concomitant with an increase in the primary use of androgen deprivation.16 This trend may be driven, in part, by the increase in PSA testing, with both patients and physicians feeling pressure to “do something” in the face of an increasing PSA level. Consistent with this hypothesis, the increasing trend in PSA testing through the 1990s closely follows the increase in androgen deprivation therapy use.43 Further evidence for the pressure to use androgen deprivation therapy comes from a national survey that showed that 53% of urologists who did not believe in the efficacy of androgen deprivation in locally advanced cancer still prescribed it.44 Finally, financial incentive may be a motivating force for some physicians to prescribe GnRH agonist therapy. During the study period, Medicare drug reimbursement was set at 95% of the average wholesale price (approximately $600 per dose of leuprolide). However, a report in 2001 by the U.S. General Accounting Office45 showed that, on average, physician providers purchased the drugs at 82% of the average wholesale price, allowing a substantial profit for every dose administered.
We noted significant variations in GnRH agonist use for localized prostate carcinoma across the different SEER regions. Geographic variability in urologists' views on prostate carcinoma has been documented previously.46, 47 Uncertainty about the benefits of androgen deprivation therapy is likely to be the major factor driving the regional differences in GnRH agonist usage.48 For example, in a national survey conducted in 1995, 95% of urologists identified radical prostatectomy as the treatment of choice for patients < 70 years with localized prostate carcinoma.49 However, for patients > 70 years, there was considerable variability in the responses (more than 1 could be selected), with 48% choosing radical prostatectomy, 31% choosing androgen deprivation, and 38% choosing observation only.
There are several limitations to our study. First, the population was limited to men ≥ 65 years. However, nearly 75% of patients diagnosed with prostate carcinoma are ≥ 65 years.50 Second, we were unable to examine GnRH agonist use in HMO patients. In a capitated setting, the economic incentive for prescribing GnRH agonists would be reversed, so it would be interesting to compare its use in a fee for service versus capitated care system. Third, no data on PSA levels were available. This is important as some of the GnRH agonist use noted in earlier-stage, low-grade cancers may have been driven by a high PSA level. However, this alone is unlikely to be responsible for the trends noted in our study, given data from a previous report indicating that only 14% of patients presenting with clinically localized prostate carcinoma have PSA levels > 20 ng/mL.47 Furthermore, data from the CaPSURE database, which include information on PSA, still show increasing use of androgen deprivation therapy in patients with low-risk disease.16 One potentially serious limitation is that > 40% of patients were listed as having an unknown AJCC stage. However, GnRH agonist use also increased similarly for patients with an unknown AJCC stage, from 14.3% in 1991 to 28.1% in 1996. Finally, because the cancer characteristics were based on tumor registry data, staging and grading of cancers were not performed in a uniform, controlled manner. However, an important strength of SEER-Medicare data is that it is population based and, therefore, is representative of cancer practice in the general population.
In conclusion, the current study demonstrates a striking increase in GnRH agonist use as androgen deprivation therapy in a large population-based sample of men with prostate carcinoma over the period 1991–1999. The increase occurred for locally advanced cancers, where clinical trials have demonstrated efficacy for GnRH agonist use, but also for localized cancers, where the benefits of this treatment are unclear. Indeed, the trend was seen even in men ≥ 80 years, with lower grade, localized cancers where the theoretic benefit of any therapy would be very small. In contrast, adjuvant use of GnRH agonists with radical prostatectomy remained at a low level throughout the study period, consistent with the absence of survival benefit in clinical trials of this regimen. Large variations in GnRH agonist use among SEER areas may be a reflection of the uncertainty about the risks and benefits of androgen deprivation therapy. Further research is needed to determine whether GnRH agonists are efficacious in localized prostate carcinoma and to better define risks related to their use.