Fax: (303) 372-9092
Marriage and ethnicity predict treatment in localized prostate carcinoma†
Article first published online: 28 MAR 2005
Copyright © 2005 American Cancer Society
Volume 103, Issue 9, pages 1819–1825, 1 May 2005
How to Cite
Denberg, T. D., Beaty, B. L., Kim, F. J. and Steiner, J. F. (2005), Marriage and ethnicity predict treatment in localized prostate carcinoma. Cancer, 103: 1819–1825. doi: 10.1002/cncr.20982
This study used the linked SEER-Medicare database. The interpretation and reporting of these data are the sole responsibility of the authors.
- Issue published online: 18 APR 2005
- Article first published online: 28 MAR 2005
- Manuscript Accepted: 20 DEC 2004
- Manuscript Revised: 24 SEP 2004
- Manuscript Received: 19 JUL 2004
- Redes en Acción research subcontract from the Baylor College of Medicine
- Special Populations Network NCI. Grant Number: U01CA86117
- prostate carcinoma;
- localized disease;
- informed decision-making;
- treatment patterns;
- health disparities;
Primary treatment for early-stage prostate carcinoma includes expectant management or, for curative intent, radical prostatectomy or radiotherapy. Treatment recommendations are generally guided by clinical factors such as Gleason grade, prostate-specific antigen level, comorbid illnesses, and patient age. Sociocultural factors may also have influences on patient and urologist treatment choices.
The authors used bivariate and multinomial logistic regression to identify medical and sociodemographic predictors of prostatectomy (compared with radiotherapy) and curative therapy (compared with expectant management) in a cohort of 27,920 non-Latino white, black, and Latino men without comorbidities in the latest linked Surveillance, Epidemiology and End Results–Medicare dataset (years 1995–1999). Predictors included tumor stage, patient age, marital status, race/ethnicity, and soscioeconomic status.
Younger age and higher tumor grade were robust predictors of curative treatment compared with expectant management and of prostatectomy compared with radiotherapy. Sociodemographic factors had an additive role in treatment choice. Marriage predicted curative treatment compared with expectant management (adjusted risk ratio [RR] = 1.28 [1.25–1.30]) and prostatectomy compared with radiotherapy (adjusted RR = 1.24 [1.20–1.28]). Although blacks and Latinos were just as likely as whites to receive curative treatment, blacks were significantly less likely, whereas Latinos were more likely, to receive prostatectomy compared with radiotherapy (adjusted RRs = 0.77 [0.72–83]) and 1.24 [1.18–1.30], respectively).
Marriage was positively associated with curative treatment in general, and with prostatectomy specifically. Blacks received prostatectomy less often than whites, although they did not receive less curative treatment overall. Latinos received prostatectomy more often than whites. Clinicians should recognize the importance of cultural and social forces as well as biomedical factors in decisions regarding the treatment of patients with early-stage prostate carcinoma. Cancer 2005. © 2005 American Cancer Society.
Prostate carcinoma is the leading cancer affecting men of all races in the U.S., and is the second most common cause of cancer death. After the introduction of the prostate-specific antigen (PSA) test in 1987, widespread screening led to a dramatic increase in the identification of patients with early-stage prostate carcinoma. With this diagnosis, many men face considerable challenges regarding treatment. In part, this is because of a lack of randomized trial evidence clearly favoring a mortality benefit for any of the main curative forms of treatment, including radical prostatectomy, external-beam radiotherapy, and brachytherapy. In addition, each of these curative options carries a high risk of late side effects, such as incontinence, impotence, and bowel urgency.1 Many men have difficulty deciding whether they can accept such risks. These difficulties are most acute among men healthy enough to have choices, whereas older and sicker men are less able to tolerate surgery, and therefore have fewer options. Indeed, because of a lack of clear-cut mortality benefit, there is also controversy regarding the merits of curative therapy versus expectant observation (sometimes known as “watchful waiting”). In summary, for men healthy enough to have treatment choices, the first decision is usually between prostatectomy and a form of radiotherapy. Another crucial choice, especially as men grow older, is between curative treatment and expectant management. In the face of such decisions, social and cultural factors may have an influence over and above purely biomedical considerations.
We hypothesized that sociocultural factors would add to the explanatory power of traditional clinical variables, such as tumor grade, PSA level, comorbid illnesses, and patient age, in predicting treatment choice in nonmetastatic prostate carcinoma. Previous studies2–6 have identified several such factors, including race/ethnicity, socioeconomic status (SES), and geographic region. Almost all studies have revealed significant regional variations in treatment patterns and substantially lower rates of prostatectomy among blacks compared with whites, possibly due to provider differences, patient preferences, or health care access barriers that constrain treatment choices. In the current study, we assess whether a number of clinical and sociodemographic variables predict treatment choice in a population of Medicare enrollees without documented comorbid illnesses. These healthier men have the greatest number of treatment options because the costs of their care are covered by Medicare. Our analysis employs the latest linked Surveillance, Epidemiology and End Results (SEER) cancer registry and Medicare inpatient records (1995–1999), and we include marital status and Latino ethnicity, variables that have not been assessed consistently in previous studies.
MATERIALS AND METHODS
We tested our hypotheses that particular clinical and sociodemographic characteristics predict receipt of 1) prostatectomy compared with some form of radiotherapy (external beam or brachytherapy) as well as 2) either form of curative therapy (prostatectomy or radiotherapy) compared with expectant management by using data from the SEER cancer registries that have been linked with Medicare hospital-discharge records. We focused our analyses on a subset of men > 65 who do not have any documented comorbidities and are, therefore, likely to be among the healthiest in the database, with the greatest number of treatment options and opportunities to make treatment decisions that are informed by a consideration of risks and benefits. The SEER-Medicare dataset has been used in several other studies to assess patterns of care for newly diagnosed patients with cancer of several sites, and its strengths and limitations for this purpose were recently reviewed.7 Covering approximately 14% of the U.S. population, the SEER registries, sponsored by the National Cancer Institute (NCI), document all incident cases of cancer in 6 urban areas (San Francisco-Oakland, San Jose, Los Angeles county, Seattle, Atlanta, and Detroit) and in 5 states (Connecticut, Hawaii, Iowa, New Mexico, and Utah). In the linked dataset, SEER includes patient age and date of diagnosis; tumor location, grade, stage, and lymph node involvement; therapy received within 4 months of diagnosis; and sociodemographic characteristics including race/ethnicity, marital status, and a variety of census-based SES measures as proxies for individual-level data not collected by SEER.
The Medicare program provides health coverage for 97% of persons ≥ 65 and collects claims for all program services. Hospitalization data, included in the Medicare Provider Analysis and Review (MEDPAR) files, contains information on all hospitalizations since 1984, including admission and discharge diagnoses and comorbid conditions.
The SEER and Medicare databases were linked to allow population-based studies of health outcomes. Data for 94% of persons ≥ 65 in SEER have been linked successfully to Medicare records. Use of the combined dataset excluded approximately 22% of prostate carcinoma cases in SEER, primarily among men < 65 and those enrolled in Health Maintenance Organizations. The remaining, linked cases, however, allowed for adjustment by comorbid conditions and eliminated the confounding effects of insurance coverage. The research described here was approved by the Colorado multiple institutional review board.
Non-Latino white, black, and Latino men with localized prostate carcinoma were included if there were matching SEER and Medicare records and the diagnosis was not made at autopsy or on a death certificate. We included all cases between the years 1995 and 1999, the most recent data available in SEER-Medicare and not analyzed in previous studies. Subjects were excluded if, at the time of diagnosis, they had one or more (nonprostate carcinoma) comorbidities or if data pertaining to patient age or tumor stage or grade were unavailable (n = 25,311). Based on these criteria, 27,290 patients were included in the current study.
Primary Treatment and Tumor Characteristics
Primary treatment was based on SEER data. In order of priority, primary treatment was defined as prostatectomy if any form of curative-intent prostatectomy was indicated by the site-directed surgery variable (transurethral resection of the prostate [TURP] procedures were excluded), as radiotherapy if indicated by any form of external-beam radiotherapy or brachytherapy, and as watchful waiting if neither of these was obtained. A variable indicating any form of curative treatment was assigned a true value if treatment included either surgery or radiotherapy.
Tumor stage was based on the SEER/American Joint Commission on Cancer extent of disease (EOD; 10 prostate pathology ext 1995+) variable corresponding to localized (T1 or T2) disease. Clinical as opposed to pathologic staging was utilized except that when the former was unavailable and prostatectomy was performed, staging was assumed to be localized because surgery is typically not indicated for clinically advanced disease. For patients who did not receive prostatectomy, staging was based only on biopsy results and radiologic criteria rather than on complete excision. In some cases, this underrepresented the true EOD but did reflect the clinical staging used to determine treatment. Tumor grade was trichotomized as low, moderate, and high (i.e., Grades 1, 2, and 3/4, respectively), corresponding to Gleason scores 2–4, 5–7, and 8–10. PSA scores were not included because these were unavailable in the SEER-Medicare database.
Demographic Characteristics and Coexisting Illnesses
Information on race and age at diagnosis was obtained from the SEER database. The SES of each patient was based on year 2000 census data. Zip code measures were utilized only if census-tract data were unavailable (21% of patients). SES measures included per capita income, percent of residents living below the poverty level (race and age specific), percent of persons ≥ 25 years with less than a high school education (race specific), and percent of persons ≥ 65 years not speaking English well. These variables were necessarily ecologic as opposed to individual-level measures. Because they were all highly correlated, we selected educational level to capture SES. Marital status was based on SEER (dichotomized as married vs. not married/unknown).
Comorbidity scores were derived from MEDPAR records, using the Deyo adaptation8 of the Charlson comorbidity index and calculated by means of a SAS macro (SAS Institute Inc., Cary, NC) available on the NCI SEER-Medicare website. All patients with scores > 0 were excluded from further analysis. If a patient received prostatectomy, we included all MEDPAR records through the date of this procedure. Otherwise, we included all MEDPAR records through the first hospitalization occurring within 6 months of the diagnosis. If there were no MEDPAR records before the diagnosis, the comorbidity score was assumed to be zero.
Using bivariate analysis, we examined separately the association between 1) prostatectomy versus a form of radiotherapy, and 2) either form of curative therapy (prostatectomy or radiotherapy) versus expectant management and the following variables: patient age, tumor stage, tumor grade, race/ethnicity, marital status, per capita income, poverty level, educational attainment, and poor use of English. Descriptive analyses and multiple logistic regression models were conducted with SAS software. The outcome variables were dichotomous: 1) prostatectomy versus radiotherapy and 2) curative therapy (prostatectomy or radiotherapy) versus expectant management. We determined a priori that the independent variables in the logistic models would be those significantly associated with initial therapy in the bivariate analyses, using a statistical significance level of 0.05. We also examined statistical interactions of age and grade, race and grade, age and marriage, age and race, age and non-high school education, race and marriage, and race and non-high school education. Results of the logistic regression models are shown as percentages of patients by race/ethnicity receiving the treatment of interest and risk ratios (RRs) associated with each independent variable adjusted for the others. RRs were computed as corrections of the adjusted odds ratios.9
The sociodemographic and clinical characteristics of the study cohort (84.4% white, 9.9% black, 5.7% Latino) are shown in Table 1. Differences by race/ethnicity were evident. Blacks were much less likely than whites and Latinos to be married. Compared with whites, blacks and Latinos were increasingly likely to live in low SES areas, as reflected by the percentage of census-tract/zip code inhabitants who had not graduated from high school. Eighty-five percent of Latino men had tumors graded as moderately or poorly differentiated compared with 89% of white men and 92% of black men. Black men were much less likely than white men to receive prostatectomy (30% vs. 38%), whereas Latino men were much more likely than white men to receive prostatectomy (45 % vs. 38%). There was, however, little difference in receipt of either form of curative therapy (compared with expectant management) among whites, blacks, and Latinos (68%, 65%, and 66%, respectively).
|All (n = 27,290)||75||48||31||21||11||70||20||37||30||32|
|White (n = 23,040)||77||55||33||12||11||70||19||38||30||32|
|Black (n = 2698)||58||13||25||62||8||71||21||30||35||35|
|Latino (n = 1552)||73||5||9||87||15||65||20||45||21||35|
|Age < 70 yrs|
|All (n = 11,107)||77||47||31||22||9||73||18||61||23||16|
|White (n = 9011)||80||56||33||11||9||74||17||63||22||15|
|Black (n = 1375)||59||14||27||58||7||72||21||45||33||22|
|Latino (n = 721)||75||4||8||88||14||67||19||64||15||20|
|Age ≥ 70 yrs|
|All (n = 16,183)||73||48||31||21||12||67||21||21||35||44|
|White (n = 14,029)||74||54||33||12||12||67||21||22||36||43|
|Black (n = 1323)||57||11||23||66||9||69||22||14||37||49|
|Latino (n = 831)||71||5||9||86||15||63||22||28||26||47|
Younger men (age < 70) received curative-intent therapy 84% of the time, with prostatectomy predominating for 73% of these men (i.e., 61% of the total). A higher percentage of married men compared with unmarried men received any form of curative therapy (86% vs. 75%) as well as prostatectomy, specifically (65% vs. 47%; data not shown). Blacks were less likely than whites to receive any form of curative therapy (78% vs. 85%) as well as prostatectomy, specifically (45% vs. 63%). whereas, among Latinos, there was greater parity with whites (79% vs. 85% and 64% vs. 63%, respectively). Table 2 depicts higher utilization of curative therapy in general, and of prostatectomy compared with radiotherapy among married men in all three racial/ethnic groups. The effect was most pronounced among younger, married Latinos, who had 20% more prostatectomies than their unmarried counterparts.
|Unmarried||Married||P valueb||Unmarried||Married||P value||Unmarried||Married||P value|
|Age < 70 n = 11,107|
|Prostatectomy||49%||66%||< 0.001||37%||51%||< 0.001||49%||69%||< 0.001|
|Any curative therapy||76%||87%||< 0.001||74%||80%||0.007||71%||83%||0.008|
|Age ≥ 70 n = 16,183|
|Any curative therapy||44%||62%||< 0.001||46%||55%||0.005||43%||57%||0.001|
Among older men (age ≥ 70), 56% received a form of curative therapy, with radiotherapy predominating 63% of the time (i.e., 35% of the total). As in the younger cohort, a higher percentage of married compared with unmarried men in the older cohort received any form of curative therapy (61% vs. 44%) as well as prostatectomy, specifically (24% vs. 14%; data not shown). Blacks were less likely than whites to receive any form of curative therapy (51% vs. 58%) as well as prostatectomy, specifically (14% vs. 22%), whereas Latinos and whites were more similar (54% vs. 58% and 28% vs. 22%, respectively; data not shown). A positive association between marriage and higher rates of prostatectomy and either form of curative therapy applied within all race/ethnic groups (Table 2). However, compared with the younger cohort, the association was less pronounced for prostatectomy, specifically, and somewhat more pronounced for any form of curative therapy.
In bivariate analyses, age, race/ethnicity, marriage, high school education, and tumor grade were predictive of prostatectomy and any form of curative therapy (P < 0.0001). However, none of the interaction terms was significant. The results of the multivariate analyses are displayed in Tables 3 and 4, showing as relative risks the adjusted predictors of prostatectomy versus radiotherapy and curative therapy versus expectant management, respectively. As expected, younger age and higher tumor grade were strong predictors of prostatectomy compared with radiotherapy, and of curative treatment compared with expectant management. Marriage was also a strong predictor of prostatectomy and curative treatment (prostatectomy or radiotherapy) in all racial/ethnic groups. Black race was negatively associated, and Latino ethnicity was positively associated, with receipt of prostatectomy, although there were no racial/ethnic differences in receipt of curative treatment overall. Lower SES predicted lower rates of prostatectomy compared with radiotherapy among blacks. Finally, we performed a subgroup analysis in the four California and New Mexico registries because Latinos are concentrated in these areas and because we wanted to exclude regional variation in procedure utilization as a possible explanation for significant differences between Latinos and whites. The results were not significantly different from those in the overall sample.
|Total no. of patients||18,444||15,685||1743||1016|
|No. of patients (%) receiving prostatectomy||10,189 (55%)||8690 (55%)||806 (46%)||693 (68%)|
|Age (yrs) ≥ 70 vs. < 70||0.50 (0.48–0.52)||0.51 (0.48–0.53)||0.47 (0.40–0.54)||0.63 (0.55–0.72)|
|Moderate vs. low||1.21 (1.14–1.27)||1.18 (1.11–1.25)||1.68 (1.27–2.09)||1.26 (1.08–1.41)|
|High vs. low||1.29 (1.22–1.36)||1.26 (1.18–1.33)||1.81 (1.37–2.25)||1.31 (1.11–1.46)|
|P = 0.003||P = 0.005|
|Married vs. unmarried||1.24 (1.20–1.28)||1.24 (1.20–1.29)||1.25 (1.12–1.38)||1.13 (1.01–1.23)|
|Percentage of persons with < 12 yrs of education|
|11–20 vs. 0–10||0.95 (0.92–0.99)||0.96 (0.92–0.99)||0.86 (0.71–1.00)||NS|
|≥ 21 vs. 0–10||0.94 (0.91–1.00)||0.95 (0.90–1.00)||0.83 (0.69–0.96)|
|P = 0.008||P = 0.015||P = 0.04|
|Black vs. white||0.77 (0.72–0.83)|
|Latino vs. white||1.24 (1.18–1.30)||—||—||—|
|Total no. of patients||27,290||23,040||2698||1552|
|No. of patients (%) receiving curative treatment||18,444 (68%)||15,685 (68%)||1743 (65%)||1016 (65%)|
|Age (yrs) ≥ 70 vs. < 70||0.67 (0.65–0.69)||0.67 (0.65–0.69)||0.66 (0.60–0.71)||0.65 (0.58–0.73)|
|Moderate vs. low||1.60 (1.56–1.64)||1.60 (1.55–1.64)||1.49 (1.34–1.63)||1.76 (1.59–1.91)|
|High vs. low||1.56 (1.52–1.61)||1.58 (1.53–1.63)||1.33 (1.16–1.50)||1.73 (1.52–1.90)|
|Married vs. unmarried||1.28 (1.25–1.30)||1.30 (1.28–1.33)||1.13 (1.07–1.19)||1.27 (1.16–1.36)|
|Percentage of persons with < 12 yrs of education|
|11–20 vs. 0–10||0.95 (0.93–0.97)||0.94 (0.92–0.97)|
|≥ 21 vs. 0–10||0.91 (0.89–0.94)||0.92 (0.89–0.95)||NS||NS|
|Black vs. white||0.96 (0.93–0.99)|
|Latino vs. white||1.00 (0.96–1.04)|
|P = 0.04|
The current study confirmed the hypothesis that sociocultural factors add significant explanatory power to traditional biomedical variables in understanding treatment patterns of early-stage prostate carcinoma. We analyzed a national sample of non-Latino white, black, and Latino Medicare beneficiaries who had no documented comorbid conditions between the years 1995 and 1999. This population is highly likely to face therapeutic options after diagnosis, including prostatectomy, external-beam radiotherapy, brachytherapy, and expectant management. We found, as anticipated, that younger age and higher tumor grade were very strong predictors of treatment. In addition to these biomedical influences, however, we also found that several sociocultural variables, such as marital status, high school education, and race/ethnicity, were also significant predictors of treatment.
That blacks were no less likely than whites to receive curative therapy, but were less likely to undergo prostatectomy compared with radiotherapy is intriguing for two reasons. First, almost all patients see urologists for confirmatory biopsies and, second, urologists in general are more likely to recommend prostatectomy rather than radiotherapy for curative intent.10 It is possible that physicians recommend surgery less often to their black compared with white patients because they perceive black men at higher risk for poor outcomes3 and also less likely to adhere to medical advice.11 The beliefs and preferences of black patients must also be considered, and could include distrust of physicians and fear of surgery,12 more concerns about impotence,13 and more pessimism than other racial groups about the curability of prostate carcinoma.14 Finally, some urologists may downplay surgery as an option because of a priori assumptions that their black patients have these sorts of concerns.
Our finding of a higher rate of prostatectomy among Latinos compared with whites was somewhat surprising. Two previous studies, using cancer registry data through 1994, found either significantly lower5 or no difference6 in prostatectomy rates among Latinos. Neither of these studies adjusted for comorbidities or measures of SES, however, and both included a large proportion of men < 65. A third investigation,2 based on 1994–1995 prostate cancer outcomes study data but also using a younger population than the one in the current study, derived results similar to our own. One possibility is that Latinos, many of whom live in rural areas or in Spanish-speaking enclaves, have less real or perceived access to radiotherapy oncology services. Alternatively, Latinos may have more favorable attitudes towards surgery (or more fear of radiotherapy) compared with whites and blacks, or perhaps they have a greater trust in the authority of physicians and, therefore, more willingness to accept prostatectomy when it is recommended.
A particularly noteworthy finding was that marital status appears to be at least as strong a predictor of prostatectomy as race/ethnicity, which has received the greatest attention in previous studies. The first study to identify higher utilization of prostatectomy among married men was conducted in 2000.5 Two subsequent studies2, 3 showed that unmarried men received more conservative therapy. It is possible that married men (or their wives) advocate therapy that they perceive as likeliest to yield cure, whereas unmarried men are more likely to lack social supports that would encourage aggressive interventions. The influence of wives on the selection of curative therapy has some plausibility, based on a qualitative study of patients with early-stage prostate carcinoma in which wives were more interested in their husbands' cure whereas husbands tended to place greater emphasis on treatment side effects.15 Finally, perhaps clinicians recommend aggressive therapy more strongly to married than unmarried men.
Given the current emphasis on improving the quality of patient involvement in decision-making, the role of spouses deserves greater attention. Little is known, for example, about how often wives accompany their husbands to the clinic, and about how their presence influences consultations. Wives may enhance the quality of decision-making by gathering information, coaching the patient to ask questions, or advising about treatment.16 Conversely, even informed spouses have the potential to impair decision-making. One study found that in consultations that include the physician, patient, and spouse, patients frequently raise fewer topics, are less assertive, engage in less shared decision-making, and are frequently excluded from conversations.17
The current study has several limitations. Provider characteristics, referral patterns, and patient/spouse preferences could not be analyzed, and ecologic rather than individual measures of SES were utilized. If there was preferential underascertainment of comorbid conditions in more socially disadvantaged groups, this would have inflated the strength of association between treatment and marital status, SES, and race/ethnicity. Regional variation in practice patterns might also influence findings of racial/ethnic differences in treatment.4 However, our findings pertaining to black men are consistent with those of previous studies that controlled for registry, and a subgroup analysis performed for Latinos concentrated in California and New Mexico did not differ significantly from results obtained in the overall sample. In addition, although the sensitivity and specificity of a designation of Latino ethnicity is unknown in the SEER-Medicare database,18 we believe misclassification is unlikely to have been a major concern in our analysis. SEER race recode B is more exact than Medicare,18 whereas Medicare appears to misreport Latino ethnicity ≤ 10% of the time.19 Finally, it should be noted that a strength of our study compared with previous research involved the inclusion of clinicopathologic staging information in the SEER-Medicare dataset beginning in 1995, thereby reducing the frequency of clinical upstaging based on surgical pathology.
As clinical uncertainty continues to surround the optimal treatment of early-stage disease, and as patient involvement in decision-making remains a key goal for ethical and quality-of-life considerations, the results of the current analysis suggest several important avenues of inquiry. How, for example, do spouses influence the nature and quality of decision-making and how might clinicians productively facilitate their involvement in this process? Do patients in certain racial/ethnic groups manifest particular beliefs and attitudes about cancer and medical therapies that complicate their ability to rationally weigh risks and benefits? Finally, which patient groups lack access to radiotherapy oncology? Assessing and ensuring the adequacy of shared decision-making in this very complex clinical scenario will depend on finding answers to these and related questions.
The authors acknowledge the efforts of the Applied Research Program, National Cancer Institute; the Office of Research, Development and Information, Centers for Medicare and Medicard Services (CMS); Information Management Services (IMS), Inc.; and the Surveillance, Epidemiology, and End Results (SEER) Program tumor registries in the creation of the SEER-Medicare database.
- 7Overview of the SEER-Medicare data: content, research applications, and generalizability to the United States elderly population. Med Care. 2002; 40(8 Suppl. ): IV-3–IV-18., , , et al.
- 12Perceptions of cancer in an African-American community: a focus group report. Ethn Dis. 2002; 12(2): 276–283., , .
- 18Patient demographic and socioeconomic characteristics in the SEER-Medicare database applications and limitations. Med Care. 2002; 40(8 Suppl. ): IV-19–IV-25., , , et al.