Transitions in health-related quality of life during the rst nine months after diagnosis with prostate cancer

Transcription

1 Prostate Cancer and Prostatic Diseases (1998) 1, 134±143 ß 1998 Stockton Press All rights reserved 1365±7852/98 $12.00 Transitions in health-related quality of life during the rst nine months after diagnosis with prostate cancer DF Penson 1, MS Litwin 1, DP Lubeck 2,3, S Flanders 4, DJ Pasta 3 and PR Carroll 5 1 Departments of Urology and Health Services, University of California, Los Angeles; 2 Department of Medicine, Division of Immunology, Stanford University, Stanford, CA; 3 Technology Assessment Group, San Francisco, CA, 4 TAP Holdings, Inc, Deer eld IL, and 6 Department of Urology, University of California, San Francisco, CA Health-Related Quality of Life (HRQOL) is an important outcome measure in the study of prostate cancer. There are few data regarding the effect of sociodemographic variables, such as insurance status, educational level, marital status or income, on HRQOL. We examined whether these or other sociodemographic and clinical variables are predictive of HRQOL outcomes using an observational database of prostate cancer patients accrued from a wide array of clinical practice settings. We studied 131 patients with newly-diagnosed prostate cancer who had been followed for at least nine months. Patients were enrolled in CaPSURE TM, a large, observational database of patients with prostate cancer. General and diseasespeci c HRQOL were measured with established, validated instruments at diagnosis and nine months later. Sociodemographic data and co-morbidity counts were recorded at baseline. Multivariate regression analysis was used to determine whether sociodemographic or clinical variables were predictive of baseline HRQOL or HRQOL changes during the study period. Several sociodemographic and clinical variables demonstrated signi cant associations with HRQOL. We found improvements in general and diseasespeci c domains of HRQOL during the nine months after diagnosis. For married patients, Emotional Well-Being and Family Functioning scores were better at baseline ( 11.8, P < 0.02), but Family Functioning declined over the nine month study period ( 18.5, P ˆ ). Older patients had slightly better baseline performance in several domains of HRQOL, but experienced greater HRQOL decrements over time than did younger patients. Increasing comorbidity was associated with worse baseline general HRQOL. Early tumor stage was predictive of better scores in general HRQOL domains at baseline. Limited palpable disease stage (T2A/T2B) was predictive of worse Sexual Function and Sexual Bother at nine months ( 8.6, P ˆ 0.04; 24, P ˆ 0.008). After initial decreases, patients appear to experience an improvement in general and disease-speci c HRQOL within nine months of initial diagnosis with prostate cancer. Marital status is associated with better HRQOL, while advancing age is associated with more signi cant HRQOL declines over time. Patients with lower stage disease were noted to have better general HRQOL at baseline, although decreases in the physical domains were noted at nine months. These data shed new light on patients' experience with prostate cancer and suggest that HRQOL outcomes over time may occur in a predictable manner. Keywords: quality of life; outcomes; prostate cancer; patient assessment; predictors Correspondence: Dr MS, Litwin, UCLA Department of Urology, Box , Los Angeles, CA , USA. Received 20 October 1997; revised and accepted 26 November 1997

2 Introduction Clinical prostate cancer research has traditionally focused on outcomes such as local recurrence, evidence of metastases and disease-free and overall survival rates. In the past ve years, however, there has been a movement towards more re ned endpoints that account for patients' subjective experience of their disease. In particular, health-related quality of life (HRQOL) has become such an important measure in clinical prostate cancer research that some authors now feel that `a modern trial of prostate cancer treatment should be regarded as insuf cient without including a validated Quality of Life Questionnaire.' 1 To this end, new instruments have been developed that measure both general and disease-speci c HRQOL in prostate cancer patients. 2,3 While several studies have examined the effect of various treatment choices on quality of life outcomes, 2,3,8±16 there is little published information regarding factors prognostic of HRQOL in men with prostate cancer. While some researchers have examined the relationship of marital status and mortality in prostate cancer, 17±18 none has addressed the effect of marital status on HRQOL outcomes. Furthermore, few studies have speci cally addressed the effect of age 19 or insurance status on quality of life outcomes in prostate cancer. This study was designed to examine whether these and other sociodemographic and clinical variables are predictive of HRQOL outcomes in prostate cancer patients. To answer these questions we utilized data from CaPSURE TM (Cancer of the Prostate Strategic Urologic Research Endeavor), a large national database of prostate cancer patients followed in a naturalistic setting. 4 Observational databases of this type have been used successfully in the past for the study of a number of other chronic illnesses, including AIDS, 5 rheumatoid arthritis, 6 and coronary artery disease. 7 CaPSURE TM is the rst database of its type to study the relationship of clinical and HRQOL information in a cohort of prostate cancer patients. Materials and methods The CaPSURE TM database Quality of life predictors in prostate cancer CaPSURE TM was initiated in 1995 as an observational database to track men with prostate cancer in a variety of clinical practice settings. Patients are enrolled at 29 clinical sites located throughout the United States representing a variety of community and academic settings. The project accrues newly and previously diagnosed patients at all stages without regard to treatment selection. CaPSURE TM incorporates many important methodological elements of other successful chronic disease databases while evaluating a variety of health outcomes (clinical, HRQOL, and economic) associated with prostate cancer and its treatment. Enrollment is offered to patients with biopsy-proven prostate cancer regardless of disease duration, treatment choice, or severity of illness. At the time of enrollment, patients complete an extensive questionnaire on baseline demographics, comorbidities and HRQOL. Subsequently, at 3-month intervals, patients are mailed follow-up questionnaires to reassess HRQOL, comorbidities, and economic impacts of prostate cancer and its treatments. Telephone calls are conducted when information is incomplete or unclear. All data and research records are maintained in a con dential database at Technology Assessment Group (San Francisco, CA). To assure standardization in data collection across the wide range of practice settings that contribute to CaPSURE TM, a research advisory panel composed of urologists, internists and outcomes researchers oversees the entire project. A detailed description of the rigorous quality control mechanisms and database maintenance techniques in CaPSURE TM is reported elsewhere. 4 Subjects For the current study, we identi ed all patients who were enrolled in CaPSURE TM from June, 1995 through November, 1995 with a new diagnosis of prostate cancer, regardless of stage. A minimum of nine months of clinical and HRQOL follow-up data following diagnosis was required for inclusion in our study. Of the 2072 patients enrolled in CaPSURE TM as of June 1997, we identi ed 131 newly diagnosed patients who had at least nine months of follow-up data. The remaining patients were excluded because they either had fewer than nine months followup or had been initially diagnosed more than six months prior to enrollment. Instruments and data collection At the time of enrollment into CaPSURE TM, patients completed a survey that assessed baseline social and demographic variables (such as age, ethnicity, marital and relationship status, employment status, education, annual household income) and a comorbidity count. Patients also completed a quality of life survey that assessed both general and disease-speci c HRQOL. General HRQOL was measured using the RAND 36-Item Health Survey 1.0, better known as the SF-36. The SF-36 has been extensively validated in a variety of patient populations and is generally considered to be the `gold standard' for assessing general HRQOL. The SF-36 measures general HRQOL in eight domains, four physical and four emotional, each of which is scored from 0±100, with higher scores representing better quality of life. 20±21 Several additional scales used to measure general HRQOL included the Health Distress Scale from the RAND Medical Outcomes Study (MOS) 22 and a scale measuring patient self esteem, developed and validated by Technology Assessment Group. 23 Disease-speci c HRQOL was measured with the UCLA Prostate Cancer Index, a 20- item instrument that encompasses urinary, bowel, and sexual disturbances. Its six domains have been previously validated in men with prostate cancer, and have been used in multiple HRQOL outcomes studies throughout the United States. 3 Two additional scales, developed and validated by Technology Assessment Group, 23 were used to assess the extent to which prostate cancer interfered with the patient's relationship with his family (Cancer Interference with Family) or with the patient's life in general (Cancer Interference). Each of the disease-speci c scales is scored on a 0±100 range, similar to the SF-36, with higher scores representing better outcomes. Patients 135

3 136 completed the same quality of life survey three, six, and nine months after baseline. Clinical data, such as tumor stage, are entered by the enrolling urologist at the time of initial visit and revised as necessary during subsequent visits. For the current study, data were analyzed from the baseline clinical, sociodemographic, and HRQOL surveys and from the 9-month HRQOL survey. Statistical analysis Descriptive statistics are reported for baseline scores in each HRQOL scale. A change score for each scale was calculated by subtracting the baseline HRQOL score from the 9-month HRQOL score. Mean scores at baseline and nine months were contrasted by comparing the change score to zero using the paired Student t-test. Differences on the 0±100 RAND scales of 6.5 points in the physical health dimensions and 7.9 points in the mental health dimensions are considered to be clinically meaningful. 38 Differences on the 0±100 UCLA scales of 10 points are considered by its authors clinically meaningful. Missing data elements were treated with pairwise deletions. Multivariate regression analysis was performed on baseline and change scores with the HRQOL domains serving as dependent variables and the sociodemographic and clinical variables as independent (predictor) variables. Categorical variables were analyzed as dummy variables as follows, with the reference category indicated in italics: marital status (unmarried vs married), insurance status (Health Maintenance Organization vs Medicare, Medicare with Supplemental, Preferred Provider Organization/Feefor-Service), education level completed (high school vs college graduate or higher, some college), race (non-white vs white), and tumor stage (T2C/T3/T4 vs T1, T2A/T2B). Age, income, and number of comorbidities were treated as continuous variables. Treatment was not included as a regression variable because it signi cantly covaried with tumor stage and because we did not have suf cient sample size in all treatments to analyze them individually. All data were analyzed using the SAS System for Windows, Version 6.12 (SAS Institute, Cary, NC). Results Quality of life predictors in prostate cancer The mean patient age at baseline was 67.7 y (s.d. ˆ 7.5; median ˆ 67.4; range ˆ 47±86). About half the patients (49%) were between 60 and 70 y old. A demographic pro le of the sample at enrollment into the study is presented in Table 1. Almost 90% of the patients identi- ed themselves as Caucasian. Although a wide range was seen in income, three-fourths of patients had annual household incomes under $ Over 90% of patients were married and co-habitating with a partner. Although 60% of patients had Medicare, various forms of insurance coverage were noted. Fully 75% of patients had one or more chronic medical conditions at the time of diagnosis. All clinical stages of prostate cancer were included in the study sample. Lastly, a variety of treatment choices is represented in our study sample: radical prostatectomy (48%), pelvic irradiation (10%), watchful waiting (12%), Table 1 Sociodemographic and clinical characteristics of sample (n ˆ 131) n % Age 40± < ± < ± < ± < Ethnicity Caucasian African±American Latino Other Missing Income $ or less $ ± $ or more Missing Marital status Never married Married Separated Divorced Widowed Missing Relationship status Living with spouse or partner In signi cant relationship but not living together Not in signi cant relationship Missing Insurance Preferred provider organization/fee for service Medicare w/supplement Medicare only Health maintenance organization Other/missing Education level Grade school, some high school, or high school graduate Some college College graduate or graduate school Tumor stage T T2A/T2B T2C/T3/T Other/missing Comorbidity count or or more Missing androgen ablation (24%), and androgen ablation with surgery or radiation (6%). HRQOL scores in the general and disease-speci c domains at baseline and at nine months are shown in Table 2. During the nine months following diagnosis, several signi cant increases were noted. Marked improvements were seen in the SF-36 domains of Role Limitations Due to Physical Problems ( 14.9, P ˆ ), Social Functioning ( 7.8, P ˆ ), and Health Change ( 14.6, P ˆ ). Signi cant improvement was also noted in Urinary Function ( 13.2, P ˆ ), Urinary Bother ( 14.4, P ˆ ), Sexual

4 Table 2 Mean health-related quality of life scores in sample population at baseline and nine months following diagnosis of prostate cancer. Baseline 9 months Change Physical Role limitations±physical ** Role limitations±emotional Energy/fatigue Emotional well-being Social ** Bodily pain General gealth perceptions Health change ** Urinary function ** Bowel function * Sexual function ** Urinary bother ** Bowel bother * Sexual bother Health distress ** Self-esteem Family Cancer interference with family ** Mean change scores were calculated by subtracting baseline scores from 9- month scores. Mean change scores were compared to zero using the paired Student t-test. * P < 0.05; ** P < Quality of life predictors in prostate cancer Function ( 6.1, P ˆ ), Bowel Function ( 3.7, P ˆ 0.03), and Bowel Bother ( 5.3, P ˆ 0.02) domains of the UCLA Prostate Cancer Index. Notable improvement was also seen in the domains of Health Distress ( 4.9, P ˆ ) and Cancer Interference with Family ( 8.0, P ˆ ). Results from the multiple regression analysis of sociodemographic predictors of baseline general and diseasespeci c HRQOL scores are shown in Tables 3 and 4. Married men were found to have better HRQOL scores than unmarried men in Emotional Well-Being ( 11.8, P ˆ 0.017) and Family Functioning ( 16.9, P ˆ 0.01). Scores on the Cancer Interference with Family domain trended higher in married men (13.1, P ˆ 0.06). Married men also scored signi cantly better than unmarried men in Bowel Function (15.5, P ˆ 0.019) and Bowel Bother (21.1, P ˆ 0.023). Analysis of health insurance coverage as a predictor of baseline HRQOL revealed that patients who had Medicare with Supplemental Insurance scored signi cantly worse in the SF-36 domains of Role Limitations Due to Emotional Problems ( 19.6, P ˆ 0.04), Energy/Fatigue ( 13.9, P ˆ 0.02), and Emotional Well-Being ( 10.5 P ˆ 0.01). In addition, those who were enrolled in health maintenance organizations (HMOs) tended to score better in the Sexual Bother domain (P ˆ 0.06) than did men with any other type of coverage. Patients with non-palpable tumors (T1) had better general HRQOL scores than did men with more locally advanced disease (T2C/T3/T4) in the domains of Role Limitations Due to Physical Problems (26.4, P ˆ 0.009) and Role Limitations Due to Emotional Problems (18.8, P ˆ 0.015). Patients with palpably limited (T2A/T2B) disease also had better baseline HRQOL scores in the domains of Role Limitations Due to Physical Problems (22.4, P ˆ 0.03), Role Limitations Due to Emotional Problems (18.8, P ˆ 0.013), and Social Functioning (12.2, P ˆ 0.04) than those with stage T2C or greater disease. Compared to men with the most locally advanced tumors (T2C/T3/T4), Urinary Function was better in men with T1 (17.2, P ˆ 0.012) or T2A/T2B disease (16.0, P ˆ 0.024). Conversely, men with T1 tumors had poorer Bowel Bother scores than did men with more advanced disease ( 12.6, P ˆ 0.04). Increasing age was associated with better baseline scores in Urinary Function (1.7 points per year, P ˆ ), Urinary Bother (1.26 points per year, P ˆ 0.036), and Bowel Function (0.8 points per year, P ˆ 0.012). No correlation was seen between age and Sexual Function or Sexual Bother. Increasing comorbidity counts were associated with poorer general HRQOL but no difference in the sexual, urinary, or bowel domains. Results from the multivariate analysis of sociodemographic and clinical predictors of the 9-month changes in general and disease-speci c HRQOL scores are shown in Tables 5 and 6. Married men experienced signi cant worsening in their Family Functioning when compared to unmarried men ( 18.5, P ˆ 0.006) during the 9-month interval. However, marital status did not correlate with changes in any other HRQOL domain. Those with some college education noted worsening in Bowel Function ( 11.0, P ˆ 0.03) compared to men without any college education, although college graduates reported similar Bowel Function changes to those without a college education. Patients with earlier stage (T1) tumors reported signi cantly more negative 9-month changes in Physical Functioning ( 10.1, P ˆ 0.048) and Role Limitations Due to Physical Problems ( 21.8, P ˆ 0.05) but greater gains in Bowel Function (11.8, P ˆ 0.006) when compared to men with larger tumors. Men with T2A/T2B disease reported greater 9-month drops in Sexual Function ( 8.6, P ˆ 0.04) and Sexual Bother ( 24.0, P ˆ 0.008) than men with more locally advanced tumors. Increasing age had an adverse effect on 9-month changes in general and disease-speci c HRQOL domains, including Emotional Well-Being ( 0.5 points per year, P ˆ 0.014), Social Functioning ( 0.9 points per year, P ˆ 0.03), Health Change ( 1.1 points per year, P ˆ 0.02), Urinary Function ( 1.1 points per year, P ˆ 0.003), and a trend toward worse Urinary Bother levels ( 1.5 points per year, P ˆ 0.055). Older patients had paradoxical ndings in the bowel domains with decreased Bowel Function ( 0.9 points per year, P ˆ 0.002) but improved Bowel Bother (1.1 points per year, P ˆ 0.02). Higher comorbidity counts were associated with improved 9- month performance on the Sexual Bother scale (7.3, P ˆ 0.01). Discussion The data presented in this study indicate that during the nine months following the initial diagnosis of prostate cancer, improvement is seen in many areas of quality of life. In particular, three general domains, Role Limitations Due to Physical Problems, Social Functioning, and Health Change, improve signi cantly. We speculate that this happens because as time progresses, patients become more comfortable with their diagnosis and treatment choice, and their perceived quality of life improves 137

5 Quality of life predictors in prostate cancer 138 Table 3 Multivariate regression analysis of baseline general health-related quality of life scores on sociodemographic variables and tumor stage Cancer interference with family Family Selfesteem Health distress Health change General health perceptions Emotional role Social Emotional well-being Energy/ fatigue Bodily pain Physical role Physical Married (vs Unmarried) ns ns ns ns 11.8 (.017) ns ns ns ns ns ns 16.9 (.01) ns Medicare ns ns ns ns ns ns ns ns ns ns ns ns ns Medicare/suppl ns ns ns 13.9 (0.02) (0.01) ns 19.6 (0.04) ns ns ns ns ns ns ns ns ns ns ns ns ns ns ns ns ns ns ns PPO/Fee (vs HMO) College ns ns ns ns ns ns ns ns ns ns ns ns ns Some college ns ns ns ns ns ns ns ns ns ns ns ns ns (vs high school) White (vs non-white) ns ns ns ns ns ns ns ns ns ns ns ns ns T1 ns 26.4 (0.009) ns ns ns ns 18.8 (0.015) ns ns ns ns ns ns T2A/T2B ns 22.4 (0.03) ns ns ns 12.2 (0.043) 19.7 (0.013) ns ns ns ns ns ns (vs T2C/T3/T4) Income ns ns ns (0.047) ns ns (0.0497) ns ns ns ns ns ns Age ns 1.5 (0.03) 0.9 (0.01) 0.95 (0.006) 0.73 (0.002) 1.65 (0.0001) 1.4 (0.01) ns ns ns 1.04 (0.002) Comorbidity 5.9 (0.0013) 8.4 (0.009) 3.6 (0.02) 4.39 (0.004) 2.13 (0.045) 5.3 (0.005) 4.85 (0.0476) 6.13 (0.0001) ns ns ns ns ns Statistically signi cant relationships are expressed as estimate of effect, with P value in parentheses. ns ˆ non-signi cant. In the case of each categorical predictor variable, a comparison was made to a reference category, indicated in parentheses for each category. Income, age and comorbidity count were treated as continuous dimensional variables, so the estimates of effect are expressed per unit change in the independent predictor variable.

6 Quality of life predictors in prostate cancer Table 4 Multivariate regression analysis of baseline disease-speci c health-related quality of life scores on sociodemographic variables and tumor stage 139 Urinary function Bowel function Sexual function Urinary bother Bowel bother Sexual bother Married (vs unmarried) ns 15.5 (0.019) ns ns 21.1 (0.023) ns Medicare ns ns ns ns ns ns Medicare/suppl ns ns ns ns ns ns PPO/Fee (vs HMO) ns ns ns ns ns ns College ns ns ns ns ns ns Some college ns ns ns ns ns ns (vs high school) White (vs non-white) ns ns ns ns ns ns T (0.0122) ns ns ns 12.6 (0.04) ns T2A/T2B 16.0 (0.024) ns ns ns ns ns (vs T2C/T3/T4) Income ns ns ns ns ns ns Age 1.7 (0.0006) 0.8 (0.012) ns 1.26 (0.036) ns ns Comorbidity ns ns ns ns ns ns Statistically signi cant relationships are expressed as estimate of effect, with P value in parentheses. ns ˆ non-signi cant. PPO, preferred provider organization; FFS, fee for service; HMO, health maintenance organization. accordingly. This is consistent with the ndings of Braslis and colleagues, 10 who used the Functional Living IndexÐ Cancer (FLIC) and the Pro le of Mood States (POMS) to compare HRQOL scores in patients one year following radical prostatectomy with a separate group of prostate cancer patients awaiting the same operation. They demonstrated that despite a deterioration in sexual function and urinary continence following surgery, there was an improvement in Emotional Tension score. The nding that our patients experienced improved Urinary Function, Sexual Function and Urinary Bother nine months after diagnosis seems counterintuitive. It is important to remember, however, that our sample includes patients undergoing a variety of treatments for various stages of disease. Hence, the improvements in the disease-speci c domains may not be unexpected as patients become accustomed to their new quality of life. Patients' families also appear to become more comfortable with the diagnosis and treatment of prostate cancer during the rst nine months as evidenced by improved Cancer Interference with Family scores. Comorbidity count was inversely associated with baseline general HRQOL, a nding that is consistent with prior studies 24 and clinical experience, as sicker patients are expected to report poorer general quality of life. This adds credence to the performance of the SF-36 in our patient population. Interestingly, there appears to be no association between increasing comorbidity and diseasespeci c domains of HRQOL at baseline. This observation is consistent with prior work in men with prostate cancer 3 and in other patient population 25, 26 and underscores the importance of using both general and disease-speci c instruments to capture a full pro le of patients' experience with illness. 27 The number of comorbidities appears to have no association with a change in general HRQOL scores at nine months, but it does correlate with improved Sexual Bother scores, suggesting that sicker patients are less concerned about sexual dysfunction. Surprisingly, there were no signi cant differences between married and unmarried men in either the sexual or urinary domains, however married men did perform better in the bowel domains. The reasons for this are unclear. Several previous reports have examined the effect of marital status on outcomes in prostate cancer and other malignancies. Krongrad et al 17 used data from the Surveillance, Epidemiology and End Results (SEER) database to show that, after controlling for age, race, tumor stage, and treatment, married patients have a lower risk of mortality from prostate cancer. Although survival may be enhanced in married patients, improved quality of life does not necessarily follow. Marital status has been shown to have little effect on quality of life in cancer in general, 28 although it does appear to have a positive effect on general and disease-speci c HRQOL in patients with head and neck cancer. 26 Perhaps the involvement of intimate organs in prostate cancer sets it apart from other cancers. Indeed, patients with a disease that causes urinary or sexual dysfunction may have added stress due to the presence of a spouse or signi cant other. 17 That married status correlates negatively with 9-month change in Family Functioning, despite better baseline Emotional Well-Being and Family Functioning, supports this explanation. Married patients may have `more to lose' and, therefore, be more profoundly affected in these domains. Our data show that HMO patients reported markedly better Sexual Bother scores (less bother) than did those with Medicare, preferred provider organization (PPO), or fee-for-service coverage. This may be because HMOs coordinate bene ts more comprehensively and hence patients feel less bothered by a dysfunction that might otherwise cause embarrassment and frustration. Another surprising trend revealed that patients who are covered by Medicare and supplemental insurance performed signi cantly worse than HMO patients in Energy/Fatigue and in both emotional domains of the SF-36. This nding

7 Quality of life predictors in prostate cancer 140 Table 5 Multivariate regression analysis of mean change in general health-related quality of life scores nine months after diagnosis on sociodemographic variables and tumor stage Cancer interference with family Family Selfesteem Health distress Health change General health perceptions Emotional role Social Emotional well-being Energy/ fatigue Bodily pain Physical role Physical Married (vs Unmarried) ns ns ns ns ns ns ns ns ns ns ns ) ns Medicare ns ns ns ns ns ns ns ns ns ns ns ns ns Medicare/suppl ns ns ns ns 8.8 (0.007) ns ns ns ns ns ns ns ns ns ns ns ns ns ns ns ns ns ns ns ns ns PRO/FFS (vs HMO) College ns ns ns ns ns ns ns ns ns ns ns ns ns Some college ns ns ns ns ns ns ns ns ns ns ns ns ns (vs high school) White vs non-white ns ns ns ns ns ns ns ns ns ns ns ns ns T (0.048) 21.8 (0.05) ns ns ns ns ns ns ns ns ns ns ns T2A/T2B ns ns ns ns ns ns ns ns ns ns ns ns ns (vs T2C/T3/T4) Income ns ns ns ns ns ns ns ns ns ns ns ns ns Age ns ns ns ns 0.46 (0.014) 0.87 (0.03) ns ns 1.12 (0.02) ns 0.6 (0.002) ns 1.12 (0.002) Comorbidity ns ns ns ns ns ns ns ns ns ns ns ns ns Statistically signi cant relationships are expressed as estimate of effect, with P value in parentheses. ns ˆ non-signi cant.

8 Quality of life predictors in prostate cancer Table 6 Multivariate regression analysis of mean change in disease-speci c health-related quality of life scores over nine months on sociodemographic variables tumor stage 141 Urinary function Bowel function Sexual function Urinary bother Bowel bother Sexual bother Married (vs unmarried) ns 15.5 (0.019) ns ns 21.1 (0.023) ns Medicare ns ns ns ns ns ns Medicare/suppl ns ns ns ns ns ns PPO/Fee (vs HMO) ns ns ns ns ns ns College ns ns ns ns ns ns Some college ns (0.03) ns ns ns ns (vs high school) White vs non-white ns ns ns ns ns ns T1 ns 11.8 (0.006) ns ns ns ns T2A/T2B ns ns 8.6 (0.04) ns ns 24 (0.008) (vs T2C/T3/T4) Income ns ns ns ns ns ns Age 1.08 (0.003) 0.94 (0.002) ns ns 1.08 (0.02) ns Comorbidity ns ns ns ns ns 7.29 (0.01) Statistically signi cant relationships are expressed as estimate of effect, with P value in parentheses. ns ˆ non-signi cant. PPO, preferred provider organization; FFS, fee for service; HMO, health maintenance organization. is curious because the effect sizes are so striking. Prior studies examining the effect of managed care on patientreported health outcome measures are equivocal. Perneger et al 29 prospectively examined SF-36 scores in a population of healthy, young patients who switched from indemnity health insurance to a managed care organization in Switzerland. They showed that insurance status had no effect on HRQOL in this patient population. Other studies have shown that in older, sicker patients with joint pain, HMO enrollees are less likely to report symptomatic relief than non-hmo enrollees. 30 Our nding that HMO patients report better general HRQOL scores than those with Medicare with supplemental insurance contradicts the previous study. Ultimately, further research in this area is needed to provide better explanations of the role of insurance status on HRQOL outcomes in prostate cancer. Increasing education level has a paradoxical effect on Bowel Function and Bowel Bother. Compared to men with a high school degree or less, those with some college education perform worse in the bowel domains nine months after treatment. However, those with a college degree or higher do not perform worse than the least educated patients. One possible explanation of these ndings is that education level may be in uencing treatment choice, which in turn impacts the level of dysfunction. Prior research has shown that patients with prostate cancer undergoing radiation therapy are more likely to have gastrointestinal complaints than aged-matched controls or patients undergoing radical prostatectomy. 2,3,31±33 This effect is most pronounced at three months following therapy, and patients often return to baseline by one year after treatment. 8 In a neural network analysis, Krongrad et al 34 found that men with prostate cancer were more likely to enjoy better HRQOL if their education level was higher than the population mean. Conversely, Mazur et al 35 reported that education level was not related to treatment choice in older patients with prostate cancer. Interestingly, in breast and other malignancies, the opposite effect was found. 36,37 Nonetheless, for unclear reasons, moderately educated men in this sample appear to have the greatest problems with their bowels. Advancing age appears to have a slightly positive effect on baseline general HRQOL domains in our sample. This is incosistent with prior work with the SF- 36, which has shown that increased age is associated with worsening general HRQOL scores in the general population. 38 The relative good health of our sample (nearly three-fourths have two or fewer comorbidities) may have diminished the expected age-related decline in baseline HRQOL. The clustering of patients in the older age range may also have attenuated the impact of age on baseline HRQOL scores. Alternatively, older patients in our sample may simply have a more positive outlook about their own general HRQOL, perhaps because of greater perceived access to health care. The salutary effect of age on baseline HRQOL in this sample does not persist over time. During the nine months following diagnosis, older patients in our study experienced a slightly greater decrease in HRQOL scores than did younger patients in the general domains of Emotional Well-Being, Social Functioning and Health Change. This could indicate that a diagnosis of prostate cancer is harder on older men despite favorable nancial access to care. Although older patients are more likely to have clinically insigni cant prostate cancer, 39 a diagnosis of prostate cancer may impose greater psychological stress on them. Indeed, Caffo et al 11 suggested that older patients had a harder time adjusting psychologically to their diagnosis, although this observation did not reach statistical signi cance. Furthermore, Krongrad's neural network model, 34 in which age was not a statistically signi cant predictor of HRQOL outcomes in prostate cancer. Tumor stage was found to be an important predictor of quality of life in this patient population at baseline. Men

9 142 Quality of life predictors in prostate cancer with lower stage disease have better general HRQOL at baseline, a nding that re ects clinical experience that those with earlier tumors are least likely to have systemic manifestations of their disease. This is consistent with previous studies. Ganz et al. 40 showed that HRQOL declines correlated directly with tumor stage in prostate and other cancers. In addition, van Andel et al 41 showed that PSA, which may act as a proxy for tumor stage in many cases, has an inverse relationship with HRQOL scores in patients with node-positive prostate cancer. Because tumor stage signi cantly covaried with treatment choice, we did not include the latter in our regression. Treatment choice will be the subject of a future research effort, in which larger numbers of patients in each treatment group will allow for a more robust analysis of this important variable. Tumor stage also predicts quality of life outcomes nine months after baseline. Patients with low stage disease (T1) reported decreases in the SF-36 Physical Function and Role-Physical domains. Because this is the group most likely to undergo radical prostatectomy, these changes may represent the lingering effects of surgery. Others have demonstrated prospectively that the physical domains of HRQOL do return to baseline with longer follow-up. 14 The presence of stage T1 disease was also predictive of better Bowel Function compared to more locally advanced stage T2C/T3/T4 disease, probably re ecting the greater likelihood more locally advanced patients receive radiation as primary therapy. This is consistent with other reports that patients undergoing radiotherapy for prostate cancer have a higher incidence of bowel dysfunction than do those treated with radical prostatectomy. 3,16,42 While T1 stage demonstrates no predictive value for Sexual Function or Sexual Bother nine months after baseline, T2A or T2B stage is associated with worsening of Sexual Bother scores when compared to T2C or greater stage. This nding may simply re ect differences in primary treatment choice. Our study has several important limitations. Firstly, although prospective, the data are not randomized. CaPSURE TM represents a heterogeneous population of prostate cancer patients in a variety of clinical practice settings. This type of research design, while representing the full spectrum of prostate cancer patients, may introduce selection bias into the results. Because of this potential for bias, it is dif cult to draw de nitive conclusions regarding the relationship of treatment choice and HRQOL outcomes. Secondly, our sample has little ethnic diversity. This may be another source of selection bias and prevents us from making any comments on the relationship of race and quality of life outcomes. Thirdly, studies such as ours may suffer from the problem of multiple comparisons, even though we de ned a number of a priori hypotheses. Fourth, tumor stage may be acting as a proxy for treatment choice, a variable we did not control for in the present analysis. While treatment choice certainly affects HRQOL, we were more interested here in describing the effects of sociodemographic variables. Including treatment choice in the regression would likely have led to spurious regression results from double counting of this variable. Finally, the explanations we provide for observed outcome differences must be considered speculative, though each provides an opportunity for future hypothesis-driven research. Although observational database methodology does not contain the rigorous control mechanisms that are found in randomized clinical trials, the heterogeneity of the CaPSURE TM population does yield new insight into patients' experience with prostate cancer across the health care system today. Observations regarding the effect of marital status, comorbidity, age, education level, and tumor stage on HRQOL can provide practitioners with important information to communicate to patients on what to expect during their disease course. Furthermore, our demonstrated ndings on the effect of insurance status on HRQOL may inform future policy debates regarding prostate cancer. Conclusion We examined the effect of prostate cancer upon HRQOL immediately following diagnosis and over a 9-month interval in a heterogeneous population of patients. No attempt was made to control for treatment choice, as the population was meant to represent the wide spectrum of disease and treatment options available to patients today. Over the 9-month interval from diagnosis, patients had signi cant improvements in their general HRQOL in both physical and emotional domains. In addition, improvements were noted in the Sexual Function and Bother, Urinary Bother and Bowel Function and Bother domains, when compared to baseline. Clinical and sociodemographic predictors of baseline HRQOL and change in HRQOL were identi ed. Married patients were found to have better scores in emotional domains at baseline, but later declined over nine month study period. In addition, insurance status was found to be a predictor of Sexual Bother at baseline, with patients in HMOs having signi cantly higher scores than other types of insurance. Patients with lower stage disease were noted to have better general HRQOL at baseline, although decreases in the physical domains were noted at nine months. Patients with early stage palpable tumors (T2A or T2B) were found to have worse outcomes in the sexual domains nine months following diagnosis. These data shed new light on patients' experience with prostate cancer and suggest that change in HRQOL outcomes over time may occur in a predictable manner. Acknowledgements Statistical and programming support were provided by Inna Shapotina. James M. Henning, MS, facilitated the study. This work was supported in part by TAP Pharmaceuticals. References 1 Altwein J et al. How is quality of life in prostate cancer patients in uenced by modern treatment? The Wallenberg Symposium. Urology : 66±76.

10 2 Borghede G, Sullivan M. Measurement of quality of life in localized prostatic cancer patients treated with radiotherapy. Development of a prostate cancer-speci c module supplementing the EORTC QLQ-C30. Qual Life Res : 212± Litwin MS et al. Quality-of-life outcomes in men treated for localized prostate cancer. JAMA : 129± Lubeck DP et al. The CaPSURE Database: A Methodlogy for Clinical Practice and Research in Prostate Cancer. Urology : 773± Lubeck D et al. Quality of life and health service use among HIVinfected patients with chronic diarrhea. J Acq Imm De c Synd : 478± Fries J, Spitz P, Young D. The dimensions of health outcomes: the Health Assessment Questionnaire, disability and pain scales. J Rheumatol : 789± Pryor DB et al. Clinical data bases: accomplishments and unrealized potential. Med Care : 623± Beard CJ et al. Complications after treatment with external-beam irradiation in early-stage prostate cancer patients: a prospective multiinstitutional outcomes study. J Clin Oncol : 223± Beckendorf V et al. Changes in sexual function after radiotherapy treatment of prostate cancer. Br J Urol : 118± Braslis KG, Santa-Cruz C, Brickman AL, Soloway MS. Quality of life 12 months after radical prostatectomy. Br J Urol : 48± Caffo O, Fellin G, Graffer U, Luciani L. Assessment of quality of life after radical radiotherapy for prostate cancer. Br J Urol : 557± Fowler FJ, Jr. et al. Effect of radical prostatectomy for prostate cancer on patient quality of life: results from a Medicare survey. Urology : 1007± Lim AJ et al. Quality of life: radical prostatectomy versus radiation therapy for prostate cancer. J Urol : 1420± Pedersen KV, Carlsson P, Rahmquist M, Varenhorst E. Quality of life after radical retropubic prostatectomy for carcinoma of the prostate. Eur Urol : 7± Roach M 3rd, Chinn DM, Holland J, Clarke M. A pilot survey of sexual function and quality of life following 3D conformal radiotherapy for clinically localized prostate cancer. Int J Rad Oncol Biol Phys : 869± Shrader-Bogen CL, Kjellberg JL, McPherson CP, Murray CL. Quality of life and treatment outcomes: prostate carcinoma patients' perspectives after prostatectomy or radiation therapy. Cancer : ± Krongrad A et al. Marriage and mortality in prostate cancer. JUrol : 1696± Kornblith AB et al. Quality of life of patients with prostate cancer and their spouses. The value of a data base in clinical care. Cancer : 2791± Jonler M et al. The effect of age, ethnicity and geographical location on impotence and quality of life. Br J Urol : 651± Hays RD, Sherbourne CD, Mazel RM. The RAND 36-item Health Survey 1.0. Health Econ : 217± Ware JE, Jr., Sherbourne CD. The MOS 36-item short-form health survey (SF-36). I. Conceptual framework and item selection. Med Care : 473± Stewart AL et al. Functional status and well-being of patients with chronic conditions. JAMA : Results from the Medical Outcomes. 907±913. Quality of life predictors in prostate cancer 23 Lubeck D, Litwin M, Henning J, Carroll P. Measurement of health-related quality of life in men with prostate cancer: Data from the CaPSURE database. Qual Life Res : 385± McHorney CA, Ware JE, Raczek AE. The MOS 36-item short-form health survey (SF-36): II: Psychometric and clinical tests of validity in measuring physical and mental health constructs. Med Care : 31± Parkerson GR, Jr. et al. Disease-speci c versus generic measurement of health-related quality of life in insulin-dependent diabetic patients. Med Care : 629± Long SA et al. Factors related to quality of life and functional status in 50 patients with head and neck cancer. Laryngoscope : 1084± Patrick DL, Deyo RA. Generic and disease-speci c measures in assessing health care status and quality of life. Med Care : (suppl): S217±S Yellen SB, Cella DF. Someone to live for: social well-being, parenthood status, and decision-making in oncology. J Clin Oncol : 1255± Perneger TV, Etter JF, Rougemont A. Switching Swiss enrollees from indemnity health insurance to managed care: the effect on health status and stisfaction with care. Am J Pub Health : 388± Clement DG, Retchin SM, Brown RS, Stegall MH. Access and outcomes of elderly patients enrolled in managed care. JAMA : 1487± Jonler M et al. Sequelae of de nitive radiation therapy for prostate cancer localized to the pelvis. Urology : 876± Ljung G et al. Adverse effects after radical external beam radiotherapy of localized prostatic adenocarcinoma using two-dimensional dose-planning and a limited eld technique. Acta Oncol : 445± Watkins-Bruner D et al. RTOG's rst quality of life studyðrtog 90-20: a phase II trial of external beam radiation with etanidazole for locally advanced prostate cancer. Int J Rad Oncol Biol Phys : 901± Krongrad A et al. Predictors of general quality of life in patients with benign prostate hyperplasia or prostate cancer. J Urol : 534± Mazur DJ, Merz JF. How older patients' treatment preferences are in uenced by disclosures about therapeutic uncertainty: surgery versus expectant management for localized prostate cancer. J Am Geriatr Soc : 934± Sen M. Communication with cancer patients. The in uence of age, gender, education, and health insurance status. Ann NY Acad Sci : 514± Graling PR, Grant JM. Demographics and patient treatment choice in stage I breast cancer. AORN J : 376± Ware JE SF-36 Health Survey: Manual and Interpretation Guide. 2nd edn The Health Institute: Boston, MA. 39 Dugan JA et al. The de nition and preoperative prediction of clinically insigni cant prostate cancer. JAMA : 288± Schag CA et al. Quality of life in adult survivors of lung, colon and prostate cancer. Qual Life Res : 127± van Andel G, Kurth KH, de Haes JCJM. Quality of life in patients with prostatic carcinoma: a review and results of a study with N disease. Urol Res : S79±S Fowler FJ, Jr. et al. Outcomes of external-beam radiation therapy for prostate cancer: a study of Medicare bene ciaries in three surveillance, epidemiology, and end results areas. J Clin Oncol : 2258±