Patients’ Survival Expectations before Localized Prostate Cancer Treatment by Treatment Status

Background: Cancer-specific mortality is projected to be only 1% in 15 years in approximately 75% of patients with screen-detected localized prostate cancer (LPC). Nearly 94% of patients choose treatment even though treatment damages health-related quality of life. No data are available regarding what survival benefit patients expected from treatment. Objectives: A self-administered mailed survey was sent to 184 men with newly diagnosed LPC to query patients about expected survival with treatment versus observation. Results: More than 90% of patients had at least a high school education and a ninth-grade health literacy. In addition, 68% patients had income of ≥$50,000. Mean cancer grade was 6.6. Twenty-three patients chose observation and 161 patients chose surgery or radiotherapy. Mean comorbidity adjusted life expectancy (CALE) without the cancer was 22.9 years. Without cancer treatment, 15.2% of patients expected to live <5 years, 48.8% 5 to 10 years, 33.5% 11 to 19 years, and 2.4% ≥20 years. With treatment, survival expectations were <5 years in 0.6%, 5 to 10 years in 6.5%, 11 to 19 years in 30.0%, and ≥20 years in 62.9% of patients. Age, prostate-specific antigen level, CALE, anxiety, depression, and social support were factors that predicted differences between CALE and patient survival expectations with and without treatment. Conclusion: LPC patients grossly underestimated their life expectancy without treatment and grossly overestimated the survival benefit of treatment.

Although around 80% of men aged 80 years and older and 15% to 30% of men aged 50 years and older have microscopic undiagnosed prostate cancer found at autopsy, only 3% men die because of prostate cancer. 1 Increasing prostate-specific antigen (PSA) screening at younger ages has increased overdiagnosis 2 and overtreatment 3 of localized prostate cancer (LPC). More than 90% of US pa-tients currently diagnosed with prostate cancer have LPC and approximaely 94% of patients with LPC choose treatment. 4 Based on data from leading studies, a model had recently projected only a 0% to 2% 15-year mortality from low-grade (Gleason score Ͻ7) screen-detected LPC in men aged 55 to 74 years if they chose observation instead of treatment. 5 By consensus, urologists and radiation oncologists recommend treatment for LPC if a patient has a further 10-year life expectancy 6 (the 10-year rule 7 ) regardless of cancer grade, even though no randomized trials have shown that treatment can improve survival in patients in whom the cancer was screen-detected. National guidelines by the American Cancer Society and the National Comprehensive Cancer Network (NCCN) also recommend treatment for most patients. 8 However, in the review by Zeliadt et al, 9 different studies had found that patients rate the sexual, urinary, and bowel side effects of treatment to be just as important as the potential benefit in survival; that if risks and benefits of treatment were explained with-This article was externally peer reviewed. out bias, 75% of patients chose a lower radiation dose despite a lower predicted survival; that 90% of physicians but fewer than 20% of patients ranked the effect of treatment on survival as one of their top 4 concerns; and that patients who chose treatment believed that treatment was guaranteed to improve survival. At a median of 6 years after treatment, health-related quality of life (HRQOL) of treated patients was worse than that of control patients. 10 Many patients regretted that they chose treatment. 11 To our knowledge, no studies of patient-physician communication have examined patients' anticipated survival benefit of treatment. Without data from randomized trials in screen-detected patients, it is difficult to counsel patients regarding their survival with and without treatment. Even with the use of multifactorial models, accuracy of predicted survival is 75% or lower. 12 Physicians are also poor at estimating baseline comorbidity adjusted life expectancy (CALE), which is critical in making an informed decision. 13 Thus, patients may accept a treatment recommendation not knowing what their baseline CALE is, how much the newly diagnosed cancer could reduce it, or how effectively treatment could minimize that reduction. Current overtreatment of LPC might be because patients do not understand the pros and cons of treatment.
In this study we surveyed newly diagnosed patients about their anticipation of survival with and without treatment. By estimating their baseline CALE without considering the newly diagnosed cancer, we calculated their perceived decrease in longevity with observation (PDLO), and their perceived increase in longevity with treatment (PILT) for the cancer.

Methods
We surveyed patients who had been newly diagnosed with LPC (stages T1a to T2c) in the preceding 6 months, had met with their urologist after the diagnosis, were scheduled to receive treatment or observation, and had not yet been treated with surgery or radiation. Patients with dementia, or those who could not read, write, or understand English, were excluded. All patients were recruited from a large, private urology practice in Norfolk, Virginia. Staff at this practice systematically contacted patients newly diagnosed with LPC between March 2005 and November 2007 regarding their interest in participation in a self-administered mailed survey. Two concomitant pretreatment selfadministered surveys were used. The first survey asked patients about expectations of survival with and without treatment, comorbid diseases, mood, social support, satisfaction with life, health, and education by physicians about treatment options. A list of health-related words in a closed envelope was mailed with the survey. Patients were requested to open the envelope and read these words on the telephone to a research assistant. This was done to estimate patient health literacy by using a brief version of the Rapid Estimation of Health Literacy in Medicine scale. 14 Patients were given a $10 stipend for completing this survey. A second pretreatment survey was a part of a longitudinal follow-up by urologists to evaluate generic HRQOL, prostate cancer related symptoms, and fear of cancer recurrence. The study methods were reviewed and approved by an Institutional Review Board.

Measures
The Charlson Comorbidity Index (CCI) is a validated measure of comorbidity. We used a patient self-reported CCI scale that asked about the presence and severity of 12 chronic conditions; the Prostate Cancer Outcomes Study used this CCI version. 15 Score categories are 0, 1, 2, and 3 or more diseases.
The NCCN practice guidelines had recommended, for the first time in 2007, 8 that the healthadjusted life expectancy of LPC patients can be estimated by weighting age-based life expectancy by 1.5 for patients in the highest health quartile, using no weighting for patients in the middle 2 health quartiles, and weighting by 0.5 for patients in the lowest health quartile. We used comorbidity scores as surrogate markers of health status because comorbidity is the main determinant of life expectancy in older patients, 16 and the most important prognostic factor for patients with LPC who are Ͻ75 years old is the comorbidity score. 17 We categorized patients into health quartiles by using their CCI score (0 disease score ϭ highest health quartile; 1 or 2 disease score ϭ middle 2 health quartiles; Ն3 disease score ϭ lowest health quartile). Our basis of equating a 1 or 2 disease score with the middle 2 health quartiles was that almost half (49%) of the 3173 patients newly diagnosed with LPC in a Prostate Cancer Outcomes Study 15 had a disease score of 1 or 2, and almost half (55.5%) of our patients had a disease score of 1 or 2. Both studies used the same version and scoring of the CCI. Patients were placed in 4 CALE categories: Ͻ5 years, 5 to 10 years, 10 to 20 years, and Ͼ20 years. These 4 categories were scored, respectively, from 1 to 4.
Patients were asked the following 2 questions: "How long do you expect you will live without any treatment for prostate cancer?" (Q1) and "How long do you expect you will live after treatment for prostate cancer?" (Q2). The possible responses to both questions were grouped into 4 categories (similar to the CALE categories): Ͻ5 years, 5 to 10 years, 10 to 20 years, and Ͼ20 years. These 4 categories were also scored from 1 to 4, respectively.
Based on Q1, Q2, and CALE scores, we calculated the patient's PDLO (which is CALE category score minus Q1 category score) and the patient's PILT (which is Q2 category score minus Q1 category score). A PDLO of Ն10 years is at least a 2-category difference between the CALE category and the Q1 category; this is only possible if the CALE was Ͼ20 years and the Q1 response was 5 to 10 years or if the CALE was 10 to 20 years and the Q1 response was Ͻ5 years. Similarly, a PILT of Ն10 years indicates that the response to Q2 was 10 years or more than the response to Q1. We conducted ordinal logistic regression analyses to identify the main sociodemographic, health, and cancer characteristics that could predict PDLO and PILT of Ն10 years.
The following validated self-administered scales were used. (1) The Short-Form 36 (SF-36, version 2) measures generic HRQOL; we calculated physical component summary and mental component summary scores from SF-36 data. 18 (2) The Prostate Cancer Index measures urinary, sexual, and bowel symptoms and how much they bother the patient. 19 (3) The Duke Activity Status Index 20 measures functional capacity in metabolic equivalents; this scale asks patients whether they could perform 12 activities which have different levels of exertion. (4) The Hospital Anxiety and Depression Scale measures the presence and severity of anxiety and depression. 21 (5) The Fear of Cancer Recurrence Scale measures the fear of possible cancer recurrence. 22 (6) The Medical Outcomes Study Social Support Survey measures social support in an overall score that includes multidimensional sub-scores. 23 (7) The Delighted-Terrible Seven Faces Scale 24 was used to measure patient satisfaction with life, health, and with education given by physicians about treatment options for LPC. (8) The Rapid Estimation of Health Literacy in Medicine scale, discussed earlier, measures health literacy; this was the only scale administered by telephone. 14

Statistical Analyses
Frequencies and relative frequencies were used to describe categorical variables. Continuous variables were described using the mean, median, and SD. 2 tests, Fisher's exact test,and independent sample t tests were used to examine bivariate associations. Unadjusted and adjusted odds ratios (ORs) and their 95% CIs were estimated using ordinal logistic regression analysis. Sociodemographic and health factors that were found to be associated with PDLO and PILT in the bivariate analysis at an ␣ level of 0.20 were kept in the multivariate models. All analyses were performed using SAS software (version 9.1, SAS Institute, Inc., Cary, NC).

Results
Surveys were mailed to 430 patients newly diagnosed with LPC, but 69 patients had already started treatment by the time the patients received the surveys, 3 patients never received the surveys, and 2 patients were found to be ineligible to participate because their cancer was not localized to the prostate. Of the 356 remaining patients, 104 patients did not return the survey because they were "not interested" in participating and 68 patients who did not return the surveys did not give a reason for not participating or could not be contacted. One hundred eighty-four of 356 patients (survey response rate of 52%) completed and returned the first pretreatment survey; 23 of these 184 patients (12.5%) patients chose observation. Table 1 shows a demographic comparison of patients who chose treatment or observation. Mean patient age was 61.5 years, and most patients reported college education and a family income of Ͼ$50,000. Table 2 shows a comparison of patients who chose treatment or observation by cancer grade, PSA, life expectancy by age, and comorbidity scores. Mean Gleason grade was 6.6. Table 2 also includes a comparison of these patient groups by SF-36 scores (Physical Component and Mental Component Scores), as well as the urinary, sexual, and bowel function scores of the 144 of 184 patients who had also returned the second pretreatment survey. Table 3 shows a comparison of patients who chose treatment or observation by CALE, anxiety and depression, function capacity, social support, and satisfaction. Mean CALE was 22.9 years. Table 4 shows the baseline CALE for the 184 patients and the responses of 170 of the 184 patients who had answered questions about their perceived life expectancy without treatment (Q1) and with treatment (Q2) of the cancer. Without treatment, perceived life expectancy was Ͻ5 years in 15.2%, 5 to 10 years in 48.8%, 11 to 19 years in 33.5% and Ն20 years in 2.4% of the patients. With treatment, it was Ͻ5 years in 0.6%, 5 to 10 years in 6.5%, 11 to 19 years in 30.0%, and Ն20 years in 62.9% of the patients. By contrast, baseline CALE was Ͻ5 years in 0.5%, 5 to 10 years in 2.2%, 11 to 19 years in 36.4%, and Ն20 years in 60.9%. A total of 170 patients had data on CALE, Q1, and Q2. As compared with CALE, 65 (38.2%) of these 170 patients expected survival to decrease by Ն10 years without treatment. As compared with their perceived survival without treatment, 81 (47.6%) of 170 patients expected their survival to increase by Ն10 years with treatment. Of the 108 patients with a baseline CALE of Ͼ20 years, only 2 (1.9%) expected to live beyond 20 years without treatment whereas 84 (77.8%) expected to live beyond 20 years with treatment (data not shown). Neither of these perceptions was significantly related to whether the patients chose treatment or observation. However, to a statistically insignificant extent, patients who chose treatment were more likely than observation patients to expect a Ն10-year reduction in survival without treatment (50% vs. 33.3%, 2 test; P ϭ .26) and a Ն10-year increase in survival with treatment (39.9% vs. 16.7%, Fisher's exact test; P ϭ .13).
Tables 5 and 6 present ordinal logistic regression models for PDLO and PILT. Age, CALE, depression, and anxiety scores predicted both PDLO and PILT. Furthermore, PSA level predicted PDLO, whereas social support predicted PILT.

Discussion
Prostate cancer is the most common solid cancer in men. Younger patients make up a fast-growing population that is being screen-detected and treated for low-risk LPC. 25 To our knowledge, this is the first study to report the perceptions of newly diagnosed patients about how the cancer or its treatment could affect their survival. The mean age of our patients (61.5 years) was similar to the range of 58 to 64 years of US patients currently undergoing radical prostatectomy. 26 The mean Gleason grade was 6.6 in our patients, similar to other series in which almost half of screen-detected cancers were "insignificant." 27 By choosing treatment,   Data provided as n (%). CALE, comorbidity adjusted life expectancy; Q1, How long do you expect you will live without any treatment for prostate cancer?; Q2, How long do you expect you will live after the treatment of your choice for prostate cancer?   34 Overtreatment would be expected if patients believe that treatment will lead to a much longer survival. Many studies have found that nearly every patient initially wants eradication of the cancer. 9 In qualitative studies, some patients accepted side effects for any gain in survival but they were convinced that treatment would improve survival. 9 Assuming that tumors would grow exponentially, urologists at the Mayo Clinic were also of the opinion that only 0.3% and 14.5% of screen-detected LPCs were "clinically insignificant." 35 Patient anxiety caused by the new diagnosis of cancer and the consensus advice of specialists that LPC patients with a CALE of Ͼ10 years should choose treatment or be offered treatment 6 will lead to high treatment rates. In 70% to 90% of patients, a treatment plan is usually made in a single visit to the urologist after a positive biopsy. 36 The mean Gleason grade of our patients was 6.6, and in 87% of patients the mean PSA was Ͻ10, both of which are low-risk categories but for which national guidelines recommend either observation or treatment. 8 Only 12.5% of patients in our study chose observation. Specialists frequently recommend treatment even in low-risk patients because over 10 to 15 years the cancer may progress. 29 To manage this risk, a strategy of active surveillance with deferred initial treatment 28 is being increasingly recommended for patients at lower risk, ie, with cancers of Gleason grade Ͻ7, cancer stages T1c to T2a, and PSA Ͻ10. Almost half of patients with screen-detected cancer possess such characteristics. 27 In conjunction with specialists, primary care physicians (PCPs) can follow patients who choose this strategy. PCPs may also offer more balanced advice because they might be more knowledgeable about the patient's preferences, comorbidities, and baseline CALE. 37 Patients with LPC may also want to review educational materials with their PCP. The American Cancer Society website was recommended because its content, accuracy, balance, and readability was rated the highest among 44 patient education materials about LPC. 38 Overtreatment can also be reduced with decreased screening, and several studies have shown that fewer patients want PSA screening if they are counseled before screening. 39 Enthusiasm for routine screening is high among specialists who treat LPC. In a random nationwide survey, 43% of 559 radiation oncologists recommended routine PSA screening in average-risk patients older than 80. 6 Primary care physicians who frequently order PSA testing without much discussion about risks and benefits of testing cited reasons of lack of time, competing demands, limited patient health literacy, and fear of liability. 40 Prescreening counseling is difficult because it is unclear what and how much discussion should occur. As yet, we cannot say that screening or treatment can improve survival. The deleterious effects of treatment on urinary, bowel, and sexual dysfunction are better known; however, their frequencies and severities after different treatment techniques have been reported in more than 800 publications, 41 vary greatly, and are difficult to balance. We can also share with patients that there is a small risk of immediate morbidity and mortality associated with prostate biopsy and cancer treatment; that a new anxiety results from a positive PSA test, whether or not it is followed by a negative biopsy; and that we cannot compare the risk of death caused by comorbid diseases with that of death caused by cancer without knowing the grade and stage of the cancer. However, patients must also know that, even if diagnosed with cancer, no randomized controlled trial has shown that treatment can or cannot improve survival in patients with screen-detected cancer. Interestingly, a study found that more African American patients wanted PSA screening after the use of a decision aid, 39 presumably because of the appreciation of the higher risk in African American patients.
In our patients, age, CALE, depression, and anxiety were the most important predictors of PDLO or PILT. Though PSA level also influenced PDLO, PILT was related to social support. PDLO and PILT were not related to other factors such as race; income; education; health literacy; physical and mental summary SF-36 scores; urinary, bowel, and sexual symptoms; choice of treatment or observation; fear of cancer recurrence; functional capacity; and satisfaction with life, health, or with education by physicians in cancer treatment options. Although the association of PDLO and PILT with continuous depression scores was statistically significant, the importance of this finding is unclear given that 96.6% patients had a depression score of Ͻ7, which indicates no depression.
Our findings may be difficult to generalize because our study sample was small. However, the differences we found between CALE and patient expectations of survival with and without treatment were large. Also, the mean age 26 and mean cancer grade 27 of patients newly diagnosed with LPC in large series and in our patients were similar. Our patients were treated by urologists in a private practice and more than 80% of urologists in the United States are in private practice. 42 Our method of equating CCI scores with NCCN recommended health quartiles to estimate CALE is new and has not been previously validated. We used this method because we could not find any other validated method to estimate long-term health-adjusted life expectancy in individual ambulatory patients. Finally, we lost accuracy in the estimation of PDLO and PILT by asking patients to predict their survival in ordinal intervals rather than in a discrete number of years. We used ordinal intervals because it might be easier for patients to predict their survival this way, and the intervals allowed an estimation of PDLO and PILT of more than or less than 10 years.

Conclusion
In patients with newly diagnosed LPC, in whom the mean cancer grade was Ͻ7 and in whom education, income, and health literacy was intermediate to high, almost 38% of our patients had expectations of a reduction in survival of Ն10 years if they chose observation, and 48.8% patients expected an improvement in survival of Ն10 years through choosing treatment. These expectations are highly unrealistic because no study has shown that the cancer or its treatment can affect survival by even 1 year, especially in screen-detected patients with a cancer Gleason grade of Ͻ7.