Fax: (404) 639-2980
Article first published online: 15 NOV 2002
Copyright © 2002 American Cancer Society
Volume 95, Issue 11, pages 2308–2315, 1 December 2002
How to Cite
Hall, H. I., Satariano, W. A., Thompson, T., Ragland, K. E., Van Den Eeden, S. K. and Selvin, S. (2002), Initial treatment for prostate carcinoma in relation to comorbidity and symptoms. Cancer, 95: 2308–2315. doi: 10.1002/cncr.10926
This article is a US Government work and, as such, is in the public domain in the United States of America.
Data collection supported by a cooperative agreement (U48/CCU909706) between the Centers for Disease Control and Prevention and the Center for Family and Community Health at the University of California, Berkeley.
- Issue published online: 19 NOV 2002
- Article first published online: 15 NOV 2002
- Manuscript Accepted: 19 JUN 2002
- Manuscript Revised: 29 MAY 2002
- Manuscript Received: 3 OCT 2001
- prostatic neoplasm;
Evidence suggests the type of treatment received for prostate carcinoma is associated with comorbidity, but little information is available on associations with specific comorbid disease or symptoms. The authors examined the relations between treatment and comorbidity, specific comorbid disease, and symptoms.
Medical records were abstracted for 1054 male members of the Kaiser Permanente medical care program diagnosed with prostate carcinoma from 1975 to 1987. Information was obtained on demographic characteristics, comorbid conditions, symptoms, tumor stage and grade, and treatment. Logistic regression was used to determine the significant predictors of treatment (radiation vs. nonaggressive treatment and surgery vs. nonaggressive treatment).
Compared to nonaggressive treatment, radiation treatment was less likely among men who had prior cancer (adjusted odds ratio [OR] 0.29, 95% confidence interval [CI] 0.09–0.90) or cerebrovascular disease (OR 0.33, 95% CI 0.13–0.83). There was a significant interaction between race and myocardial infarction (P = 0.02). Surgery, compared to nonaggressive treatment, was less common among men with a prior cancer (OR 0.21, 95% CI 0.07–0.63) or congestive heart failure (OR 0.29, 95% CI 0.09–0.90). Significant interactions were observed between race and myocardial infarction (P = 0.01), diabetes and dysuria or hematuria (P = 0.02), and para- or hemiplegia and urinary frequency or nocturia (P = 0.01).
Specific symptoms and comorbidity appear to influence treatment for prostate carcinoma. More research is needed on treatment differences by race. Cancer 2002;95:2308–15. Published 2002 by the American Cancer Society.
Since the mid-1980s, aggressive treatment of prostate carcinoma, usually with surgery but often with radiation therapy, has become more common.1, 2 Even so, there is insufficient evidence to date as to whether treatment such as prostatectomy decreases mortality when compared with expectant management.3, 4
Several factors may influence treatment decisions, including the tumor stage and grade, the patient age and life expectancy, clinical factors, and personal preferences. Treatment has also been found to vary by race and geographic area.2, 5, 6 In addition, aggressive treatment, defined as surgery or radiation, has been found to be more likely to be administered to men who lacked comorbid conditions and less likely among men with certain symptoms.6, 7 However, overall there is little information available on the relations between specific treatments and specific comorbid disease or symptoms of prostate disease.
In the current study, we examine the relations between initial treatment for prostate carcinoma and comorbidity, with both a comorbidity index and specific disease, and symptoms of prostate disease.
PATIENTS AND METHODS
Men diagnosed with prostate carcinoma from 1975 to 1987 who were members of the Kaiser Permanente Medical Care Program in the San Francisco Bay area were eligible for the current study. Cases were identified through the computerized cancer incidence files of the Kaiser Division of Research, the Greater Bay Area Cancer Registry, and the Surveillance, Epidemiology, and End Results (SEER) program for the San Francisco Bay area. The study sample included all black men diagnosed from 1975 to 1987 and a sample of white men randomly selected for each month within the time period 1975–1987. In all, 1054 white and black men were included in the current study; other races or ethnicities were excluded because of low numbers.
Medical Record Review
Information was obtained from electronic files of the SEER registry (stage, grade, treatment) and Kaiser Permanente medical records. All patients had a medical record containing all contacts (hospital and ambulatory). Hospital discharge electronic records were used to obtain information on comorbidity. After standardized chart abstraction forms were developed, trained abstractors reviewed the complete medical records for additional information. Information obtained from the records included race, marital status, performance of diagnostic tests (digital rectal exam, prostate specific antigen [PSA] test, bone scan), performance of chemotherapy or hormone therapy, and symptoms of prostate disease recorded up to 12 months before or at the time of diagnosis. Because only one man had a PSA test before diagnosis, this variable was not included in the analyses.
Stage and grade at diagnosis of prostate carcinoma were obtained from SEER data. The four categories for stage were local (confined to the prostate), regional (spread to adjacent tissue or lymph nodes), distant (spread to distant tissue or lymph nodes), and unstaged. The four categories of grades were 1 (well differentiated; Gleason score 2, 3, 4), 2 (moderately differentiated; Gleason score 5, 6, 7), 3 and 4 (poorly differentiated; Gleason score 8, 9, 10 [Grade 3]; and undifferentiated [Grade 4]), and undetermined. Grades 3 and 4 were combined because of the small number of patients with Grade 4 cancer.
Information on first course of treatment (treatment within four months of diagnosis) was obtained from the SEER registry, and additional information on chemotherapy and hormone therapy was obtained from the medical record review. For analysis, two treatment categories (surgery or surgery and radiation therapy; radiation therapy only) were created and compared to nonaggressive treatment (no treatment, hormone therapy, or chemotherapy). Participants who received hormone therapy or chemotherapy in addition to surgery or radiation were classified as having had surgery or radiation in the current analyses regardless of whether the hormone therapy or chemotherapy preceded, followed, or was contemporaneous with the more aggressive treatment. Earlier studies6–8 combined surgical and radiation therapy into an aggressive treatment category; however, the current data indicate that the effect of predictors on these treatments may be different.
Information on comorbidity was obtained from computerized hospital discharge summaries for all hospitalizations beginning five years before and including the time of prostate carcinoma diagnosis. Comorbid conditions related to prognosis (i.e., short term mortality) were assigned scores, and these scores were summed to derive a weighted index of comorbidity (Charlson index).9, 10 Scores were assigned according to the method of Deyo et al.10 to conditions derived from hospital discharge summary reports, which were coded in the diagnoses and procedures codes of the ninth revision of the Manual of the International Statistical Classification of Diseases, Injuries, and Causes of Death (ICD-9).11 Comorbid conditions that were coded to the ICD-8 in the medical records were recoded by ICD-9. In earlier research, Satariano et al. found that hospital discharge summary reports represent the leading comorbid conditions.12 In the current analyses, the comorbid conditions were first analyzed using the Deyo comorbidity index for descriptive purposes and then by disease in logistic regression modeling.
Demographic, clinical, and comorbidity data were compared according to the type of treatment received. Continuous variables are presented as medians, 25th, and 75th percentiles, and discrete variables as frequencies and percentages. Statistical testing was performed using the Pearson chi-square test or Fisher exact test (two-sided) for discrete variables. The Wilcoxon rank-sum test was used to test for differences among continuous variables.
Multivariable logistic regression techniques were used to determine the significant predictors of treatment. Two models were developed, the first comparing radiation versus nonaggressive treatment and the second comparing surgery or both surgery and radiation versus nonaggressive treatment. Binary logistic regression models were used because the sample sizes were too small to examine the data together using a generalized logits model for a polytomous response. An efficient multiple imputation procedure, implemented in S-plus (version 6.0; Insightful Corp., Seattle, WA), was used to impute the missing covariate values. This procedure makes use of relationships among all variables to impute the missing values and computes imputation-corrected variances of the parameter estimates to account for the added variability that arises from imputation. The relationship between age and each outcome was tested for linearity using restricted cubic spline functions. Restricted cubic spline plots were used to determine appropriate transformations needed to induce linearity. For both outcomes, age was transformed in the final model with a quadratic polynomial. The two-way interactions between the comorbid conditions and age, race, stage, and symptoms were examined. The comorbid conditions examined were myocardial infarction, congestive heart failure, cerebrovascular disease, chronic pulmonary disease, rheumatologic disease, peptic ulcer disease, liver disease, diabetes, hemiplegia or paraplegia, renal disease, and cancer or metastatic solid tumor. Statistical testing for both models was performed using the Wald chi-square test. Results are also presented as odds ratios and 95% confidence intervals.
A total of 1054 men with prostate carcinoma were identified. There were 636 (60.3%) white men in the sample and 418 (39.7%) black men. The median age at diagnosis was 69 years (range, 35–95 years). More than half of the men were diagnosed with local cancer (n = 570, 54.1%), 129 (12.2%) had regional cancer, and 229 (21.7%) had distant cancer. For 126 (12.0%) men the cancer was not staged. The majority of men (63.6%) received aggressive treatment: 21.7% received surgery, 30.8% radiation, and 11.0% both surgery and radiation. The remaining 36.4% received no treatment or received hormonal treatment or chemotherapy (i.e., nonaggressive treatment). A substantial proportion (61.1%) of men had no comorbid conditions (comorbidity summary score of 0); 18.2% had a score of 1; 7.5% had a score of 2; and 13.2% had a score of ≥3.
Men who were older, who had higher stage and grade disease, or who had a bone scan were less likely to receive aggressive treatment (Table 1). Married men were also more likely to receive radiation treatment. Men with a high comorbidity score or urinary incontinence were less likely to receive radiation therapy than men who received surgery or nonaggressive treatment. Surgery patients were more likely to have presented with dysuria, urinary frequency or hesitancy, hematuria, or nocturia. There was little difference in treatment by race or receipt of a digital rectal exam before diagnosis. We found some year-to-year variation in the proportion of men receiving aggressive treatment, and on average the proportion of men treated increased from 1975–1979 to 1984–1987, but none of the changes were significant.
|Characteristic||Nonaggressive (n = 384)||Radiation (n = 325)||Surgery/both (n = 345)||P value|
|Age, median (25th and 75th percentile)||73 (66, 78)||66 (61, 70)||69 (63, 74)||< 0.0001|
|3 or 4||44%||27%||28%|
|Digital rectal exam||92%||95%||92%||0.2513|
|Urinary frequency||53%||45%||63%||< 0.0001|
|Urinary hesitancy||26%||17%||60%||< 0.0001|
Men with congestive heart failure, cerebrovascular disease, chronic pulmonary disease, or diabetes were less likely to receive radiation therapy than surgery or nonaggressive treatment (Table 2).
|Characteristic||Nonaggressive (n = 384)||Radiation (n = 325)||Surgery/both (n = 345)||P value|
|Congestive heart failure||5%||1%||3%||0.0036|
|Peripheral vascular disease||1%||1%||1%||1.00|
|Chronic pulmonary disease||10%||4%||8%||0.0079|
|Peptic ulcer disease||8%||3%||5%||0.0207|
|Hemiplegia or paraplegia||7%||3%||7%||0.0970|
|Cancer or metastatic solid tumor||4%||2%||3%||0.1721|
Radiation Compared with Nonaggressive Treatment
Cerebrovascular disease, prior cancer or metastatic tumor, and distant or unstaged cancer were all associated with decreased likelihood of radiation therapy (Table 3). Age was also a significant predictor of treatment, but this effect was nonlinear (Fig. 1). Age did not have an effect on treatment for patients below an age of around 60. For ages above 60, increased age was associated with decreased radiation. There was a significant interaction between race and myocardial infarction (P = 0.018). Overall, black men were less likely to receive radiation treatment compared to white men. However, this effect was more noticeable among patients with myocardial infarction (OR = 0.06, 95% CI = 0.01–0.41) compared to those without a myocardial infarction (OR = 0.64, 95% CI = 0.42–0.96). Similar results were found when excluding patients with late stage prostate carcinoma (data not shown).
|Characteristic||Chi-square||d.f.||P value||Odds ratio||95% CI|
|Regional vs. local||.65||.37–1.17|
|Distant vs. local||.04||.02–.07|
|Unstaged vs. local||.28||.17–.48|
|Age (nonlinear)||86.82||2||< .001||-||-|
|Black vs. white|
|Cancer or metastatic tumor||4.57||1||.033||.29||.09–.90|
Surgery Compared to Nonaggressive Treatment
Men who had congestive heart failure, late stage or unstaged cancer, a digital rectal exam, or cancer or metastatic tumor were less likely to receive surgery (Table 4). Urinary hesitancy was associated with increased surgery. Age was a significant nonlinear predictor of treatment, with a relationship similar to the one observed for radiation versus nonaggressive treatment (Fig. 2). There was a significant interaction between diabetes and dysuria or hematuria (P = 0.021). Among patients without dysuria or hematuria, diabetics were more likely to receive surgery (OR = 3.02, 95% CI 1.68–5.43). Among patients with dysuria or hematuria, there were no treatment differences between diabetics and nondiabetics. There was also a significant interaction between urinary frequency/nocturia and paraplegia/hemiplegia (P = 0.008). Among patients without urinary frequency, patients with para- or hemiplegia were less likely to receive surgery (OR = 0.08, 95% CI = 0.01–0.90) than patients without this comorbid condition. Among patients with urinary frequency, para- or hemiplegia was associated with higher rates of surgery (OR = 2.82, 95% CI = 1.01–7.83). An interaction between race and myocardial infarction (P = 0.014) showed a decreased likelihood for black men with this comorbidity to receive surgery. Similar results were found when excluding patients with late stage prostate carcinoma (data not shown).
|Characteristic||Chi-square||d.f.||P value||Odds ratio||95% CI|
|Regional vs. local||.54||.30–.98|
|Distant vs. local||.05||.03–.08|
|Unstaged vs. local||.03||.01–.08|
|Urinary hesitancy||42.71||1||< .001||4.77||2.98–7.61|
|Digital rectal exam||13.29||1||< .001||.22||.10–.50|
|Urinary frequency or nocturia||10.31||2||.006|
|Cancer or metastatic tumor||7.59||1||.006||.21||.07–.63|
|Black vs. white|
|Paraplegia or hemiplegia||8.22||2||.016|
|With urinary frequency||2.82||1.01–7.83|
|No urinary frequency||.08||.01–.90|
|Congestive heart failure||4.63||1||.032||.29||.09–.90|
The current study suggests that men who have certain comorbidities are less likely to receive aggressive treatment for prostate carcinoma. These results confirm earlier findings6, 7 that receiving aggressive treatment for prostate carcinoma is less likely with certain comorbidities or increasing comorbidity scores; a study that compared any treatment and no treatment found a similar relation.13 However, we found that surgery and radiation treatment may be related to different factors and should therefore be assessed separately. We also found that the effects of comorbid conditions may vary according to other risk factors, such as presenting symptoms.
Treatment decisions may be related to the presence of comorbidity because such problems can limit the success of therapies and shorten life expectancy. Among deceased men with prostate carcinoma, higher comorbidity scores were associated with shorter time to death14 for both men who died of the cancer and those who died of other causes. Men with prostate carcinoma were also more likely to die from other causes than prostate carcinoma when comorbidity was present.12, 15
Particular comorbid conditions may have a relatively greater impact on treatment decisions. The current results of a decreased likelihood of surgery or radiation treatment in the presence of other nonprostate neoplastic disease confirm earlier findings6 and may reflect concerns about life expectancy or treatment complications. The current results are also similar to those reported by Harlan et al.,7 who found less frequent aggressive treatment among men with heart attacks. The reasons for the lower likelihood of radiation treatment with age > 60 years and the age-treatment relation found in earlier studies12, 16 may also relate to considerations of treatment outcomes and life expectancy. Comorbidity increases with age,17 and both factors were predictors of treatment. Likewise, palliative care may be more appropriate with diagnosis at a later stage.
Among the symptoms of prostate disease assessed in the current study, urinary hesitancy was associated with an increased likelihood of surgery. Urinary hesitancy may be considered to require attention because of potentially severe consequences if left untreated. In fact, urinary hesitancy, dysuria, hematuria, and urinary frequency may all be considered serious symptoms that affect treatment decisions. Symptoms that have been found to be associated with conservative therapy among men with localized prostate carcinoma include impotence and poor bladder control.7 Before the advent of PSA testing to screen for prostate carcinoma, urinary hesitancy was likely to be treated by transurethral resection of the prostate, which often led to a diagnosis of prostate carcinoma. In the current study, information was not collected on transurethral resection before prostate carcinoma diagnosis.
Treatment decisions may then appear complex when considering multiple factors, such as comorbidity, symptoms, and age. For example, among those with hemi- or paraplegia, we found that those presenting with urinary frequency or nocturia were more likely to receive surgery than those without. While in some cases those who did not receive surgery may still receive aggressive treatment (i.e., radiation treatment), among patients who presented without urinary frequency or nocturia, those with hemi- or paraplegia were less likely to receive aggressive treatment than those without hemi- or paraplegia (data not shown). Among patients with para- or hemiplegia, symptoms of a neurogenic bladder are common. Severity of symptoms was not measured in the current study.
Although univariate analyses did not reveal significant differences in treatment between white and black men, interactions were found between race and myocardial infarction. This suggests that black men with this comorbidity were less likely to receive radiation treatment or surgery for prostate carcinoma in the health care organization we studied. The men in the current study were drawn from an equal-access medical care system; a previous study in such a system (the military) found no stage-specific differences in treatment between blacks and whites,18 but there was no adjustment for comorbidity in that study. In addition, an interview study of black and white men in North Carolina failed to find an association between treatment and race.19 Conversely, other studies have found lower rates of aggressive treatment among black men than among white men.8, 20 The latter studies included men covered by various health insurance systems, with the results not adjusted for insurance status. In a study of Medicare recipients, an interaction was found between race and transurethal resection of the prostate in predicting aggressive treatment.6 Finally, one study found an interaction between age and race, with a decrease in aggressive treatment with age among black men ≥ 60 years.
Earlier studies6, 7, 13, 21 compared aggressive treatment to nonaggressive treatment or surgery to radiation. This does not allow for an assessment of differences in relations between risk factors and treatment type and complicates assessments of type of treatment chosen. Future studies need to further examine whether differences exist in the treatment-comorbidity relationship between surgery and radiation and whether grouping surgery and radiation into an aggressive treatment category is appropriate. Researchers might also seek to confirm the current finding of interactions between symptoms and comorbidity.
The strengths of the current study include selection of subjects from an equal-access medical care system, availability of information on comorbid conditions, and an assessment of symptoms of prostate disease. Data from the SEER program are generally of very good quality,22 and data on aggressive treatment modalities —surgery and radiation— are generally found to be of good accuracy.23–25 Among the limitations of the current study was the modest sample size, which at times resulted in wide confidence intervals. In addition, although we used data from an equal-access health care system, data on other socioeconomic factors that may be related to treatment, such as education or income, were not available, nor were data available on knowledge of treatment options or the choices offered to patients. Adjustment for stage may have been incomplete in part because of incorrect staging26 or because assignment of stage was based on surgery-related findings among men who received surgery. We used a validated comorbidity index and specific comorbid conditions in the current study. However, there may be additional factors that influence treatment decisions. Finally, the study cohort (diagnosis years 1975–1987) was diagnosed before PSA testing became widespread, which led to a higher number of men diagnosed with early stage cancer. Future studies need to clarify the relations between treatment and race.
The use of comorbidity indices has been found to be useful in predicting survival14 among prostate carcinoma patients. The results from the current study and previous investigations suggest that comorbidity may be a strong factor in decision making about treatment for prostate carcinoma. It has been suggested that comorbidity indices may help clinicians in the decision making process;14 however, individual comorbidity conditions and symptoms appear to be important as well.
- 3The prostate cancer intervention versus observation trial. A randomized trial comparing radical prostatectomy versus expectant management for the treatment of clinically localized prostate cancer. Cancer. 1995; 75: 1963–1968., .
- 5The National Cancer Data Base report on prostate carcinoma after the peak in incidence rates in the U.S. The American College of Surgeons Committee on Cancer and the American Cancer Society. Cancer. 1998; 83: 1679–1684., , , .
- 11World Health Organization. Manual of the international statistical classification of diseases, injuries, and causes of death, based on the recommendations of the ninth revision conference, 1975. Geneva (Switzerland): World Health Organization, 1977.
- 22ChenVW, HoweHL, WuXC, HotesJL, CorreaCN, editors. Cancer in North America, 1993–1997. Vol. 1: incidence. Springfield, IL: North American Association of Central Cancer Registries, 2000.
- 25Cancer treatment: registry data versus claims data. Presented at the Annual Meeting of the North American Association of Central Cancer Registries, New Orleans, LA, April 2000., .