The Relation of Surgery for Prostatic Hypertrophy to Carcinoma of the Prostate

American Journal of Epidemiology Vol. 138, No. 5 Copyright C 1993 by The Johns Hopkins University School of Hygiene and Public Health Printed in U.SA. All rights reserved The Relation of Surgery for Prostatic Hypertrophy to Carcinoma of the Prostate Brad D. Simons, 1 Alan S. Morrison, 1 ' Robert H. Young, 2 and Wendy Verhoek-Oftedahl 1 The authors evaluated the risk of death from prostate cancer in Rhode Island men who had had surgery for prostatic hypertrophy during the years 1959-1970. The retrospective follow-up method was used. Among 4,853 men who had had either a transurethral resection or a prostatectomy, age- and time-standardized mortality from prostate cancer was 1.14 (95% confidence interval (Cl) 0.96-1.33) times the rate in the general population of Rhode Island men. The age- and time-standardized mortality ratio was 1.01 (95% Cl 0.77-1.31) for men who had had a suprapubic prostatectomy and 1.18 (95% Cl 0.94-1.47) for men who had had a transurethral resection. The authors used the case-control method to relate the histologic features of prostatic hypertrophy to subsequent prostate cancer. The observed associations were weak and inconsistent. The results of this study suggest that surgically treated prostatic hypertrophy is not an important determinant of prostate cancer. Am J Epidemiol 1993; 138:294-300. histology; mortality; prostatic hypertrophy; prostatic neoplasms Cancer of the prostate accounts for approximately 11 percent of cancer deaths in US men (1). This disease is the secondleading cause of cancer death in US men aged 55 years or more (1). The epidemiology of cancer of the prostate is not well understood. There may be an association of prostate cancer with prostatic hypertrophy, but the evidence is inconsistent (2, 3). We conducted a retrospective follow-up study to investigate this relation. Our primary goal was to evaluate the risk of death from prostate cancer in men who had had surgery for prostatic hypertrophy. Secondary goals were to relate the occurrence of prostate cancer to the type of operation for prostatic Received for publication March 18,1991, and in final form May 21, 1993. Abbreviation: Cl, confidence interval. 1 Department of Community Health, Brown University, Providence, Rl. 2 Department of Pathology, Massachusetts General Hospital, and Harvard Medical School, Boston, MA. Deceased. Reprint requests to Dr. Wendy Verhoek-Oftedahl, Department of Community Health, Box G-A420, Brown University, Providence, Rl 02912. 294 hypertrophy and to the histologic characteristics of the benign prostatic tissue. MATERIALS AND METHODS Subjects consisted of male residents of Rhode Island under 75 years of age who were hospitalized in Rhode Island during the period 1959-1970 and who had prostatic surgery without cancer being found. Eligibility was limited to those men who had had no previous prostatic surgery and who did not have prostate cancer or bladder cancer diagnosed before hospital admission or in the subsequent 3 months. An original aim of the project was to identify patients statewide, but three small hospitals refused to participate, and pertinent records had been lost or destroyed for various periods at some of the hospitals that did participate. Subjects were identified initially by review of original pathology reports of surgical specimens. The medical record was retrieved for each potential subject, and a follow-up form and a record abstract were initiated if the individual was found to be

Prostatic Hypertrophy and Prostate Cancer 295 eligible. Identifying data were recorded on the follow-up form. The abstract included the patient's date of birth, the date of the "index" admission (admission for initial surgery), the urologic procedure, the pathologic diagnosis, and the weight of the prostatic tissue removed. The abstract also contained provisions for recording data on subsequent prostate-related hospitalizations, the date of the most recent observation, the subject's vital status, and the causes of death of subjects who had died. The latter data items were obtained either from medical records at the "index" hospital (hospital of initial surgery) or from other follow-up procedures described below. There were 5,113 eligible subjects identified. The mean age at entry into the study was 65.0 years; 50 percent were aged 60-69. Twenty-one percent were under 60, and 29 percent were aged 70 years or more. Rhode Island death certificates were searched manually, beginning with the last date on which the subject was known to be alive. The minimum matching criteria were name, date of birth, and one of the following: address, place of birth, Social Security number (for some out-of-state deaths, the state of origin of the Social Security number was used), name of next of kin, or occupation. In evaluating potential matches, we allowed for possible errors in spelling or in transcription of numbers. For men with prostate cancer listed as a cause of death on their death certificates, we sought medical records in order to record autopsy findings and to determine the date of diagnosis of prostate cancer, if it was not already known. We used the National Death Index (4) for the years 1979-1987 and the Social Security Administration's Death Master File (5) for the years 1962-1987 to search for deaths among men for whom a Rhode Island death certificate was not found. For each apparently correct match, the death certificate was requested from the respective state. If prostate cancer was listed as a cause of death, autopsy findings and the date of diagnosis were requested from the hospital in which the death occurred. To identify cases of prostate cancer among subjects not found to have died of the disease, we searched pathology records or medical records at each participating hospital for evidence of prostate cancer in men whose index surgery occurred at another hospital. Two additional resources were used in follow-up: lists of holders of a Rhode Island driver's license, and directories (such as the Polk directories (R. L. Polk & Company, Maiden, Massachusetts)) for cities and towns in Rhode Island and neighboring areas of Massachusetts. Name, date of birth, and address were the matching criteria for a holder of a driver's license. Name, address, and name of next of kin were used to match subjects in the city and town directories. A subject for whom no death certificate had been found was considered alive if his name was located in one of these resources. Our analysis considered observations made through 1985. Of the 5,113 eligible subjects, 3,931 (76.9 percent) were known to have died on or before December 31, 1985. In total, 4,869 subjects (95.2 percent) were known to have died, or were followed past that date. The mean number of years of observation per subject was 10.8. Among 127 men who had prostate cancer listed on the death certificate, a record of a previous diagnosis of prostate cancer was found for 91 (71.7 percent). Of these, a record of histologic confirmation was found for 74 (81.3 percent). Among 3,810 men who died without prostate cancer being given as a cause of death, 110 (2.9 percent) were identified as having had a diagnosis of prostate cancer. Of these, a record of histologic confirmation was found for 94 (85.4 percent). Among 1,184 men not known to have died, 44 (3.7 percent) were identified as having had prostate cancer diagnosed, and 43 (97.7 percent) of these were found to have a histologically confirmed diagnosis.

296 Simons et al. The expected number of deaths was derived from mortality rates for Rhode Island men (J. Fulton, Rhode Island Department of Health, unpublished data). Adjustments for age and calendar time were made using distributions of person-years of observation for the specific group under study. The expected number of newly diagnosed cases of cancer was derived from ageand time-specific rates for Connecticut (6; L. Ries, unpublished data). A specialized histology review was carried out by means of the case-control method within the study group (7). For each death or incident case of prostate cancer identified, three controls were selected, insofar as possible, from the remaining index series of subjects. Cases and controls were matched according to age (within 5 years), as well as hospital and time of surgery for benign prostatic disease. Slides of the specimen from the index surgery were retrieved, when possible, for each case and control, and a single slide was chosen for review according to the amount of tissue and the adequacy of staining. The slides were reviewed by a pathologist (R. Y.) who did not know which subjects had developed prostate cancer. Slides were evaluated with respect to type of hyperplasia (glandular vs. fibromuscular), severity and extent of architectural atypia, and cytologic dysplasia. Slides were reviewed for 249 (88.6 percent) of 281 cases identified and for 754 (92.2 percent) of 818 selected controls. There were 23 cases and 11 controls for whom slides were reviewed who were not included in the analysis: Eleven cases were members of the study group who were known to have first developed prostate cancer after 1985; six cases had disease that was diagnosed as benign by the original pathologist but were considered by the reviewing pathologist to have prostate cancer present on the index slide; four cases were members of the study group who had been classified erroneously as cases; and two cases and 11 controls were found to be ineligible for the study. The respective 64 control slides that were reviewed for the 23 cases were retained in the study. Slides selected for 17 cases and for 30 controls could not be reviewed because the slides were not of adequate quality or did not contain prostatic tissue. Associations between histologic features of prostatic hyperplasia and prostate cancer were evaluated by means of odds ratios (7, 8). Age stratification (8) was incorporated into the analysis to accommodate the matching of cases and controls. Preliminary analysis indicated that it was not necessary to adjust for hospital and time of surgery. RESULTS Of the 5,113 subjects identified, 4,853 had had either a transurethral resection or a prostatectomy, and 236 had had a prostatic biopsy. Type of operation could not be determined for 24 men. Subjects who underwent a transurethral resection or a prostatectomy were considered to have had surgery for urinary obstruction from prostatic hypertrophy. The results described below are based primarily on this group. Of these 4,853 men, 2,896 (59.7 percent) had a transurethral resection during the index hospitalization, 1,906 (39.3 percent) had a suprapubic prostatectomy, and 51 (1.1 percent) had an operation that was described as "subtotal prostatectomy," "total prostatectomy," or simply "prostatectomy." Since there were no deaths from prostate cancer in subjects under 50 years of age, the analyses presented here are based on prostate cancer deaths, incident cases, and person-years of observation for men aged 50 years or more. Data on men who had biopsies were analyzed separately. Table 1 presents the prostate cancer mortality rate by age for men who had prostatic surgery for urinary obstruction. The rate increased strongly with age. The trend with age was similar to that seen in the general population of Rhode Island men (J. Fulton, Rhode Island Department of Health, unpublished data). The prostate cancer mortality rate in men who had surgery for urinary obstruction was estimated as 1.14 (95 percent confidence interval (CI) 0.96-1.33 (9))

Prostatic Hypertrophy and Prostate Cancer 297 TABLE 1. Prostate cancer mortality among Rhode Island men who had prostatic surgery for urinary obstruction and all Rhode Island men aged 50 years or more, by age group, 1959-1970 Age group (years) 50-64 65-69 70-74 75-79 80-84 2:85 All ages Men who had prostatic surgery deaths 7 11 19 33 18 17 105 Personyears of observation 10,192.3 10,161.6 13,447.0 10,796.2 5,547.1 2,307.2 52,451.41: Mortality rate* 68.7 108.3 141.3 305.7 324.5 736.8 200.2 Mortality rate among Rhode Island men'.t 14.5 67.5 131.7 259.1 385.2 608.7 176.4 Per 100,000 person-years of observation, t Standardized for calendar time t Excludes 974 pefson-years of observation (no deaths) In men under 50 years of age. Standardized for age and calendar time. times the rate in the general population of Rhode Island men, standardized for age and calendar time. The relation of prostate cancer mortality to duration of follow-up could be evaluated only among men aged 70-79 years at observation. The results are shown in table 2. The trend was irregular. For the age group 70-79 years, the mortality ratio, standardized for age and calendar time, was 1.14 (95 percent Cl 0.90-1.44) for all years of observation. The prostate cancer mortality ratio, standardized for age and calendar time, was 1.01 (95 percent Cl 0.77-1.31) for men who had TABLE 2. Prostate cancer mortality among Rhode Island men aged 70-79 years who had prostatic surgery for urinary obstruction, by duration of follow-up, 1959-1970 Duration of follow-up (years) 0-2 3-4 5-9 10-14 2:15 All years deaths 12 11 10 11 8 52 Person-years of observation 4,749.1 3,690.1 8,192.8 5,136.5 2,474.7 24,243.2 MortaHty rate* 274.6 321.5 117.7 206.4 361.2 214.5 PeMOO.OOO person-years of observation; standardized for age. a suprapubic prostatectomy and 1.18 (95 percent Cl 0.94-1.47) for men who had a transurethral resection of the prostate. The age-adjusted mortality ratio for prostate cancer in men who had a transurethral resection as compared with those who had a suprapubic prostatectomy was estimated (7, 10) as 1.16 (95 percent Cl 0.77-1.72). Among the men who had a prostatic biopsy, the prostate cancer mortality ratio, standardized for age and calendar time, was 4.50 (95 percent Cl 2.72-7.04). The weight of the prostatic tissue removed was known for 4,103 (84.6 percent) of the men who had prostatic surgery for urinary obstruction. Age-standardized prostate cancer mortality was 258.0 per 100,000 person-years of observation (based on 19 cases) for men who had ^7 g of tissue removed, 162.3 (13 cases) for men who had 8-13 g removed, 189.8 (17 cases) for men who had 14-22 g removed, 174.6 (19 cases) for men who had 23-43 g removed, and 174.9 (19 cases) for men who had >:44 g removed. Thus, the trend with weight is irregular, but the highest rate occurred in men who had ^7 g of tissue removed. Type of operation was very highly correlated with the weight of tissue removed; much less tissue tends to be removed in a transurethral resection than in a suprapubic prostatectomy. Therefore, the associations of type of surgery and weight of tissue removed to risk of prostate cancer could not be evaluated simultaneously. Based on age- and time-specific incidence rates of prostate cancer for Connecticut (6; L. Ries, unpublished data), the observed incidence in men who had surgery for urinary obstruction was close to that expected: The age- and time-standardized incidence ratio was 1.03 (95 percent Cl 0.91-1.15). The incidence ratio was slightly lower than the mortality ratio, which may be indicative of incomplete ascertainment of cases not followed by prostate cancer death. The observed mortality rate from all causes of death other than prostate cancer in men who had prostatic surgery for urinary

298 Simons et al. obstruction was 69.0 per 1,000 personyears. This rate is 1.20 (95 percent Cl 1.16-1.23) times the rate for Rhode Island men, with standardization for age and calendar time (J. Fulton, Rhode Island Department of Health, unpublished data). The agestandardized death rate for causes other than prostate cancer was 72.9 per 1,000 personyears for men who had a transurethral resection and 63.6 for men who had a suprapubic prostatectomy. The age-adjusted mortality ratio (7, 10) was 1.16 (95 percent CI 1.08-1.24). The reviewing pathologist considered prostate cancer to be present on the slides of 11 subjects who had been diagnosed by the original pathologist as having benign disease. Five of these subjects were known to have had a subsequent death certificate diagnosis of prostate cancer. One of the 11, who did not die of the disease, was known to have had a pathologic diagnosis of prostate cancer. We estimate that exclusion of men who had prostatic surgery for urinary obstruction who were thought to have prostate cancer by the reviewing pathologist would have led to about a 3 percent relative reduction in the observed mortality ratio for the disease in men who had prostatic surgery for urinary obstruction. The histologic analyses were limited to 192 cases (86.9 percent of cases eligible for histologic study for whom slides were reviewed) and 735 controls (98.9 percent of controls eligible for histologic study for whom slides were reviewed). These cases and controls had prostatic surgery for urinary obstruction and were considered by the reviewing pathologist to have a benign prostatic abnormality. The odds ratio for subsequent prostate cancer was 1.43 (95 percent CI 0.72-2.82) in men who had glandular hyperplasia as compared with men who had fibromuscular hyperplasia (table 3). Additional histologic features were evaluated in the 182 cases and 682 controls with glandular hyperplasia. Both cellular dysplasia and architectural atypia were found to have weak inverse (statistically nonsignifi- TABLE 3. Odds ratios for prostate cancer among Rhode Island men who had prostatic surgery for urinary obstruction, according to specified histologic characteristics of prostatic hyperplasia, 1959-1970 Histologic characteristic Glandular hyperplasiat Cellular dysplasiai Architectural atypia Extent Severity cases 182 8 9 10 controls 682 32 43 49 Odds ratio* 1.43 0.91 0.76 0.75 95% confidoncs \^\^i 11nil" ^^^^ Interval 0.72-2.82 0.42-1.96 0.37-1.57 0.38-1.49 * Adjusted for age. t Reference group was men with fibromuscular hyperplasia (10 cases, 53 controls). t Reference group was men with no cellular dysplasia (174 cases, 650 controls). Rated as more extensive than focal; reference group was men with focal atypia (173 cases, 639 controls). Rated as greater than mild; reference group was men with mild atypia (172 cases, 633 controls). cant) associations with prostate cancer. DISCUSSION Two previous reports have provided estimates of the association between prostatic hyperplasia and risk of prostate cancer (2, 3). Greenwald et al. (2) studied 838 patients with surgically confirmed benign prostatic hyperplasia and 802 age-matched comparison patients without cancer or prostatic disease who had surgery on an organ other than the prostate. They found no overall increase in risk of prostate cancer in patients with benign hyperplasia (incidence rates were 3.0 and 3.1 cases per 1,000 person-years of observation in study and comparison groups, respectively; both mortality rates were 2.3 deaths per 1,000 person-years). Armenian et al. (3) reported results from both follow-up and case-control studies of the relation of benign prostatic hypertrophy to prostate cancer. Subjects in the prospective study included 296 men with prostatic hypertrophy and 299 age-matched controls who had no diagnosis of cancer. The series with prostatic hypertrophy included 219 men whose diagnosis had been made on

Prostatic Hypertrophy and Prostate Cancer 299 clinical grounds or by biopsy and 77 who had had a prostatectomy. After exclusion of prostate cancer cases that were diagnosed during the first 6 years of follow-up, the death rate from prostate cancer was found to be 11.3 per 1,000 person-years in the series of men with prostatic hypertrophy diagnosed clinically or by biopsy, 5.1 per 1,000 person-years in the series with prostatectomy, and 2.7 per 1,000 person-years in the comparison group. The case-control study included 290 men with prostate cancer and 290 age-matched controls without a diagnosis of cancer. The relative risk of prostate cancer was estimated to be 3.7 in men who had been hospitalized for prostatic disease more than 6 years earlier, as compared with men without such a hospitalization. Since most of the men in the prospective study by Armenian et al. (3) had only a clinical diagnosis of prostatic hypertrophy, some of their urinary symptoms may have been caused by cancer. Therefore, a stronger observed association would be expected in their study than in the study by Greenwald et al. (2), which was limited to men with surgically confirmed benign disease. Although men who had a prostatectomy had a smaller increase in risk than did men with prostatic hypertrophy who did not have a prostatectomy, the elevation of risk of prostate cancer in the period 6 years or more after diagnosis of prostatic hypertrophy in both of the studies by Armenian et al. suggests that initially undiagnosed prostate cancer was not the only reason for the observed positive association. Nevertheless, the prospective study by Armenian et al. did not clearly distinguish preclinical prostate cancer from prostatic hypertrophy at entry in most subjects. Like the study by Greenwald et al. (2), our study was limited to men who had prostatic surgery without cancer being found in the specimen. We found only a small (and not statistically significant) elevation of prostate cancer mortality in men who had surgery for urinary obstruction compared with the general population of Rhode Island men. Together, these studies indicate that surgically treated prostatic hypertrophy is not an important determinant of the risk of prostate cancer. Prostate cancer may create symptoms of urinary obstruction. The cancer may not be discovered if a partial prostatectomy is performed, but it may be diagnosed some time later. Our results provide some support for this possibility. The prostate cancer mortality rate was relatively high among men who had surgery in which only small amounts of prostatic tissue were removed. Inapparent prostate cancer might have been relatively common in this group. Furthermore, we found the prostate cancer mortality ratio to be quite high among men who had biopsies but not among men who had a transurethral resection or a prostatectomy. Biopsies seem likely to have been done because prostate cancer was suspected (although it was not found at the time). On the other hand, the prostate cancer mortality associated with prostatic hypertrophy might have been underestimated. Other things being equal, men who have had urologic evaluation and prostatic surgery without cancer being found seem less likely to harbor an inapparent cancer, and therefore less likely to die of prostate cancer in the future, than men in the general population who have not had such an evaluation. An error from this source, however, may be small: Prostate cancer tends to arise in the outer part of the prostate (11), which is usually not removed in surgery for prostatic hypertrophy. Histologically premalignant lesions of the prostate have been described. In general, these lesions occur in the peripheral zone of the prostate (11). In our study, only a few subjects had specimens that showed cellular dysplasia or architectural atypia, and we found only weak and inconsistent associations of prostate cancer with these characteristics and with hyperplasia that was glandular (versus fibromuscular). Like the results described above, these findings suggest that prostatic hypertrophy and cancer

300 Simons et al. are not closely related. Roos et al. (12) reported that mortality from all causes was 45 percent higher following a transurethral resection, as compared with a suprapubic prostatectomy, for prostatic hyperplasia. We observed a trend in the same direction, but the difference was much smaller: The death rate for all causes other than prostate cancer was 16 percent higher for men who had had a transurethral resection than for men who had had a suprapubic prostatectomy. ACKNOWLEDGMENTS This study was supported by US Public Health Service grant R01 CA38707 through the National Cancer Institute. The authors are indebted to Joyce Coutu Babcock for her contributions to this project. The study could not have been conducted without the cooperation of staff members of pathology and medical records departments in Rhode Island hospitals and the Division of Vital Statistics, Rhode Island Department of Health. REFERENCES 1. Silverberg E, Boring CC, Squires TS. Cancer statistics, 1990. CA 1990;40:9-26. 2. Greenwald P, Kirmss V, Polan AK, et al. Cancer of the prostate among men with benign prostatic hyperplasia. J Natl Cancer Inst 1974;53:335-40. 3. Armenian HK, LUienfeld AM, Diamond EL, et al. Relation between prostatic hyperplasia and cancer of the prostate. Lancet 1974;2:115-17. 4. Patterson BH, Bilgrad R. Use of the National Death Index in cancer studies. JNCI 1986;77: 877-81. 5. The Death Master File. (Communication). Am J Public Health 1987;77:1548. 6. Young JL Jr, Percy CL, Asire AJ, eds. Surveillance, Epidemiology, and End Results: incidence and mortality data, 1973-77. Bethesda, MD: National Cancer Institute, 1981. (National Cancer Institute Monograph no. 57). 7. Kleinbaum DG, Kupper LL, Morgenstern H. Epidemiologic research: principles and quantitative methods. Belmont, CA: Lifetime Learning Publications, 1982. 8. Miettinen O. Estimability and estimation in casereferent studies. Am J Epidemiol 1976;103:226-35. 9. Mulder PGH. An exact method for calculating a confidence interval of a Poisson parameter. (Letter). Am J Epidemiol 1983;117:377. 10. Mantel N, Haenszel W. Statistical aspects of the analysis of data from retrospective studies of disease. J Natl Cancer Inst 1959;22:719-48. 11. Bostwick DG. Pathology of the prostate. New York: Churchill Livingstone, Inc, 1990. 12. Roos NP, Wennberg JE, Malenka DJ, et al. Mortality and reoperation after open and transurethral resection of the prostate for benign prostatic hyperplasia. N Engl J Med 1989;320:1120-4.