AMERICAN JOURNAL OF EPIDEMIOLOGY Vol. 121, No.

1
Copyright © 1985 by The Johns Hopkins University School of Hygiene and Public Health Printed in U.SA.
All rights reserved

THE DEVELOPMENT OF BENIGN PROSTATIC HYPERPLASIA

AMONG VOLUNTEERS IN THE NORMATIVE AGING STUDY

ROBERT J. GLYNN,1 EDWARD W. CAMPION,*3 GLEN R. BOUCHARD1 AND

JEREMIAH E. SILBERT1-4

Glynn, R. J. (VA Outpatient Clinic, Boston, MA 02108), E. W. Campion, G. R.

Bouchard and J. E. Silbert The development of benign prostatic hyperplasia

Downloaded from http://aje.oxfordjournals.org/ at Pennsylvania State University on September 18, 2016

among volunteers in the Normative Aging Study. Am J Epidemiol 1985;121:78-90.
This study describes the development of benign prostatic hyperplasia among
2,036 volunteers in the Veterans Administration Normative Aging Study, a longi-
tudinal study of human aging situated in Boston. Men were followed from enroll-
ment in the study (between 1961 and 1970) until their last examination prior to
May 15, 1982. Two indications of benign prostatic hyperplasia were considered:
1) a clinical diagnosis made at a uniform physical examination, and 2) surgical
treatment Incidence rates for both a clinical diagnosis and surgery for benign
prostatic hyperplasia increased through the eighth decade. Life table analysis
estimated the lifetime probability of surgical treatment to be 0.29. Known risk
factors for cardiovascular disease and diabetes as well as marital and socioec-
onomic status, religion, cigarette smoking and alcohol and coffee consumption
were evaluated as risk factors. Controlling for age in proportional hazards models,
statistically significant predictors of surgery were prior clinical diagnosis, lower
socioeconomic status, Jewish religion, and not currently smoking cigarettes;
whereas only body mass index was a significant predictor of a clinical diagnosis.
Although a prior clinical diagnosis was an important predictor of surgery (adjusted
odds ratio 3.52, 95% confidence interval = 1.93-6.42), this diagnosis is neither
sensitive nor specific In its association with surgery.

hyperplasia; longitudinal studies; prostate

Operations on the prostate are the most tate operations performed in 1979, giving a
frequently performed surgical procedures in rate of 22.6 per 1,000. Hyperplasia of the
men aged 65 years and over, and hyperpla- prostate is the fourth leading discharge di-
sia of the prostate is the leading reason for agnosis from non-Federal short stay hos-
prostate surgery (1, 2). In American males pitals for men 65 and older, with an annual
aged 65 and older there were 218,000 pros- r a t e of 16.5 discharges per 1,000 population
(1). In spite of its clinical and economic
Received for publication January 16, 1984, and in significance, there is remarkably Little in-
finalformApril 24,1984. formation relating precursor variables to
1
Normative Aging Study, Veterans Administration t n e onget ^ j t n e surgical treatment of
Outpatient Clinic, 17 Court St, Boston, MA 02108. . . . . 7 . .„ „. T
(Reprint requests to Dr. Glynn.) benign prostatic hyperplasia (2, 3). In par-
1
Geriatrics Unit, Massachusetts General Hospital, ticular, there have been no longitudinal
Boston, MA. afiidipa
3 slucue8
Division on Aging, Harvard Medical School, Bos- -
ton,4 MA. Previous case-control studies, examina-
Department of Medicine, Harvard Medical tions of clinical series, and comparisons of
School, Boston, MA. . , ., . ., , , r ,
Supported by the Medical Research Service of the worldwide incidence rates have found some
Veterans Administration. important characteristics of benign pros-
78
 DEVELOPMENT OF BENIGN PROSTATIC HYPERPLASIA 79
tatic hyperplasia. Benign prostatic hyper- nosis and by surgery for benign prostatic
plasia is seldom found in men under age 40 hyperplasia? 2) Are Jewish men at greater
but beginning at age 40 there is a very risk than non-Jewish men? 3) Are marital
strong age-related increase, estimated by status, socioeconomic status, as well as cof-
Moore (4) to reach a prevalence of 75 per fee, alcohol and cigarette consumption re-
cent among men aged 80-90. Lytton et al. lated to the development of benign pros-
(5) have calculated the probability of a 40- tatic hyperplasia? 4) Are known risk factors
year-old man requiring an operation for for coronary heart disease and diabetes also
benign prostatic hyperplasia, if he lives to risk factors for benign prostatic hyperpla-
age 80, to be approximately 10 per cent. sia? 5) What is the risk of surgery among
The prevalence is low among Japanese and those men with a prior clinical diagnosis

Downloaded from http://aje.oxfordjournals.org/ at Pennsylvania State University on September 18, 2016

Chinese men, high among whites, and as relative to the risk of men with no such
high or higher among blacks (3, 6). Two prior diagnosis?
studies (2, 5) reported that Jewish men
have a higher incidence than non-Jewish MATERIALS AND METHODS
men. Thus, there are apparent differences
in rates of benign prostatic hyperplasia by The Normative Aging Study is a longi-
ethnicity and religion. No association has tudinal study of human aging established
been found for marital status (2, 7, 8). by the Veterans Administration in 1961.
Benign prostatic hyperplasia has been Six thousand male volunteers from the
found to be related to lifestyle variables; Greater Boston area were screened to pro-
coffee drinkers are reportedly at somewhat vide an initially healthy population.
greater risk than men who do not drink Screening was based on laboratory, clinical,
coffee and smokers are at reduced risk com- radiologic, and electrocardiographic crite-
pared with nonsmokers (2, 7). Benign pros- ria. Men were disqualified from participa-
tatic hyperplasia is more prevalent in men tion in the study if they had a history of
with hypertension, coronary heart disease, conditions such as heart disease, cancer,
and diabetes (9-11); but these may be spu- gout, diabetes, cirrhosis, peptic ulcer, re-
rious relationships existing because of the current asthma, bronchitis or sinusitis.
common association of all these diseases Also disqualifying was either systolic
with age. Other epidemiologic findings, in- blood pressure greater than 140 mmHg or
cluding comparison of international mor- diastolic blood pressure greater than 90
tality rates and possible relationships with mmHg. Acceptable conditions included
prostate cancer, have been described (3, 7, childhood and other generally acute condi-
8). tions from which there were no sequelae.
No volunteer was excluded because of a
The present study examines prospec- clinical diagnosis or prior surgical treat-
tively the development of benign prostatic ment of benign prostatic hyperplasia.
hyperplasia among male volunteers in the With these criteria, 2,280 men were ac-
Normative Aging Study. Two outcomes are cepted into the Normative Aging Study,
considered: 1) the diagnosis of benign pros- ranging in age at entry from 21-81 with a
tatic hyperplasia at a regularly scheduled mean of 42 years. The educational level of
Normative Aging Study physical examina- the population as of 1971 was as follows: 14
tion, and 2) surgery for benign prostatic per cent less than high school, 25 per cent
hyperplasia as reported by participants as high school graduates, 35 per cent beyond
part of a medical history during the exam- high school and 26 per cent college gradu-
ination. We consider five specific questions ates. By occupation, 52 per cent were
suggested by the findings summarized white-collar workers (professional, mana-
above: 1) What are the age-specific inci- gerial, clerical and sales). Ethnically, the
dence rates determined by a clinical diag- population accorded fairly closely with the
80 GLYNN ET AL.

distribution found in the Boston area. amination were excluded. These excluded
However, only 2 per cent of the volunteers men were younger at baseline (mean ±
were black, which is lower than the 3.7 per standard deviation (SD) = 40.4 ± 10.7
cent in the Boston metropolitan area (12). years) than the included men (mean ± SD
Additional information about the popula- = 41.9 ± 9.3 years). This slight but statis-
tion and a description of early dropouts tically significant difference (p < 0.05) in-
from the study have been published (13, dicates a tendency for younger men to dis-
14). continue participation in the study. In ex-
Screening and initial examination oc- amining prospectively the development of
curred between 1961 and 1970. Subse- a clinical diagnosis of benign prostatic hy-
quently, men under age 52 have reported perplasia we further excluded the 101 men

Downloaded from http://aje.oxfordjournals.org/ at Pennsylvania State University on September 18, 2016

for examinations every five years; after age who had such a diagnosis at their first
52 they have reported every three years. examination. Not surprisingly, these 101
Each examination contains a uniform med- men were older (mean ± SD = 48.7 ± 11.0
ical history and physical examination, in years) than the included men. In examining
addition to blood and urine tests. In partic- the risk of surgery for benign prostatic hy-
ular, at each examination the examining perplasia the 12 men who had such surgery
physician inquires about symptoms of prior to entry into the Normative Aging
lower urinary tract obstruction including Study were excluded; however, men with a
hesitancy in initiating voiding, decreased prior clinical diagnosis were not excluded
urinary stream, and postvoiding dribbling. from analyses of the risk of surgery. Since
Such obstructive symptoms are distin- the exact date of onset of a clinical diag-
guished from irritative symptoms such as nosis could not be determined, onset was
dysuria, frequency or urgency (15). This assumed to occur at the midpoint of the
physician also performs a digital examina- interval between the examination of first
tion of the prostate. A clinical diagnosis of diagnosis and the prior examination. Par-
benign prostatic hyperplasia is based upon ticipants were followed from their initial
a physical finding of an enlarged or abnor- examination until the development of be-
mally firm prostate, or by a history of nign prostatic hyperplasia or until their last
symptoms if the symptoms can not be at- examination before May 15,1982. Age-spe-
tributed to another cause such as prostatitis cific incidence rates for both diagnosis and
or cancer of the prostate. surgery were estimated nonparametrically
Because most men over age 60 have some using the Cutler-Ederer life table approach
evidence of prostatic enlargement there is (16).
clearly a subjective component to this clin- Variables examined as possible predic-
ical diagnosis. The need for surgery is a tors of either a clinical diagnosis or surgery
more reliable indication of the presence of for benign prostatic hyperplasia were: age,
significant benign prostatic hyperplasia. At systolic and diastolic blood pressure, serum
each examination participants report cholesterol, uric acid, fasting and two-hour
whether they had surgery for benign pros- glucose, body mass index, religion, socio-
tatic hyperplasia, either by transurethral economic status, marital status, as well as
resection or by one of the open methods of coffee, cigarette and alcohol consumption.
surgical treatment. Surgery for benign At each physical examination blood pres-
prostatic hyperplasia is distinguished from sure is measured by an examining physician
prostatectomy for prostate cancer. using a standard mercury sphygmomanom-
Men eligible for the current study were eter and a 14-cm cuff. With the subject
the 2,049 Normative Aging Study partici- seated, systolic blood pressure and fifth-
pants who had at least two examinations. phase diastolic blood pressure are measured
Thus, the 231 men who had only one ex- to the nearest 2 mmHg in each arm. Values
 DEVELOPMENT OF BENIGN PROSTATIC HYPERPLASIA 81
used for the current study are the average = 41.9 ± 11.0 years) almost identical to the
systolic and diastolic levels for the two arms age distribution of the remaining 1,758 par-
at each examination. At each examination ticipants (mean ± SD = 41.9 ± 9.0 years).
subjects report to the study center at 8:00 There were no significant differences in the
A.M. after an overnight fast and abstinence rates of either clinical diagnosis or surgery
from smoking. Serum is collected by veni- for benign prostatic hyperplasia between
puncture and analyzed for cholesterol, uric the group of men with some missing data
acid and fasting glucose. Two-hour glucose and those with complete data.
level is measured following a 100 g glucose Associations of putative risk factors with
load. Body mass index is the ratio of weight both the diagnosis and surgery for benign
in kilograms to the square of height in prostatic hyperplasia were estimated as-

Downloaded from http://aje.oxfordjournals.org/ at Pennsylvania State University on September 18, 2016

meters. Details of measurement and the suming the proportional hazards model
relationship of these variables with cardio- stratified by age groups (21). In additional
vascular disease in this population have unstratified models, age was considered as
been described (17-19). a separate risk factor. Full models consid-
Cigarette smoking information and mar- ering all putative risk factors as well as
ital status have been assessed five times stepwise models were estimated. Variables
during the course of the study. The level of entered the stepwise model if the p value
these variables at a given physical exami- associated with their chi-square to enter
nation was taken to be the measure nearest was less than 0.10 and remained in the
in time to that examination. The Duncan model if the p value associated with their
Index of socioeconomic status (20) and the chi-square to remove was less than 0.15. By
usual number of cups of coffee and drinks considering each individual's most recent
of alcoholic beverages per day were assessed risk factor levels, risk factors were time-
by mailed questionnaire in 1973. An indi- varying. Only the most recent levels of pu-
vidual's levels of these variables were as- tative risk factors were considered in each
sumed constant throughout the study. proportional hazards model. Parameters in
the proportional hazards model were esti-
Upon admission into the study men were
mated by maximizing the partial likelihood
asked about their ethnicity and religious
as implemented by the BMDP statistical
background. All Jewish participants in the
package (22). For all multivariate models
study were of European ancestry.
the alcohol, cigarette and coffee consump-
Each individual's risk of a diagnosis or tion variables were transformed by log (1 +
of surgery for benign prostatic hyperplasia daily consumption) to normalize their dis-
was considered to depend on his most re- tributions.
cent assessment of these predictors. For
example, an individual's risk of having a
clinical diagnosis at his third Normative RESULTS
Aging Study examination was considered Age-specific incidence rates for both a
to depend upon his characteristics assessed clinical diagnosis and surgery for benign
at his second examination. In analyses of prostatic hyperplasia are shown in tables 1
possible risk factors other than age, 291 and 2. Each man was classified into an age
men were excluded because of missing data, group according to his age at his previous
which in almost all cases was due to non- physical examination. Since most Norma-
response to the socioeconomic status, al- tive Aging Study participants have had
cohol and coffee consumption questions. three or four examinations, an individual
Among the 2,049 Normative Aging Study might contribute person-years of experi-
participants with more than one examina- ence to several different age groups. The
tion, these 291 men with some incomplete incidence rates of both a clinical diagnosis
data had an age distribution (mean ± SD and surgery were markedly greater in sue-
82 GLYNN ET AL.

TABLE 1 O.OT

Incidence of benign prostatic hyperplasia. Clinical

diagnosis: 1,948 Normative Aging Study participants
with no clinical diagnosis of benign prostatic
hyperplasia at entry into the study, followed from
entry (between 1961 and 1970) until their last
examination prior to May 15, 1982
Age
(years)
at Man-years Cases Incidence
previous rate
examination
21-39 6,780.3 25 0.0037

Downloaded from http://aje.oxfordjournals.org/ at Pennsylvania State University on September 18, 2016

40-49 8,927.8 84 0.0094
50-59 4,977.1 156 0.0313
1.0 1.5 2.0 2.5 3.0 J.S H.O H.i 5.0 5.5
60-69 1,268.1 65 0.0513
70-87 219.6 13 0.0592 TERRS OF FOLLOH-UP
FIGURE 1. The cumulative probability of surgery
TABLE 2 for benign prostatic hyperplasia over an interval equal
Incidence of benign prostatic hyperplasia. Surgical to the time between examinations for men in the
treatment: 2,037 Normative Aging Study participants Normative Aging Study. Using the Cutler-Ederer life
with no surgical treatment for benign prostatic table method (16), estimates are based on 2,037 par-
hyperplasia prior to entry into the study, followed ticipants with no surgical treatment for benign proa-
from entry (between 1961 and 1970) until their last tatic hyperplasia prior to entry into the study, followed
examination prior to May 15, 1982 from entry into the study (between 1961 and 1970)
until their last examination prior to May 15, 1982.
Age
(years)
at Man-years Caws Incidence than 3.5 years. For men aged 70 or older
rate
previous there was a probability of 0.058 of having
examination surgery at some time during a 3.5-year in-
21-39 7,021.1 0 0 terval. The 3.5-year probabilities for men
40-49 9,767.1 2 0.00020 aged 60-69, 50-59 and 40-49 were 0.041,
50-59 6,135.7 25 0.00407
0.015 and 0.0005, respectively. Again, it is
60-69 1,904.1 23 0.01208
70-87 361.8 7 0.01935 clear that the risk was markedly greater
with each increasing decade of life. Using
cessively older age groups. The greatest these data, the probability that a 40-year-
increase between age groups in the inci- old man who lives to age 80 will have sur-
dence rate of a clinical diagnosis occurred gery for benign prostatic hyperplasia by age
at age 50. The per cent increase in diagnosis 80 is
among men older than 70 compared with
men 60-69 was 15 per cent. By contrast, 1 - (0.9995)2 (0.9796)2 (0.9591)3
the greatest increase in the incidence rate • (0.9425)3 = 0.292.
of surgery occurred at age 60 and the per
cent increase in men older than 70 com- A similar graph describing the probabil-
pared with men 60-69 was 60 per cent. ity of developing a clinical diagnosis of be-
The probability of having first-time sur- nign prostatic hyperplasia could not be
gery for benign prostatic hyperplasia, esti- drawn because the exact date of onset was
mated by the Cutler-Ederer method (16), is unknown. Assuming that the symptoms or
shown in figure 1. Estimates for men under physical finding developed at the midpoint
age 60 were based on 5.5 years of follow-up of the interval between examinations, the
because the great majority of men in this 3.5-year probability of a man initially aged
age group had intervals between examina- 70-87 having a first diagnosis of benign
tions of less than 5.5 years; for most men prostatic hyperplasia was 0.178. The 3.5-
over 60, examination intervals were less year probabilities for men aged 60-69, 50-
 DEVELOPMENT OF BENIGN PROSTATIC HYPERPLASIA 83
59 and 40-49 were 0.171, 0.110 and 0.042, nign prostatic hyperplasia are shown in
respectively. The cumulative probability table 5. In assessing the relationship of each
that a 40-year-old man who is initially free variable to surgery an individual's most
of benign prostatic hyperplasia and lives to recent previous measure of each variable
age 80 will develop symptoms or a physical was considered. Thus, for example, an in-
finding is 0.777. dividual's risk of having surgery for benign
Putative risk factors for benign prostatic prostatic hyperplasia between his second
hyperplasia other than age were first ex- and third examinations was considered to
amined categorically within age groups. depend on his age and other factors as
This was done because relationships might assessed at his second examination. Age
not be monotonic and might differ with was by far the most important predictor

Downloaded from http://aje.oxfordjournals.org/ at Pennsylvania State University on September 18, 2016

age. Incidence rates of surgery for benign of surgery. With age in the model, a prior
prostatic hyperplasia within categories of clinical diagnosis was an important inde-
putative risk factors and age are shown in pendent predictor of surgery. Controlling
table 3. Not shown are diastolic blood pres- for age and prior clinical diagnosis, three
sure and fasting glucose which gave results additional variables were significant inde-
similar to systolic blood pressure and two- pendent predictors and the signs of their
hour glucose, respectively. There were some estimated coefficients indicate the direc-
clear trends in incidence rates across cate- tion of their association with surgical treat-
gories. Men smoking one pack of cigarettes ment. Men of higher socioeconomic status
or more per day had markedly lower rates and cigarette smokers were less likely to
of surgery than both nonsmokers and for- have surgery whereas Jewish men were
mer smokers. Jewish men had higher rates more likely than non-Jews. Controlling for
of surgery than non-Jews, and men with a these variables, no other variable showed a
prior clinical diagnosis had much higher significant prospective association with
rates of surgery than men with no prior surgical treatment. In particular, married
diagnosis. Yet about half (26 out of 53) of men were no more likely to have surgery
the men having surgery had no prior clini- than unmarried men and none of the stan-
cal diagnosis at an examination within five dard risk factors for cardiovascular disease
years of their operation. None of the labo- and diabetes were positively associated
ratory or physical variables considered with surgery. For variables entering the
showed any clear association with surgery. stepwise model, estimated coefficients and
There were also no apparent associations standard errors from this model were quite
with marital status, coffee or alcoholic bev- similar to the estimates from the full model.
erage consumption. Estimated standard errors were slightly
In table 4 are shown incidence rates of a larger in the full model. To potentially con-
clinical diagnosis of benign prostatic hy- trol more effectively for the effects of age,
perplasia within categories of putative risk a proportional hazards model stratified
factors and age. There was a trend for men over four age groups (40-49, 50-59, 60-69
with lower cholesterol levels and lower body and 70-87 years) was considered. The es-
mass indices to have lower incidence rates timated coefficients and standard errors for
of a clinical diagnosis than men with higher variables other than age in this model were
levels of these variables. No other variables nearly identical to the coefficients in the
showed any marked associations with the unstratified model presented.
diagnosis. Not shown are diastolic blood The coefficients in Cox's proportional
pressure and fasting glucose which also hazards model can be interpreted in a man-
showed no association. ner analogous to the interpretation of coef-
Multivariate associations of putative risk ficients in a logistic regression model. For
factors other than age with surgery for be- instance, the estimated coefficient for age
84 GLYNN ET AL.

TABLE 3
Incidence of surgery for benign prostatic hyperplasia by age within specific categories of putative risk factors:
1,747 Normative Aging Study participants with no surgical treatment for benign prostatic hyperplasia prior to
entry into the study and with complete covariate information, followed from entry (between 1961 and 1970) until
their last examination prior to May 15, 1982

Age (years) at previous examination

4CM9 60-69 60-«9 70-87
No. No. No. No.
of IR* of IR of IR of IR
cases cases cases cases
Systolic blood pressure
<120 mmHg 2 0.5 6 2.9 7 13.3 1 10.4

Downloaded from http://aje.oxfordjournals.org/ at Pennsylvania State University on September 18, 2016

120-139 mmHg 0 0.0 13 5.0 10 12.8 4 33.6
£140 mmHg 0 0.0 4 4.6 4 10.6 2 25.2
Uric acid
<5.0 mg/100 ml 0 0.0 1 1.9 4 32.1 0 0.0
5.0-6.9 mg/100 ml 2 0.4 16 4.8 11 10.8 5 25.1
£7.0 mg/100 ml 0 0.0 6 3.6 6 11.2 2 38.4
Serum cholesterol
<200 mg/100 ml 1 0.3 12 7.5 5 10.2 1 10.5
200-249 mg/100 ml 0 0.0 8 3.5 8 12.2 3 24.4
£250 mg/100 ml 1 0.5 3 1.8 8 14.9 3 39.2
Two-hour glucose
<90 mg/100 ml 1 0.7 2 2.7 2 7.9 1 26.5
90-139 mg/100 ml 1 0.2 18 4.7 16 14.7 5 26.4
£140 mg/100 ml 0 0.0 3 3.1 3 8.8 1 14.9
Body mass index
<23 kg/m 2 0 0.0 2 2.7 6 16.2 0 0.0
23-27.9 kg/m 2 1 0.2 16 4.5 12 11.6 6 32.2
£ 2 8 kg/m 2 . 1 0.5 5 4.0 4 11.8 1 17.2
Alcoholic beverages
Nondrinker 0 0.0 0 6.0 2 11.1 3 45.7
1-729 drinks/year 2 0.4 21 5.4 14 12.5 4 22.6
£ 2 drinks/day 0 0.0 2 1.6 5 12.9 0 0.0
Coffee consumption
No coffee 0 0.0 2 3.4 2 8.7 0 0.0
< 3 cups/day 1 0.4 12 6.1 10 13.8 3 26.2
£ 3 cups/day 1 0.2 9 3.0 9 3.0 4 27.2
Cigarette smoking
Never 0 0.0 11 5.7 7 9.8 6 32.1
Former 2 0.6 9 3.7 12 16.4 1 11.1
< 1 pack/day 0 0.0 2 7.5 1 13.9 0 0.0
£ 1 pack/day 0 0.0 1 1.1 1 6.1 0 0.0
Marital status
Never married 0 O.O 0 0.0 2 36.5 1 90.2
Married 2 0.3 22 4.3 17 11.8 5 24.7
Other 0 0.0 1 3.7 2 10.8 1 12.4
Socioeconomic status
Duncan index <40 1 0.3 8 4.7 5 11.1 3 70.1
40 < Duncan <70 0 0.0 9 4.5 10 15.0 2 12.3
Duncan £ 7 0 1 0.4 6 3.2 6 10.5 2 22.5
Jewish religion
No 2 0.3 17 3.4 17 11.7 4 17.2
Yes 0 O.O 6 10.0 4 17.6 3 48.2
Clinical benign prostatic
hyperplasia
No 1 0.1 15 3.2 10 8.7 0 0.0
Yes 1 1.8 8 9.7 11 20.5 7 54.5
' Incidence rate x 103.
 DEVELOPMENT OF BENIGN PROSTATIC HYPERPLASIA 85
TABLE 4
Incidence of a clinical diagnosis of benign prostatic hyperplasia by age within specific categories of putative risk
factors: 1,702 Normative Aging Study participants with no clinical diagnosis of benign prostatic hyperplasia at
entry into the study and with complete covariate information, followed from entry (between 1961 and 1970) until
their last examination prior to May 15, 1982
Age (years) at previous examination
20-39 40-49 60-69 60-69 70-87
No. No, No. No No
of IR* of IR of IR of IR of IR
case* cases cases cases can s

Systolic blood pressure

<120 mmHg 9 3.9 28 8.2 53 30.8 19 55.9 2 56.7

Downloaded from http://aje.oxfordjournals.org/ at Pennsylvania State University on September 18, 2016

120-139 mmHg 12 3.9 36 9.4 68 32.3 32 59.5 5 62.6
£140 mmHg 1 2.8 8 13.9 22 32.0 6 23.9 2 29.0
Uric acid
<5.0 mg/100 ml 3 5.0 9 11.9 16 37.4 4 50.6 2 74.5
5.0-6.9 mg/100 ml 15 3.7 47 8.9 87 31.6 35 49.9 4 29.9
£7.0 mg/100 ml 4 3.8 16 8.9 40 29.9 18 51.7 3 128.3
Serum cholesterol
<200 mg/100 ml 13 4.8 28 9.9 42 31.5 18 50.2 5 117.7
200-249 mg/100 ml 7 3.3 29 9.1 60 32.1 27 67.7 2 21.5
£250 mg/100 ml 2 2.2 15 8.1 41 31.3 12 32.3 2 41.0
Two-hour glucose
<90 mg/100 ml 9 7.7 12 8.7 24 38.4 11 72.1 2 99.2
90-139 mg/100 ml 11 2.6 55 9.5 97 31.1 36 48.8 5 41.3
£140 mg/100 ml 2 5.6 5 7.1 22 28.5 10 42.1 2 46.6
Body mass index
<23kg/m' 3 3.5 9 9.4 24 41.5 11 56.2 3 146.8
23-27.9 kg/m 1 15 4.1 46 9.2 94 32.3 38 55.0 4 34.3
£28 kg/m2 4 3.3 17 9.0 25 24.4 8 33.2 2 42.6
Alcoholic beverages
Nondrinker 2 6.8 5 7.8 9 27.6 11 107.4 2 36.2
1-729 drinks/year 16 4.2 44 8.2 101 31.3 33 43.2 6 56.1
£2 drinks/day 4 2.5 23 12.3 33 34.2 13 49.8 1 45.8
Coffee consumption
No coffee 3 3.9 8 8.9 19 42.2 7 56.2 1 54.0
<3 cups/day 7 5.3 20 8.9 50 29.1 26 46.0 2 24.6
£3 cups/day 12 3.3 44 9.4 74 31.5 24 54.8 6 71.1
Cigarette smoking
Never 8 5.9 25 11.3 47 29.1 26 55.5 6 49.4
Former 7 4.0 33 10.6 63 31.9 23 44.1 3 58.9
<1 pack/day 1 1.6 1 1.8 8 35.9 1 20.8 0 0.0
£ l pack/day 6 3.0 13 6.6 25 35.7 7 77.4 0 0.0
Marital status
Never married 3 7.7 0 0.0 3 24.3 3 85.2 1 621.1
Married 18 3.5 71 9.6 135 32.4 47 48.0 5 38.0
Other 1 6.3 1 3.8 5 21.8 7 62.0 3 60.6
Socioeconomic status
Duncan index <40 8 4.2 24 8.1 40 28.1 16 47.4 2 59.7
40 s Duncan <70 6 2.9 24 9.3 56 34.1 23 48.9 3 32.4
Duncan £70 8 4.6 24 10.4 47 32.5 18 56.2 4 68.8
Jewish religion
No 22 4.0 67 9.1 122 30.2 47 47.6 8 53.3
Yes 0 0.0 5 10.7 21 44.3 10 71.5 1 29.3
* Incidence rate x 103.
86 GLYNN ET AL.

TABLE 5
Cox regression analysis of surgery for benign prostatic hyperplasia in relation to previous examination
assessments of putative risk factors: 1,747 Normative Aging Study participants with no surgical treatment for
benign prostatic hyperplasia prior to entry into the study and with complete covariate information, followed from
entry (between 1961 and 1970) until their last examination prior to May 15, 1982
St«pwise model Full model
Variable* Standard x'to Standard
Coefficient Coefficient
error enter error
Entering the stepwise model
Age 0.0880 0.0154 67.71 0.0888 0.0166
Prior diagnosis of benign prostatic 1.2582 0.3064 15.42 1.2644 0.3097
hyperplasia

Downloaded from http://aje.oxfordjournals.org/ at Pennsylvania State University on September 18, 2016

Socioeconomic status -0.0176 0.0066 4.50 -0.0169 0.0067
Jewish religion 0.7955 0.3438 5.13 0.7669 0.3625
Cigarette packs/day (log) -1.2153 0.6980 4.10 -1.2162 0.7039
Not entering the stepwise model
Diastolic blood pressuret 2.66
Uric acid 0.76 -0.1295 0.1281
Body mass index 0.39 0.0346 0.0490
Fasting glucose 0.30 0.0064 0.0114
Coffee cups/day (log) 0.15 0.0480 0.2577
Alcoholic drinks/day (log) 0.06 -0.0439 0.2962
Two-hour glucoset 0.05
Marital status 0.01 0.0638 0.4254
Systolic blood pressure 0.01 -0.0004 0.0088
Serum cholesterol 0.00 0.0001 0.0031
* Variables listed in order of entry into the stepwise model.
t Diastolic blood pressure and two-hour glucose were not considered in the full model because of high
correlations with systolic blood pressure and fasting glucose, respectively.

of 0.0880 means that, for two groups of men (95 per cent confidence interval = 0.90-
differing in age by 10 years but identical in 5.99).
all other important risk factors, older men The results of a Cox regression analysis
have an estimated incidence of surgery for considering predictors of a first clinical di-
benign prostatic hyperplasia approximately agnosis of benign prostatic hyperplasia are
2.41 times the incidence of men 10 years shown in table 6. The variables considered
younger. The factor 2.41 is exp(0.880 X 10). as predictors were identical to the variables
A 95 per cent confidence interval for this in the previous regression with the excep-
estimate is 1.78-3.20. Similarly, the ad- tion of a prior clinical diagnosis. Although
justed rate ratio for men with a prior clin- four variables entered the stepwise model,
ical diagnosis compared with men with no only two, age and body mass index, could
prior diagnosis is 3.52 (95 per cent confi- be considered statistically significant. Age
dence interval = 1.93-6.42). Other adjusted was by far the most important predictor.
rate ratios are: 0.70 for men 20 units higher Controlling for age, men with lower body
in socioeconomic status (on a 100 point mass indices were somewhat more likely to
scale) compared with men at any specified have a clinical diagnosis than heavier men.
level (95 per cent confidence interval = Adjusted rate ratios from this model are:
0.54-0.91); 2.22 for Jewish men compared 1.99 for men 10 years older than men of
with non-Jews (95 per cent confidence in- any specified age (95 per cent confidence
terval = 1.13-4.35); and 2.32 for men cur- interval = 1.78-2.22); 0.82 for men 5 kg/m2
rently not smoking compared with current heavier than men with any specified body
smokers of one pack of cigarettes per day mass index (95 per cent confidence interval
 DEVELOPMENT OF BENIGN PROSTATIC HYPERPLASIA 87
TABLE 6
Cox regression analysis of a clinical diagnosis of benign prostatic hyperplasia in relation to previous examination
assessments of putative risk factors: 1,702 Normative Aging Study participants with no clinical diagnosis of
benign prostatic hyperplasia at entry into the study and with complete covariate information, followed from
entry (between 1961 and 1970) until their last examination prior to May 15, 1982
Stepwise model Full model
Variable*
Coefficient Standard x1 to Coefficient Standard
error enter error
Entering the stepwise model
Age 0.0687 0.0057 145.05 0.0712 0.0063
Body mass index -0.0404 0.0202 4.47 -0.0410 0.0215
Serum cholesterol -0.0022 0.0013 2.77 -0.0023 0.0013

Downloaded from http://aje.oxfordjournals.org/ at Pennsylvania State University on September 18, 2016

Jewish religion 0.3011 0.1778 2.67 0.2689 0.1855
Not entering the stepwise model
Systolic blood pressure 2.19 -0.0058 0.0042
Uric acid 1.06 0.0688 0.0549
Two-hour glucoset 1.05
Fasting glucose 0.93 -0.0055 0.0053
Socioeconomic status 0.85 0.0022 0.0028
Diastolic blood pressuret 0.77
Cigarette packs/day (log) 0.44 -0.1253 0.1903
Married 0.35 0.1196 0.2045
Alcoholic drinks/day (log) 0.22 -0.0435 0.1143
Coffee cups/day (log) 0.00 0.0212 0.1010
* Variables listed in order of entry into the stepwise model.
t Diastolic blood pressure and two-hour glucose were not considered in the full model because of high
correlations with systolic blood pressure and fasting glucose, respectively.

= 0.67-1.00); 0.90 for men 50 mg/100 ml to the eighth decade. The sharpest increase
higher in serum cholesterol than men with in rate of first diagnosis occurred among
any specified level (95 per cent confidence men in their fifties, whereas the sharpest
interval = 0.79-1.02); and 1.35 for Jewish increase in rate of first surgery occurred
men compared with non-Jews (95 per cent among men in their sixties. The probability
confidence interval = 0.95-1.91). For vari- that a 40-year-old man who lives to age 80
ables entering the stepwise model, esti- will have surgery for benign prostatic hy-
mated coefficients and standard errors perplasia at some time during his life is
from this model were similar to the esti- estimated from this population to be about
mates from the full model; estimated stan- 0.29. This is almost three times the lifetime
dard errors in the full model were slightly estimate obtained by Lytton et al. (5) in
larger. When the proportional hazards their survey of prostate surgery in New
analysis was stratified by age, identical re- Haven between 1953 and 1961.
sults were obtained. No other risk factor could approach age
in the strength of its association with either
DISCUSSION a clinical diagnosis or surgery for benign
In spite of the increasing number of el- prostatic hyperplasia. A prior clinical di-
derly men in the population, and the cost agnosis was a strong predictor of surgery
to society of prostatic disorders, very little but there were many men with clinically
is known about the epidemiology of benign diagnosed benign prostatic hyperplasia at
prostatic hyperplasia. Among volunteers in entry into the study who did not have sur-
the Normative Aging Study, rates of both gery during the approximately 15 years of
first diagnosis and first surgery for benign follow-up, while other men had surgery
prostatic hyperplasia increased with age up with no prior clinical diagnosis. Of the 57
88 GLYNN ET AL.

men having surgery, 31 had a clinical di- creased risk of surgery for benign prostatic
agnosis at a prior examination. However, hyperplasia. The adjusted relative rate for
only nine of these 31 men had a clinical Jewish men compared with non-Jews of 2.2
diagnosis upon entry into the Normative found in the current study is quite consis-
Aging Study. Other than these nine, there tent with the relative risks of 2.4 reported
were 92 men with a clinical diagnosis at by Morrison (2) and 2.5 reported by Lytton
entry to the study who reported for at least et al. (5). The study by Armenian et al. (8)
one subsequent examination and did not reported identical percentages of Jewish
have surgery during follow-up. This sub- men in both their case and control series.
stantiates the variable clinical course of Lytton et al. noted the international differ-
benign prostatic hyperplasia, with some ences in incidence rates of benign prostatic

Downloaded from http://aje.oxfordjournals.org/ at Pennsylvania State University on September 18, 2016

men proceeding rapidly to a stage where hyperplasia and speculated that the higher
surgery is required and others able to tol- rates among Jewish men might be because
erate the clinically apparent condition for the majority of Jewish men in their study
intervals over 10 years without surgical were foreign-born. Of the 13 Jewish men in
treatment. However, the description of the the Normative Aging Study who had sur-
degree of variability in the progression of gery for benign prostatic hyperplasia, only
benign prostatic hyperplasia is limited by one was born outside the United States.
the reliability of the clinical diagnosis. Incidence rates of surgery among foreign-
The limited reliability and validity of the born participants in the Normative Aging
clinical diagnosis of benign prostatic hy- Study were roughly comparable to the rates
perplasia have been described elsewhere. among USA-born men, but could not be
Meyhoff and Hald (23) reported consider- accurately estimated because of small num-
able interobserver variation in estimates of bers (only four foreign-born men had sur-
preoperative prostatic weight. However, gery for benign prostatic hyperplasia).
they did find a significant correlation be- The adjusted relative risk for smokers of
tween estimated and actual prostatic one pack of cigarettes per day compared
weight. Castro et al. (24) and Andersen et with nonsmokers was 0.4. This is consistent
al. (25) found no correlation between the with the relative risk of 0.5 reported by
severity of symptoms and the degree of Morrison (2) and the finding by Greenwald
infravesical obstruction as assessed urody- et al. (7) of a lower proportion of smokers
namically. These studies did not consider among men having surgery for benign pros-
men in the general population. The corre- tatic hyperplasia than among controls.
lation between the size of the prostate as Rates of surgery for former smokers were
palpated on rectal examination and the de- similar to rates for never smokers (table 3),
gree of obstruction of the urinary tract may so these groups were combined in determin-
be slight. The present finding of a clinical ing overall rates (table 5). Morrison re-
diagnosis as a significant predictor of sub- ported that former smokers had rates in-
sequent surgery lends some support to the termediate between current and never
validity of the clinical diagnosis. The limi- smokers.
tations of this diagnosis are also clear in The finding in the current study of a
the Normative Aging Study data. There are higher incidence of surgery among men of
newer noninvasive techniques for assessing lower socioeconomic status was unantici-
postvoiding residual urine, and these pated. Richardson (27) described social
should lead to more reliable clinical diag- class differences in several measures of be-
noses (26). nign prostatic hyperplasia among Scottish
In addition to age and a clinical diagno- men in the early 1960s. Men of higher social
sis, Jewish men, nonsmokers and men of status were found to have higher rates of
lower socioeconomic status had an in- mortality and surgery due to benign pros-
 DEVELOPMENT OF BENIGN PROSTATIC HYPERPLASIA 89
tatic hyperplasia. However, a higher per- tionally, there were no indications that any
centage of men in the lower social classes of the major risk factors for cardiovascular
reported symptoms of nocturia, and among disease were predictive of benign prostatic
all men admitted to the hospital for benign hyperplasia. Cigarette smokers had lower
prostatic hyperplasia, men of lower social rates of surgery than nonsmokers, and
status were more likely to be in acute reten- heavier men as well as men with higher
tion. Differences in rates of prostate sur- cholesterol levels had lower rates of clinical
gery between men of different social status diagnosis. The directions of these relation-
may be due to differences in health insur- ships are opposite to what would be ex-
ance plans or medical care utilization pat- pected under the hypothesis of an associa-
terns. We are not encouraged to conclude tion between cardiovascular disease and be-

Downloaded from http://aje.oxfordjournals.org/ at Pennsylvania State University on September 18, 2016

that the differences between men of differ- nign prostatic hyperplasia. The large num-
ent social status in rates of surgery are due ber of man-years of follow-up gives
to lifestyle differences which have a direct strength to these findings of no association.
biologic effect. Greenwald et al. (7) found equal numbers
A clinical assessment of benign prostatic of diabetics in their case and control series
hyperplasia remains a part of most stan- as well as equal numbers of men using
dard physical examinations and so it is of antihypertensive medications. Morrison
interest to compare the clinical diagnosis (2) reported that cases were more likely to
with surgery for benign prostatic hyperpla- be using rauwolfia (relative risk for rau-
sia. Risk factors for palpable enlargement wolfia use = 2.2). He found no strong as-
may not be the same as those for the hy- sociation with other antihypertensive med-
pertrophied median lobe. Older age and ications and concluded that the risk with
Jewish religion were predictive and had rauwolfia use is not secondary to a relation-
similar effect estimates for both a clinical ship of benign prostatic hyperplasia to hy-
diagnosis and surgery among men in the pertension.
Normative Aging Study. Men with lower There is currently considerable contro-
body mass indices and lower cholesterol versy about the possible relationship of be-
levels were more likely to have a clinical nign prostatic hyperplasia to prostatic can-
diagnosis but were not more likely to have cer (7, 8). Data from the Normative Aging
surgery. This difference may be due to a Study are too sparse to address this contro-
detection bias since a clinical diagnosis may versy directly. As of May 15, 1982, 10 of
be more difficult to make in a heavier man. the volunteers in the Normative Aging
In two case-control studies of men Study had developed prostatic cancer.
undergoing surgery for benign prostatic hy- None of these men had prior surgery, but
perplasia, Bourke and Griffin (10, 11) six of the 10 had a prior clinical diagnosis
found strong positive relationships between of benign prostatic hyperplasia. A compar-
hypertension, diabetes mellitus and benign ison of risk factors for prostatic cancer and
prostatic hyperplasia. They concluded that for benign prostatic hyperplasia shows
an age-related increase in estrogens was a some similarities and some differences. Age
common link in all three diseases. However, is the most important risk factor for both
these studies assessed blood pressure and diseases and for both the incidence in-
glucose levels at the time of surgery and creases monotonically with age (28). In
thus did not control for the stress of the contrast to their higher rates of benign
operation. The current study found no dif- prostatic hyperplasia, Jewish men in both
ferences in the distributions of both fasting Israel and New York are reported to have
and two-hour glucose between men devel- lower rates of prostatic cancer (28). We
oping benign prostatic hyperplasia and know of no evidence that cigarette smoking
those remaining free of the disorder. Addi- is protective against prostatic cancer.
90 GLYNN ET AL.

REFERENCES of scientific objectives to population selection and

attrition in longitudinal studies: the case of
1. National Center for Health Services Research.
the Normative Aging Study. Gerontologist
Health, United States 1981. Hyattsville, MD: Na- 1976;16:508-16.
tional Center for Health Statistics, 1981. (Data
15. Walsh PC. Benign prostatic hyperplasia. In: Har-
from the National Hospital Discharge Survey.) rison JH, Gittes RF, Perlmutter AD, et al, eds.
DHHS publication no. (PHS)82-1232. Campbell's urology. 4th ed. Philadelphia: WB
2. Morrison AS. Prostatic hypertrophy in Greater Saunders, 1978.
Boston. J Chronic Dis 1978;31:357-62. 16. Cutler SJ, Ederer F. Maximum utilization of the
3. Rotkin ID. Epidemiology of benign prostatic hy- life table method in analyzing survival. J Chronic
pertrophy: review and speculations. In: Grayhack Dis 1968;8:699-713.
JT, Wilson JD, Scherbenske MJ, eds. Benign 17. Sparrow D, Garvey AJ, Rosner B, et aL Factors
prostatic hyperplasia. Washington, DC: US De- in predicting blood pressure change. Circulation
partment of Health, Education and Welfare, 1975. 1982;65:789-94.
DHEW publication no. (NIH)76-1113. 18. Glynn RJ, Rosner B, Silbert JE. Changes in cho-

Downloaded from http://aje.oxfordjournals.org/ at Pennsylvania State University on September 18, 2016

4. Moore RA. Benign hypertrophy and carcinoma of lesterol and triglyceride as predictors of ischemic
the prostate. Surgery 1944;16:152-67. heart disease in men. Circulation 1982;66:724-31.
5. Lytton B, Emery JM, Harvard BM. The incidence 19. Glynn RJ, Campion EW, Silbert JE. Trends in
of benign prostatic obstruction. J Urol serum uric acid levels: 1961-1980. Arthritis
1968;99:639-45. Rheum 1983;26:87-93.
6. Derbes V de P, Leche SM, Hooker CW. The 20. Riess A, Duncan O, Hatt P, et al. Occupations
incidence of benign prostatic hypertrophy among and their social status. New York: Free Press,
the whites and Negroes in New Orleans. J Urol 1961.
1937;38:383-8. 21. Cox DR. Regression models and life tables (with
7. Greenwald P, Kinnss V, Polan AK, et al. Cancer discussion). J Roy Statist Soc Series B
of the prostate among men with benign prostatic 1972;34:187-220.
hyperplasia. JNCI 1974;53:335-4O. 22. Hopkins A. Regression with incomplete survival
8. Armenian HK, Lilienfeld AM, Diamond EL, et al. data. In: Dixon WJ, Brown MB, Engelman L, et
Epidemiologic characteristics of patients with al, eds. BMDP statistical software 1981. Berkeley:
prostatic neoplasms. Am J Epidemiol 1975;102:47- University of California, 1981.
54. 23. Meyhoff HH, Hald T. Are doctors able to assess
9. Van der Ries L. Benign prostatic hypertrophy and prostatic size? Scand J Urol Nephrol 1978;12:219-
coronary heart disease: correlation or coinci- 21.
dence? J Am Geriatric Soc 1959;7:866-9. 24. Castro JE, Griffiths HJL, Schackman R. Signifi-
10. Bourke JB, Griffin JP. Diabetes mellitus in pa- cance of signs and symptoms in benign prostatic
hypertrophy. Br Med J 1969;2:598-601.
tients with benign prostatic hyperplasia. Br Med
25. Andersen JT, Nordling J, Walter S. Prostatism:
J 1968;4:492-3.
1. The correlation between symptoms, cystometric
11. Bourke JB, Griffin JP. Hypertension, diabetes and urodynamic findings. Scan J Urol Nephrol
mellitus, and blood groups in benign prostatic 1979;13:229-36.
hypertrophy. Br J Urol 1966;38:18-23. 26. Szabo V, Sobel M, Pajor L, et al. Ultrasonic mea-
12. Bureau of the Census. 1970 Census of Population. surement of residua] urine: a method requiring no
Vol. 1. Characteristics of the population. Wash- catheter. Int Urol Nephrol 1981;13:345-51.
ington, DC: US GPO, 1973. 27. Richardson IM. Prostatic hyperplasia and social
13. Bell B, Rose CL, Damon A. The Normative Aging class. Br J Prev Soc Med 1964;18:157-62.
Study: an interdisciplinary and longitudinal study 28. Greenwald P. Prostate. In: Schottenfeld D, Frau-
of health and aging. Aging Hum Dev 1972;3:5-17. meni JF, eds. Cancer epidemiology and preven-
14. Rose CL, Bosse R, Szretter W. The relationship tion. Philadelphia: WB Saunders, 1982.