Symposium Early Detection of Prostate Cancer PETER T. SCARDINO, MD, ROBERT WEAVER, AND M’LISS A. HUDSON, MD Prostate cancer is unique among the potentially lethal human malignancies in the wide discrepancy histologic changes recognizable
between the high prevalence
The concept that adenocarcinoma of the prostate exists in a latent and a clinical form is supported by epidemiologic, pathologic, and clinicall evidence. Although these divergent manifestations of prostate cancel have common architectural and cytologic features, the), can be distinguished from each other to some degree by differences in certain pathologic features, such as the volume, grade, and invasiveness of the lesion. To appl) the available diagnostic tests rationallv wr need to know the nature of cancers detected with’these tests to determine whether these cancers are “clinically important” or capable of causing morbid or lethal complications if left untreated in men with a sufficiently long life expectancy. ‘We will review the epidemiologic, pathologic, and clinical features of latent and clinical prostate cancel and describe the outcome of treatment trials and the available information about the natural history of the disease to show that prostate canc‘er. when detected clinically, is most often a morbid and potentially lethal tumor. We will then describe the features of cancers detected with each of the available diagnostic tests. estimate the proportion of those cancers rhat are “cliniand review the preliminary experience cally important,” with the use of these tests in early detection trials. Thus, our analysis will attempt to provide a rational basis for the early delection that is essential if the morrality rate from prostate cancer is to decline.
of
as cancer and the much lower prev-
alence of the clinical disease. Despite the availability
of effective
tests for early detection and of effective treatment for cancers so detected, the diagnosis is locally advanced
usually is not established
or metastatic. Yet. physicians
until the tumor hesitate to use
these tests for fear that many cancers found would
be latent, of
little threat to the life or health of the host, and treatment could introduce inappropriate
morbidity.
Latent or “clinically unimpor-
tant” cancers can be distinguished
from those that are clinically
important
by the larger volume, higher
grade,
and greater inva-
siveness of the latter. The available tests can detect only those cancers large enough to be palpable. visible on ultrasound, or capable of elevating: the serum level of prostate-specific
antigen. Such can-
cers are clinically important and should be treated for cure if the
life expectancy
of the patient
is sufficiently
long and the morbidity
rate of therapy is low. Early detection of prostate cancer using the tests that are available today may widen the window of opportunity so that treatment indeed becomes possible necessary. HUM PATHOL
23:211-222.
in those for whom it is
Copyright
MD,
G 1992 by W.B.
Saunders Company I’I-ost:ite cancer is unique among the potentially leI hat human rnali~mancies in the wide discrepancy becween th’r high prevalence of histologic changes recognizable as cancer and the much lower prevalence of ( linical disease. Despite the availability of diagnostic tests to detect prostafe cancer (digital rectal examination /IIRE]. prostate-specific antigen [PSA], and transrectal ~tltrasonogral)~~y [TRUS]) and of effective treatment for c’ancers detected early (radical prostatectomy or radiotherapy), the diagnosis usually is not established until rhe turrlor is locally advanced or metastatic.. Prostate c’ancer will kill over 30,000 American men this year, and the age-specific morrality rate is increasing. Yet, physitians hesitate to apply these diagnostic tests in a vigorous program of early detection because of their concern that man\ cancers found would be “latent,” ie, of no real threat to the life or well-being of the host. Treatment of such tumors could result in inappropriate morbidity, md the ‘healrh of the community as a whole could be reduced by suc11 efforts.
CLINICAL
AND LATENT
PROSTATE
CANCER
Clinical Cancer Epidemiology
Prostate cancer has become the most common GLIBcer among American men, and only lung cancer is responsible for more cancer deaths.’ The ape-spec.ific mortality rate has slowly increased ol;tar thcb past 50 years,’ and in black American men is ne,lrly double the rate found in white men.’ Prostate cancer is responsible for nearly 35% of all deaths in men over the age of 55 years.” Since the incidence of prostate I.‘Lmcer increases more rapidly with age than any other cancel, and the average age of American men is rising, the number of patients with prostate cancer is expected to increase dramatically over the next decade.” Estimates are that in 199 1 o&r 122,000 men will be diagnosed with prostate cancer and 32,000 will die of the disease in the United States.’ The mortality rate may underestimate the morbidity, since endocrine therapy allows 10% of men with metastatic prostate cancer to live for 10 years 211
HUMAN PATHOLOGY TABLE 1.
Percentage Patients
Estimated Percentage of Patients in Each Clinical Stage at Diagnosis, Risk of Death From Prostate Cancer, and Approximate “Cure“ Rate (Lifelong Freedom From Prostate Cancer) Associated With Each Stage
of Stage*
JO
M+
“0
N+
10
T~LMI,
30 10
T,,,.,N,,W, T,,A,M,,
100
Volume 23, No. 3 (March 1992)
Prognosis Incurable Rarely turahle Occasionally curable Often curable ‘l‘reatment “unnecessary”
All Exe-ludina ‘I‘,, (stage A I)
* Based on clinical stage. pelvic lymph node dissection.
bone scan, and acid phosphatase.’
and 50% to live for 3 years.“.” These patients require frequent therapeutic intervention for local and systemic complications of the disease.’ Oufcome of Treatment
Trials
Approximately 30% of men with prostate cancer have distant metastases at the time of diagnosis.’ Despite the impressive symptomatic response of metastases to hormonal manipulation (androgen deprivation), the survival rate for these patients is dismal: the median duration of survival is less than 3 years.“.“,s,” By 5 years, over 75% and by 10 years, more than 90% of these patients die of their cancer rather than with it.“.” Approximately 30% of patients with apparently localized prostate cancer (To_d NxM,,) have pelvic lymph node metastases.“.‘” Virtually all patients found to have positive pelvic lymph nodes treated with local or regional therapy show clinical evidence of progression of tumor within 10 years”~“~‘-5 unless additionally treated with systemic therapy (androgen deprivation), which has not been shown to alter the long-term survival rate (reviewed by Austenfield and Davis’“). Prostate-specific antigen (PSA) levels are elevated within the first 2 years after radical prostatectomy and pelvic lymphadenectomy in more than 85% of patients with positive nodes’“-‘s and in 100% of these patients followed for longer periods.lg The cancer-specific mortality rate for these patients, whether hormonal therapy is administered early or late, is approximately 25% at 5 years and 60% at 10 years. Even patients with one single microscopic focus of tumor in one lymph node seem to fare as poorly in the long run as patients with more extensive nodal metastases, although long-term data are not available after some forms of therapy.” Once prostate cancer has spread to the regional nodes, it can rarely be controlled with any form of local or regional therapy. At least 70% of such patients will eventually die of their cancer rather than with it.” Approximately 15% of patients have a locally extensive tumor at the time of diagnosis (stage C or T:s..Js.’ ’ and half of these have pelvic lymph node metastases.‘z.‘3 The 5-year relative survival rate for patients with clinical sta e C cancer (T3_,‘NS) was 66% in a large national survey 8 and the risk of death from prostate cancer is approximately 65% to 75% at 15 years.“-” In 212
1O-Year CancerSpecific Survival Rate (%)
Estimated “Cure” Rate (%)
10 40 60 NJ 05
50 years old with cancer present (prevalence of autopsy cancer) Estimated proportion of autopsy cancer cases that are “clinically important” (0.2 x 30%) Actual t-ate of detection with DRE, PSA, +TRUS Annual incidence rate in men 250 years old Annual mortality rate in men >50 years old
30
Clinically Important 20%
6
linical Ca 100,000
l-3 0.31 0.09
Note. The estimated prevalence varres more than 3004old depending on the presumed prevalence of clinically important cancer. Abbreviations: DRE, digital rectal examination; PSA, prostatespecific antigen; TRUS, transrectal ultrasound.
malignancy in human beings.“* The age-specific prevalence of prostate cancer in cystoprostatectom mens confirms the findings in autopsy studies. Uqt; the Stanford series, 25 of 66 prostates (38%) contained cancer.“” Yet, the estimated number of men diagnosed with clinical prostate cancers in 1985 was only approximately 86,000, which represents only 1.05% (86,000 of 8,193,OOO) of the men estimated to have cancer present in the prostate (Fig 1).5? There were an estimated 25,500 deaths from prostate cancer in 1985, which is only 0.3 1% of these men. Thus, only one of 95 men with cancer was diagnosed with the disease, and only one in 323 died of the disease that year (Table 2). This enormous discrepancy between the high prevalence of the disease at autopsy and the low incidence of the disease clinically has provided a theoretical basis for a policy of “no treatment” and has stymied efforts at early detection and screening,“8,5”,“4,5” This apparent discrepancy, however, has been exaggerated by the failure to consider the effect of time. Although the clinical incidence of prostate cancer is low (approximately 0.31% per year for men more than 50 years old), the lifetime risk of developing prostate cancer for a man 50 years old in 1985 was 9.51%3 (Table 3). Similarly, the annual mortality rate for men more than 50 years old in 1985 was only 0.09%, but the lifetime risk of dying of prostate cancer was 2.89%. In 1985 a 50-year-old man was expected to live to the age of 75. Autopsy data confirm that prostate cancer is not found in every 75-year-old man but in only approximately 42%.5’ For a 50-year-old man, therefore, we can estimate that the lifetime risk of developing cancer in the prostate is approximately 42%, the risk of developing the disease clinically is approximately 9.5%, and the risk of dying from the disease is approximately 2.9%. Over their lifetimes, then, one of every 4.4 men who harbors cancer in his prostate will develop the disease clinically and one 214
FIGURE 1. Distribution of prostate cancers If the prevalence of adenocarcinoma among men more than 50 years of age is 30%‘ then approximately 8 million American men harbor foci of cancer in their prostate. Yet, in 1990 only approximately 100,000 men were diagnosed with the disease (“clinical cancer”) and the other 7.9 million were considered to have incidental or “autopsy” cancer. Based on the pathologic features of prostate cancer found at autopsy46 and on the lifetime risk of developing prostate cancer.52 we estimate that approximately 20% of these cancers are “clinically important” (that is, they will threaten the life or well-being of the host if not treated effectively), and 80% are truly latent cancer (“clinically unimportant”).
of every 14.5 will die of prostate cancer (Table 3).‘” Although the ratio of prostate cancer deaths to the development of histologic evidence of the disease is much greater (one in 14.5 rather than one in 323) when viewed in this perspective, it is apparent that most histologic (“autopsy”) cancers do not progress within the normal life expectancy of the host. Yet, these cancers found at autopsy have all of the cytologic and architectural features of the disease detected clinically.““-“‘.“‘,“” Pathologic Features Detailed pathologic examination of cancers found at autopsy and of clinically detected cancers do, however, show important differences between autopsy-detected and clinically detected cancers in the volume, grade, and invasiveness of the tumor.~6,48,4’1..51,.57 Most cancers found at autopsy are small, well differentiated, and show no tendency to invade normal structures ( rostatic capsule, seminal vesicles) (reviewed by Dhom” B). Others are larger, less well differentiated, and do invade normal structures. The prevalence and proportion of these larger, proliferative cancers are greater in older men,
TABLE 3.
The Lifetime Risk of Developing or of Dying of Prostate Cancer for a 50-Year-Old White Man in 1985 Lifetime Risk (Yo)
Lifetime Lifetime Lifetime
risk of developing “autopsy” cancer5p: risk of developing clinical cancer’: risk of dying of prostate came?:
Ratio
42 9.51 2.80
Note. The ratios indicate that for every death from prostatecancer, 3.3 men will he diagnosed with the clinical disease and 14.5will develop histologic evidence of cancel- in the prostate dur-ing their lifetime.
EARLY DETECTION
TABLE 4.
Geographic
in the Incidence
Variations
and Mortality
OF PROSTATE
(Scardino et al)
TABLE 5. Features That Help Distinguish Ciinicalty Important From Unimportant Prostate Cancers
(8.4- to 10.8-Fold) Rates
CANCER
of Clinical
--__
Prostate Cancer (Per 100,000 Men&r) (Iinicall~ Important
Clinically C:nimportant :small Well-diflerentiated ((;lea\on gwlr I or2) Soninwsiv~~ Dipicki lormal ‘1 rmsitirm
~onr
cer in men over the age of 50 years would be approximately 20% of the 30% of men more than 50 years of age with prostate cancer, or 6% rather tlhan 30% (prevalence at autopsy) or 0.31% (annual incidence) (Table 2). These are the clinically important (IFig 2) but currently undetected cancers that would be the appropriate targets of an early detection program and could represent as many as 1.6 million men in the United States alone, a staggering figure unless one considers that 2.0% of 50-vear-old men alive today (or approximatelv 790,OOiI in this country alone) will eventually die oi prostate cancer.” To prevent these deaths we may have to detect and treat two to three patients for every death prevented. The proper goal of an early detection program is not to detect all prostate cancers, but only IO detect those that are potentially morbid or lethal (the clinically important cancers), which we estimate have a prevalence of approximately 6% (20% of cancers prevalent in the prostate in men more than 50 veal-s old., or all the cancers, except To,, that will even&allv he detected as clin-
ill
black Americans. and in geographic regions with a higher clinical incidence of and mortality rate from prostate ‘cancer, while the smaller, nonproliferative c.mcers are more uniform with age, ethnic group, and geographic region (Table 4)” (reviewed by Dhom’s and Scardino”‘). These observations strongly support the c’c)ncept of a multistep process in the pathogenesis of prostate cancer in which latent cancers have progressed through s~ome but not all of the steps necessary for full malignant expression.“’ Since some cancers found at autopsy have the features of clinical cancers”.4H.‘,1.6” and some cancers detrcted clinically have the features of autopsy or latent cancers (for example, stage A, or To,),““-“” we have found il more helpful to describe prostate cancers as “clinically important” (ie, threatening the life or well being of the host within his remaining life expectancy) or “clinically unimportant” (ie, a latent cancer of no threat to the host).“’ At the light microscope level, tumor volume, grade, am1 invasiveness are some of the features that help to distinguish clinically important from unimportant cancers (Table 5). Llnfortunately, no objective nlarkers of progression in human prostate can&r have been established,7” hut additional features that seem to he strongly associated with the biologic behavior of prostate cancer are the DNA ploidy status of the tumor, the degre’r of PSA production (serum PSA level), and, possibly, the Lone of origin of the turnor within the prostate (‘Table 5). The proportion of prostate cancers present in the population of men over 50 years old that are clinically importanl is difficult to determine precisely, but information available from detailed morphometnc studies of prostate cancers found at autopsy and in cystoprostatectomy series suggests that approximately 20% of these currently undetected cancers are clinically important based on the features listed in Table 5 (I;ig l).’ 1,;‘-IH4!1.5I..-,:!.r,7.l>ll
DIRT
IBUTI~N
I I
OF
PROSTATE
N=8,000,000
LANCERS I ;
i AutoDsY ca
FIGURE 2. Of the clinically detected cancers nearly half are advanced (Table 1). Approximately 40% are early stage and may be curable, and 10% are focal, well-differentiated stage A, (T,,) and may be “clinically unimportant.” Of the large pool of undetected but clinically important cancers, similar proportions are estimated to apply.
Goal of a,? Early Detection Program If OUI‘ estimates are correct, the prevalence of clinic-ally impl:)rtant but presently undetected prostate can215
HUMAN PATHOLOGY
Volume 23, No. 3 (March 1992)
TABLE6. Patients With Moderate or Severe Obstructive Voiding Symptoms at the Time of Diagnosis of Prostate Cancer (Stages B and C, or T,_J Were More Likely to Have Pelvic Lymph Node Metastases (P < .OOOl) Results of Pelvic Lymph Dissection Degree of Obstructive Voiding Symptoms None Mild Moderate Severe Total
N (%)* 171 55 93 50
(46) (15) (25) (14)
369 (100)
Negative
Positive (%)t
135 35 58 ?I ?49
* Percentage of patients with each degree of voiding t Percentage of patients with lymph node metastases. Data from Meacham et aLa
36 20 35 29
(21) (36) (38) (58)
6
c
8
120 (33)
c
20
i P
symptoms.
I
p=.OOl l
--* Negllglble Sx (226)
M
’
’
’
’
’
’
’
’
5
0
Time
TECHNIQUES FOR EARLY DETECTION AND CHARACTERISTICS OF CANCERS DETECTED
Substantlal Sx (n=143)
t
01
ical cancers) (Fig 2) (Table 2). We will review below the pathologic features of cancers detected with the available noninvasive diagnostic tests (symptoms, DRE, PSA, ultrasound, and combinations of these) and estimate the proportion that are clinically important.
I
’
’
10
(Years)
FIGURE 3. Freedom from distant metastasis after treatment (radiotherapy) for clinically localized (stages B and C, or T,. ,N,M,) prostate cancer based on the degree of obstructive voiding symptoms at the time of diagnosis. Patients with substantial (moderate or severe) obstructive symptoms had a significantly worse prognosis. (Reproduced with permissionBO: Meacham RB, Scardino PT. Hoffman GS, et al: The risk of distant metastases after TURP vs needle biopsy in patients with localized prostate cancer. J Urol 142:320-325, 1989, @ by American Urological Association, Inc.)
Symptoms
Prostate cancer is a notoriously silent disease with few early symptoms. Symptoms associated with bladder outlet obstruction are commonly present in men over the age of 50 years and are often ascribed to benign prostatic hyperplasia (BPH). Indeed, BPH is more common than prostate cancer: the lifetime risk of a 50-yearold man requiring a prostatectomy for symptomatic BPH is 20% to 25%.* However, symptoms of bladder outlet obstruction also develop in men with prostate cancer. In a series of 300 consecutive men hospitalized for obstructive voiding symptoms, Guinan et al diagnosed prostate cancer in 23%?” In our own series of 5 10 men treated with radioactive gold seed implantation and external beam irradiation for TOb-TJ cancer (stages AB, B, and C), 234 (46%) had a transurethral resection of the pstate for symptoms of bladder outlet obstruction. ’ Of the patients with a palpable (T,.,) cancer, 143 (39%) had moderate or severe obstructive voiding symptoms at the time of diagnosis. These patients had a greater chance of having lymph node metastases (Table 6) and developing distant metastases (Fig 3) than comparably staged patients without obstructive voiding symptoms, confirming the clinical observation that patients with prostate cancer who have obstructive voiding symptoms at presentation are more likely to have locally advanced disease and a poor prognosis. s1 While there are no objective data describing the detailed pathologic characteristics of prostate cancers that are detected because of the development of urinary symptoms, such symptoms are likely to be of little use in the early detection of prostate cancer. 216
Digital
Rectal
Examination
The DRE traditionally has been considered the most accurate test for the detection of prostate cancer. Digital rectal examination has been demonstrated to be more sensitive, more specific, and to have a greater efficiency than a variety of laboratory tests available in the 1970s7” However, few of these laboratory tests are still in clinical use today. In one study patients were hospitalized for urinary symptoms and were not asymptomatic subjects of a screening program; prostate cancer was detected in 23%, considerably more than expected in a screening trial. Digital rectal examination has been used to detect prostate cancer in several large series, and the detection rates range from 0.8% to 1.4% (Table 7).““m56.8” Although 67% to 88% of these tumors are clinically localized,
TABLE7.
Results of Early Detection Programs Using Each Diagnostic Test17.53.56.68.76.82-86 DRE (%)
Positive predictive Detection Clinically
value
rate localized
Pathologically
confined
TRW
(%)
22-36
15-41
0.8-1.4 53-88
2.3-3.1 58-100
37-77
56-68
PSA (%) “‘L-35 (>4)* 65-67 (>lO) 2.2 96 (4- 10) 77 (>lO) 56-72 (4-10) 8-33 (10)
Abbreviations: DRE, digital rectal examination; TRUS, tramrectal ultrasound; PSA, prostate-specific antigen. * Numbers in parentheses refer to the level of PSA.
EARLY DETECTION
OF PROSTATE
TABLE 8. Characteristics of Cancers Detected With Each Diagnostic Test in Patients With Clinically Localized Prostate Cancers Treated With Radical Prostatectomy, and for Latent Cancers Detected Incidentally at Autopsy or in Cystoprostatectomy Specimens __.__
CANCER
(Scardino et al)
palpable induration correlates well with the total volume of tumor”” nor with the area of cancer on the posterior surface of the prostate.!“’ Asymmetry and induration, as well as a discrete hard nodule, are included in the tindings on I)RE that may be judged suspicious for prostate CYIIK~I‘. but these findings lack specificity for cancer. The positive predictive value of a palpable abnormalit\; is on,lk 22%, to 36Yo (Table 7). Yet, the most serious 1in;itation of DRE is its lack of sensitivity (false-negative results). For example, approximately 10% to 20% of transurethral resections performed for benign prostatic hypertroph~ in patients with no palpable abnormality suggestive of cancer uncover an incidental cancer of the prostate. Digital rectal examination detected only 12 of 22 cancers found in ;I screening study, while TRUS found 20.““ Goner et al detected cancer by TRUS in 12.4% of 22.5 men, none of whom was ‘:juclged by the urologist tl:) have a suspiciously palpable lesion.““g Goner et at later reported that 25% of the cancers found in men examined h> TRUS were not palpable (Table 9).“” Thus, DRE is relatively insensitive and nonspecific. Cancers detected by palpation are relatively large, man) are late in their- development and no longer curable. and some are so small they would fit OUI definition of clinically unimportant cancers.
l~;~tl~ologi;lc~ staging revealed that many were, in fact, not confined to the prostate gland.9”,54 In our own series of 9X clinical stage B (T,.,) radical prostatectomy specimens, 63% extended through the capsule and 18% into the seminal vesicles.x7 These cancers ranged in size from 0.45 to 19.2 cc (mean, 4.04 cc), although we occasionally have encountered palpable cancers as small as 0.04 cc.88 Stanley et al reported that palpable clinical stage B cancers in patients with negative lymph nodes ranged from 0.01 to 25 cc (mean, 5.4 cc).‘” Nonpalpable cancers discovered in cystoprostatectomy specimens by these same investigators ranged from 0.00 1 to 1.1 C‘C(mean, 0.1 cc) and all were confined to the prostate.‘” Thus, palpable cancers clinically confined to the prostate are one to two orders of magnitude larger, on average, than latent cancers (Table 8) and most already extend outside the l,rostate_:‘“.;;.“7.~~~
Prostate-Specific
Antigen
The on]!. study using PSA as the primarv test to screen for prostate cancer was conducted by Catalona et al, who determined serum levels in 4,293 self-referred patients,‘t’,x’l Overall, serum levels higher than 4.0 ng/ mL (Hybritech monoclonal I-adioimrnunoassa~; Hybritech, San Diego, CA) were detected in .516 patients (12%). These patients were asked to ret.urn for a DRE and TRUS to determine the need for a Ileedle biopsy. The overall cancer detection rate in the published series was 2.2% (Table i).” Cancer was detected in 26% of the men with a PSA level of -1.0 to 10.0 ng/ml.. Nearly all of these tumors (96%) were clinically localir.ed, and the tumor was confined to the prostate in 72% of these patients who had a radical prostatectomv. Callcer was detected in 66% of patients with a serum level higher than 10.0 ng/mL. While 77% of these cancers were clinically localized, only 33% of the patients who had a rad-
Not only does DRE detect cancer relatively late, there is only a weak correlation between the size of the cancer estimated by DRE and the actual volume of cancer present. Tumors palpable as small nodules (1.5 cm or half of one lobe) are generally smaller than larger palpable tumors, but the overlap is substantial.“’ Neither the absolute nor the relative dimensions of an area of
TABLE 9. The Detection of Prostate Cancer in Studies Using a Combination of Digital Rectal Examination. Prostate-Specific Antigen, and Transrectal Ultrasound With Transrectal Ultrasound-Guided Eliopsy of Any Suspicious Hypoechoic Lesion: Influence of Abnormal Digital Rectal Examination or Prostate-Specific Antigen (>4.0 ng/mL) on Results _______~ ____
Abnormal ~rRtIS N (%) DRE+. PSAt DRE+. PSADRE-. PSAt DRE-, PSA-~ All patients __--
438 (I 7) 446 (17) 487 (IX)
Biopsy Performed
PPV TRUS
0!I ti7
3i 15
19
“7
Puwnragc With Cancc~
No. of Patknts With Gmrer and Percentage of All
between the high prevalence
The concept that adenocarcinoma of the prostate exists in a latent and a clinical form is supported by epidemiologic, pathologic, and clinicall evidence. Although these divergent manifestations of prostate cancel have common architectural and cytologic features, the), can be distinguished from each other to some degree by differences in certain pathologic features, such as the volume, grade, and invasiveness of the lesion. To appl) the available diagnostic tests rationallv wr need to know the nature of cancers detected with’these tests to determine whether these cancers are “clinically important” or capable of causing morbid or lethal complications if left untreated in men with a sufficiently long life expectancy. ‘We will review the epidemiologic, pathologic, and clinical features of latent and clinical prostate cancel and describe the outcome of treatment trials and the available information about the natural history of the disease to show that prostate canc‘er. when detected clinically, is most often a morbid and potentially lethal tumor. We will then describe the features of cancers detected with each of the available diagnostic tests. estimate the proportion of those cancers rhat are “cliniand review the preliminary experience cally important,” with the use of these tests in early detection trials. Thus, our analysis will attempt to provide a rational basis for the early delection that is essential if the morrality rate from prostate cancer is to decline.
of
as cancer and the much lower prev-
alence of the clinical disease. Despite the availability
of effective
tests for early detection and of effective treatment for cancers so detected, the diagnosis is locally advanced
usually is not established
or metastatic. Yet. physicians
until the tumor hesitate to use
these tests for fear that many cancers found would
be latent, of
little threat to the life or health of the host, and treatment could introduce inappropriate
morbidity.
Latent or “clinically unimpor-
tant” cancers can be distinguished
from those that are clinically
important
by the larger volume, higher
grade,
and greater inva-
siveness of the latter. The available tests can detect only those cancers large enough to be palpable. visible on ultrasound, or capable of elevating: the serum level of prostate-specific
antigen. Such can-
cers are clinically important and should be treated for cure if the
life expectancy
of the patient
is sufficiently
long and the morbidity
rate of therapy is low. Early detection of prostate cancer using the tests that are available today may widen the window of opportunity so that treatment indeed becomes possible necessary. HUM PATHOL
23:211-222.
in those for whom it is
Copyright
MD,
G 1992 by W.B.
Saunders Company I’I-ost:ite cancer is unique among the potentially leI hat human rnali~mancies in the wide discrepancy becween th’r high prevalence of histologic changes recognizable as cancer and the much lower prevalence of ( linical disease. Despite the availability of diagnostic tests to detect prostafe cancer (digital rectal examination /IIRE]. prostate-specific antigen [PSA], and transrectal ~tltrasonogral)~~y [TRUS]) and of effective treatment for c’ancers detected early (radical prostatectomy or radiotherapy), the diagnosis usually is not established until rhe turrlor is locally advanced or metastatic.. Prostate c’ancer will kill over 30,000 American men this year, and the age-specific morrality rate is increasing. Yet, physitians hesitate to apply these diagnostic tests in a vigorous program of early detection because of their concern that man\ cancers found would be “latent,” ie, of no real threat to the life or well-being of the host. Treatment of such tumors could result in inappropriate morbidity, md the ‘healrh of the community as a whole could be reduced by suc11 efforts.
CLINICAL
AND LATENT
PROSTATE
CANCER
Clinical Cancer Epidemiology
Prostate cancer has become the most common GLIBcer among American men, and only lung cancer is responsible for more cancer deaths.’ The ape-spec.ific mortality rate has slowly increased ol;tar thcb past 50 years,’ and in black American men is ne,lrly double the rate found in white men.’ Prostate cancer is responsible for nearly 35% of all deaths in men over the age of 55 years.” Since the incidence of prostate I.‘Lmcer increases more rapidly with age than any other cancel, and the average age of American men is rising, the number of patients with prostate cancer is expected to increase dramatically over the next decade.” Estimates are that in 199 1 o&r 122,000 men will be diagnosed with prostate cancer and 32,000 will die of the disease in the United States.’ The mortality rate may underestimate the morbidity, since endocrine therapy allows 10% of men with metastatic prostate cancer to live for 10 years 211
HUMAN PATHOLOGY TABLE 1.
Percentage Patients
Estimated Percentage of Patients in Each Clinical Stage at Diagnosis, Risk of Death From Prostate Cancer, and Approximate “Cure“ Rate (Lifelong Freedom From Prostate Cancer) Associated With Each Stage
of Stage*
JO
M+
“0
N+
10
T~LMI,
30 10
T,,,.,N,,W, T,,A,M,,
100
Volume 23, No. 3 (March 1992)
Prognosis Incurable Rarely turahle Occasionally curable Often curable ‘l‘reatment “unnecessary”
All Exe-ludina ‘I‘,, (stage A I)
* Based on clinical stage. pelvic lymph node dissection.
bone scan, and acid phosphatase.’
and 50% to live for 3 years.“.” These patients require frequent therapeutic intervention for local and systemic complications of the disease.’ Oufcome of Treatment
Trials
Approximately 30% of men with prostate cancer have distant metastases at the time of diagnosis.’ Despite the impressive symptomatic response of metastases to hormonal manipulation (androgen deprivation), the survival rate for these patients is dismal: the median duration of survival is less than 3 years.“.“,s,” By 5 years, over 75% and by 10 years, more than 90% of these patients die of their cancer rather than with it.“.” Approximately 30% of patients with apparently localized prostate cancer (To_d NxM,,) have pelvic lymph node metastases.“.‘” Virtually all patients found to have positive pelvic lymph nodes treated with local or regional therapy show clinical evidence of progression of tumor within 10 years”~“~‘-5 unless additionally treated with systemic therapy (androgen deprivation), which has not been shown to alter the long-term survival rate (reviewed by Austenfield and Davis’“). Prostate-specific antigen (PSA) levels are elevated within the first 2 years after radical prostatectomy and pelvic lymphadenectomy in more than 85% of patients with positive nodes’“-‘s and in 100% of these patients followed for longer periods.lg The cancer-specific mortality rate for these patients, whether hormonal therapy is administered early or late, is approximately 25% at 5 years and 60% at 10 years. Even patients with one single microscopic focus of tumor in one lymph node seem to fare as poorly in the long run as patients with more extensive nodal metastases, although long-term data are not available after some forms of therapy.” Once prostate cancer has spread to the regional nodes, it can rarely be controlled with any form of local or regional therapy. At least 70% of such patients will eventually die of their cancer rather than with it.” Approximately 15% of patients have a locally extensive tumor at the time of diagnosis (stage C or T:s..Js.’ ’ and half of these have pelvic lymph node metastases.‘z.‘3 The 5-year relative survival rate for patients with clinical sta e C cancer (T3_,‘NS) was 66% in a large national survey 8 and the risk of death from prostate cancer is approximately 65% to 75% at 15 years.“-” In 212
1O-Year CancerSpecific Survival Rate (%)
Estimated “Cure” Rate (%)
10 40 60 NJ 05
50 years old with cancer present (prevalence of autopsy cancer) Estimated proportion of autopsy cancer cases that are “clinically important” (0.2 x 30%) Actual t-ate of detection with DRE, PSA, +TRUS Annual incidence rate in men 250 years old Annual mortality rate in men >50 years old
30
Clinically Important 20%
6
linical Ca 100,000
l-3 0.31 0.09
Note. The estimated prevalence varres more than 3004old depending on the presumed prevalence of clinically important cancer. Abbreviations: DRE, digital rectal examination; PSA, prostatespecific antigen; TRUS, transrectal ultrasound.
malignancy in human beings.“* The age-specific prevalence of prostate cancer in cystoprostatectom mens confirms the findings in autopsy studies. Uqt; the Stanford series, 25 of 66 prostates (38%) contained cancer.“” Yet, the estimated number of men diagnosed with clinical prostate cancers in 1985 was only approximately 86,000, which represents only 1.05% (86,000 of 8,193,OOO) of the men estimated to have cancer present in the prostate (Fig 1).5? There were an estimated 25,500 deaths from prostate cancer in 1985, which is only 0.3 1% of these men. Thus, only one of 95 men with cancer was diagnosed with the disease, and only one in 323 died of the disease that year (Table 2). This enormous discrepancy between the high prevalence of the disease at autopsy and the low incidence of the disease clinically has provided a theoretical basis for a policy of “no treatment” and has stymied efforts at early detection and screening,“8,5”,“4,5” This apparent discrepancy, however, has been exaggerated by the failure to consider the effect of time. Although the clinical incidence of prostate cancer is low (approximately 0.31% per year for men more than 50 years old), the lifetime risk of developing prostate cancer for a man 50 years old in 1985 was 9.51%3 (Table 3). Similarly, the annual mortality rate for men more than 50 years old in 1985 was only 0.09%, but the lifetime risk of dying of prostate cancer was 2.89%. In 1985 a 50-year-old man was expected to live to the age of 75. Autopsy data confirm that prostate cancer is not found in every 75-year-old man but in only approximately 42%.5’ For a 50-year-old man, therefore, we can estimate that the lifetime risk of developing cancer in the prostate is approximately 42%, the risk of developing the disease clinically is approximately 9.5%, and the risk of dying from the disease is approximately 2.9%. Over their lifetimes, then, one of every 4.4 men who harbors cancer in his prostate will develop the disease clinically and one 214
FIGURE 1. Distribution of prostate cancers If the prevalence of adenocarcinoma among men more than 50 years of age is 30%‘ then approximately 8 million American men harbor foci of cancer in their prostate. Yet, in 1990 only approximately 100,000 men were diagnosed with the disease (“clinical cancer”) and the other 7.9 million were considered to have incidental or “autopsy” cancer. Based on the pathologic features of prostate cancer found at autopsy46 and on the lifetime risk of developing prostate cancer.52 we estimate that approximately 20% of these cancers are “clinically important” (that is, they will threaten the life or well-being of the host if not treated effectively), and 80% are truly latent cancer (“clinically unimportant”).
of every 14.5 will die of prostate cancer (Table 3).‘” Although the ratio of prostate cancer deaths to the development of histologic evidence of the disease is much greater (one in 14.5 rather than one in 323) when viewed in this perspective, it is apparent that most histologic (“autopsy”) cancers do not progress within the normal life expectancy of the host. Yet, these cancers found at autopsy have all of the cytologic and architectural features of the disease detected clinically.““-“‘.“‘,“” Pathologic Features Detailed pathologic examination of cancers found at autopsy and of clinically detected cancers do, however, show important differences between autopsy-detected and clinically detected cancers in the volume, grade, and invasiveness of the tumor.~6,48,4’1..51,.57 Most cancers found at autopsy are small, well differentiated, and show no tendency to invade normal structures ( rostatic capsule, seminal vesicles) (reviewed by Dhom” B). Others are larger, less well differentiated, and do invade normal structures. The prevalence and proportion of these larger, proliferative cancers are greater in older men,
TABLE 3.
The Lifetime Risk of Developing or of Dying of Prostate Cancer for a 50-Year-Old White Man in 1985 Lifetime Risk (Yo)
Lifetime Lifetime Lifetime
risk of developing “autopsy” cancer5p: risk of developing clinical cancer’: risk of dying of prostate came?:
Ratio
42 9.51 2.80
Note. The ratios indicate that for every death from prostatecancer, 3.3 men will he diagnosed with the clinical disease and 14.5will develop histologic evidence of cancel- in the prostate dur-ing their lifetime.
EARLY DETECTION
TABLE 4.
Geographic
in the Incidence
Variations
and Mortality
OF PROSTATE
(Scardino et al)
TABLE 5. Features That Help Distinguish Ciinicalty Important From Unimportant Prostate Cancers
(8.4- to 10.8-Fold) Rates
CANCER
of Clinical
--__
Prostate Cancer (Per 100,000 Men&r) (Iinicall~ Important
Clinically C:nimportant :small Well-diflerentiated ((;lea\on gwlr I or2) Soninwsiv~~ Dipicki lormal ‘1 rmsitirm
~onr
cer in men over the age of 50 years would be approximately 20% of the 30% of men more than 50 years of age with prostate cancer, or 6% rather tlhan 30% (prevalence at autopsy) or 0.31% (annual incidence) (Table 2). These are the clinically important (IFig 2) but currently undetected cancers that would be the appropriate targets of an early detection program and could represent as many as 1.6 million men in the United States alone, a staggering figure unless one considers that 2.0% of 50-vear-old men alive today (or approximatelv 790,OOiI in this country alone) will eventually die oi prostate cancer.” To prevent these deaths we may have to detect and treat two to three patients for every death prevented. The proper goal of an early detection program is not to detect all prostate cancers, but only IO detect those that are potentially morbid or lethal (the clinically important cancers), which we estimate have a prevalence of approximately 6% (20% of cancers prevalent in the prostate in men more than 50 veal-s old., or all the cancers, except To,, that will even&allv he detected as clin-
ill
black Americans. and in geographic regions with a higher clinical incidence of and mortality rate from prostate ‘cancer, while the smaller, nonproliferative c.mcers are more uniform with age, ethnic group, and geographic region (Table 4)” (reviewed by Dhom’s and Scardino”‘). These observations strongly support the c’c)ncept of a multistep process in the pathogenesis of prostate cancer in which latent cancers have progressed through s~ome but not all of the steps necessary for full malignant expression.“’ Since some cancers found at autopsy have the features of clinical cancers”.4H.‘,1.6” and some cancers detrcted clinically have the features of autopsy or latent cancers (for example, stage A, or To,),““-“” we have found il more helpful to describe prostate cancers as “clinically important” (ie, threatening the life or well being of the host within his remaining life expectancy) or “clinically unimportant” (ie, a latent cancer of no threat to the host).“’ At the light microscope level, tumor volume, grade, am1 invasiveness are some of the features that help to distinguish clinically important from unimportant cancers (Table 5). Llnfortunately, no objective nlarkers of progression in human prostate can&r have been established,7” hut additional features that seem to he strongly associated with the biologic behavior of prostate cancer are the DNA ploidy status of the tumor, the degre’r of PSA production (serum PSA level), and, possibly, the Lone of origin of the turnor within the prostate (‘Table 5). The proportion of prostate cancers present in the population of men over 50 years old that are clinically importanl is difficult to determine precisely, but information available from detailed morphometnc studies of prostate cancers found at autopsy and in cystoprostatectomy series suggests that approximately 20% of these currently undetected cancers are clinically important based on the features listed in Table 5 (I;ig l).’ 1,;‘-IH4!1.5I..-,:!.r,7.l>ll
DIRT
IBUTI~N
I I
OF
PROSTATE
N=8,000,000
LANCERS I ;
i AutoDsY ca
FIGURE 2. Of the clinically detected cancers nearly half are advanced (Table 1). Approximately 40% are early stage and may be curable, and 10% are focal, well-differentiated stage A, (T,,) and may be “clinically unimportant.” Of the large pool of undetected but clinically important cancers, similar proportions are estimated to apply.
Goal of a,? Early Detection Program If OUI‘ estimates are correct, the prevalence of clinic-ally impl:)rtant but presently undetected prostate can215
HUMAN PATHOLOGY
Volume 23, No. 3 (March 1992)
TABLE6. Patients With Moderate or Severe Obstructive Voiding Symptoms at the Time of Diagnosis of Prostate Cancer (Stages B and C, or T,_J Were More Likely to Have Pelvic Lymph Node Metastases (P < .OOOl) Results of Pelvic Lymph Dissection Degree of Obstructive Voiding Symptoms None Mild Moderate Severe Total
N (%)* 171 55 93 50
(46) (15) (25) (14)
369 (100)
Negative
Positive (%)t
135 35 58 ?I ?49
* Percentage of patients with each degree of voiding t Percentage of patients with lymph node metastases. Data from Meacham et aLa
36 20 35 29
(21) (36) (38) (58)
6
c
8
120 (33)
c
20
i P
symptoms.
I
p=.OOl l
--* Negllglble Sx (226)
M
’
’
’
’
’
’
’
’
5
0
Time
TECHNIQUES FOR EARLY DETECTION AND CHARACTERISTICS OF CANCERS DETECTED
Substantlal Sx (n=143)
t
01
ical cancers) (Fig 2) (Table 2). We will review below the pathologic features of cancers detected with the available noninvasive diagnostic tests (symptoms, DRE, PSA, ultrasound, and combinations of these) and estimate the proportion that are clinically important.
I
’
’
10
(Years)
FIGURE 3. Freedom from distant metastasis after treatment (radiotherapy) for clinically localized (stages B and C, or T,. ,N,M,) prostate cancer based on the degree of obstructive voiding symptoms at the time of diagnosis. Patients with substantial (moderate or severe) obstructive symptoms had a significantly worse prognosis. (Reproduced with permissionBO: Meacham RB, Scardino PT. Hoffman GS, et al: The risk of distant metastases after TURP vs needle biopsy in patients with localized prostate cancer. J Urol 142:320-325, 1989, @ by American Urological Association, Inc.)
Symptoms
Prostate cancer is a notoriously silent disease with few early symptoms. Symptoms associated with bladder outlet obstruction are commonly present in men over the age of 50 years and are often ascribed to benign prostatic hyperplasia (BPH). Indeed, BPH is more common than prostate cancer: the lifetime risk of a 50-yearold man requiring a prostatectomy for symptomatic BPH is 20% to 25%.* However, symptoms of bladder outlet obstruction also develop in men with prostate cancer. In a series of 300 consecutive men hospitalized for obstructive voiding symptoms, Guinan et al diagnosed prostate cancer in 23%?” In our own series of 5 10 men treated with radioactive gold seed implantation and external beam irradiation for TOb-TJ cancer (stages AB, B, and C), 234 (46%) had a transurethral resection of the pstate for symptoms of bladder outlet obstruction. ’ Of the patients with a palpable (T,.,) cancer, 143 (39%) had moderate or severe obstructive voiding symptoms at the time of diagnosis. These patients had a greater chance of having lymph node metastases (Table 6) and developing distant metastases (Fig 3) than comparably staged patients without obstructive voiding symptoms, confirming the clinical observation that patients with prostate cancer who have obstructive voiding symptoms at presentation are more likely to have locally advanced disease and a poor prognosis. s1 While there are no objective data describing the detailed pathologic characteristics of prostate cancers that are detected because of the development of urinary symptoms, such symptoms are likely to be of little use in the early detection of prostate cancer. 216
Digital
Rectal
Examination
The DRE traditionally has been considered the most accurate test for the detection of prostate cancer. Digital rectal examination has been demonstrated to be more sensitive, more specific, and to have a greater efficiency than a variety of laboratory tests available in the 1970s7” However, few of these laboratory tests are still in clinical use today. In one study patients were hospitalized for urinary symptoms and were not asymptomatic subjects of a screening program; prostate cancer was detected in 23%, considerably more than expected in a screening trial. Digital rectal examination has been used to detect prostate cancer in several large series, and the detection rates range from 0.8% to 1.4% (Table 7).““m56.8” Although 67% to 88% of these tumors are clinically localized,
TABLE7.
Results of Early Detection Programs Using Each Diagnostic Test17.53.56.68.76.82-86 DRE (%)
Positive predictive Detection Clinically
value
rate localized
Pathologically
confined
TRW
(%)
22-36
15-41
0.8-1.4 53-88
2.3-3.1 58-100
37-77
56-68
PSA (%) “‘L-35 (>4)* 65-67 (>lO) 2.2 96 (4- 10) 77 (>lO) 56-72 (4-10) 8-33 (10)
Abbreviations: DRE, digital rectal examination; TRUS, tramrectal ultrasound; PSA, prostate-specific antigen. * Numbers in parentheses refer to the level of PSA.
EARLY DETECTION
OF PROSTATE
TABLE 8. Characteristics of Cancers Detected With Each Diagnostic Test in Patients With Clinically Localized Prostate Cancers Treated With Radical Prostatectomy, and for Latent Cancers Detected Incidentally at Autopsy or in Cystoprostatectomy Specimens __.__
CANCER
(Scardino et al)
palpable induration correlates well with the total volume of tumor”” nor with the area of cancer on the posterior surface of the prostate.!“’ Asymmetry and induration, as well as a discrete hard nodule, are included in the tindings on I)RE that may be judged suspicious for prostate CYIIK~I‘. but these findings lack specificity for cancer. The positive predictive value of a palpable abnormalit\; is on,lk 22%, to 36Yo (Table 7). Yet, the most serious 1in;itation of DRE is its lack of sensitivity (false-negative results). For example, approximately 10% to 20% of transurethral resections performed for benign prostatic hypertroph~ in patients with no palpable abnormality suggestive of cancer uncover an incidental cancer of the prostate. Digital rectal examination detected only 12 of 22 cancers found in ;I screening study, while TRUS found 20.““ Goner et al detected cancer by TRUS in 12.4% of 22.5 men, none of whom was ‘:juclged by the urologist tl:) have a suspiciously palpable lesion.““g Goner et at later reported that 25% of the cancers found in men examined h> TRUS were not palpable (Table 9).“” Thus, DRE is relatively insensitive and nonspecific. Cancers detected by palpation are relatively large, man) are late in their- development and no longer curable. and some are so small they would fit OUI definition of clinically unimportant cancers.
l~;~tl~ologi;lc~ staging revealed that many were, in fact, not confined to the prostate gland.9”,54 In our own series of 9X clinical stage B (T,.,) radical prostatectomy specimens, 63% extended through the capsule and 18% into the seminal vesicles.x7 These cancers ranged in size from 0.45 to 19.2 cc (mean, 4.04 cc), although we occasionally have encountered palpable cancers as small as 0.04 cc.88 Stanley et al reported that palpable clinical stage B cancers in patients with negative lymph nodes ranged from 0.01 to 25 cc (mean, 5.4 cc).‘” Nonpalpable cancers discovered in cystoprostatectomy specimens by these same investigators ranged from 0.00 1 to 1.1 C‘C(mean, 0.1 cc) and all were confined to the prostate.‘” Thus, palpable cancers clinically confined to the prostate are one to two orders of magnitude larger, on average, than latent cancers (Table 8) and most already extend outside the l,rostate_:‘“.;;.“7.~~~
Prostate-Specific
Antigen
The on]!. study using PSA as the primarv test to screen for prostate cancer was conducted by Catalona et al, who determined serum levels in 4,293 self-referred patients,‘t’,x’l Overall, serum levels higher than 4.0 ng/ mL (Hybritech monoclonal I-adioimrnunoassa~; Hybritech, San Diego, CA) were detected in .516 patients (12%). These patients were asked to ret.urn for a DRE and TRUS to determine the need for a Ileedle biopsy. The overall cancer detection rate in the published series was 2.2% (Table i).” Cancer was detected in 26% of the men with a PSA level of -1.0 to 10.0 ng/ml.. Nearly all of these tumors (96%) were clinically localir.ed, and the tumor was confined to the prostate in 72% of these patients who had a radical prostatectomv. Callcer was detected in 66% of patients with a serum level higher than 10.0 ng/mL. While 77% of these cancers were clinically localized, only 33% of the patients who had a rad-
Not only does DRE detect cancer relatively late, there is only a weak correlation between the size of the cancer estimated by DRE and the actual volume of cancer present. Tumors palpable as small nodules (1.5 cm or half of one lobe) are generally smaller than larger palpable tumors, but the overlap is substantial.“’ Neither the absolute nor the relative dimensions of an area of
TABLE 9. The Detection of Prostate Cancer in Studies Using a Combination of Digital Rectal Examination. Prostate-Specific Antigen, and Transrectal Ultrasound With Transrectal Ultrasound-Guided Eliopsy of Any Suspicious Hypoechoic Lesion: Influence of Abnormal Digital Rectal Examination or Prostate-Specific Antigen (>4.0 ng/mL) on Results _______~ ____
Abnormal ~rRtIS N (%) DRE+. PSAt DRE+. PSADRE-. PSAt DRE-, PSA-~ All patients __--
438 (I 7) 446 (17) 487 (IX)
Biopsy Performed
PPV TRUS
0!I ti7
3i 15
19
“7
Puwnragc With Cancc~
No. of Patknts With Gmrer and Percentage of All
