Int. J. Cancer: 124, 547–552 (2009) ' 2008 Wiley-Liss, Inc.
Biological aggressiveness of prostate cancer in the Finnish screening trial Marita Laurila1*, Teuvo L.J. Tammela2, Anssi Auvinen3,4, Jorma Isola5, Tapio Visakorpi5, Tiina Luukkaala6, Liisa M€a€att€anen7, Mirja Ruutu8, Martti Ala-Opas8, Markus Mildh8 and Paula Martikainen1 1 Department of Pathology, Tampere University Hospital, Tampere, Finland 2 Department of Urology, Tampere University Hospital and University of Tampere, Tampere, Finland 3 Finnish Cancer Institute, Helsinki, Finland 4 School of Public Health, University of Tampere, Tampere, Finland 5 Institute of Medical Technology, University of Tampere and Tampere University Hospital, Tampere, Finland 6 Research Unit, Pirkanmaa Hospital District and Tampere School of Public Health, University of Tampere, Tampere, Finland 7 Finnish Cancer Registry, Helsinki, Finland 8 Department of Urology, Helsinki University Hospital, Helsinki, Finland Prostate cancer aggressiveness was evaluated based on pathologic characterization of cases detected in the Finnish prostate cancer screening trial. The trial population consists of 80,458 men aged 55–67 years. A total of 32,000 men were randomized to the screening arm. The remaining 48,000 men formed the control arm. The interval cases and cancers among nonparticipants and in the control arm were identified from the Finnish Cancer Registry. Random samples were selected from screen-detected cases (126 of 543 in the first and 133 of 508 in the second round) and control arm cancers (133 out of 863), in addition to all 92 interval cancers and 106 cases among nonparticipants. All the biopsies were regraded according to the Gleason system. The expression of the proliferation antigen Ki-67 was determined in 479 cases (72%). More than half of the tumors diagnosed in the first round of screening were high-grade cancers (Gleason 7 or higher). In the second round, the proportion of low-grade cancers increased from 47% to 70%. Cancers in the screening arm were more commonly focal and fewer bilateral cancers were detected. The cancers among nonparticipants were the most aggressive group. The aggressiveness of the interval cancers was between the cancers detected in the first and the second round. Our results indicate that prostate cancers detected through screening are less biologically aggressive. This was most notable after the first screening round. Nonparticipants had more aggressive cancers. ' 2008 Wiley-Liss, Inc. Key words: prostate neoplasms; prostate cancer; screening; prostatespecific antigen; proliferation antigen
Prostate cancer is the most commonly diagnosed noncutaneous cancer in men in the Western world and is the second most frequent cause of cancer-related deaths in men.1 More than 95% of malignant prostate neoplasms are adenocarcinomas, and in general, the term prostate cancer is used to designate primary adenocarcinoma. However, prostate cancer comprises a wide spectrum of biological and clinical behavior, ranging from a minute latent tumor to a highly aggressive, life-threatening disease.2 Today, majority of prostate cancers are detected based on serum prostatespecific antigen (PSA). The ultimate goal of prostate cancer screening is to reduce the disease-specific mortality. However, because the biological behavior and the natural course of prostate cancer and its relation to PSA is not known, controlled randomized screening studies are needed to evaluate the effects of early detection and treatment of cancer in mortality and in quality of life.3 The European Randomized Study of Screening for Prostate Cancer (ERSPC) is a multicenter, randomized trial, which started in 1994.4 Within the ERSPC, approximately 193,000 men from 8 European countries have been randomized, of whom approximately 40% are from Finland with a sample size of approximately 80,000 men.5 Pathological stage and tumor grade together with PSA are the most important indicators of outcome.6 However, additional markers for growth potential are needed, because of variation of growth potential and prognosis among tumors with similar stage or grade. The proliferation marker Ki-67 (also known as MIB-1) Publication of the International Union Against Cancer
is a protein that recognizes nuclear antigens in the cell cycle from the late G1 phase, in S and G2 phases to M phase but not in noncycling cells in G0 phase.7 Ki-67 expression is associated with the prognosis of many malignant tumors, such as breast cancer,8 colorectal adenocarcinoma,9 and also prostate carcinoma.10 Ki-67 expression is associated with pretreatment prostate-specific antigen (PSA) level, tumor grade (Gleason score), pathologic stage and progression of cancer.11,12 Patients with higher Ki-67 staining have worse prognosis than those with a lower expression.13 We have recently shown that Ki-67 staining is an independent prognostic marker in prostatectomy treated patients.14 The aim of this study was to characterize prostate cancers diagnosed during the first 2 rounds of the screening trial and compare their tumor characteristics with other cancers in the trial population, including interval cancers and cancers among nonparticipants. The characterization was performed from the diagnostic biopsies and the indicators of biologic aggressiveness were Gleason score, focality or bilaterality of tumor and Ki-67 expression.
Material and methods Subjects The Finnish prostate cancer screening trial is the largest component of the ERSPC with a cohort of 80,458 men aged 55–67 years at entry. The study population was identified from the Population Register of Finland, and those men with previously diagnosed prostate cancer were excluded before randomization (n 5 1,570). Annually, during 1996–1999, 8000 men aged 55, 59, 63 and 67 years were randomized into the screening arm, and the remaining men formed the control arm. At the time of invitation, 30,403 men were eligible and finally, 20,793 participated in the first and 19,107 men participated in the second screening round after 4 years. For the present study, we randomly selected 128 cancers out of a total of 543 screen-detected cases from the first round. Of these 126 (98%) histologic samples were available. Of the 508 screendetected cases in the second round, 125 cancers were selected and 123 samples were available. During 1996–2003, a total of 863 cancer were diagnosed in the control arm and a random sample of 135 cancers was selected for the study (133 were available). All the prostate cancers diagnosed among nonparticipants in screening (92 cancers, of which 89 histologic samples were available) and the cancers diagnosed between the screening rounds (106 cancers, of which 99 samples were available) were included in the study. The cancers were defined as interval cancers, if they were diag-
*Correspondence to: Department of Pathology, Tampere University Hospital, Tampere, Finland. E-mail:
[email protected] Received 15 February 2008; Accepted after revision 5 June 2008 DOI 10.1002/ijc.23873 Published online 11 November 2008 in Wiley InterScience (www.interscience. wiley.com).
548
LAURILA ET AL.
nosed during the screening interval following either normal PSA or elevated PSA, but negative biopsies. The cancers diagnosed in the control arm and among nonparticipants were identified from the Finnish Cancer Registry. Screening algorithm Men with serum PSA level 4.0 ng/ml or higher were referred for diagnostic examination [a digital rectal examination (DRE), a transrectal ultrasound (TRUS) and sextant needle biopsy with 18 G needles]. Additional biopsies could be taken from suspicious nodules. For men with PSA of 3.0–3.9 ng/ml, an ancillary screening test was performed (DRE in 1996–1998 and determination of the proportion of free PSA with a cut-off value of 0.16 since 1999). If the first biopsy indicated a cancer suspicion or high-grade prostatic intraepithelial neoplasia or if the material was insufficient, 1 repeat biopsy was suggested. If the serum PSA was above or equal to 10 ng/ml, a repeat biopsy was also recommended regardless of the benign histological finding in the first biopsy. Histopathological analysis The histologic samples were available from 570 prostate cancers, of which 540 were biopsies and 30 were resection materials. Material for the first and the second screening rounds was classic sextant biopsy cores throughout the study. The samples were fixed in buffered formalin and embedded into paraffin blocks according to routine protocol. Each slide was stained with hematoxylin and eosin. The primary cancer diagnosis was made by general pathologists in several laboratories. All the samples with a primary diagnosis of prostate cancer were reviewed and the diagnoses were confirmed by a single uropathologist (M. Laurila). All the tumors were regraded according to the Gleason system. Tumors were classified in 3 groups: cancers with Gleason score 7. The number of the needle biopsy cores and the number of the cores containing cancer were counted. Focal cancers, with a malignant focus of maximum 3 mm and a Gleason score of maximum 6 in a single biopsy core, were reported. The cancers in biopsies from both prostate lobes were reported as bilateral cancers. Immunohistochemical analysis The block with the most representative slide containing the largest tumor volume from each case was selected for the immunohistochemical staining. The volume of cancer in biopsy cores was small in some cases, especially in focal cancers, and therefore, no cancer tissue was available for immunostaining in 91 samples. The proliferation antigen Ki-67 (MIB-1) could be analyzed in 479 cases (84%) of 570 cancers. A 5-micron section was cut from the representative block of formalin-fixed, paraffin-embedded tissue and placed on silinated slides. For the antibody, the sections were deparaffinized in graded alcohols and immunostained with the monoclonal antibodies to Ki-67 antigen (MIB-1, clone MM-1, Novocastra Laboratories, Newcastle, UK) according to manufacturer’s instructions. Hightemperature antigen retrieval was performed with 20-min incubation in 1 mM ethylenediaminetetra-acetic acid, pH 8.5. The bound antibody was visualized using a standard avidin-biotin technique (Vectastain Elite, Vector Laboratories, Burlingame, CA). The Ki-67 stainings were graded by a single uropathologist (M. Laurila). Most of the specimens showed a heterogeneous distribution of immunostaining. For Ki-67 staining interpretation, brown, granular nuclear reactivity was considered positive. Tumor areas showing the highest density of cells were chosen for analysis. Positively stained tumor nuclei were calculated under 400-fold magnification. The whole tumor, or at least 5 areas were selected, and a minimum of 1,000 tumor cells were scored. The mean percentage of positive tumor nuclei per counted tumor cells was characterized as low (occasional or up to 5 percent positive cells), moderate (5 to 10 percent positive cells) or high (more than 10 percent positive cells).
TABLE I – DISTRIBUTION, NUMBER OF CASES (n) WITH PERCENTAGES (%), OF FOCAL AND BILATERAL CANCERS IN DIFFERENT SUBGROUPS (n 5 570)
1st round 2nd round Control arm Nonparticipants Interval cancers Total
All
Focal
Bilateral
n
n (%)
n (%)
126 123 133 89 99 570
15 (12) 25 (20) 10 (8) 12 (13) 20 (20) 82 (14) p 5 0.039
27 (21) 27 (22) 57 (43) 31 (35) 15 (15) 157 (28) p < 0.001
Differences between subgroups were tested by Pearson v2 test.
Data analyses Comparisons of proportions of focal and bilateral cancers, Gleason score and proliferation antigen Ki-67 between patient groups, as well as comparison of proliferation with Gleason score were done by Pearson v2 test. Logistic regression models with odds ratios (OR) and 95% confidence intervals (CI) were used to examine the association of focal and bilateral cancers in 5 prostate cancer subgroups. Statistical analyses were performed using SPSS for Windows, version 14.0.2. A p-value less than 0.05 was considered statistically significant.
Results The total number of prostate cancers in the 5 subgroups (the first round, second round, nonparticipant, interval cancers and control arm), including focal and bilateral cancers in each subgroup, are presented in Table I. Of all 570 cancers 82 (14%) were focal tumors. Most of them (89%) were Gleason score 6 tumors; only 2 focal cancers were graded as Gleason score 4 and 7 as Gleason score 5. The proportion of focal tumors differed significantly between the subgroups (p 5 0.04). In the second round, the proportion of focal cancers was 2.8 times higher than in cases detected than in the control arm. Focal cancers were also more than twice as common in the interval cases compared to the control group OR 2.63 [95% CI 1.19–5.23]. Of the cancers, 157 (28%) were bilateral and the proportion of bilateral cancers differed significantly between the groups (Table I). In the first and second round of screening as well as interval cancers significantly fewer bilateral cancers were detected than in the controls and nonparticipants (p < 0.001). In the control arm, the proportion of bilateral cancers was 2-fold higher (OR 2.88 [95% CI 1.66–5.01]) compared to the first round and second screening round (OR 2.67 [95% CI 1.54–4.61]), and nearly 3-fold compared to the interval cases (OR 4.20 [95% CI 2.20–8.03]). More than half of the tumors diagnosed in the first round were high-grade cancers (Gleason 7 or higher) (Fig. 1a). In the second round, the proportion of high grade cancers decreased and lowgrade tumors increased from 47% to 67%. The Gleason score in interval cancers was intermediate between the first and the second round. The highest Gleason scores were in the nonparticipants and in the control arm. Bilateral cancers had high Gleason score in every subgroup (Fig. 1b, p 5 0.002). The proliferation antigen Ki-67 (MIB-1) could be analyzed in 479 (84%) of 570 cancers (from 78% of nonparticipants’ to 90% of first round cancers). Ki-67 was associated with the Gleason score: tumors with a low Gleason score had a low degree of Ki-67 expression, and correspondingly, tumors with a high Gleason score showed most frequently high Ki-67 staining (Fig. 2). Ki-67 could be analyzed in 46 (56%) of 82 focal cancers and in 135 (86%) of 157 bilateral cancers. Ki-67 distribution differed significantly between the subgroups in total material (p 5 0.039, Fig. 3a). Focal cancers generally expressed Ki-67 at low level, whereas more than half of the bilateral cancers expressed Ki-67 5%. Ten focal cancers of 46 (22%) had proliferation index more than 10%.
BIOLOGICAL AGGRESSIVENESS OF PROSTATE CANCER
549
FIGURE 1 – Gleason score distribution in subgroups of all (a) and in bilateral (b) cancers. Percentages of grouped Gleason score for all (n 5 570) in (a) and with bilateral cancers (n 5 155) in (b). Differences between groups were significant (p < 0.001 for all cancers; p 5 0.002 for bilateral cancers; tested by Pearson v2 test). NP, Nonparticipants.
number of focal cancers in the screening would thus be 192 including the 20 focal cancers in the interval cases and the 12 in the nonparticipants. In the control arm 10 cancers of 133 randomized cases were focal cancers. Based on a sampling ratio of 15%, a total of 67 focal cancers in the whole control arm could be estimated. The cumulative incidence of the focal cancer was 6.3 per 1,000 (192 focal cancers in 30,430 men in screening arm) and 1.4 per 1,000 men in the control arm (67 focal cancers of 48,000 controls) corresponding to a rate ratio of 4.5 (95% CI 3.4–5.9). For the bilateral cancers, the detection rates were 8.6 per 1,000 men in screening arm (n 5 262) and 7.9 per 1,000 men in the control arm (n 5 378). This corresponds to a rate ratio of 1.1 (95% CI 0.9– 1.3). Discussion
FIGURE 2 – Number of cases by MIB score (1–3) and Gleason score (n 5 475). MIB score 1: < 5%; MIB score 2: 5–10%; MIB score 3: > 10% positively stained nuclei. Differences of MIB score distributions between Gleason score groups were tested by Pearson v2 test.
Ki-67 expression was most common in cancers among nonparticipants and in the control group (Fig. 3). Cancers diagnosed in the second round were positive for Ki-67 slightly more commonly than those detected in the first round. The interval cancers had the lowest Ki-67 expression. Our material of 249 randomly selected cancers diagnosed in the first and the second round represented approximately 1 quarter of the total number of cancer cases in each screening round, and 40 focal cancers of the screen detected cases were found. This should correspond to 160 focal cancers, found in the first and in the second screening rounds, with the given sampling ratio. The total
We found that in a population-based screening program, the biological aggressiveness of the cancers diagnosed in the second screening round was lower than in the first round or that of the interval cancers, while the nonparticipants had the most aggressive cancers. Two possible explanations could account for the findings: either early detection results in diagnosis of cancers before they develop more malignant features or, alternatively, the lower degree of aggressiveness is due to detection of indolent tumors. Both explanations could account for the more aggressive features in the control group and nonparticipants. The low aggressiveness of interval cancers suggests that a large proportion of them was not true de novo cancers arising during the screening interval, but commonly detected after repeated biopsy rounds in screen-positive men. Screening is aimed at reducing mortality by detecting cancers in their early, asymptomatic phase, while they are still curable. Because of the wide spectrum of clinical behavior in prostate cancer, it is crucial to distinguish cancers that would progress from those that would remain indolent. In our earlier study, we have shown that biological aggressiveness of screen detected cancers is intermediate between clinically detected cancers and latent cancers verified in autopsy.15 However, the results of present study suggest that a proportion of screen-detected cancers are clinically significant. This poses a major challenge as we should know
550
LAURILA ET AL.
FIGURE 3 – Distribution of proliferation marker MIB-1 in total material (a), in bilateral cancers (b) and in focal cancers (c). Significant differences between subgroups were detected (Pearson v2 test p 5 0.042). NP, Nonparticipants.
whether the cancer is a relatively harmless and does not require treatment, manageable by active surveillance or watchful waiting, or potentially aggressive demanding more radical treatment modalities. A meaningful proportion of screen-detected cancers are minimal and could remain indolent during the rest of the man’s lifetime without progressing into clinical cancer. It is, however, difficult to predict the natural course of a cancer due to uncertainties in estimating both life expectancy and prognosis of the tumor. Our primary results were given as proportions of cancer in each group fulfilling criteria for aggressiveness such as Gleason score, MIB-1 staining, etc. These results reflect the relative frequencies of cancers with varying aggressiveness. As the screen-detected cancers represent a mixture of indolent and clinically relevant cancers, the former dilute the aggressiveness of those cancers that would be relevant for mortality. To overcome this limitation, we also give the findings as cumulative incidence. Our results indicate a rate ratio of 4.5 for focal cancer, i.e., screening increases the detection of minimal tumors. This gives a crude estimate of overdiagnosis. However, due to leadtime (anticipation on earlier detection of tumors by screening, estimated as 6–12 years for prostate cancer) the control group is likely to catch up with further follow-up to some extent and the difference is therefore likely to decrease. In particular, the expected effect of screening is an eventual reversal of the ratio between the trial arms in the occurrence of clinically
progressive tumors (such as bilateral cancers), because effective screening should decrease the risk of cancer with poor outcome. Only in a randomized trial, the characteristics of screendetected cancers can be compared with control cancers in a valid fashion, because randomization creates comparable groups and the control arm represents the situation in a similar population in the absence of screening. Volume of prostate cancer in biopsies correlates with the volume of prostate cancer in radical prostatectomy, and 12 biopsies have better predictive value compared to 6 biopsies.16 However, tools to predict prostate cancer characteristics and extent are still limited, and it is uncertain, how volume and biopsy characteristics predict the clinical behavior of the cancer. In prostate biopsy, a small single focus of cancer with a low Gleason grade suggests probably clinically insignificant, organdefined cancer with low tumor volume. Different definitions for small cancers in biopsy have been used. In the present study, we used the definition of Weldon et al.17 to define focal cancer, i.e., as a single cancer positive biopsy core with a small focus, 3 mm or less, of adenocarcinoma without high grade cancer (Gleason grade 4 or 5). This definition has also been used in the analysis of biopsy results from Dutch screening centre of ERSPC.18 By this definition, our results showed that focal cancer is more frequent in screen-detected cancers than in cancers among controls or nonattendants. Furthermore, the proportion of focal cancers increases further at the second screening round. These results are in accord-
551
BIOLOGICAL AGGRESSIVENESS OF PROSTATE CANCER 18,19
ance with those reported by the Dutch group as well as the general ERSPC experience.20 In the study by Postma and others, the proportion of focal cancer increased significantly from 16% in the first round to 29% in the second screening round. This finding is similar to the results of Swedish section of ERSPC in subsequent screening rounds at a 2-year interval.21 Also, in our study the proportion of focal cancer increased from 12% in the first round to 20% in the second screening round. Interestingly, the interval cancers had similar frequency of focal cancers as detected in the second screening round. This indicates that few aggressive cancers develop during the screening interval. At present we do not yet know, what the real outcome of these focal cancers is. In earlier studies,22 focal cancer in biopsy predicted minimal cancer (less than 0.5 ml) correctly in 53% of subsequent radical prostatectomy specimens. Stamey et al. suggested that tumors smaller than 0.5 ml, regardless of Gleason grade, should be considered as clinically insignificant. However, in 15% of the cases, surgical margins were positive and extraprostatic growth was diagnosed in 9% of specimens. In addition to the definition of focal cancer used in these studies, further criteria for small cancer lesions have been proposed. Allan et al. defined focal cancer as 0.5 mm tumor in 1 biopsy core and no Gleason grade 4 or 5.23 Based on prostatectomy findings, Epstein et al. proposed even smaller volume, 0.2 ml as indication of clinically insignificant or minimal cancer, because some prostate cancers with the volume of 0.2–0.5 ml were not organ-confined.24 Postma et al.25 compared PSA progression after radical prostatectomy to PSA doubling time of watchful waiting policy among patients with focal cancer. The PSA doubling time is used as a variable to observe prostate cancer. Patients with focal cancer in biopsy have a small risk of PSA progression after radical surgery (4.6%). However, using a PSA doubling time
