Incidence of Cancer in Finnish Families with Clinically Aggressive and Nonaggressive Prostate Cancer

Background: Clinical features of familial prostate cancer (PCa) and other malignancies associated with PCa are poorly described. Using a large family-based data registry of histologically confirmed cancers with a 40-year follow-up, we sought to determine incidence of cancer in Finnish PCa families, separately for clinically aggressive and clinically nonaggressive PCa.

Methods: We calculated standardized incidence ratios (SIR) for 5,523 members of 202 families by dividing the number of observed cancers (altogether 497 cases) by the number of expected cancers. The number of expected cancers is based on the national cancer incidence rates.

Results: SIR for overall cancer risk, excluding PCa, for male relatives in clinically nonaggressive families was 0.7 [95% confidence interval (95% CI), 0.6-0.8] and in clinically aggressive families 0.8 (95% CI, 0.6-1.0). The respective SIRs for women were 1.0 (95% CI, 0.8-1.1) and 1.1 (95% CI, 0.8-1.3). The incidence of lung cancer among men was significantly lower than in the general population. The SIR for gastric cancer among women was 1.9 in both clinically nonaggressive and clinically aggressive families. In clinically aggressive families, there was borderline significant excess of cancer of the gallbladder in men and liver cancer in women.

Conclusions: The incidence of non-PCa cancers is not increased in clinically aggressive or clinically nonaggressive PCa families except for stomach cancer among women. (Cancer Epidemiol Biomarkers Prev 2009;18(11):3049–56)

Prostate cancer (PCa) is the most frequently diagnosed malignancy among men in the Western world.⁶

Largely due to the earlier diagnosis provided by the prostate-specific antigen (PSA) test, introduced in the 1990s, the incidence of PCa is no longer increasing. The highest incidence (115/100,000) of PCa in Finland was observed in 2005. In 2007, there were 4,188 newly diagnosed PCa cases in Finland with an age-adjusted incidence rate of 85.7/100,000.⁷ Similar patterns of decline in PCa incidence have been observed in most Western industrialized countries (1).

The etiology of PCa has remained poorly understood. Ethnicity, age, and family history are considered major risk factors for PCa. The familial aggregation of PCa has been observed as early as the 1950s (2). Men with a positive family history of PCa have a 2- to 10-fold higher risk of getting PCa compared with those with no family history. The risk of PCa is highest in families with multiple PCa cases and in those with a low age of cancer diagnosis (3, 4).

Extensive efforts have been made to find genetic causes for PCa susceptibility. By linkage analysis, several chromosomal regions have been associated to PCa (5). Fine mapping of promising loci has identified three highly penetrant candidate genes ELAC2, RNASEL, and MSRI (6-8). However, mutations in these genes seem to be extremely rare, especially in the Finnish population (9-11). Several low penetrant polymorphisms have been found to be associated with PCa; however, recent studies suggest that the detected associations vary in different populations and ethnic groups (12-14), indicating genetic heterogeneity among people with PCa.

Based on the first reports on familial aggregation of PCa, hereditary PCa was considered to be site-specific (15). Hereditary PCa has been associated with a number of cancers, including gastric, breast, central nervous system, colon, multiple myeloma, gallbladder, non–Hodgkins lymphoma, skin melanoma, and kidney cancer (15-17). In population-based study associations between PCa and other cancers, such as gastric, colon, rectal, kidney, breast, ovarian, bladder, thyroid, and brain cancers, melanoma and non–Hodgkin lymphoma, have been reported, but the findings have been inconsistent between studies and only a few of these neoplasms have been reported even twice (4, 18-23). Studies of families with hereditary breast cancer, however, have reported that male carriers of BRCA2 mutations are at increased risk for PCa (24-26).

Being a genetically heterogeneous disease, analysis of PCa families with incidences of similar cancer types could be highly informative in genetic linkage and association analysis toward the identification of PCa predisposition genes. The use of clinically defined phenotypes could simplify locus-heterogeneity problems that confound the search for PCa susceptibility genes (27, 28). Further, alternative phenotypes, such as tumor aggressiveness, may be a solution for overcoming the apparent heterogeneity that has hindered the identification of PCa genes. The odds of finding clinically relevant mutations should be hypotentially greater among clinically aggressive PCa families (29).

We have done the first complete epidemiologic evaluation of cancer risks in Finnish families with two or more cases of PCa across the generations studied. The Finnish genetic population structure is relatively homogenous due to a small founder population some 100 generations ago that has expanded into 5.3 million people today, with immigration contributing little to the genetic variation (30). Moreover, unbiased epidemiologic data on familial cancer clustering can be accessed through nation-wide population and cancer registries. The aim of this study was to assess whether primary non-PCa tumors are associated with Finnish families having two or more PCa cases.

A detailed description of the Finnish PCa family data has been previously described (31). Since then, additional families from 1999 to 2008 have been found and integrated into the database using the same methods. In the present study, we used data from 202 families with two or more first-degree relatives with PCa. Genealogic information on family members was confirmed from records kept by the Finnish Population Register Centre, and medical information regarding cancer incidence was obtained from hospital records and the Finnish Cancer Registry. Known cancer syndromes were excluded by pedigree analysis. Previously, BRCA1 and BRCA2 Finnish founder mutations and multiple other genes, including CDH1, CHEK2, MLH1, NBS1, and FH have been screened and found to have only a minor role in predisposition to PCa (32-37).

The genealogic data of 202 families were collected through church and parish registries and the Finnish Population Register Centre. Complete pedigrees were constructed for each family. We traced back the parents of the index person, the earliest person being born at the end of 19th century and then traced all their descendants, through parish registries, until death or until further offspring were considered unlikely (female age, 55 y; male age, 70 y). The 844 persons (13%) who had died before the beginning of follow-up (1967) were excluded.

Because PCa is a common disease, it is likely that even when strict hereditary cancer family criteria, the so-called Carter criteria (3), is followed, a family with sporadic disease aggregation can become classified as a hereditary PCa family. To avoid this, we analyzed the clinical characteristics of the PCa cases in each of the families. All familial PCa cases were obtained from hospital records, including age of diagnosis, primary PSA value, WHO grade, Gleason score, and tumor-node-metastasis stage at diagnosis. In addition, the proportion of PCa cases in the family was calculated by dividing the number of PCa cases with the number of males in the family. Hierarchical cluster analysis by the agglomerative method and average distance was used to classify families with clinically aggressive disease into a separate cohort. Diagnostic age, primary PSA and the proportion of PCa cases in the family were standardized to equalize variables. As the incidences of PCa in these families were diagnosed between 1962 and 2006, Gleason grading was introduced in 1990s, Gleason grading was available for only 204 patients. In addition, because the criteria for Gleason grading have changed remarkably from 1990 to 2000, this variable was excluded from our analyses, relying instead on WHO grading.

The mean number of affected men with PCa was 2.9 per family (range, 2-8). The mean year of PCa diagnosis was 1992 (SD, 9.3; range 1962-2006), and the mean age of onset of PCa was 68 y (SD, 9.0; range, 43-98). The primary PSA median value was 16.0 [Quartile deviation (QD), 17; range, 0.8-11000]. The histopathologic WHO grade was 1 for 25%, 2 for 43%, and 3 for 9% (missing 23%).

Primary PSA values were available for 420 patients. To avoid the effect of outliers in hierarchical cluster analysis, the value of primary PSA was downgraded to a value of 101 for the analysis if it was initially measured to be over 100. We assumed that these extreme PSA values over 100 are uninformative. Cluster analysis was done for all variables collected (except Gleason score) and in different combinations. Primary PSA showed significant dependence on WHO grading, tumor size, and metastasis (P_{Kruskall-Wallis} < 0.0005 for all except WHO grading P_Mann-Whitney < 0.0005). Overall, the analyses showed that primary PSA alone was sufficient to well group the families compared with the use of all the variables or different combinations of the variables.

Using the collected clinical data, patients were sorted into one of two groups, group 1 (clinically nonaggressive PCa) or group 2 (clinically aggressive PCa). Fifty-nine of the 202 families were classified as a family with clinically aggressive PCa and 143 as nonaggressive PCa. The clinically aggressive PCa men had a primary PSA median value of 230 (QD, 280; range, 76-11,000) compared with 13.3 (QD, 16; range, 1.0-70) of the clinically nonaggressive men. When using the downgraded PSA values in clinically aggressive families the median value was 101 (QD, 101; range, 80-101) and 13.1 (QD, 16.2; range, 0.8-70) in clinically nonaggressive families, respectively. In addition, 71% of the clinically aggressive PCa men had metastases at the time of diagnosis as this was only 10% in clinically nonaggressive PCa men. Eighty-three percent of the clinically aggressive PCa men had extraprostatic tumor growth compared with only 34% in clinically nonaggressive PCa men. Consequently, with clinically aggressive disease, a histologically poorly differentiated and/or more advanced disease is meant showing in either a high tumor grade or stage (38, 39).

Follow-up analysis for cancer incidence among family members was done using medical information obtained by personal identity code–based automatic record linkage from the Finnish Cancer Registry. The follow-up period was from January 1, 1967 to December 31, 2004. However, for index patients, the follow-up started from their first PCa diagnosis. For the parents of the index patient, the follow-up started from the birth of the index. For the other cohort members, the follow-up started from their date of birth, if this was later than January 1, 1967. If the person had emigrated or deceased before December 31, 2004, the calculation of person-years ended at that earlier date.

When calculating family data and trying to avoid the bias inherent in defining a PCa-family, the index person should be the first PCa male from any given family recruited into the study. However, due to the large amount of variability in this information or because the information was simply not available, we calculated familial SIRs in two ways. In the single-index data method (method 1), the follow-up to the date of PCa diagnosis was excluded for one PCa case, but all other family members were followed-up from the beginning. In the multiple index data method (method 2), all known PCa males were considered as index persons and follow-up data before PCa diagnosis was excluded from the analysis. The standardized incidence ratios (SIR) calculated by these two methods provide the lowest and highest possible estimate of the true relative risk for PCa and for overall cancer risk. The number of people analyzed using single index data were 5,523 (169,700 person-years) and 5,504 (161,500 person-years) in the multiple index data (Table 1).

Expected numbers of malignancies is based on person-years at risk, gender, age, and calendar-period of the specific incidence rates in the general population. SIRs were calculated by dividing the observed numbers of malignancies by the expected numbers. Exact 95% confidence intervals (CI) were defined, assuming that the numbers of the observed cases followed a Poisson distribution. The SIRs were calculated separately for clinically aggressive and clinically nonaggressive families using method 1 and 2, respectively, as described above.

Cancers belonging to hereditary cancer syndromes tend to have an earlier age of onset compared with sporadic cancers cases in a population; for example, the mean age of onset of breast cancer for BRCA1 carriers is 35 y (40), so for breast cancer, we looked at the cancer risk for women under the age of 40 y.

There were 497 malignancies among 3,902 members of the clinically nonaggressive PCa families. Using analysis method 1, 364 malignancies were observed among males, 211 cases were expected, yielding a SIR of 1.7 (95% CI, 1.6-1.9; Tables 2 and 3). Using analysis method 2, the number of observed malignancies was 151, whereas 180 cases were expected, yielding a SIR of 0.8 (95% CI, 0.7-1.0). The difference in values is mainly due to the different numbers of PCa cases in analysis. Among females, there were 133 observed malignancies compared with 140 expected cases (SIR, 1.0; 95% CI, 0.8-1.1; Table 4). Increased risk for gastric cancer among women in clinically nonaggressive PCa families was statistically significantly (SIR, 1.9; 95% CI, 1.0-3.2; Table 4). Among men, there was less lung cancer than expected (for method 1: SIR, 0.4; 95% CI, 0.2-0.6; and for method 2: SIR, 0.5; 95% CI, 0.4-1.1; Tables 2 and 3).

We analyzed 1,621 members of 59 clinically aggressive PCa families using analysis method 1. We observed 188 malignancies (52,849 person-years) among males, 97 malignancies were expected, yielding a SIR of 1.9 (95% CI, 1.7-2.2). Using analysis method 2 to analyze 1,619 members with 50,143 person-years, we observed 83 malignancies, 79 malignancies were expected, yielding a SIR of 1.1 (95% CI, 0.8-1.3). Among females, there were 72 observed cancer cases compared with 68.6 expected cases (SIR, 1.1; 95% CI, 0.8-1.3; Tables 1-4). Analyzing males and females separately, males had more small intestine (observed, 2; SIR, 5.9; 95% CI, 0.7-21.2; method 1: observed, 2: SIR, 6.9; 95% CI, 0.8-24.9) and gallbladder cancer (observed, 3; SIR, 4.1; 95% CI, 0.8-11.8; method 1: observed, 3; SIR, 5.1; 95%, CI 1.0-14.9), whereas females had more liver cancer (observed, 3; SIR, 4.8; 95% CI, 1.0-14.0) and multiple myeloma (observed, 3; SIR, 3.5; 95% CI, 0.7-10.1; Table 4). Women under 40 years had slightly increased risk of breast cancer (observed, 4; SIR, 3.1; 95% CI, 0.8-7.9) and ovarian cancer (observed, 3; SIR, 3.9; 95% CI, 0.8-11.4).

The 479 cancer cases observed during the follow-up were in 425 individuals. Sixty individuals had two malignancies, and six individuals had three. Forty-one had PCa as one of the cancers. Other common cancers found included urinary bladder (9), kidney (4), pancreas (4), gastric (3), and rectal (3). In Fig. 1, we present an example of a family (257) with individuals having multiple malignancies, including two men with three malignancies and a woman with two malignancies. Table 5 lists families with individuals having both PCa and gastric cancer, and Fig. 1 shows a selected sample of a family with gastric cancer (143). Family 62 is an example of a clinically aggressive family showing both gallbladder and small intestine cancers.

The incidence of non-PCa cancers other than women's gastric cancer was not higher than in the reference population. Out of the numerous associations tested in our study, only the risks of liver cancer in women and gallbladder cancer in men were increased in clinically aggressive PCa families. There was no statistically significant difference in risk of any cancer type studied between clinically aggressive and clinically nonaggressive families.

Unlike this study, to date, there are only a few studies on the overall cancer risk in PCa families where: (a) all cancers of the relatives have been identified, (b) family pedigrees have been genealogically confirmed, and (c) family material is population based and data are registry-based. In a study by Isaacs et al. (15), PCa was suggested to be relatively site specific and only central nervous system tumors were statistically increased. In 2000, Grönberg et al. (17) published a study on 62 hereditary PCa families, which found a significant aggregation of PCa together with breast and/or gastric cancer, suggesting a common germ line mutation in a cancer susceptibility gene. Valeri et al. (41) reported a significant increase in breast cancer risk. Conflicting results were published in 2005, in an American study on 1,238 Utah hereditary prostate cancer cases, supporting the existence of heritable PCa syndromes that included other malignancies. The Utah hereditary prostate cancer families had increased risk for colon, breast, rectum, gallbladder, and kidney cancers and also had increased risk to multiple myeloma, non–Hodgkin lymphoma, and melanoma (16). Of above-mentioned studies, only the study by Albright et al. (16) was done in a population-based manner. However, only 55% to 60% of the Utah Cancer Registry records can be linked to an individual in the Utah Population Data Base genealogy (16).

The different results in the studies done here can be explained in part by the difficulty in identifying hereditary prostate cancer pedigree populations and all malignancies among relatives. In addition, differences in cancer aggregation and clinical phenotype among the studied populations could reflect different genetic backgrounds and heterogeneity of germ line mutations among the different analyzed populations.

In accordance with findings of others, we observed an increase in gastric cancers among female members of the PCa families (42, 43). In a previous population-based Finnish study (18), an increased risk for gastric cancer was only seen among male relatives of early onset PCa patients, whereas in the present analysis, the risk for gastric cancer was increased among female family members. In the study by Grönberg et al. (17), a gender difference was seen in the SIR for gastric cancer; male gastric cancer SIR was 3.7 (1.9-6.2) and female was 1.4 (0.3-4.0). E-cadherin (CDH1) had been suggested as a potential gene explaining prostate and gastric cancer association (32), but subsequent studies did not confirm the hypothesis (43). Our findings on the association of familial PCa and gastric cancer support the need for further studies to find a candidate gene.

Our results are to certain extent congruent with the study by Albright et al. (16) showing an increased risk of gallbladder cancer, breast cancers, and multiple myeloma with familial PCa. We also found weak suggestions of increased incidence for liver and small intestine cancers that have not been published previously in any family data. The risk for liver cancer has been found to be elevated among the relatives of PCa patients in population-based settings but not in PCa families (16, 23).

Malignancies in many hereditary cancer syndromes often have an early age of onset and/or a clinically aggressive form of the disease compared with sporadic malignancies of the same cancer type. Due to the heterogeneity of PCa, subgrouping patients based on clinical characteristics will enhance detection of clinically relevant genes and mutations. In comparing clinicopathologic features and progression-free survival among sporadic and familial PCa cases, Roehl et al. (44) observed no clear differences, but sibling pairs had a trend toward less favorable tumor features and progression-free survival.

In our analysis of clinically aggressive PCa families, females under the age of 40 years had (nonsignificantly) increased risk for breast cancer and ovarian cancer, supporting the epidemiologic findings of coaggregation of prostate and breast cancers (45, 46). Reported associations of PCa and breast cancer may be partially explained by the presence of recognized cancer predisposition genes; BRCA2 has been implicated in PCa cancer predisposition in some populations (47-49). Significant elevations in PCa risk were found for BRCA2 mutation carriers (24, 50) in the relatively homogenous Icelandic population. Likewise, in a report from the Breast Cancer Consortium, the risk for PCa in BRCA2 mutation carriers was RR = 4.65 (51). Although it seems that known Finnish BRCA1/2 founder mutations do not associate with PCa predisposition among Finnish PCa patients (33), an increased risk for PCa in Finnish breast cancer families carrying BRCA2 mutations has been observed (52). Other negative findings for an association between BRCA2 mutations and susceptibility to hereditary PCa in high-risk families have been reported (53). Most likely, mutations in BRCA1/2 pathway genes can cause malignancies at multiple cancer sites (51). Likewise, several studies of cancer clusters within families have reported cooccurrence of PCa with breast, ovarian and endometrial cancers suggesting a single gene or a limited number of genes could be responsible for association with cancers in hormonal tissue (44, 54, 55).

From our data, male members of PCa families have significantly less lung cancer than men in the general population. One possible explanation is that to get PCa, you have to live to a relatively healthy life-style and old age (the average age of onset is 71 years in Finland). A study by Pukkala et al. (56) revealed that PCa is most common among the males in the highest social class. It is therefore possible that families with PCa are a selected group of people from higher social class with healthier life-style such as lower prevalence of smoking.

The strengths of this study are that it is based on a large family-based data registry (202 families) with total of 5,523 family members with confirmed genealogy within the homogenous Finnish population with a follow-up time of almost 40 years. All malignancies in all family members are verified from the Finnish Cancer Registry with unique linked personal identity codes. Cancer diagnoses were also confirmed from medical records. When calculating SIRs, we compared the observed cases of malignancies to the expected cases in the whole Finnish population and standardized these values with gender, 10-year calendar periods, and age. To our knowledge, this is the first study where clinical data of all PCa cases within families has been collected, enabling separation of families into two groups based on the clinical aggressiveness of the cancers.

In summary, members of the studied 202 Finnish PCa families had no general elevation in non-PCa malignancies. However, females in both clinically nonaggressive and clinically aggressive families had elevated incidence of stomach cancer, indicating that the risk of developing other cancers seems not to be related to the clinical characteristics of familial-PCa cases. The profile of related cancers and risk genes for hereditary PCa families may differ between populations; therefore, further studies of PCa in other populations are warranted.

No potential conflicts of interest were disclosed.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.

We thank Riitta Vaalavuo and Ha Nati for their excellent assistance.