SEARCH

SEARCH BY CITATION

Keywords:

  • prostate cancer;
  • family history;
  • disease severity

Abstract

  1. Top of page
  2. Abstract
  3. INTRODUCTION
  4. PATIENTS AND METHODS
  5. RESULTS
  6. DISCUSSION
  7. REFERENCES

OBJECTIVE

To estimate the risk of prostate and other types of cancer among relatives of Icelandic men diagnosed with prostate cancer over a 5-year period.

PATIENTS AND METHODS

The risk ratio (RR) was used to estimate the risk among relatives of 371 patients with prostate cancer, all of whom lived in Iceland and were diagnosed when alive over a 5-year interval (1983–7). Information on cancer incidence was obtained from the population-based Icelandic Cancer Registry, and information on families from a comprehensive genealogical database covering the population of Iceland.

RESULTS

First-degree male relatives were at a 1.7-fold greater age-adjusted risk of prostate cancer (1832 men; 95% confidence interval, CI, 1.28–2.34). The risk was independent of proband's age at diagnosis. First-degree male relatives of patients who died from prostate cancer were at a statistically significantly greater risk of the disease (784 men; RR 2.17; 95% CI 1.34–3.53) and relatives of patients with incidental disease (T1a) were at a greater risk but not statistically significant so (261; RR 1.86; 95% CI 0.75–4.58). Female first-degree relatives were not at greater risk of breast cancer. The risk of kidney cancer was higher in first- and second-degree female relatives, with an RR (n, CI) of 2.50 (1780, 1.10–5.66) and 2.67 (5534, 1.04–6.81), respectively. The risk of kidney cancer was not statistically significantly greater in male relatives.

CONCLUSION

Family history is a risk factor for prostate cancer in Icelandic men. The risk is potentially higher for relatives of patients who die from the disease. Female relatives are not at greater risk of breast cancer but they may be at greater risk of kidney cancer.


Abbreviations
FDMR

first-degree male relative

RR

risk ratio

HPC

hereditary prostate cancer.

INTRODUCTION

  1. Top of page
  2. Abstract
  3. INTRODUCTION
  4. PATIENTS AND METHODS
  5. RESULTS
  6. DISCUSSION
  7. REFERENCES

Prostate cancer is the most common cancer in Icelandic men; the age-standardized (world standard) incidence per 100 000 person-years was 46.9 during 1981–5 and rose to 59.8 during 1986–90 [1], and is higher than in other European countries [2]. During the interval 1993–7 the incidence was 75.9 in Iceland, ≈ 60 in Finland, Norway and Sweden, and 30.0 in Denmark [3]. The incidence of prostate cancer has risen dramatically in Iceland during the past 30 years following the introduction of TURP in the early 1980s and better diagnostic methods, e.g. PSA, biopsy guns and TRUS in the late 1990s.

The risk factors most strongly associated with the disease are family history, ethnicity, and diet [4,5]. Prostate cancer is probably a polygenic disease, as several possible loci have been reported [6–9], primarily involving chromosome 1 and the X chromosome. There is also evidence for loci on chromosome 8 involved in prostate carcinogenesis [10]. No single locus seems to explain a large part of familial prostate cancer, suggesting that locus heterogeneity is an important variable in understanding the cause of the disease [11].

An increased risk of prostate cancer among relatives of patients with prostate cancer has been reported in epidemiological studies [12–14]. The presence of a population-based cancer registry and a genealogical database covering all the Icelandic nation permits an unbiased estimate of familial risk. In this study, the cancer risk in relatives of patients with prostate cancer in Iceland, diagnosed when alive over a 5-year period, was estimated.

PATIENTS AND METHODS

  1. Top of page
  2. Abstract
  3. INTRODUCTION
  4. PATIENTS AND METHODS
  5. RESULTS
  6. DISCUSSION
  7. REFERENCES

All Icelandic men diagnosed with prostate cancer during a 5-year interval (1983–7) were included in the study. Information on cancer in relatives was obtained by record linkage of the population-based genealogical and cancer registries. The Icelandic Cancer Registry has been in operation since 1954, and covers all cancers diagnosed in the country [15]. The information on cancer incidence in Iceland is therefore complete since 1955 for cancer at all sites. Stage, treatment and outcome information were obtained from medical records. Information on each proband's family up to and including third-degree relatives, with first-degree male relatives (FDMRs) including father, brothers and sons, was obtained from the computerized records of the Genealogical Committee of the University of Iceland.

The calendar year from 1955 up to and including 1999 and patient age were used as stratification variables when calculating person-years. Both variables were defined by 5-year strata. If, for each cancer, c ij denotes the incidence (per 100 000) based on the population rates in Iceland, nij denotes the person-years, oij denotes the number of observed cases, and

  • image

for calendar stratum i (i = 1, . . . ,9) and age stratum j (j = 1, . . . ,20), the expected number of cases was estimated by:

  • image

The risk of cancer was estimated as the ratio between the observed and expected number of cases (standardized mortality ratio, risk ratio, RR), which compares the observed number of cases in a cohort with an expected number obtained by applying calendar- and age-specific standard rates to the cohort age structure [16]. The expected number of cases was calculated for only those age- and calendar periods with an observed number of cases. If N denotes the total number of relatives in question, and O denotes the observed number of cases, then

  • image

is an estimate of the standard error of ln(RR). The 100 (1 –α)% CI of RR is then estimated by

  • image

where z1−α/2 denotes the 100 (1 − α/2)% fractile of the standard normal distribution. The risk among spouses can be considered as representative of the risk in the general population. Significance tests with P < 0.05 were considered statistically significant. Consequently, one test of 20 would be classified as significant by chance, when no real association exists.

RESULTS

  1. Top of page
  2. Abstract
  3. INTRODUCTION
  4. PATIENTS AND METHODS
  5. RESULTS
  6. DISCUSSION
  7. REFERENCES

In all, 433 men were diagnosed with prostate cancer in 1983–7, including 28 incidentally diagnosed at autopsy and six by death certificates only. Eighteen individuals had other histopathology than adenocarcinoma of the prostate. The stage was unknown in 10. This left 371 men with adenocarcinoma of the prostate diagnosed alive during the 5-year interval (Table 1) who formed the proband group in the study. The proband group included two father-son pairs, four families each had two brothers as probands, and in one family there were three brothers who were all probands (Table 2).

Table 1.  Grouping of probands by stage
StageNAgeDOD, %F/U
  1. N, number of individuals; age, mean age at diagnosis; DOD, dead of disease; F/U, follow-up (months).

T1aN0–XM0–X 5071 498
T1b, T2a–bN0–XM0–X167742182
T3N0–XM0–X 32775353
T4N0–XM0–X, TX−4 N1 and/or M1122748933
Total371744466
Table 2.  The number of families with at least two first-degree relatives with prostate cancer by relatedness between relatives and proband
RelationshipNo. of families
Proband + proband's brothers15
Proband's father + proband + proband's brother(s) 1
Proband + proband's brother(s) + proband's son(s) 1
Proband + proband's sons 2
Total19

First- and third-degree male relatives were at a significantly greater risk of prostate cancer (Table 3), but the risk decreased with increased distance in relatedness from the proband. The probands’ mean (range) age at diagnosis of adenocarcinoma of the prostate was 74.4 (53–94) years. The first quartile (at 68 years) of the distribution of the age at diagnosis of prostate cancer in the probands was chosen as a criterion for distinguishing two groups; one included male relatives of 92 probands diagnosed at ≤ 68 years old and the other included all other male relatives. The RR of prostate cancer was slightly but not significantly higher for relatives in the former group (Table 3). The RR shows a similar pattern for the second-degree relatives.

Table 3.  The RR of prostate cancer among various groups
GroupNObsExpObs/Exp95% CI
  1. N, number of individuals; Obs, observed number of cases; Exp, expected number of cases.

Male relatives and spouses
First degree1832109 63.41.72(1.28–2.34)
Second degree5604 85 67.61.25(0.91–1.72)
Third degree10649241197.81.22(1.01–1.47)
Spouses4816108111.00.97(0.75–1.26)
Relatives and spouses by the age at diagnosis of proband (<68, 92 families; > 68, 279 families)
<68 years
First degree483 29 12.42.42(1.25–4.68)
Second degree1544 32 18.51.68(0.96–2.96)
Third degree2504 58 51.01.14(0.78–1.65)
Spouses1234 32 21.91.45(0.85–2.49)
> 68 years
First degree1349 80 42.21.90(1.32–2.75)
Second degree4060 53 37.71.39(0.92–2.11)
Third degree8145183128.41.43(1.14–1.79)
Spouses3582 76 79.20.96(0.70–1.31)
First-degree relatives by disease severity of the proband
Incidental (T1a)261 13  7.21.86(0.75–4.58)
Lethal784 50 23.42.17(1.34–3.53)
RR of breast and kidney cancer among female relatives and spouses
Breast
Mother of proband355  8  3.112.67(0.71–9.97)
Sister of proband884 32 43.10.74(0.48–1.16)
Daughter of proband545 16  11.951.33(0.64–2.79)
Sum1784 56 67.50.83(0.59–1.18)
Second degree5534101100.341.01(0.77–1.33)
Third degree9838190194.380.98(0.80–1.19)
Kidney
First degree1780 20  7.672.50(1.10–5.66)
Second degree5534 16  5.932.67(1.04–6.81)
Third degree9838 43 30.021.43(0.90–2.28)
Spouses5165 20 18.381.11(0.59–2.10)

Nineteen families had three or more FDMRs diagnosed with prostate cancer (proband included), of which 16 had three FDMRs, two had four FDMRs and one had five FDMRs diagnosed with prostate cancer. All of the FDMRs in these families were aged > 60 years when diagnosed with prostate cancer. In most of these families only brothers were diagnosed with prostate cancer (Table 2).

Two groups were formed based on disease severity in the proband; in one were 48 probands who had untreated incidental disease (T1a) that did not progress, and in the other were 159 probands who died from the disease. FDMRs of probands who died from the disease were at a statistically significantly greater risk of prostate cancer (RR 2.17, 95% CI 1.34–3.53), but FDMRs of probands in the former group were not (1.86, 0.75–4.58; Table 3).

Female first-degree relatives (FFDRs) were at no greater risk of breast cancer (Table 3). The risk of cancer in the bladder, cervix or uterus was also not significantly greater in female relatives, but the risk of kidney cancer was significantly higher in female first- and second-degree relatives (Table 3), while in males it was not, at 1.67 (0.73–3.80) and 1.64 (0.85–3.19) for first- and second-degree, respectively. Male relatives were not at a significantly increased risk of bladder or stomach cancer. Spouses (of probands or relatives) were not at a statistically significantly greater risk of any of the types of cancer investigated.

DISCUSSION

  1. Top of page
  2. Abstract
  3. INTRODUCTION
  4. PATIENTS AND METHODS
  5. RESULTS
  6. DISCUSSION
  7. REFERENCES

Family history is a well known risk factor for prostate cancer [13,14], but the clustering of prostate cancer cases in families alone does not allow any inference of the factors potentially involved. An inference of a genetic component might be justifiable if the clustering showed a pattern consistent with Mendelian inheritance and higher risk among relatives of patients diagnosed at an early age [17]. Excess clustering in families could be a result of several factors, including common genes, shared environment, enhanced awareness among family members, or chance. The present data suggest increased clustering of cases among first-and third-degree relatives of patients with prostate cancer (Table 3). The RR of third-degree relatives is lower than for first- and second-degree relatives, and the significant outcome among third-degree relatives may be explained by their large number.

Excessive clustering of prostate cancer in families is not prominent in this sample, despite an increased risk of the disease among FDMRs (Table 2). Strong family clustering is a major criterion of hereditary prostate cancer (HPC), which is proposed to be defined in families with a minimum of three first-degree relatives with prostate carcinoma or in families with a minimum of two first-degree relatives diagnosed with prostate cancer at < 55 years old or where prostate cancer occurs in three successive generations [18]. This criterion is based on the hypothesis that genes predisposing to cancer are most likely segregating in families with many affected family members and early onset of the disease. However, this definition is without reference to family size, or the number of years family members have been at risk of acquiring the disease. A definition of the ‘density’ or ‘concentration’ of cases in each family, instead of only the number of cases, is lacking. Such a definition should take into account the number of person-years at risk lived by family members. A suitable next step would be to characterize families with a high ‘density’ of prostate cancer for genetic analysis, and to estimate their frequency in the population.

However, the risk in relatives of men with prostate cancer appears to be independent of the age of the proband (Table 3). Grönberg et al.[19] reported that in families potentially linked to the HPC1 locus, the mean age at onset of prostate cancer appeared to be lower than in families potentially unlinked to the HPC1 locus. Thus, genes predisposing to early onset of the disease may not be important risk factors in the present sample, or that the risk in relatives of patients with prostate cancer diagnosed at an earlier age (e.g. <55 years) may need to be considered. The range of the age at diagnosis of prostate cancer in the probands (53–94 years) in the present sample precludes conditioning on a very low age at diagnosis.

The reason why we did not detect a greater risk in families with younger probands might be that during the years selected (1983–7) the mean age at diagnosis was relatively high (74.4 years, sem 0.46). Only four (1.1%) probands were diagnosed at ≤ 55 years old. Furthermore, during this period the diagnostic and treatment approach was relatively conservative and patient awareness of prostate cancer was much less than at present. This ‘conservative attitude’ might have delayed diagnosis and moved patients from younger to older groups. The RR might thus be influenced by this and the effect of early-onset disease reduced in the cohort.

Carter et al.[20] reported that early age at diagnosis of prostate cancer in the proband is an important determinant of the risk of prostate cancer in relatives. However, the mean age at onset of prostate cancer in the probands in that study was 59.3 years, which is 15 years lower than the mean age at diagnosis in the probands in the present sample (P << 0.001, t-test for the difference of means). Also, the highest quintile of age at onset of prostate cancer in the probands in that study [20] occurred at 65 years, 3 years lower than the lowest quartile of age at diagnosis in the present probands. Thus > 80% of the probands in that study acquired prostate cancer at the same age or younger than the quarter of the present probands who were diagnosed with prostate cancer at the youngest age. Both studies together indicate that a statistically significant effect of the proband's age at diagnosis may be detected when considering families of probands diagnosed when young (e.g. <55 years), while no effects will be detected when only considering families of probands diagnosed when older. Thus, despite a high incidence of prostate cancer in Iceland, significant genetic heterogeneity and absence of strong predisposing prostate-cancer genes in Iceland is supported by this evidence.

Prostate cancer is a synonym for a spectrum of diseases, ranging from an incidental asymptomatic state to a lethal disease [21]. The present data suggest that FDMRs of patients with a severe form of the disease may be at a higher risk than relatives of patients with incidental prostate cancer (Table 3). However, the insignificance of the RR for the latter group may be a result of small sample size. Also, death from prostate cancer is the most visible sign of the disease in a given family. Relatives of men who dies from the disease might be more likely to seek medical attention and screening than relatives of men with milder forms of prostate cancer.

Greater aggressiveness of familial prostate cancer than in sporadic prostate cancer has been suggested [22,23]. Kupelian et al.[23] speculate that a single gene might be involved in both predisposition to prostate cancer and the pathophysiology of the disease. An excess of prostate cancer incidence among relatives of patients with a severe disease is an indication of genetic susceptibility. Epidemiological studies investigating the association between family history and biological aggressiveness of prostate cancer may thus help in elucidating the genetic contribution to the disease.

A greater risk of prostate cancer among relatives of patients with breast cancer has been proposed [24–26], and vice versa, i.e. an increased risk of breast cancer among female relatives in families with a history of prostate cancer [13,27], and in families with a history of both breast and prostate carcinoma [28]. A recent study in Iceland suggests that a single BRCA2 mutation (the Icelandic founder mutation 999del5 on chromosome 13) explains all the increased risk of prostate cancer in first-and second-degree relatives of patients with breast cancer [29]. First- and second-degree relatives of BRCA2 positive breast cancer probands have a prostate cancer risk of 4.79 (3.27–6.32) and 2.21  (1.57–3.02), respectively, while the risk is 1.08 (0.89–1.29) and 1.15 (0.98–1.32) for first- and second-degree relatives of BRCA2-negative breast cancer probands [29]. Sigurdsson et al.[30] sought the BRCA2 mutation in patients with prostate cancer in families of BRCA2-positive breast cancer patients. Of 26 with prostate cancer 12 were analysed, of whom eight had the mutation, and all eight died from the disease [30]. Edwards et al.[31] found protein-truncating mutations in the BRCA2 gene in six of 263 men diagnosed with prostate cancer when ≤ 55 years old, which suggests that 2% of men with early-onset prostate cancer harbour germline mutations in the BRCA2 gene. These are similar findings to those of Johannesdottir et al.[32], who found that 2.7% of men diagnosed with prostate cancer before age 65 years were BRCA2 mutation carriers, compared with an estimated frequency of 0.5% in the general population. However, four of the cases with mutations in the earlier study [31] did not have a family history of breast or ovarian cancer, and only one had a family history of prostate cancer.

In the light of these results it is unlikely that the frequency of BRCA2 mutations would be much higher among the present probands than in the general population, given the distribution of the age at diagnosis of prostate cancer of the probands in our study. Also, the findings of Edwards et al. indicate that BRCA2 protein-truncating mutations would not be found at a high frequency in families with a history of prostate cancer, which underscores the genetic complexity of the disease. Nevertheless, these studies indicate that BRCA2 is a strong risk factor for early-onset prostate cancer, and for prostate cancer in relatives with BRCA2-positive breast cancer. However, the present results indicate that female relatives of men with prostate cancer are not at greater risk of breast cancer (Table 3).

As the BRCA2 status of the probands was unknown we could not control for the presence/absence of the mutation, which could give valuable information. Again, given the distribution of the age at diagnosis of prostate cancer in the present probands, we could not estimate the risk of breast cancer among female relatives of early-onset (e.g. <55 years) prostate cancer probands. The evidence suggests therefore that males in breast cancer families are at greater risk of prostate cancer, but females in prostate cancer families are not at greater risk of breast cancer.

Among first- and second-degree female relatives there was a significantly greater risk of kidney cancer (Table 3). Grönberg et al.[33] also reported a greater risk of kidney cancer in female relatives in families with hereditary prostate carcinoma. In a study from Finland [34], the risk of kidney cancer was slightly, but not statistically significantly, greater among first-degree relatives. Five of the present families had two first-degree relatives with kidney cancer, the most first-degree relatives with the disease in one family.

The risk of cervix, uterus, ovary, bladder or stomach cancer was not significantly greater in female relatives. In particular there was no evidence of a greater risk of stomach cancer among FDMRs, in contrast to Grönberg et al.[33], who reported a significant RR of 3.65 (1.94–6.24). However, in the present study a minority of the families could be classified as having hereditary prostate carcinoma by the criterion used by Grönberg et al.[33]. Matikainen et al.[34] also reported a high risk of stomach cancer among FDMRs of patients with prostate cancer diagnosed at ≤ 55 years old (RR 5.0, 95% CI 2.8–8.2). Different study populations and/or heterogeneity of stomach and prostate cancer-predisposing genes may explain the different findings in Iceland, Sweden and Finland.

The present spouses considered include those of probands and of relatives, and may therefore be considered as representing the general population. The absence of a statistically significantly greater risk of cancer among the spouses precludes shared environmental risk factors for cancer. It also suggests that heritable factors contribute to the familial incidence observed, rather than environmental factors.

In summary, family history is a risk factor for prostate cancer in Icelandic men. Relatives of patients with a severe form of the disease are at significantly greater risk of prostate cancer, supporting genetic susceptibility as a cause, but the few families with excessive clustering of the disease suggests genetic heterogeneity. Female relatives are not at greater risk of breast cancer; there was a greater risk of kidney cancer only in female relatives, despite little observed familial clustering of kidney cancer.

REFERENCES

  1. Top of page
  2. Abstract
  3. INTRODUCTION
  4. PATIENTS AND METHODS
  5. RESULTS
  6. DISCUSSION
  7. REFERENCES