Age specific and attributable risks of familial prostate carcinoma from the family-cancer database
Article first published online: 5 SEP 2002
Copyright © 2002 American Cancer Society
Volume 95, Issue 6, pages 1346–1353, 15 September 2002
How to Cite
Hemminki, K. and Czene, K. (2002), Age specific and attributable risks of familial prostate carcinoma from the family-cancer database. Cancer, 95: 1346–1353. doi: 10.1002/cncr.10819
- Issue published online: 5 SEP 2002
- Article first published online: 5 SEP 2002
- Manuscript Accepted: 15 APR 2002
- Manuscript Revised: 11 APR 2002
- Manuscript Received: 18 FEB 2002
- David and Astrid Hagelen Foundation
- Swedish Cancer Society
- King Gustaf V Jubilee Fund
- familial risk;
- age of onset;
- recessive effect;
- heritable effect
Familial risks by proband status and age are useful for clinical counseling, and they can be used to calculate population-attributable fractions (PAFs), which show the proportion of disease that could be prevented if the cause could be removed.
The authors used the nationwide Swedish Family-Cancer Database on 10.2 million individuals and 182,104 fathers and 3710 sons with medically verified prostate carcinoma to calculate age specific familial standardized incidence ratios (SIRs) with 95% confidence intervals (95%CI) and familial PAFs for prostate carcinoma in sons ages 0–66 years.
The incidence of prostate carcinoma was doubled between the years 1961 and 1998. The familial SIRs for prostate carcinoma were 2.38 (95%CI, 2.18–2.59) for men with prostate carcinoma in the father only, 3.75 (95%CI, 2.73–4.95) for men with prostate carcinoma in a brother only, and 9.44 (95%CI, 5.76–14.03) for men with prostate carcinoma in both a father and a brother. The corresponding familial PAFs were 8.86%, 1.78%, and 0.99%, respectively, yielding a total PAF of 11.63%. Age specific risks were shown for the same proband histories. The SIR was 8.05 for prostate carcinoma before age 55 if a brother had been diagnosed before that age. If, in addition, a father was diagnosed at any age, then the SIR was 33.09.
The authors recommend that having a brother who is diagnosed with prostate carcinoma before age 55 years or having a brother and father who are diagnosed at any age are indications to screen for prostate carcinoma. The familial PAF of prostate carcinoma among a population of sons ages 0–66 years was 11.63%. Cancer 2002;95:1346–53. © 2002 American Cancer Society.
Familial prostate carcinoma has been the subject of numerous studies, including over 30 case-control studies and cohort studies that focused specifically on prostate carcinoma.1–3 In addition, patients with prostate carcinoma have been included among patients with other malignancies in large cohort studies.4–8 A population-based family study from Sweden of males ages 0–61 years showed a risk of 2.71 for men with a father who had prostate carcinoma, 4.91 for men with a brother who had prostate carcinoma, and 23.7 for men with both a father and a brother who had prostate carcinoma.7 The higher risks from a brother than from a father and the more than doubled risks among monozygotic twins compared with dizygotic twins suggest the possible involvement of recessive or X-linked genes in prostate carcinoma.1, 9–12 Estimates of the proportion of heritable prostate carcinoma range from 5% to 9%, and from 10% to 25% of patients who may have a relative with the same disease.1, 13 Twin studies have estimated the heritability for prostate carcinoma at 33–57%.10, 14, 15 The family studies and segregation analyses described above indicate that prostate carcinoma may be explained in part by a dominant inheritance pattern;16–21 locus heterogeneity indicates that many genes may be involved.22–25 There have been major efforts to identify genes that predispose to prostate carcinoma, although the results have remained inconclusive.22, 23, 25–27 The late age of onset of the disease suggests a lack of strong predisposing factors for most forms of this disease. Indeed, apart from age, androgenic hormones, and family history, no other established risk factors are known; however, the large global differences in incidence and migrant studies suggest a role for lifestyle-related environmental factors.28–34
Thus, it may be questionable whether further studies on familial prostate carcinoma are needed. Almost all previous studies were based on interview data on disease in relatives, and the reliability of such data is not very good.35, 36 By using the nationwide Swedish Family-Cancer Database with national coverage of families and patients from registered sources, we can provide nonbiased data.37, 38 No systematic screening programs for patients with prostate carcinoma have been initiated in Sweden.39 However, opportunistic screening takes place, and it may contribute to the increasing incidence in prostate carcinoma.40 To make the data comparable with the clinical situation, we defined familial risks in sons by three mutually exclusive family relations, using father only, brother only, or both as probands. Previous studies have considered proband status without concern about other affected family members. Estimates of the population-attributable proportion (PAF) of familial prostate carcinoma are not available in the literature, and these would indicate the proportion of prostate carcinoma that may be ascribed to familial causes. We used the 2001 update of the Database, which covers 10.2 million individuals and over 1 million patients with diagnosed malignancies, to answer these questions in the largest study yet on familial prostate carcinoma.
MATERIALS AND METHODS
The Swedish Family-Cancer Database, which was updated in 2001, includes persons born in Sweden after 1931 with their biologic parents, totaling over 10.2 million individuals.37, 38 The Database is organized into 3.2 million families, with parents and offspring. Data on men who were diagnosed with prostate carcinoma were retrieved from the nationwide Swedish Cancer Registry for the years 1958–1998. In the 2001 update, the number of men diagnosed with invasive carcinomas in the second generation, offspring, was 157,000. A three-digit diagnostic code according to the seventh revision of the International Classification of Diseases was used (code 177 for prostate carcinoma).
The completeness of cancer registration in the 1970s has been estimated at > 95% and is now considered to be close to 100%, and the percentage of cytologically or histologically verified diagnoses of prostate carcinoma is close to 100%.40 The Family-Cancer Database has incomplete linkage data from deceased offspring to parents, particularly among offspring who were born between 1932 and 1940 and died before 1990. Of a total of 7.0 million offspring, 216,000 died by the end of follow-up. Parental information was missing from 15,600 dead offspring who had a diagnosis of carcinoma (9.9% of all offspring carcinomas). It is unlikely that this deficit caused any appreciable effect on familial risk estimates in the current study, because the mortality among men with prostate carcinoma before age 60 years is very low.
Age-standardized incidence trends were calculated using the European standard population as a reference with data from the Swedish Cancer Registry and from the Family-Cancer Database.41 For all other calculations, only the Family-Cancer Database was used. Follow-up started at birth, immigration, or January 1, 1961, whichever came latest. Follow-up was terminated at time of the first diagnosis of carcinoma, death, emigration, or the closing date of the study (December 31, 1998). Age specific incidence rates in offspring were calculated according to 5-year diagnosis ages. The age of fathers was not limited, but sons were limited to ages 0–66 years. Some analyses were limited to fathers up to age 66 years to make them comparable in age to the offspring population. Standardized incidence ratios (SIRs) were calculated as the ratio of observed to expected numbers of men diagnosed with prostate carcinoma. SIRs always were calculated for sons using a father, or a brother, or both as probands. No more than two affected brothers were found in any family. The expected numbers were calculated from 5-year standard incidence rates that were specific for age, gender, tumor type, period (10-year bands), socioeconomic status (4 groups), and residential area (2 groups) for sons with no family history of prostate carcinoma in the Family-Cancer Database.42 The 95% confidence intervals (95%CIs) were calculated assuming a Poisson distribution.42 Risks for brothers were calculated using the cohort method, which as been described elsewhere.8 The PAF for men with a family history of prostate carcinoma was estimated as follows: proportion of familial cases × (familial SIR-1)/familial SIR.43, 44 For brothers, the SIRs was calculated only in families with two or more brothers, as described previously;8 the proportion was calculated by taking the number of affected men in families with two affected members and dividing this by the number of affected men in families with at least one affected brother. The calculation of SIRs for men with prostate carcinoma in all families or in families with two or more brothers yielded essentially similar results. The PAF was calculated separately for each mutually exclusive family history.
The age-standardized incidence of prostate carcinoma among all Swedish men is shown in Figure 1 for the period 1961–1998. Separate graphs are shown for data from the Swedish Cancer Registry and the Family-Cancer Database. There was good agreement between these sources, and they showed a doubling of the incidence of prostate carcinoma during the 38-year period, with the steepest increase toward the end of the period, reaching a rate of > 100 men per 100,000 person-years.
In the Database, there were 182,104 fathers and 3710 offspring with prostate carcinoma. There were 505 affected father-son pairs with only 1 affected son, 22 brother pairs without an affected father, and 10 triplets of an affected father and 2 affected sons (Table 1). The SIRs are shown for sons according to their diagnostic age in these three proband groups. For men with only an affected father, the SIR declined modestly from 2.78 among sons age 50–54 years to 2.22 among sons age 60–66 years, with an overall SIR of 2.38. The SIRs for men with only an affected brother were higher, yielding an overall SIR of 3.75. For men with both an affected father and an affected brother, the overall SIR was very high at 9.44, and, in the youngest age group, it was 33.74. The familial proportions were 15.28%, 2.43%, and 1.10% for men with an affected father, an affected brother, or both an affected father and an affected brother, respectively. The PAFs derived from SIRs and proportions were 8.86%, 1.78%, and 0.99% for the three proband groups, respectively. The sum of PAF values is shown in the far right column of Table 1. The overall PAF was 11.63%. The PAF was 1.56% for sons with fathers who were diagnosed before age 67 years.
|Son's age at diagnosis (yrs)||O||SIR||95%CI||Proportion (%)||PAF (%)||PAF A+B+C|
|Father only (PAF A)|
|Brother only (PAF B)|
|Father and brother (PAF C)|
The data on SIRs are plotted in Figure 2 on a logarithmic scale. In an additional graph, data are shown for sons with fathers who were diagnosed before age 67 years to make the data fully comparable with data from the men who had affected brothers, because both brothers were diagnosed before age 67 years. The truncation of the paternal age at diagnosis to 66 years produced a graph that was between brother-brother and father-son graphs. The overall SIR in this group was 2.81.
The dependence of the son's age specific risk on the age of onset of the father's prostate carcinoma in families with only one affected son is shown in Table 2. The SIRs were highest in the group of men with the youngest age at the time of onset: 8.08 among sons age 40–49 years by fathers age 50–59 years and 5.18 among sons age 50–54 years by fathers age 50–59 years. In some of the oldest groups, the SIRs were < 2.00. In Table 3, the risk for the second brother is shown according to age group. When both brothers were diagnosed before age 55 years, the SIR was 8.05. The SIR was about 5.00 when the diagnostic age was < 55 years or when the brother was diagnosed before that age. In Table 4, the risks for sons are shown according to the history of affected fathers and brothers. Only 10 triplets were available, and no attempt was made to control for the father's age at diagnosis in the analysis. The SIR was 33 for sons who were diagnosed before age 55 years when a brother was diagnosed before that age.
|Fathers's age at diagnosis (yrs)||Son's age at diagnosis (yrs)|
|Brothers's age at diagnosis (yrs)||Age at diagnosis (yrs)|
|Brothers's age at diagnosis (yrs)||Age at diagnosis (yrs)|
The incidence of prostate carcinoma is higher in Sweden than in any other European country, although it remains somewhat lower than that among white males in the United States population.32, 41 All first-generation immigrant groups to Sweden from Europe, Asia, and Africa had lower SIRs for prostate carcinoma compared with Swedes; the lowest SIR (0.16) was among Turkish men.33 Men from the United States showed a marginally increased SIR of 1.05. The incidence of prostate carcinoma doubled during the current 38-year study period, which, in part, may be the result of opportunistic prostate carcinoma screening. The extent of screening is unknown, but no organized screening schemes have been instituted in Sweden.39 Opportunistic screening probably would be used preferentially by the well-educated population, as observed previously for patients with carcinoid tumors.45 However, the socioeconomic group professionals showed an SIR of only 1.20 (n = 644 men; 95%CI, 1.10–1.29; data not shown), suggesting that opportunistic screening still may be a minor source of new diagnoses of prostate carcinoma in Sweden.
Prostate carcinoma is a disease of older men, and the men among the second generation in the Swedish Family-Cancer Database, with the oldest age 66 years, have entered the critical period for prostate carcinoma. The total of 525 father-son pairs in the current study was almost 3 times larger than the total from the previous version of the Database, in which the oldest age was 61 years.2 In the previous version, only two brother pairs without an affected father and three brother pairs with an affected father were recorded7 compared with the current 22 and 10 pairs, respectively, indicating a large increase in statistical power in the current analysis. The available number of patients allowed analysis by specific proband status and by age at onset. For clinical purposes, it is important to define proband status exactly: The physician is seeing the brother or the son of a proband, not his first-degree relative. Unfortunately, the term first-degree relative has been used widely in discussions of the genetic epidemiology of cancer, even for prostate carcinoma, indicating a lack of familiarity with the clinical situation and an a priori disregard of genetic mechanisms, such as recessive inheritance.
Table 2 presents age specific data on the risk that a son will be affected, depending on the father's age at the time of diagnosis. If the father was diagnosed before age 60 years, then the son had a high risk of 3.55. When the father was diagnosed later, the son's risk was relatively stable at ≈ 2.5. The father's history of prostate carcinoma is an indication that the son should seek medical advice; however, a brother's diagnosis is even more worrisome for the unaffected brothers. Having a brother who was diagnosed before age 55 years was associated with a high risk of 8.05; however, if the brother was diagnosed at an older age, then there was a relatively stable risk of ≈ 3.5. The most dramatic risk, 33.09, was observed for a son before age 55 years when a brother was diagnosed before that age and a father also was affected. Even the overall SIR for this family history was 9.44. Although no definite limits have been proposed for the follow-up of individuals who are at a risk of prostate carcinoma, it appears that a risk of 8.0 definitely should initiate such action well before the diagnostic age of the proband.13, 46 Thus, having a brother who was diagnosed before age 55 years or having both a brother and a father who were diagnosed at any age should be indications for screening by means of prostate specific antigen determination and digital rectal examination.47–49
The specific proband status defined in the current study not only is clinically useful, but it also may have genetic connotations. The history of an affected father may be due to dominant heritable effects, the history of an affected brother (without an affected father) may be due to recessive or X-chromosome-linked effects, and the history of an affected father and an affected brother may be due to high-penetrant dominant effects. An alternative interpretation may be that the first two family histories show low-penetrant dominant effects. A segregation analysis of patients with prostate carcinoma has suggested an autosomal-dominant mode of inheritance, but studies have shown a wide range in the proportion of population affected.16–21 Family studies also have provided evidence for recessive or X-linked effects because of higher risks from brothers than from fathers.1, 7, 8 Also, some linkage studies have suggested loci on the X-chromosome; however, these and other suggested loci seem to be difficult to confirm, and the putative genes evade identification.22, 23, 25–27 The current data on men with a history of an affected father and brother are consistent with a high-penetrant dominant effect. A history of an affected brother (without an affected father) may indicate a recessive or X-linked effect. The SIR for men who had an affected brother was higher (3.75) compared with men who had an affected father (2.38). However, when the age of the father was restricted to age ≤ 66 years, compatible with the age of the brother, then the SIR for men with an affected father increased to 2.81. This still was below the SIR of men with an affected brother; however, due to various age truncations in the Database, we have refrained from making strong statements regarding the evidence for recessive or X-linked effects. The alternative explanation would be low-penetrant dominant effects observed from affected fathers and brothers. Such low penetrance, which probably is modified by environmental effects, together with locus heterogeneity, may be an explanation for the difficulties with gene mapping and identification.
PAFs have been useful in the quantification of population burdens of carcinogenic exposures.43, 50 PAFs also have been defined for individuals with a family history of many types of malignancies, although not for those with a family history of prostate carcinoma.12 Table 1 shows that PAFs depend on the age of the population, and the current data apply only to the population under study: fathers of any age but sons age < 67 years. Although the familial PAF was highest (8.86%) through father probands, it decreased to 1.56% when the father population was truncated to age < 67 years. This percentage was not different from the 1.78% PAF obtained for brothers who also were age < 67 years. The overall PAF by any proband status was 11.63%. This value was higher than that found for any other malignancies in the Database;12 the second highest was for female breast carcinoma, with a PAF of 7.05%.51 In light of these data, it is interesting to refer to a recent twin study in which prostate carcinoma showed the highest heritability, 42%, of all primary carcinomas.15 The result from that twin study was based on a multifactorial threshold model, the assumptions of which have been discussed elsewhere.11, 12 The PAFs calculated in the current study provided estimates of the heritable effects of prostate carcinoma when only two-generation families are studied for single-locus dominant or additive effects.11, 12 In this population of men who were relatively young for prostate carcinoma, the total PAF was 11.63%. The main reason for the large difference in these estimates may be the involvement of low-penetrant and polygenic effects, as discussed above. Low-penetrant gene effects will make familial patterns difficult to observe and may hamper clinical counseling. However, the risk figures shown in Tables 2–4 were empiric, and they should be useful in the clinical situation.
- 12Re: genetic epidemiology of cancer: interpreting family and twin studies and their implications for molecular genetic approaches. Cancer Epidemiol Biomarkers Prevent. 2002; 11: 423..
- 28IARC. Cancer: causes, occurrence and control, vol 100. Lyon: IARC, 1990.
- 29Basic pathology. Philadelphia: WB Saunders, 1997., , .
- 30World Cancer Research Fund. Food, nutrition and the prevention of cancer: a global perspective. Washington, D.C.: American Institute of Cancer Research, 1997.
- 32FerlayJ, BrayF, PisaniP, ParlinM, editors. GLOBOCAN 2000. Cancer incidence, mortality and prevalence worldwide. Lyon: IARC Press, 2001.
- 40Centre for Epidemiology. Cancer incidence in Sweden 1998. Stockholm: Centre for Epidemiology, 2000.
- 41IARC. EUCAN, cancer incidence, mortality and prevalence in the European Union. Lyon: IARC Press, 1999.
- 42Statistical methods in cancer research, vol 128. Lyon: IARC, 1994., , .
- 44Modern epidemiology, 2nd ed. Philadelphia: Lippincott-Raven, 1998., .
- 49Large-scale randomized prostate cancer screening trials: program performances in the European randomized screening for prostate cancer trial and the prostate, lung, colorectal and ovary cancer trial. Int J Cancer. 2002; 97: 237–244., , , et al.
- 50Cancer epidemiology: principles and methods. Lyon: IARC, 1999..