Family history of prostate cancer and prostate cancer risk in the Alpha-Tocopherol, Beta-Carotene Cancer Prevention (ATBC) Study

Abstract

Prostate cancer family history has been associated with increased risk of the malignancy. Most prior studies have been retrospective and subject to recall bias, however, and data evaluating interactions with other important risk factors are limited. We examined the relationship between a family history of prostate cancer and prostate cancer risk in relation to body size, micronutrients and other exposures in the Alpha-Tocopherol, Beta-Carotene Cancer Prevention (ATBC) Study cohort of Finnish male smokers. Family history of cancer information was self-reported once during the study in 1991, and anthropometry was measured by trained personnel. Among 19,652 men with complete data, 1,111 incident cases were identified during up to 12.3 years of follow-up. A first-degree family history of prostate cancer was associated with an overall relative risk (RR) of 1.91 (95% CI = 1.49–2.47) and a RR of 4.16 (95% CI = 2.67–6.49) for advanced disease (stage ≥ 3), adjusted for age and trial intervention. Our data also suggest that to some degree, height, body mass index, and serum α-tocopherol and β-carotene modify the family history and prostate cancer association, although the interactions were not statistically significant. Supplementation with vitamin E or β-carotene did not modify the family history-prostate cancer association. This study provides additional evidence that family history is a significant risk factor for prostate cancer. Published 2008 Wiley-Liss, Inc.

Prostate cancer is the most common non-skin cancer and the second leading cause of cancer death among men in the United States.1 Although little is known about the underlying causes of prostate cancer, family history of the disease has been consistently associated with increased risk. A recent meta-analysis estimated that a family history of prostate cancer in first-degree relatives was associated with a relative risk of 2.5 [3.5 for risk of early onset (<60 years old) disease].2 However, most existing studies are limited to retrospective case-control studies3–7 and some existing cohort studies are based on family history probands.8–10 Additionally, 2 prospective studies reported that associations between body mass index and prostate cancer were more pronounced among men with a positive family history,11, 12 but data evaluating interactions with other important prostate cancer risk factors are limited.

Family history is an important public health screening tool for identification of high-risk groups, particularly in the post-genomic era with the discovery of inherited causes of many diseases including cancers.13 Since prostate cancer is a result of complex genetic and environmental interactions,14 family history can be a personalized genomic tool that captures many of these interactions13 and is essential to individualized cancer prevention strategies. Thus, accurate risk assessment of a family history as well as evaluation of interaction with known or potential risk factors of prostate cancers are particularly important and merit investigation using multiple study designs.

In this paper, we hypothesized that a family history of prostate cancer in first-degree relatives may predispose men to prostate cancer and may enhance risk conferred by other known or potential prostate cancer risk factors. Using a prospective design, we investigated the association of family history, assessed once during the study in 1991, with prostate cancer risk in the Alpha-Tocopherol, Beta-Carotene Cancer Prevention (ATBC) Study cohort. We also assessed the modifying influences of known or potential prostate cancer risk factors on risk relationships between family history and prostate cancer risk.

Material and methods

Study population

The ATBC Study was a placebo-controlled, double-blinded primary prevention trial with a 2 × 2 factorial design that tested the hypothesis of whether α-tocopherol or β-carotene supplementation would reduce the incidence of lung or other cancers in male smokers. Study rationale, design, and methods have been previously described.15 Between 1985 and 1988, 29,133 eligible men aged 50–69 years in southwestern Finland who smoked at least 5 cigarettes per day were randomized to receive supplements (50 mg/day of α-tocopheryl acetate, 20 mg/day of β-carotene or both) or a placebo. Criteria for exclusion from the study included a history of malignancy other than nonmelanoma cancer of the skin or carcinoma in situ, severe angina upon exertion, chronic renal insufficiency, liver cirrhosis, chronic alcoholism, receipt of anticoagulant therapy, other medical problems which might limit long-term participation, and current use of supplements containing vitamin E (>20 mg/day), vitamin A (>20,000 IU/day), or β-carotene (>6 mg/day). The trial ended on April 30, 1993, and follow-up continued after randomization for the present study until death or through April 30, 2003. The study was approved by the institutional review boards of both the U.S. National Cancer Institute and the Finnish National Public Health Institute. All study participants provided written informed consent prior to the study's initiation.

Case ascertainment

Nearly all cases were ascertained by Finish Cancer Registry during the study follow-up, with additional centralized record reviews through April 1999. These reviews and other data provided strong support for the robustness of the Finnish Cancer Registry data regarding our cohort.16 By design for the postintervention analysis,17 the medical records of each prostate cancer case diagnosed through April 1999 were centrally reviewed by 2 study oncologists to confirm the diagnosis and stage. In addition, the histopathologic and cytologic specimens of the cases were reviewed by study pathologists during the trial and post-trial periods (56% of total cases). Prostate cancer cases diagnosed after April 1999 had only Finnish Cancer Registry confirmation of the diagnosis, histology, and diagnosis date. Stage of prostate cancer as defined by the American Joint Committee on Cancer 199218 was available for cases diagnosed through April 1999 (n = 571). Stage III and IV cases were considered advanced disease.

Study data

At their baseline visit, the study participants completed baseline questionnaires on general background characteristics including medical, smoking, and lifestyle histories. Trained study staff using standard methods measured height and weight. Body mass index (BMI) was calculated from measured height and weight (kg/m²). Physical activity was categorized based on combined occupational and leisure time activity with those sedentary in both activity types serving as the lowest level. Fasting blood samples were collected and stored at −70°C, and serum α-tocopherol, β-carotene, and retinol concentrations were determined by reverse-phase high-performance liquid chromatography, as described previously.19 The between-run coefficients of variation were 2.2 for α-tocopherol, 3.6% for β-carotene, and 2.4 for retinol. Family history was assessed by a self-administered questionnaire completed in 1991 with detailed information queried for diagnoses of 9 common cancers (i.e., prostate, breast, lung, colon, rectal, bladder, stomach, pancreas, skin, and other cancers) among parents and siblings. Participants were also specifically asked how many brothers and sisters they had.

Statistical analysis

Person-time was calculated from the return date of the family history questionnaire (1991) to the date of prostate cancer diagnosis, death, or April 30, 2003, whichever came first. Only subjects with complete data were included in the analyses. Participants were ineligible for this study of family history if they did not complete the family history questionnaire (n = 8,257), food frequency questionnaire (FFQ; n = 1,149) or serum vitamin measurements (n = 21). After exclusions, the analytic cohort consisted of 19,652 men with 1,111 cases. There were no significant differences in prostate cancer risk factors between the whole ATBC cohort and the analytic cohort for this study of family history (data not shown). For example, the prostate cancer risk increase per 1 year of age were 1.09 (95% CI = 1.08–1.11) and 1.09 % (95% CI = 1.08–1.10) for the whole and analytical cohorts, respectively.

Age-adjusted baseline descriptive characteristics of participants were calculated using general linear models (GLM), and p-values for the difference by family history status were also calculated. Relative risks (RR) and 95% confidence intervals (95% CI) were determined using Cox proportional hazards models. All multivariable models were adjusted for age at randomization (continuous) and trial intervention (α-tocopherol and β-carotene supplement, yes/no). Factors that did not confound the family history associations included the following: number of brothers, BMI (<25, 25.0–29.9, 30+, kg/m²), height (tertile), benign prostatic hyperplasia (yes/no), intakes of total energy (continuous), carbohydrate, protein, fat, linoleic acids, meat, total vitamin E and β-carotene (diet and supplement), lycopene, and alcohol, number of cigarettes smoked per day, number of years of smoking, physical activity (low, medium, high), urban residence (yes/no), education (elementary school higher/not higher), and marital status (yes/no). When those variables were added to individual models in a stepwise fashion, each factor produced a nonsignificant, less than 10% change in family history relative risk.

To evaluate the joint effects of family history and known and potential prostate cancer risk factors, variables reflecting the combined exposures were analyzed using a common referent category (e.g., no family history combined with the lowest category of the risk factors tested). Tertiles of anthropometry, diet and lifestyle variables were created. To test multiplicative interactions, cross-product terms were added to a fully-adjusted model. Absolute rates of prostate cancer were standardized to the age distribution of person-years experienced by all study participants, using 5-year age categories. All statistical analyses were performed using SAS software (SAS Institute, Cary, North Carolina), and statistical tests were 2-sided.

Results

The mean age at baseline of the study participants was 56.9 years (median 56, range 49–70 years). Age-adjusted characteristics of the study participants by prostate cancer family history status are shown in the Table I. Five hundred eighty-seven men, or 3.0%, had a family history of prostate cancer in first-degree relatives (i.e., father or brother). Men with a family history were slightly older, taller, and more educated, compared with men with no family history. The number of brothers was lower among those with no family history of prostate cancer. In addition, men with a family history had higher average serum retinol and a higher prevalence of benign prostatic hyperplasia. However, variables which may be related to health-conscious behavior, such as physical activity and alcohol consumption, did not significantly differ by family history status.

Table I. Age-Adjusted Baseline Characteristics of Study Participants by Reported First-Degree Family History of Prostate Cancer, Alpha-Tocopherol, Beta-Carotene Cancer Prevention Study, Finland, 1991–2003

Characteristics12	Family history of prostate cancer (mean ± SD)
	No (N = 19,065, 97%)	Yes (N = 587, 3%)	p-value
1 All values other than age were adjusted for age. 2 Dietary intakes were adjusted for energy intake; serum antioxidant biomarkers were adjusted for total cholesterol. 3 Low, sedentary in leisure and occupational activity; medium, sedentary in either leisure or occupational activity, and high, greater than sedentary activity in leisure or moderate/heavy occupational activity.
Age, years	56.9 ± 4.95	57.3 ± 5.06	0.03
Number of brothers	1.8 ± 1.8	2.2 ± 1.8	<0.0001
Number of sisters	1.7 ± 1.7	1.9 ± 1.7	0.29
Height, cm	173.8 ± 6.1	174.7 ± 6.1	0.001
Weight, kg	79.6 ± 12.5	80.0 ± 12.5	0.41
Body mass index, kg/m^2	26.3 ± 3.7	26.2 ± 3.7	0.47
Daily dietary intake
Total energy, kcal	2825.8 ± 773.1	2810.2 ± 795.1	0.45
Carbohydrate, g	307.5 ± 42.8	306.1 ± 42.8	0.38
Protein, g	103.6 ± 13.6	104.3 ± 13.6	0.28
Saturated fat, g	52.7 ± 13.3	52.9 ± 13.3	0.73
Linoleic acid, g	9.4 ± 6.0	9.4 ± 6.0	0.99
Total vitamin E, mg	14.6 ± 15.8	15.2 ± 15.9	0.39
α –tocopherol, mg	10.5 ± 4.3	10.5 ± 4.3	0.78
Lycopene, μg	820.5 ± 719.1	824.2 ± 719.2	0.99
Selenium, μg	90.0 ± 15.2	90.8 ± 15.2	0.24
Vitamin C, mg	100.5 ± 42.8	99.2 ± 42.8	0.49
Biomarker serum concentration
α –tocopherol, mg/L	12.1 ± 2.8	12.0 ± 2.8	0.73
β-carotene, μg/L	222.5 ± 190.7	222.4 ± 190.9	0.86
Retinol, μg/L	589.9 ± 122.6	604.8 ± 122.8	0.003
Total cholesterol, mmol/L	6.3 ± 1.15	6.2 ± 1.09	0.32
HDL cholesterol, mmol/L	1.2 ± 0.30	1.2 ± 0.30	0.90
Alcohol use, g	17.0 ± 19.6	16.7 ± 19.6	0.69
Numbers of cigarette/day	20.2 ± 8.6	20.0 ± 8.6	0.79
Vitamin supplement use, %	23.3	24.9	0.18
Married, %	83.9	83.7	0.92
Urban residence, %	41.5	39.2	0.27
Education, %> elementary school	23.6	28.4	0.001
Benign prostatic hyperplasia, %	3.8	5.7	0.01
Physical activity, % low3	13.3	12.7	0.61
Physical activity, % medium3	54.9	57.4
Physical activity, % high3	32.0	29.9
Intervention, % AT supplement	50.0	52.2	0.29
Intervention, % BC supplement	49.5	52.1	0.23

Among 19,652 men with complete data, 1,111 incident prostate cancer cases were identified during the follow-up period. Age at diagnosis was somewhat earlier among men with a family history of prostate cancer, compared with men with no family history, although distributions of age at diagnosis and time to diagnosis did not statistically differ by family history status (Table II). Men with a family history were more likely to have advanced disease (58% for men with family history versus 31% for men without family history; p = 0.0005). However, the method of diagnosis and percent receiving oncological reviews did not differ by family history.

Table II. Descriptive Characteristics of the Incident Prostate Cancers by First-Degree Family History of Prostate Cancer Status, Alpha-Tocopherol, Beta-Carotene Cancer Prevention Study, Finland, 1991–2003

	No family history (N = 1,048) (%)	Family history (N = 63)	p-value
1 Method of diagnosis, tumor grade, and stage were available for 646 cases, 616 cases and 644 cases, respectively. 2 Advanced disease was defined as TNM stage III or IV.
Age at diagnosis
≤60 years	33 (3.2)	3 (4.7)	0.89
61–65 years	203 (19.4)	13 (20.6)
66–70 years	336 (32.2)	20 (37.8)
71+ years	476 (45.4)	27 (42.9)
Time to diagnosis
0–5 years	281 (26.8)	16 (25.4)	0.09
6–10 years	504 (48.1)	37 (58.7)
11–15 years	263 (25.0)	10 (15.9)
Method of diagnosis (n, %)1
Histology	522 (94.2)	35 (84.6)	0.09
Cytology	30 (5.4)	5 (12.8)
Clinical	2 (0.4)	1 (2.6)
Tumor grade (n, %)1			0.27
Well differentiated	128 (23.8)	9 (25.7)
Moderately differentiated	283 (52.6)	14 (40.0)
Poorly differentiated	127 (23.6)	12 (34.3)
TNM Stage at diagnosis (n, %)1
0–II	380 (69.3)	16 (42.1)	0.001
III	71 (12.8)	13 (34.2)
IV	99 (17.9)	9 (23.7)
Advanced disease2	170 (30.7)	22 (57.9)	0.0005

A family history of prostate cancer in first-degree relatives was associated with a nearly doubling of prostate cancer risk (Table III). Adjustment for number of brothers, height, serum retinol, education, benign prostatic hyperplasia, red meat intake, and lycopene intake did not alter the risk estimates (data not shown). We had conducted a sensitively analysis that used the full 19 years of follow-up from original study entry, which yielded RR (95% CI) = 1.91(1.49–2.47) for analytic cohort and RR 1.90 (1.48–2.46) for the whole cohort. Men whose fathers had been diagnosed with prostate cancer showed a slightly greater risk increase (RR = 1.99) and men who had brother(s) with the disease had a slightly smaller risk increase (RR = 1.70) compared with those with no family history. For cases with early onset of disease (i.e., < 65 years), family history of prostate cancer was associated with a RR of 2.40, based on 16 and 236 cases with and without family history, respectively. By contrast, for cases with later onset disease, the RR associated with family history was 1.81 (95% CI 1.34–2.43). Because most men with a family history of prostate cancer in our cohort had either a father or only 1 brother affected with only 5 cases having more than one primary relative affected, we were not able to evaluate whether having additional first-degree relatives conferred even higher risk. Men with a family history of prostate cancer had over 4 times the risk of advanced prostate cancer compared with men with a negative family history (Table III). A tangential evaluation of the possible role of shared environments in the prostate cancer family history assessment showed that positive family histories of other common cancers (i.e., breast, lung, or colorectal cancer) were not associated with increased prostate cancer risks. Family history of other cancers was not associated with prostate cancer risk (data not shown).

Table III. Multivariate-Adjusted RRs of Prostate Cancer by First-Degree Family History of Cancer Status, α-Tocopherol, β-Carotene Cancer Prevention Study, Finland, 1991–2003

Family history of cancers	Total prostate cancer cases (N = 1,111)			Advanced prostate cases only (N = 192)
	Cases	RR1	95% CI	Cases	RR1	95% CI
1 RRs and 95% CIs calculated by Cox proportional hazard regression, adjusted for age, and trial intervention (α-tocopherol and β-carotene supplement). 2 RRs for sibling history were assessed among men with at least one sibling.
Prostate cancer
No family history	1,048	1.00		182	1.00
Father or brother(s)	63	1.91	1.49–2.47	24	4.16	2.67–6.49
Father	46	1.99	1.48–2.67	17	4.02	2.37–6.83
Brother(s)2	17	1.70	1.05–2.75	7	4.44	2.07–0.53
Breast cancer
No family history	900	1.00		162	1.00
Mother or sister(s)	86	1.16	0.93–1.48	16	1.21	0.73–2.06
Mother	37	1.24	0.89–1.72	4	0.76	0.28–2.05
Sister(s)2	52	1.13	0.85–1.49	12	1.74	0.78–2.66
Lung cancer
No family history	879	1.00		156	1.00
Any first-degree relative	108	0.97	0.80–1.19	16	0.79	0.48–1.31
Parent(s)	67	0.96	0.75–1.23	7	0.59	0.29–1.20
Sibling(s)2	46	1.03	0.76–1.38	10	1.15	0.61–2.19
Colorectal cancer
No family history	949	1.00		164	1.00
Any first-degree relative	31	1.09	0.82–1.46	4	0.93	0.49–2.03
Parent(s)	20	1.14	0.81–1.62	4	1.00	0.49–2.03
Sibling(s)2	12	1.08	0.69–1.70	1	–	–

Risks associated with a first degree of family history were somewhat more pronounced among taller, heavier, and more active men, as well as those with lower serum α-tocopherol and higher β-carotene and retinol concentrations (Table IV); however, these interactions were not statistically significant. There was a suggestion that at low serum β-carotene levels, family history increased prostate cancer risk, little if at all, whereas at higher levels the risk was double (p interaction=0.08). α-Tocopherol and β-carotene supplementation intervention did not modify the family history-prostate cancer association, although it appeared stronger in the α-tocopherol study arm. These same characteristics showed no association with prostate risk among men with no family history, and weak to moderate positive associations among men having a family history (with the exception of an inverse association for serum alpha-tocopherol in this group). We were not able to evaluate interactions for advanced disease given the sample size, since the number of advanced cases was limited.

Table IV. Multivariate Adjusted RRs of Prostate Cancer by Family History of Prostate Cancer in First Degree Relatives Stratified by Body Size, Diet, Lifestyle Factors, Serum Nutrients and Intervention Group, α-Tocopherol, β-Carotene Cancer Prevention Study, Finland, 1991–2003

	No family history				Family history				p interaction
	Cases	Incidence rate1	RR2	95% CI	Cases	Incidence rate1	RR2	95% CI
1 Incidence rates are per 100,000 person-years, directly standardized to the age distribution of the cohort. 2 RRs and 95% CIs calculated by Cox proportional hazard regression, adjusted for age, and trial intervention (α-tocopherol and ß-carotene supplement). 3 Height: additionally adjusted for weight. 4 Physical activity: low (less than 1 time exercise per week in leisure time and sedentary occupational activity), medium (less than 1 time exercise per week in leisure time or sedentary occupational activity), and high (more than 1 time exercise per week in leisure time or moderate/heavy occupational activity). 5 Additionally adjusted for serum cholesterol.
Height (cm)3									0.28
<171	364	372	1.00	Referent	16	604	1.62	0.99–2.68
172–176	340	405	1.09	0.94–1.27	20	743	1.97	1.26–3.09
>176	344	450	1.12	0.97–1.30	27	889	2.50	1.69–3.69
BMI (kg/m^2)									0.61
<24.6	345	378	1.00	Referent	19	642	1.76	1.11–2.79
24.6–27.6	387	445	1.12	0.97–1.30	22	801	2.17	1.41–3.34
>27.6	316	391	0.96	0.82–1.12	22	815	2.00	1.30–3.07
Physical activity4									0.79
Low	164	368	1.00	Referent	7	400	1.50	0.70–3.19
Moderate	594	408	0.94	0.79–1.12	34	731	1.65	1.14–2.39
High	290	411	0.85	0.70–1.04	22	900	2.11	1.35–3.30
Alcohol (g/day)
<5.2	365	383	1.00	Referent	20	715	2.01	1.29–3.16	0.55
5.2–19.5	360	420	1.04	0.90–1.21	21	679	1.67	1.09–2.63
>19.5	323	405	1.02	0.88–1.18	22	893	2.22	1.44–3.41
Serum α-tocopherol (mg/L)5
<10.5	342	391	1.00	Referent	22	844	2.21	1.44–3.41	0.17
10.5–12.9	378	437	1.09	0.94–1.27	28	917	2.32	1.57–3.42
>12.9	328	388	0.96	0.80–1.15	13	470	1.26	0.72–2.20
Serum β-carotene (ug/L)5
<137	337	412	1.00	Referent	14	476	1.30	0.76–2.22	0.08
138–234	345	395	0.96	0.83–1.12	24	846	2.15	1.42–3.26
>234	366	409	0.99	0.85–1.15	25	894	2.16	1.44–3.25
Serum retinol (ug/L)5
<532	335	368	1.00	Referent	17	610	1.67	1.03–2.73	0.34
533–632	357	414	1.07	0.92–1.24	19	727	1.84	1.13–2.99
>632	356	435	1.14	0.98–1.33	29	917	2.54	1.74–3.72
Intervention									0.16
α-tocopherol (50 mg/day)
No	568	449	1.00	Referent	28	695	1.70	1.09–2.33
Yes	480	364	0.83	0.74–0.94	35	804	1.90	1.35–2.68
β-carotene (20 mg/day)									0.80
No	510	389	1.00	Referent	30	718	1.98	1.37–1.86
Yes	538	423	1.09	0.97–1.23	33	793	2.02	1.42–2.88

Discussion

In this large prospective cohort study, with a long follow-up period, having a family history of prostate cancer in first-degree relatives was associated with a nearly doubling of risk for incident prostate cancer and of early onset disease (<65 years) in particular. In addition, men with a positive family history were over 4 times more likely to have an advanced disease when diagnosed. Our data also suggest that to some degree, height, serum α-tocopherol and β-carotene may modify the family history–prostate cancer association. The study provides additional prospective evidence and quantitative details regarding family history as a significant risk factor for prostate cancer.

Although the family history–prostate cancer association is fairly established, few of the prior studies have been prospective, which raises the issue of recall bias and its impact on the magnitude of the previously observed associations. Our risk estimates are very similar to registry-based, prospective investigations of incident prostate cancer in Sweden (RR = 1.65)20 and the Netherlands (RR = 1.77),21 as well as to prostate cancer mortality in the American Cancer Society Cancer Prevention Study II cohort (RR = 1.6).22 These risk estimates are substantially lower than those from case–control studies, which ascertained data on family history retrospectively and may have overestimated risk.23–27 Given that family history reflects not only shared genes but also shared environments and common behaviors,14 lower risk estimates could in part result from less variability in the latter factors, particularly in a relatively homogeneous population such as in the present study. In this regard, it is interesting to note that having a family history of other cancers that do share some of the same risk factors with prostate cancer (i.e., red meat consumption with colorectal cancer)28, 29 was not associated with prostate cancer risk in our study, although 1 study reported that family history of breast cancer was associated with a prostate cancer relative risk of 1.7.30

Although it has been thought that 5–10% of prostate cancer is accounted for by genetic susceptibility,31 an analysis of monozygotic and dizygotic twin pairs in Scandinavia concluded that 42% (95% CI, 29–50%) of prostate cancer risk may be heritable.32 More recently, linkage analysis combined with genome-wide scan data from 1,233 families identified 5 suggestive regions (i.e., 5q12, 8p21, 15q11, 17q21 and 22q12),33 with significant linkage at 22q12 (logarithm of the odds score = 3.57) as well as 8q24.34 Suggestive linkage was also observed at 1q25, 8q13, 13q14, 16p13 and 17q21, among 269 families with at least 5 affected members. Although high penetrance germline mutations have not been established for prostate cancer (as, e.g., with BRCA1 and BRCA2 and breast cancer), data from such pedigree and genome-wide scan studies may lead to identification of loci underlying genetic susceptibility and prioritization of chromosomal regions for further study.

Several2, 21, 30, 35 but not all,22 studies have found that a history of prostate cancer in a sibling confers greater risk than that of paternal history, which would be consistent with the additional contribution to risk of a shared environment in childhood and adolescence. We observed a somewhat stronger association with paternal history, although the number of cases was small, and the confidence intervals for paternal versus sibling history were not mutually exclusive. Since we queried family history only once during the study, history of prostate cancer among brothers may not have been fully manifested, as compared with that for fathers. This probably influenced our risk estimates for the sibling history.

Although the interactions evaluated were not statistically significant in our study, risk associated with family history was somewhat pronounced among taller and overweight men, a finding supported by other studies of heredity-anthropometry interactions. For example, Sellers and colleagues36 reported that the increased risks of breast cancer associated with a high waist-to-hip ratio, low parity, and greater age at first birth were more pronounced among women with a family history of breast cancer. In line with the present findings, Rohrmann et al.12 reported stronger positive BMI-prostate cancer associations among men with younger age at onset of the disease. By contrast, Giovannucci et al.11 reported a stronger and significant inverse association between BMI and prostate cancer among men with a family history and for early disease onset (i.e., before 60 years of age) and suggested that this might be due to lower androgen status in obese men and a higher ratio of sporadic to hereditary cases among them. These data indicate that the development of prostate cancer may differ between men with and without a hereditary predisposition to the disease.

We observed the modest (nonsignificant) physical activity effect modification, suggesting little or no activity impact among men with no family history and a small, nonsignificant risk increase in those with an FDR, although the role of activity in prostate cancer is not clear. Findings could be due to chance, and further studies are warranted. To our knowledge, ours is the first prospective study to examine the modifying influences of serum micronutrients on the association between family history and prostate cancer risk. How serum α-tocopherol, β-carotene, or retinol might modify the risk associated with family history is not clear, particularly when α-tocopherol and β-carotene supplementation did not. Even though these interactions were not formally significant, other studies of family history and prostate cancer risk in relation to these serum micronutrients would be informative.

Our investigation benefits from its prospective nature and approximately 100% cancer ascertainment for the cohort but also has some limitations. Assessment of family history was based on self-report (as in most previous studies), and validation of reported cancers through the cancer registry or family members was not feasible. The prevalence of family history of prostate cancer we observed (3%) was, however, similar to that in the Dutch cohort (3.0%),21 and only slightly lower than in the U.S. (4.6%)30 and Swedish cohorts (6.4%)20 (although the latter encompassed 30 years of registry follow-up). Underreporting of prostate cancer family history is possible (sensitivity, 70% and specificity, 94%37), though it is unlikely to differ by prostate cancer status since it was ascertained years before diagnosis. Such nondifferential misclassification would be expected to attenuate the risk estimates toward the null, with the true risk estimates being larger than those observed.38 Because family history information was collected only once during the study and because participants are relatively young age for the group at entry, it is also possible that additional family members (particularly brothers) were diagnosed with prostate cancers during the subsequent 12 years of follow-up, with a resulting underestimation of the prevalence of family history. Nonetheless, the incomplete ascertainment of FDRs should in large part be nondifferential with respect to prostate cancer status. Detection bias resulting from more vigilant screening also appears unlikely, however, because men with a prostate family history did have higher disease stage at diagnosis than men with no family history, which is consistent with findings regarding breast cancer.39 Information on Gleason score was available from only a small number of the cases diagnosed during the original ATBC Study trial period (n = 320), and thus our advanced disease definition was based on stage information. Finally, since this study was based on smokers, our results may not be generalizable to nonsmoking populations.

In summary, this prospective study of Finnish men having up to 12 years of follow-up provides additional prospective evidence that having a family history of prostate cancer is a significant risk factor for developing the disease, with a doubling of risk overall (quadrupling of risk for advanced disease). Possible interactions between family history and body size and serum antioxidant vitamins were also suggested and should be reevaluated in other studies. Based on the increased risk resulting from family history, recommendations for early detection practices are advocated by, for example, the American Cancer Society Cancer Detection Guidelines, the Centers for Disease Control and Prevention, and the National Cancer Institute.