Download Metabolic Syndrome Predicts Prostate Cancer in a Cohort of Middle

American Journal of Epidemiology
Copyright ª 2006 by the Johns Hopkins Bloomberg School of Public Health
All rights reserved; printed in U.S.A.
Vol. 164, No. 8
DOI: 10.1093/aje/kwj284
Advance Access publication September 4, 2006
Original Contribution
Metabolic Syndrome Predicts Prostate Cancer in a Cohort of Middle-aged
Norwegian Men Followed for 27 Years
L. Lund Håheim1, T. F. Wisløff1,2, I. Holme3, and P. Nafstad2,4
1
Norwegian Knowledge Centre for Health Services, Oslo, Norway.
Norwegian Institute of Public Health, Oslo, Norway.
3
Ulleval University Hospital, Oslo, Norway.
4
University of Oslo, Oslo, Norway.
2
Received for publication June 27, 2005; accepted for publication April 5, 2006.
The aim of the study was to establish whether metabolic syndrome predicts the incidence of prostate cancer. The
hypothesis was tested using the 27-year follow-up of the prospective cohort of 16,209 men aged 40–49 years who
participated in the Oslo Study in 1972–1973. Men with established diabetes and men with cancer diagnosed before
screening were excluded, leaving 15,933 for analyses. Metabolic syndrome is here composed of body mass index,
nonfasting glucose, triglycerides, and blood pressure or drug-treated hypertension. Two analytical approaches
were compared, namely, predefined (adjusted from National Cholesterol Education Program) and quartile values
of risk factors. Age, body mass index, and sedentary versus intermediate physical activity at work were significant
predictors in univariate proportional hazards regression analyses. Combinations of any two (relative risk ¼ 1.23;
p ¼ 0.04) or any three (relative risk ¼ 1.56; p ¼ 0.00) factors of the metabolic syndrome using quartile values of risk
factors were predictive of prostate cancer. The number of cases for four factors was too small for analyses.
Predefined values of the risk factors were not found to be predictive. In conclusion, metabolic syndrome was
found to predict prostate cancer during 27 years of follow-up, indicating an association between insulin resistance
and the incidence of prostate cancer.
incidence; insulin resistance; metabolic syndrome X; prospective studies; prostatic neoplasms
Abbreviations: CI, confidence interval; IGF-1, insulin-like growth factor-1; IGFBP, insulin-like growth factor binding protein;
NCEP, National Cholesterol Education Program; RR, relative risk.
No major cause of prostate cancer has been established,
although genes, dietary factors, and lifestyle-related factors
are believed to contribute to the development of prostate
cancer (1–3). Prostate cancer occurs mainly after the age of
50 years, and the incidence increases sharply with advancing
age. This is the most common cancer in men in Norway and
the incidence is increasing (4). In the period from 1988 to
1997, the age-adjusted incidence rate increased from 44.6 to
67.2 per 100,000. The incidence rate has to some extent been
influenced by an increase in screening by prostate-specific
antigen testing, but the high rates call for further etiologic
study.
Norwegian men aged 40–42 years are reported to have
increased their mean weight by 9.1 kg (from 76.9 to 86.0
kg) during the period of 1963–1999 (5). The mean height
also increased. The weight increase observed during this
period coincides with the present increase in the incidence
of prostate cancer. Metabolic syndrome, of which body mass
index constitutes an important factor, is established as being
associated with increased risk of cardiovascular disease (6).
Correspondence to Dr. Lise Lund Håheim, Norwegian Knowledge Centre for Health Services, P.O. Box 7004, St. Olavs Plass, N-0130 Oslo,
Norway (e-mail: [email protected]).
769
Am J Epidemiol 2006;164:769–774
770 Håheim et al.
More uncertainty exists with regard to prostate cancer (7, 8).
Insulin resistance and compensatory hyperinsulinemia are
thought to be an underlying pathway by stimulating production of insulin-like growth factor-1 (IGF-1) and related factors (9–11). Metabolic syndrome has been posited as a cause
of several forms of cancer, such as prostate cancer (9–12)
and cancers of the breast (13, 14), pancreas (15), and colon
(16, 17). The predictiveness of metabolic syndrome for cancer of the prostate is studied in this prospective cohort of
Norwegian men initially aged 40–49 years who were followed for 27 years.
MATERIALS AND METHODS
The Oslo Study cohort of men was screened in 1972–1973
(18). The main objective was to study the prevention and
epidemiology of cardiovascular diseases. The participants
were aged 20–49 years in 1972. This presentation concerns
the men aged 40–49 years (n ¼ 25,915) at the time of the
screening, of whom 16,209 men attended the screening. In
short, the men had to answer a questionnaire on their history
of cardiovascular symptoms or diseases, diabetes, medication for hypertension, smoking, mental stress, and physical
activity at work and at leisure. Physical activity was coded as
sedentary, intermediate, moderate, and high. A blood sample
was taken while the participant was in a nonfasting state to
estimate the serum total cholesterol level, triglycerides, and
glucose levels. Height and weight were measured. Blood
pressure was measured with a mercury sphygmomanometer.
Information on level of education achieved, cancer incidence, and cause-specific mortality was added to the screening data. The level of education achieved was coded to nine
levels as follows: no education or per-school education, primary, three levels of secondary, postsecondary nontertiary,
and three levels of graduate education. Information on the
level of education achieved and cause of death was given by
Statistics Norway (Oslo, Norway). Incident cases of any
cancer diagnosis were provided by the Cancer Registry of
Norway (Oslo, Norway). The last day of follow-up was
December 31, 1998.
All examinations were done according to the Helsinki
Declaration (http://www.wma.net/e/policy/b3.htm), and the
necessary permits were granted by the Norwegian Data
Directorate and the Norwegian Board of Health (Statens
Helsetilsyn, Oslo, Norway).
Men with a history of cancer (n ¼ 123) or diabetes (n ¼
153) at the time of screening were excluded from the analyses, leaving 15,933 eligible for analysis. The incidence rate
is the number of cases per 1,000 person-years. Descriptive
values of risk factors are presented as the mean or percentage
and standard deviation. Statistical significance is indicated
by p < 0.05 (two sided). The main statistical analysis of an
association between risk factors and prostate cancer was
proportional hazards regression (Cox regression).
The identification of cases was through obligatory reporting to the Cancer Registry of Norway of all newly diagnosed
cancers since 1954. The diagnoses are validated by questionnaire information, pathology reports, or other diagnostic
tests. The register is believed to include close to 100 percent
of all cancer cases. The 11-digit personal identification number of all Norwegians allows for a complete follow-up on
cancer and mortality. Men who were diagnosed with other
cancer or died from any cause during follow-up were censored in the Cox regression analyses.
The metabolic syndrome was composed of the risk factors body mass index, diastolic blood pressure, nonfasting
glucose, and triglycerides based on the factors used by the
National Cholesterol Education Program (NCEP) (19). Cutpoints for risk factors of the metabolic syndrome were tested
by use of either the upper quartile values or the predefined
values of NCEP. Their standard code includes five factors of
which a person must have a risk level in three of these five
factors to have the metabolic syndrome. As this data set did
not include exactly the same factors, adjustments were
made. Body mass index was used instead of waist measurement for overweight. High density lipoprotein cholesterol
was not measured. The predefined values used were blood
pressure (systolic/diastolic blood pressure of 130/85 mmHg
or above) or drug-treated hypertension, nonfasting glucose
of 11.0 mmol/liter or above, triglycerides of 2.0 mmol/liter
or above, and body mass index of 30 kg/m2 or above. The
upper quartile values were as follows: systolic/diastolic
blood pressure ¼ 144/92 mmHg, nonfasting glucose ¼
6.10 mmol/liter, triglycerides ¼ 3.01 mmol/liter, and body
mass index ¼ 26.31 kg/m2. The analyses were adjusted for
time since the last meal in the Cox analyses to compensate
for the serum samples taken in the nonfasting state. Interaction analyses were performed for any two risk factors of the
metabolic syndrome.
A sequential analysis of metabolic syndrome was made by
four multivariate proportional hazards regression models including any one of the four factors, then any of two, then any
of three, and finally all four factors. The quartile values used
are those of the fourth quartile compared with those of quartiles 0, 1, and 3 together. Every comparison is with those that
did not have that many risk factors; that is, ‘‘any three factors’’ are those with three factors compared with all others.
In addition, these models were compared by calculating the
proportion attributable risk, which is defined as the proportion of the incidence in these models that is due to the additional factor (compared with the models with one less
factor): proportion attributable risk ¼ (relative risk – 1)/
relative risk. The proportional hazard models were tested
for proportionality by use of Schoenfeld residuals. Analyses
were done mainly with SPSS, version 12.0, software (SPSS,
Inc., Chicago, Illinois), but splines and the test for proportionality were done in S-PLUS, version 6.1, software
(Insightful Corporation, Seattle, Washington).
RESULTS
The incidence of prostate cancer in this cohort was 1.45
per 1,000 person-years. In all, 507 cases were identified
among the 15,933 men at risk. Comparing cases with noncases, we found that risk factor levels by mean values were
not significantly different except for physical activity at
work on a scale from 1 to 4 (highest) (cases ¼ 1.69, standard
deviation: 0.87; noncases ¼ 1.76, standard deviation: 0.87)
(table 1).
Am J Epidemiol 2006;164:769–774
Metabolic Syndrome and Prostate Cancer
TABLE 1. Mean (standard deviation) levels or percentage
of risk factors in cases of prostate cancer and men with no
prostate cancer or diabetes among 15,933 middle-aged men at
screening, Norway, 1972–1973
TABLE 2. Age and age-adjusted univariate proportional
hazards regression analyses of risk factors for prostate cancer
presented with regression coefficient and relative risk with 95%
confidence interval, Norway, 1972–1998
No prostate
cancer
(n ¼ 15,426)
Prostate
cancer
(n ¼ 507)
Risk factor
771
Risk factor
Relative
risk
95%
confidence
interval
Age (years)
46.49 (2.74)
45.77 (2.89)
Age (years)
1.12
1.08, 1.15
Triglycerides (mmol/liter)
2.53 (1.24)
2.55 (1.60)
Triglycerides (mmol/liter)*
1.02
0.97, 1.07
Total cholesterol (mmol/liter)
6.38 (1.09)
6.38 (1.17)
Total serum cholesterol (mmol/liter)
1.02
0.94, 1.10
Body mass index (kg/m2)
24.95 (2.79)
24.73 (2.97)
Body mass index (kg/m2)
1.02
1.001, 1.06
Height (cm)
177.57 (6.17)
177.40 (6.51)
Height (cm)
1.004
0.99, 1.02
Weight (kg)
78.74 (10.17)
77.87 (10.58)
Nonfasting glucose (mmol/liter)*
1.01
0.92, 1.10
Glucose (mmol/liter)
5.73 (0.87)
5.74 (1.05)
Diastolic blood pressure (mmHg)
1.002
0.99, 1.01
Systolic blood pressure (mmHg)
1.002
0.99, 1.01
Diastolic blood pressure
(mmHg)
86.64 (10.43)
86.83 (11.04)
Systolic blood pressure
(mmHg)
135.44 (16.89)
135.68 (16.30)
Physical activity at leisure
(levels 1–4)
2.00 (0.64)
2.00 (0.67)
Physical activity at work
(levels 1–4)
1.68 (0.87)*
1.76 (0.87)
Daily smoker (%)
55
56
Level of education
(grades 0–9)
3.78 (1.66)
3.65 (1.64)
Blood pressure
medication (%)
2
3
Blood pressure, 130/85
mmHg (%)
39
42
Metabolic syndrome
(predefined) (%)
39
43
Metabolic syndrome
(quartile values) (%)
31
30
* p < 0.001 (Kruskal-Wallis test).
Physical activity at leisure
(levels 1–4 (highest))
ptrend ¼ 0.47
Intermediate vs. sedentary
0.98
0.79, 1.23
Moderate vs. sedentary
0.96
0.72, 1.27
High vs. sedentary
0.45
0.17, 1.22
Physical activity at work
(levels 1–4 (highest))
ptrend ¼ 0.13
Intermediate vs. sedentary
0.77
0.63, 0.96
Moderate vs. sedentary
0.88
0.69, 1.13
0.53, 1.38
High vs. sedentary
0.86
Daily smoking (yes/no)
1.12
0.94, 1.34
Education (grades 0–9)
1.02
0.97, 1.08
Blood pressure medication (yes/no)
1.12
0.94, 2.14
Blood pressure, 130/85 mmHg
(predefined values)
0.99
0.99, 1.14
Blood pressure, 144/92 mmHg
(quartile values)
0.99
0.99, 1.47
* Adjusted for time since last meal.
Results of age-adjusted univariate proportional hazards
regression analyses for risk factors predicting prostate cancer showed significantly increasing risk by age per year (relative risk (RR) ¼ 1.12, 95 percent confidence interval (CI):
1.08, 1.15) and body mass index (weight (kg)/height (m)2)
(RR ¼ 1.03, 95 percent CI: 1.001, 1.06) (table 2). Trend
analyses of physical activity at leisure and at work were
inversely related. Intermediate versus sedentary activity at
work was significantly different (RR ¼ 0.78, 95 percent CI:
0.63, 0.96). Former analyses on this data set found nitrogen
oxide (NOx) to be a risk factor for lung cancer (20), but these
analyses did not find air pollution to be associated with
prostate cancer. Age was not found to fulfill the criteria of
proportionality. However, analysis with splines did not give
ideas for alternative parameterizations.
The predictiveness of metabolic syndrome was analyzed
by using one of two strategies: 1) the upper quartile as risk
factor or 2) the predefined risk factor levels (table 3). The
number of persons at risk in the quartile or predefined categorizations of metabolic syndrome varied for each model.
In all, 317 cancer cases had one of the metabolic risk
factors at the upper quartile value, 133 had two, 72 had three,
Am J Epidemiol 2006;164:769–774
and 10 had all four risk factors. The equivalent number of
cases for the predefined values was 430 cases for one risk
factor, 127 cases for two, 17 for three, and none for four
factors.
Any one of the four factors by upper quartile value was not
predictive, but any two (RR ¼ 1.23; p ¼ 0.04) or three (RR ¼
1.56; p ¼ 0.00) of the factors were predictive. The model of
all four factors of the metabolic syndrome was not significant because of low power, as only 10 cases had this risk
factor profile. The attributable risk estimates for any two
quartile criteria were 10 percent and for any three quartile
criteria were 17 percent. None of the four models with predefined values was predictive. There were no interactions
between any two of the risk factors of the metabolic syndrome using quartile values (data not shown).
DISCUSSION
The knowledge of the increase in incidence of prostate
cancer, the increase in mean weight of middle-aged men
772 Håheim et al.
TABLE 3. Multivariate proportional hazards regression analyses (Cox analyses) of risk factors of metabolic syndrome for prostate
cancer in four models, Norway, 1972–1998*
Population
Definition of
analytical modelsy
No. of persons
with metabolic
syndrome
No. of
prostate
cancer cases
Proportional hazards regression analyses
Proportion
with prostate
cancer
Relative risk
95%
confidence interval
p value
Model 1 (any one factor)
Quartile
9,979
317
0.032
1.05
0.87, 1.26
0.60
13,234
430
0.032
1.16
0.91, 1.49
0.22
Quartile
3,766
133
0.035
1.23
1.01, 1.50
0.04
Predefined
4,501
127
0.028
0.92
0.75, 1.13
0.44
1,702
72
0.042
1.56
1.21, 2.00
0.00
573
17
0.030
1.05
0.65, 1.70
0.85
357
10
0.028
1.13
0.60, 2.11
0.71
10
0
0.000
0.00
NAz
0.94
Predefined
Model 2 (any two factors)
Model 3 (any three factors)
Quartile
Predefined
Model 4 (all four factors)
Quartile
Predefined
* The risk level was set to the upper quartile value versus the lower values.
y All the analyses were adjusted for age and time since the last meal.
z NA, not available.
in Norway, and the research information on the association of
IGF-1 and related factors to prostate cancer is the background for this paper’s focus on metabolic syndrome as a
predictor of prostate cancer. The incidence of prostate cancer
is, however, influenced by opportunistic screening. To what
extent this has influenced the Norwegian rate of incidence
of prostate cancer is uncertain. Simulation modeling (21) of
Dutch screening data gives an estimated mean lead time of
12.3 years for a single screening at age 55 years. Overweight
is a major factor of metabolic syndrome and is also found to
be a predictor in these analyses. Insulin resistance and the
compensatory hyperinsulinemia are the underlying factors of
the metabolic syndrome (9–11). Hyperinsulinemia stimulates the production of IGF-1 and suppresses the production
of sex-hormone binding globulin and insulin-like growth
factor binding protein (IGFBP)-1 and -2. Through these
modulations the insulin resistance is believed to affect
hormone-related cancers, such as prostate cancer. Circulating
levels of IGF-1 and IGFBP-3 have been found to be predictive for advanced-stage prostate cancer in a case-control
study (10).
The present cohort study examines the risk of prostate
cancer over 27 years. This allows for studying risk factors
with a long latency period for prostate cancer presentation.
During this period, there was a continuous validated registration of all cancers diagnosed. No second screening was performed for the full cohort during this period of follow-up.
Included are, however, men that had participated in two intervention studies aimed at reducing their cardiovascular
risk status. In all, 1,232 men participated in a randomized
controlled trial on diet and smoking over 5 years, and 785
men participated in a randomized controlled trial on hypertension over 5 years. Our previous publications show that the
screened population has a lower mortality rate than does the
nonscreened population (22) and the general population of
comparative Norwegian men (23). However, comparative
statistics are not available for cancer. The consequence is
rather an underestimation of risk associations for mortality and, for this reason, we believe the external validity is
adequate.
The men with diagnosed prostate cancer were very similar
to the other men of the study when comparing the risk factors measured at screening, except the cancer patients reported lower physical activity at work. The screening results
showed the body mass index of the men of the total cohort
to be close (24.73 kg/m2 for noncases and 24.95 kg/m2 for
cases of prostate cancer) as for daily smoking (55 percent for
noncases and 56 percent for cases of prostate cancer). The
clustering effect of the risk factors comprising the metabolic
syndrome was therefore of strong interest. Few factors of
this data set were found to be predictive in the univariate
analyses of the proportional hazards analyses. The wellknown age gradient was confirmed and, in addition, body
mass index, low physical activity at work, and a subsidiary
question on recorded mental stress were identified. Of these
factors, only body mass index is one of the factors constituting metabolic syndrome. As noticeable are the factors that
are not predictive. Although there was a protective trend
with increasing physical activity at leisure, physical activity
was not significant at any level compared with the sedentary level of activity. Two Norwegian prospective studies
have not found any association between anthropometry
(24) or dietary fat intake and risk of prostate cancer (25). However, a more recent prospective study found height but
not body mass index to predict prostate cancer (26). The effect of overweight on prostate cancer has been studied
Am J Epidemiol 2006;164:769–774
Metabolic Syndrome and Prostate Cancer
prospectively, but risk has been found to be an independent
risk factor (27), has not been confirmed (28–30), or has been
related to age and familial risk (30). A prospective Finnish
study, however, found metabolic syndrome at baseline to be
associated with a 1.9 times increased risk for prostate cancer
(12). The study was on a population-based sample of 1,880
men, of whom 56 men got prostate cancer. It strengthens the
hypothesis of metabolic syndrome’s possible association with
prostate cancer.
The predefined values used for the analyses of the metabolic syndrome were modeled after those in the NCEP.
NCEP uses systolic/diastolic blood pressure of 130/85
mmHg or above or blood pressure medication, waist measured to 102 cm or above for men, fasting glucose of 6.1
mmol/liter or more, high density lipoprotein cholesterol to
be less than 1.0 mmol/liter for men, and triglycerides greater
than 1.7 mmol/liter. However, this database did not have all
the factors measured, such as high density lipoprotein cholesterol and waist measurement, nor were the serum samples
taken in the fasting state. This is a limitation and reduces
comparability with other studies. For this reason, upper
quartile values were used in comparative analyses. The last
approach of using upper quartile values gave, however, the
significant findings of increasing risk estimates with two or
alternatively with three factors added in the analytical model
except for the full model of four factors. It seems as if there is
a trend toward higher impact of the metabolic syndrome with
more factors included in its definition. For four risk factors,
however, the data set seems too small. Lakka et al. (6) used
the same approach as did NCEP and the World Health Organization standard criteria and upper quartile values in
studying metabolic syndrome and predicting cardiovascular
disease and total mortality. The NCEP codes gave less consistent results than did the upper quartile values. The analyses are sensitive to the number of cases in the analyses,
which explains why having all four factors present was not
predictive in our study. One should also note the different
numbers of persons falling into the risk categories in the four
models. This reflects a possible discrepancy between a Norwegian cohort and cohorts of other countries. These results
indicate that a cautious approach should be held when extrapolating data from international data.
Information on familial risk was not obtained at the
screening. This would have given an indication of genetic
susceptibility for prostate cancer. It has been found that polymorphism of the insulin gene is associated with increased
prostate cancer risk, indicating a possible linkage between
metabolic syndrome and cancer development (31). The interplay of androgens on IGF-1, insulin, and leptin has been
studied in a nested case-control study, indicating a complex
association among these factors (32). However, in a nested
case-control study, no causal association between IGF-1 and
prostate cancer was found (33).
This prospective cohort study gives evidence for an association between metabolic syndrome (consisting of the risk
factors body mass index, blood pressure, triglycerides, and
glucose) and prostate cancer over a 27-year period. The
results of this study indicate that metabolic syndrome may
be associated with the increasing Norwegian national incidence rate of prostate cancer.
Am J Epidemiol 2006;164:769–774
773
ACKNOWLEDGMENTS
Conflict of interest: none declared.
REFERENCES
1. Grönberg H. Prostate cancer epidemiology. Lancet 2003;361:
859–64.
2. Nelson WG, De Marzo AM, Isaacs WB. Prostate cancer.
Mechanisms of disease. N Engl J Med 2003;349:366–81.
3. Harvei SI. Epidemiology of prostate cancer. (In Norwegian).
Tidsskr Nor Laegeforen 1999;119:3589–94.
4. The Cancer Registry of Norway. Cancer in Norway 1997.
Oslo, Norway: The Cancer Registry of Norway, 2000.
5. Statens råd for ernæring og fysisk aktivitet. Vekt helse. (In
Norwegian). Oslo, Norway: Norwegian Directorate for Health
and Social Affairs, 2000. (Report 1/2000).
6. Lakka HM, Laaksonen DE, Lakka TA, et al. The metabolic
syndrome and total and cardiovascular disease mortality in
middle-aged men. JAMA 2002;288:2709–16.
7. Laukkanen JA, Laaksonen DE, Niskanen L, et al. Metabolic
syndrome and the risk of prostate cancer in Finnish men:
a population-based study. Cancer Epidemiol Biomarkers Prev
2004;13:1646–50.
8. Hubbard JS, Rohrmann S, Landis PK, et al. Association of
prostate cancer risk with insulin, glucose, and anthropometry
in the Baltimore Longitudinal Study of Aging. Urology 2004;
63:253–8.
9. Kaaks R, Lukanova A. Energy balance and cancer: the role of
insulin and insulin-like growth factor-I. Proc Nutr Soc 2001;
60:91–106.
10. Chan JM, Stampfer MJ, Ma J, et al. Insulin-like growth factor-I
(IGF-I) and IGF binding protein as predictors of advancedstage prostate cancer. J Natl Cancer Inst 2002;94:1099–106.
11. Barnard RJ, Aronson WJ, Tymchuk CN, et al. Prostate cancer:
another aspect of the insulin-resistance syndrome? Obes Rev
2002;3:303–8.
12. Laukkanen JA, Laaksonen DE, Niskanen L, et al. Metabolic
syndrome and the risk of prostate cancer in Finnish men:
a population-based study. Cancer Epidemiol Biomarkers Prev
2004;13:1646–50.
13. Furberg AS, Veierod MB, Wilsgaard T, et al. Serum highdensity lipoprotein cholesterol, metabolic profile, and breast
cancer risk. J Natl Cancer Inst 2004;96:1152–60.
14. Manjer J, Kaaks R, Riboli E, et al. Risk of breast cancer in
relation to anthropometry, blood pressure, blood lipids and
glucose metabolism: a prospective study within the Malmo
Preventive Project. Eur J Cancer Prev 2001;10:33–42.
15. Michaud DS, Liu S, Giovannucci E, et al. Dietary sugar, glycemic load, and pancreatic cancer risk in a prospective study.
J Natl Cancer Inst 2002;94:1293–300.
16. Colangelo LA, Gapstur SM, Gann PH, et al. Colorectal cancer mortality and factors related to the insulin resistance syndrome. Cancer Epidemiol Biomarkers Prev 2002;11:385–91.
17. Nilsen TI, Vatten LJ. Prospective study of colorectal cancer
risk and physical activity, diabetes, blood glucose and BMI:
exploring the hyperinsulinaemia hypothesis. Br J Cancer
2001;84:417–22.
18. Leren P, Askevold EM, Foss OP, et al. The Oslo Study.
Cardiovascular disease in middle-aged and young Oslo men.
Acta Med Scand Suppl 1975;588:1–38.
19. Standard code from the National Cholesterol Education Program. JAMA 2001;285:2486–97.
774 Håheim et al.
20. Nafstad P, Lund Håheim L, Oftedal B, et al. Lung cancer
and air pollution. A 27 year follow up of 16,209 Norwegian
men. Thorax 2003;58:1071–6.
21. Draisma G, Boer R, Otto SJ, et al. Lead times and overdetection due to prostate-specific antigen screening: estimates
from the European Randomized Study of Screening for Prostate Cancer. J Natl Cancer Inst 2003;95:868–78.
22. Lund Håheim L, Holme I, Hjermann I, et al. The predictability
of risk factors with respect to incidence and mortality of
myocardial infarction and total mortality. A 12-year follow-up
of the Oslo Study, Norway. J Int Med 1993;234:17–24.
23. Lund Håheim L, Holme I, Hjermann I, et al. Trends in the
incidence of acute myocardial infarction and stroke: a 21-year
follow-up of the Oslo Study. Scand Cardiovascular J 2004;38:
216–21.
24. Nilsen TI, Vatten LJ. Anthropometry and prostate cancer risk:
a prospective study of 22,248 Norwegian men. Cancer Causes
Control 1999;10:269–75.
25. Veierod MB, Laake P, Thelle DS. Dietary fat intake and risk of
prostate cancer: a prospective study of 25,708 Norwegian men.
Int J Cancer 1997;73:634–8.
26. Engeland A, Tretli S, Bjorge T. Height, body mass index,
and prostate cancer: a follow-up of 950,000 Norwegian men.
Br J Cancer 2003;89:1237–42.
27. Gronberg H, Damber L, Damber JE. Total food consumption
and body mass index in relation to prostate cancer risk: a casecontrol study in Sweden with prospectively collected exposure
data. J Urol 1996;155:969–74.
28. Johnsson F, Wolk A, Pedersen NL, et al. Obesity and hormonedependent tumors: cohort and co-twin control studies based
on the Swedish Twin Registry. Int J Cancer 2003;106:594–9.
29. Lee IM, Sesso HD, Paffenberg RS Jr. A prospective cohort
study of physical activity and body size in relation to prostate
cancer risk (United States). Cancer Causes Control 2001;12:
187–93.
30. Giovannucci E, Rimm EB, Liu Y, et al. Body mass index and
risk of prostate cancer in U.S. health professionals. J Natl
Cancer Inst 2003;95:1240–4.
31. Ho GY, Melman A, Liu SM, et al. Polymorphism of the insulin
gene is associated with increased prostate cancer risk. Br J
Cancer 2003;88:263–9.
32. Kaaks R, Lukanova A, Sommersberg B. Plasma androgens,
IGF-1, body size, and prostate cancer risk. A systematic review. Prostate Cancer Prostatic Dis 2000;3:157–72.
33. Woodson K, Tangrea JA, Pollak M, et al. Serum insulinlike growth factor I: tumor marker or etiologic factor?
A prospective study of prostate cancer among Finnish men.
Cancer Res 2003;63:3991–4.
Am J Epidemiol 2006;164:769–774