plasma insulin-like growth factor-i, insulin-like growth factor-binding proteins, and prostate...

Plasma Insulin-Like Growth Factor-I, Insulin-LikeGrowth Factor-Binding Proteins, and Prostate CancerRisk: a Prospective Study

Par Stattin, Annika Bylund, Sabina Rinaldi, Carine Biessy, Henri Dechaud,Ulf-Håkan Stenman, Lars Egevad, Elio Riboli, Goran Hallmans, Rudolf Kaaks

Background: Recent studies have suggested that men withelevated plasma levels of insulin-like growth factor-I (IGF-I)may have an increased risk of prostate cancer. Furthermore,IGF-binding proteins (IGFBPs) and insulin can modulatethe activity of IGF-I. In this study, we sought to determinethe role of IGF-I as well as IGFBPs-1, -2, and -3 and insulinas possible etiologic factors for prostate cancer. Methods: Weconducted a nested case–control study within the NorthernSweden Health and Disease Cohort Study. We measuredlevels of IGF-I, IGFBP-1, IGFBP-2, IGFBP-3, and insulin inplasma samples from 149 men who had a diagnosis of pros-tate cancer between 1 month and 10 years after blood col-lection and among 298 control men. All statistical tests aretwo-sided. Results: Case subjects had statistically signifi-cantly higher mean levels of IGF-I than control subjects(229 ng/mL; 95% confidence interval [CI] = 218–240 ng/mL]versus 214 ng/mL [95% CI = 208–221 ng/mL]; P = .02) andIGFBP-3 (2611 ng/mL [95% CI = 2518–2704 ng/mL] versus2498 ng/mL [95% CI = 2437–2560 ng/mL]; P = .04). Condi-tional logistic regression analyses showed increases in pros-tate cancer risk with rising levels of IGF-I (Pfor trend = .02)and IGFBP-3 (Pfor trend = .03). In case subjects younger than59 years at the time of blood collection, the risk associatedwith increased IGF-I was higher (Pfor trend = .01), whereasthe risk associated with increased IGFBP-3 was lower(Pfor trend = .44) than the corresponding risks in the fullcohort. Prostate cancer risk was not associated with levels ofIGFBP-1, IGFBP-2, or insulin. Conclusions: Prostate cancerrisk is increased in men with elevated plasma IGF-I. Thisassociation was particularly strong in younger men in thisstudy, suggesting that circulating IGF-I may be specificallyinvolved in the early pathogenesis of prostate cancer. [J NatlCancer Inst 2000;92:1910–7]

Insulin-like growth factor-I (IGF-I) may stimulate the devel-opment of prostate cancer by stimulating cell proliferation andby inhibiting apoptosis (1,2). In vitro studies (3,4) have demon-strated that prostatic epithelial cells respond to the mitogenicactivity of IGF-I. Moreover, in vivo studies have shown thattumors of the prostate cancer cell line PC-3 have a significantlylower proliferation rate in IGF-I-deficient hosts than in IGF-I-expressing hosts (5). In addition, tumor formation induced byinjection of fibroblasts in nude mice is inhibited when the fibro-blasts have been transfected with an inactivated human IGF-Ireceptor (6). In men, one prospective cohort study (7) and twocase–control studies (8,9) have shown positive associations be-tween prostate cancer risk and circulating IGF-I level.

The bioactivity of IGF-I is determined by circulating levels,as well as the production within tissues, of IGF-I and at least sixdifferent IGF-binding proteins (IGFBPs) (1,10). Most of the

circulating IGF-I and the IGFBPs-1, -2, and -3 are produced inthe liver, and at least 90% of circulating IGF-I is bound toIGFBP-3, the major binding protein in plasma. The main stimu-lus for synthesis of IGF-I and IGFBP-3 in the liver as well as inother tissues is provided by growth hormone (GH) (1,11). Asecond level of regulation of IGF-I bioactivity is provided byinsulin. Insulin enhances the GH-stimulated synthesis of IGF-Iand IGFBP-3. Furthermore, insulin can increase IGF-I bioactiv-ity by decreasing the synthesis and plasma levels of IGFBP-1and IGFBP-2.

The circulating levels of IGFBPs-1, -2, and -3 vary in re-sponse to nutritional status and in response to changes in energymetabolism (11). In the energy-restricted state, which is stronglyprotective against many forms of tumors, including prostate can-cer (12), circulating levels of IGF-I, IGFBP-3, and insulin aredecreased, whereas levels of IGFBPs-1 and -2 are increased(11,13). Overeating and obesity, on the other hand, lead to hy-perinsulinemia and decreased levels of IGFBPs-1 and -2 (11,14–16) but have little effect on IGF-I and IGFBP-3 levels. In arecent experimental study in nude mice (17), high caloric intakeincreased both circulating IGF-I levels and the growth of pros-tatic tumor implants. In addition to nutritional lifestyle factors,genetic predisposition also plays an important role in determin-ing circulating IGF-I and IGFBP-3 levels (18).

In this prospective, population-based, cohort study, wesought to determine the association between prostate cancer riskand plasma levels of IGF-I, IGFBP-1, IGFBP-2, IGFBP-3, andinsulin.

METHODS

Northern Sweden Health and Disease Cohort Study

Men are recruited to the Northern Sweden Health and Disease Cohort Studythrough the Vasterbotten Intervention Project (VIP) and the Northern Swedenpart of the World Health Organization (WHO) study for Monitoring of Trendsand Cardiovascular Disease Study (MONICA). VIP is an ongoing population-based, intervention study initiated in 1985 that aims to decrease mortality due tocardiovascular disease and cancer by advocating a healthy diet and lifestyle tothe general public. VIP invites all persons residing in the county of Vasterbotten

Affiliations of authors: P. Stattin (Department of Urology and Andrology),A. Bylund (Department of Geriatrics), G. Hallmans (Department of PublicHealth and Clinical Medicine), Umeå University Hospital, Sweden; S. Rinaldi,C. Biessy, E. Riboli, R. Kaaks, International Agency for Research on Cancer,Lyon, France; H. Dechaud, Central Laboratory for Biochemistry, Hopital del’Antiquaille, Lyon; U.-H. Stenman, Department of Clinical Chemistry, HelsinkiUniversity Central Hospital, Finland; L. Egevad, Department of Pathology,Karolinska Hospital, Stockholm, Sweden.

Correspondence to: Par Stattin, M.D., Ph.D., Department of Urology andAndrology, Umeå University Hospital, 901 85 Umeå, Sweden (e-mail: [email protected]).

See “Notes” following “References.”

© Oxford University Press

1910 ARTICLES Journal of the National Cancer Institute, Vol. 92, No. 23, December 6, 2000

at Univeristy of South A

ustralia on August 23, 2012

http://jnci.oxfordjournals.org/D

ownloaded from

http://jnci.oxfordjournals.org/

(total population, 260 000) to participate in a health survey when they reach theages of 30, 40, 50, and 60 years. In March 1999, a total of 26 856 men hadparticipated in VIP (19). MONICA includes 2704 men, recruited in 1986, 1990,and 1994, who are also a population-based sample from the counties of Vaster-botten and Norrbotten (19).

In both projects, subjects were asked to complete a self-administered ques-tionnaire that included questions about demographic, medical, and lifestyle char-acteristics. In addition, we recorded the subjects’ height and weight (recorded tothe closest 0.2 kg and cm, respectively) and drew a 20-mL blood sample fromeach subject at study entry. Because blood samples were collected from mostparticipants in the morning, fasting time before blood donation was more than 8hours in 60% of the cohort, 4–8 hours in 36%, and less than 4 hours in 4%. Theblood was collected in one heparin tube and one EDTA tube, centrifuged at1500g for 15 minutes at ambient temperature, separated as plasma and buffycoat, frozen at −20 °C or −80 °C, and transferred within 1 week of collection toa −80 °C central storage facility. The methods for collection of anthropometricmeasurements, as well as for the collection, processing, and storage of bloodsamples, were identical in the MONICA and VIP projects. All participantssigned an informed consent form. The study was approved by the Ethical Com-mittee of Umeå University Hospital. So far, no additional blood samples havebeen collected for analysis from any of the study subjects.

Case Ascertainment and Control Selection

All incident cases of prostate cancer and deaths that occurred in the cohortwere identified by linkage with cancer and all-cause mortality registries, respec-tively, using an individual identification number as the identity link. Approxi-mately 97% of the cases have been estimated to be ascertained through suchregistries (20). We identified 166 incident cases of prostate cancer by March1999, of which 17 were excluded because insufficient plasma samples remained.The remaining 149 prostate cancer case subjects were included in this study. Thetime between blood donation and cancer diagnosis ranged from less than 1month to 10 years (median, 3.85 years). Of the 149 case subjects, 126 (85%)were diagnosed more than 1 year after blood donation, 115 (77%) were diag-nosed more than 2 years, and 70 (47%) were diagnosed more than 4 years. Twocontrol subjects per case subject were randomly selected from all of the cohortmembers who were still alive and free of cancer at the time of diagnosis of thecase subject. Control subjects were matched to the case subject by age (±1 year),date (±1 year) the health survey was completed, and town or village of residency.

Additional information on tumor classification was obtained from the LocalPrimary Prostate Cancer Registry (21). However, because this registry has onlybeen in operation since 1992 and has a delay of approximately 1 year from thetime of diagnosis to final registration, clinical data on tumor classification wereextracted directly from patient medical records for tumors diagnosed before 1992or after 1997. Tumor stage and differentiation were evaluated according to theclassifications of the Union Internationale Contre le Cancer in 1992 (22) and theWHO, respectively (23). The presence of lymph node metastases was evaluatedby histologic examination of obturator lymph nodes obtained by surgery. Thepresence of bone metastases was evaluated by a radionuclide scan. In most casesubjects, tumors were localized and either highly or intermediately differenti-ated. Among the case subjects, 10 men had nonpalpable tumors detected bytransurethral resection (seven T1a and three T1b tumors), 53 had nonpalpabletumors detected by prostate-specific antigen (PSA)-driven biopsies (T1c), 61 hadpalpable tumors that were confined to the prostate (T2), 16 had locally advancedtumors (T3 and T4), and five had tumors that were not locally staged (Tx).Lymph node metastases were present in six case subjects (4%), bone metastaseswere present in 14 (9%), and poorly differentiated tumors were present in 19(13%). No formal screening for prostate cancer was performed during the studyperiod in the region. However, the distribution of tumors by stage suggests thatmany of the tumors may have been found through informal screening on theinitiative of the patient himself or of his physician.

Biochemical Assays

Serum levels of PSA in blood drawn shortly before the diagnostic biopsieswere determined by either the Tandem-R PSA assay (Hybritech, Inc, San Diego,CA) or the IMx PSA assay (Abbot Laboratories, Abbott Park, IL). The corre-lation coefficient between the two assays was .990 [IMx PSA assay value �

(1.22 × Tandem-R PSA value) − 2.80) (24).The concentrations of insulin, IGF-I, IGFBP-1, IGFBP-2, and IGFBP-3 were

measured in EDTA-treated plasma from baseline blood collection. Insulin, IGF-I,and IGFBPs-1 and -3 were measured by double-antibody, immunoradiometric

assays, while IGFBP-2 was measured by a single-antibody radioimmunoassay.Reagents were obtained from Sanofi Diagnostics Pasteur (Marnes la Coquette,France) for insulin assays, from Immunotech (Marseille, France) for IGF-I andIGFBP-3 assays, and from Diagnostic Systems Laboratories (Webster, TX) forIGFBP-1 and IGFBP-2 assays. The protocol for the IGF-I assay included anacid–ethanol extraction step to release IGF-I from its binding proteins. All assayswere performed by laboratory personnel who were blinded as to the case–controlstatus of the plasma samples. Samples from matched study subjects were alwaysanalyzed together in the same batch. For quality control, all batches includedthree control samples containing known amounts of the specific peptide. Themean intra-assay coefficients of variation calculated from these control sampleswere 4.2%, 13.5%, 2.9%, 2.5%, and 3.3% for insulin, IGF-I, and IGFBPs-1, -2,and -3, respectively.

The baseline PSA levels in plasma samples of both case and control subjectswere determined by time-resolved immunofluorometric assays (Prostatus PSA;Wallac, Turku, Finland). The analytic detection limit of the assay was 0.01ng/mL; for values between 0.2 and 100 ng/mL, the inter-assay and intra-assaycoefficients of variation were between 2% and 4%.

Statistical Analyses

Age-adjusted Pearson coefficients of correlation were used to examine thecross-sectional relationships between serum insulin, IGF-I, and the IGFBPs andbetween each of those peptides and body height, weight, and body mass index(BMI) (BMI � weight/height2).

A paired Student’s t test was used to test for mean differences betweenanthropometric measurements and hormone levels of the case subjects and themean values of the two control subjects who had been matched to each casesubject. Conditional logistic regression models were used to calculate odds ratios(ORs) for disease by quartile or tertile levels of insulin, IGF-I, and IGFBPs-1, -2,and -3. Quartiles were used for full cohort analyses, whereas tertiles were usedfor the subgroup analysis of young patients (<59 years) because of the relativelysmall number of men in this subgroup. Cut points for quartiles and tertiles weredetermined on variable distributions of case and control subjects combined.Ninety-five percent confidence intervals (CIs) were computed using the standarderrors of the pertinent regression coefficients, assuming a normal probabilitydistribution for the estimated coefficients. Likelihood ratio tests for linear trendsin risk with increasing levels of the various peptides were performed on theoriginal, continuous variables. All statistical tests and corresponding P valueswere two-sided.

Multivariate conditional logistic regression analysis was used to estimate ORsadjusted for possible confounding factors other than those controlled for bymatching. Potential confounding factors included BMI and smoking status at thetime of blood donation. In addition, risk associated with different levels of IGF-Iwas estimated after adjustment for levels of each of the IGFBPs. Each of thevarious adjustments for potential confounding factors was made by a two-stepprocedure. First, in a multiple linear regression model, values for a specificpeptide were regressed on the confounding variables (smoking status, BMI, andIGFBP level), and the residuals of this regression were categorized into quartilesor tertiles. Second, ORs were estimated for the quartiles or tertiles of the re-siduals by multivariate conditional logistic regression models. The same two-step procedure was also used to estimate ORs for each of the IGFBPs adjustedfor levels of IGF-I. The “PHREG” procedure for proportional hazards’ regres-sion in the Statistical Analysis System (SAS Institute, Cary, NC) was used toperform all logistic regression analyses.

RESULTS

Baseline Characteristics

Case and control subjects differed in several baseline char-acteristics (Table 1). Mean plasma levels for case subjects werestatistically significantly higher than those for control subjectsfor IGF-I (229.0 [95% CI � 217.7–240.3] versus 214.4 [95% CI� 207.8–220.9] ng/mL; P � .02) and IGFBP-3 (2611 [95% CI� 2518–2704] versus 2498 [95% CI � 2437–2560] ng/mL;P � .04). Case subjects also had higher mean levels of insulin(8.41 [95% CI � 6.58–10.24] versus 7.93 [95% CI � 6.92–8.94] pmol/mL; P � .62) and lower mean levels of IGFBP-1(39.50 [95% CI � 35.61–43.39] versus 43.02 [95% CI �

Journal of the National Cancer Institute, Vol. 92, No. 23, December 6, 2000 ARTICLES 1911




ownloaded from


39.80–46.24] ng/mL; P � .19) and IGFBP-2 (653.3 [95% CI �529.3–714.2] versus 671.9 [95% CI � 621.7–722.1] ng/mL;P � .55) than control subjects, but none of these differenceswere statistically significant. Case subjects were statistically sig-nificantly taller than control subjects; the ORs for prostate can-cer risk over increasing quartiles of height (Pfor trend � .048)were 1.00 (referent), 1.27 (95% CI � 0.74–2.21), 0.73 (95% CI� 0.40–1.34), and 1.48 (95% CI � 0.87–2.50). However, incontrast to some previous studies (25,26), we found that bothgroups of subjects had a similar BMI; the ORs for prostatecancer risk for increasing quartiles of BMI (Pfor trend � .69) were1.00 (referent), 1.38 (95% CI � .79–2.42), 1.18 (95% CI �0.68–2.05), and 1.32 (95% CI � 0.77–2.26). Although severalprevious studies (27–29) have shown weak associations betweenprostate cancer risk and smoking, we found only modest andstatistically nonsignificant differences between the case and con-trol subjects in this study with regard to the percentages ofcurrent smokers, ex-smokers, and never smokers (Table 1).

At initial blood donation, plasma PSA levels were statisti-cally significantly higher (P � .012) in case subjects than incontrol subjects (Table 1). Eighty percent of the case subjects

and 20% of the control subjects had plasma PSA levels above4.0 ng/mL at the time of initial blood donation. At the time ofprostate cancer diagnosis, plasma PSA values for case subjectshad risen further (25th, 50th, and 75th percentiles were 6.3, 12.0,and 22.8 ng/mL, respectively).

Cross-sectional Interrelationships

We combined the data from case and control subjects toexamine the cross-sectional interrelationships between theplasma peptide levels, body height, and BMI after adjustmentfor age at sampling and case–control status (Table 2). All cor-relation coefficients were similar in the case and control groupswith the exception of height, which was weakly correlated withIGF-I (r � .19) and IGFBP-3 (r � .20) levels in the casesubjects but not in the control subjects (r � −.01 and .04),respectively. In both groups combined, there were positive cor-relations between plasma levels of IGF-I and IGFBP-3 (r �.59), between plasma levels of IGFBP-1 and IGFBP-2 (r � .36),and between BMI and plasma levels of insulin (r � .27). How-ever, BMI and insulin levels both correlated inversely withplasma levels of IGFBP-1 (r � −.38 and r � −.21, respectively)

Table 1. Selected baseline characteristics of prostate cancer case subjects and control subjects*

Characteristic Case subjects (n � 149) Control subjects (n � 298) Pdifference

Median age at recruitment, y 59.7 59.6Median age at diagnosis, y 63.0Height, cm† 176.3 (6.50) 174.8 (5.97) .02Weight, kg† 81.92 (10.98) 80.91 (13.01) .42Body mass index, kg/m2†,‡ 26.30 (2.92) 26.27 (4.10) .65Insulin, pmol/mL† 8.41 (11.41) 7.93 (8.83) .62IGF-I, ng/mL† 229.0 (70.27) 214.4 (57.50) .02IGFBP-1, ng/mL† 39.50 (24.22) 43.02 (28.15) .19IGFBP-2, ng/mL† 653.3 (371.8) 671.9 (428.3) .55IGFBP-3, ng/mL† 2611 (578.1) 2498.5 (539.5) .04Prostate-specific antigen at baseline, ng/mL§ 5.30 (3.39 and 9.34) 1.14 (0.65 and 2.23) .012Prostate-specific antigen at diagnosis, ng/mL§ 12.0 (6.3 and 22.8)Smoking status�

Current smoker 30 (20) 51 (18)Ex-smoker 39 (27) 96 (34)Never smoker 79 (53) 138 (48)

*IGF-I � insulin-like growth factor-I; IGFBP-1 � insulin-like growth factor-binding protein-1; IGFBP-2 � insulin-like growth factor-binding protein-2;IGFBP-3 � insulin-like growth factor-binding protein-3.

†Data shown as means (population standard deviations).‡Body mass index � (weight in kilograms)/(height in meters)2.§Data shown as median (25th and 75th percentile).�Data shown as number of subjects (percentage); information on smoking status was missing for three case subjects and 13 control subjects.

Table 2. Cross-sectional correlations* between insulin, IGF-I, IGFBPs, and anthropometric variables†

Body height BMI Insulin IGFBP-1 IGFBP-2 IGFBP-3 IGF-I PSA at baseline

BMI −0.02Insulin −0.03 0.27‡IGFBP-1 −0.10‡ −0.38‡ −0.21‡IGFBP-2 −0.01 −0.46‡ −0.25‡ 0.36‡IGFBP-3 0.10 0.06 −0.04 −0.18‡ −0.17‡IGF-I 0.07 −0.08 −0.05 −0.04 −0.11‡ 0.59‡PSA at baseline −0.01 −0.02 0.01 0.01 −0.17‡ 0.02 0.01PSA at diagnosis§ −0.08 0.02 0.01 0.14 0.04 −0.13 −0.07 0.83‡

*Pearson correlation coefficients adjusted for age at recruitment and case–control status; case subjects and control subjects combined for this analysis except whereindicated.

†BMI � body mass index; IGF-I � insulin-like growth factor-I; IGFBP-1 � insulin-like growth factor-binding protein-1; IGFBP-2 � insulin-like growthfactor-binding protein-2; IGFBP-3 � insulin-like growth factor-binding protein-3; PSA � prostate-specific antigen.

‡Correlation coefficients significantly different from 0 (P<.05).§Case subjects only.





ownloaded from


and IGFBP-2 (r � −.46 and r � −.25, respectively). PlasmaPSA levels at baseline for case and control subjects and at thetime of diagnosis (case subjects only) did not correlate withheight, BMI, or levels of insulin, IGF-I, or IGFBPs.

Smoking affects many metabolic pathways. We found thatsmoking status was statistically significantly associated withBMI (25.3 kg/m2 ± 3.2 in current smokers, 27.2 kg/m2 ± 3.6 inex-smokers, and 26.4 kg/m2 ± 3.8 in never smokers; P � .002)and plasma insulin levels (9.9 pmol/mL ± 14.1 in current smok-ers, 9.1 pmol/mL ± 12.1 in ex-smokers, and 6.8 pmol/mL ± 5.1in never smokers; P � .02). In addition, smoking status was alsostatistically significantly associated with levels of IGFBP-2 (833± 473 ng/mL in current smokers, 611 ± 376 ng/mL in ex-smokers, and 633 ± 392 ng/mL in never smokers; P � .0002)but not with levels of IGFBP-1 (45.1 ± 26.1 ng/mL in currentsmokers, 41.9 ± 27.6 ng/mL in ex-smokers, and 39.8 ± 25.1ng/mL in never smokers; P � .30). Smoking status did not showany associations at all with levels of IGF-I or IGFBP-3.

Logistic Regression Analyses of Prostate Cancer Risk

Prostate cancer risk in relation to plasma levels of IGF-I,IGFBP1–3, and insulin was analyzed by logistic regressionanalyses. We found a statistically significant trend of prostatecancer risk with increasing levels of IGF-I (Pfor trend � .02) andIGFBP-3 (Pfor trend � .03) as continuous variables. These in-creases in risk were reflected in ORs of 1.57 (95% CI � 0.88–2.81) and 1.56 (95% CI � 0.86–2.83) for top quartiles relativeto the bottom quartiles of IGF-I and IGFBP-3, respectively(Table 3). Plasma levels of insulin and the insulin-regulatedIGFBPs-1 and -2 did not show any statistically significant rela-tionship with prostate cancer risk. Adjustment for BMI andsmoking status did not materially alter the associations of IGF-Iand IGFBP-3 with prostate cancer risk (Table 3). The associa-tion of prostate cancer risk with IGF-I and IGFBP-3 levels ascontinuous variables remained statistically significant after ad-justing for height (Pfor trend � .03 and Pfor trend � .04, respec-tively). However, the OR estimates, after adjusting for height,moving from the bottom to the top quartile were somewhatlower for IGF-I (1.00 (referent), 0.80 [95% CI � 0.48–1.45],1.27 [95% CI � 0.72–2.23], and 1.39 [95% CI � 0.75–2.55])but did not change substantially for IGFBP-3 (1.00 (referent),1.52 [95% CI � 0.85–2.72], 1.45 [95% CI � 0.79–2.65], and1.60 [95% CI � 0.86–2.95]). Adjustment for IGFBP-3 levels,which were positively related to both IGF-I levels and prostatecancer risk, also practically cancelled the association of IGF-Ilevels to risk (Table 3). Conversely, adjustment for IGF-I levelsabolished the association of prostate cancer risk with IGFBP-3levels.

Lagtime, Tumor Characteristics, and Age

To investigate the influence of time between blood collectionand diagnosis on OR estimates, we excluded the 34 case subjectswhose diagnosis of prostate cancer was made less than 2 yearsafter the time of blood collection (and their matched controlsubjects). In this restricted analysis, the associations of risk withtotal IGF-I and IGFBP-3 levels was stronger compared withthe full-cohort analysis. OR estimates for increasing quartilesof IGF-I were 1.00 (referent), 1.18 (95% CI � 0.61–2.26), 1.68(95% CI � 0.88–3.21), and 2.18 (95% CI � 1.09–4.36); Pfor trend

� .003); for increasing quartiles of IGFBP-3, OR estimateswere 1.00 (referent), 1.53 (95% CI � 0.78–3.00), 1.68 (95% CI

� 0.87–3.26), and 1.74 (95% CI � 0.85–3.52); Pfor trend � .02).Excluding the 39 control subjects who had baseline plasma PSAlevels above 4.0 ng/mL did not materially alter any of the aboveOR estimates.

To examine whether the presence of patients with large tu-mors had a strong influence on the OR estimates, we excludedfrom our analysis the 44 case subjects who had large (T3–T4)tumors, lymph node or bone metastases, or poorly differentiatedtumors. We found that the exclusion of these subjects did notsubstantially alter OR estimates for IGF-I levels compared withthe full cohort analysis (ORs � 1.00 (referent), 1.24 [95% CI �0.61–2.49], 1.98 [95% CI � 1.01–3.85], and 1.66 [95% CI �0.83–3.31]; Pfor trend � .05) but led to lower OR estimates forIGFBP-3 (1.00 [referent], 1.58 [95% CI � 0.81–3.05], 1.24[95% CI � 0.64–2.40], and 1.35 [95% CI � 0.69–2.64];Pfor trend � .11). In each of these restricted analyses, levels ofinsulin, IGFBP-1, and IGFBP-2 remained unassociated withprostate cancer risk.

Previous studies (7,8) have shown contradictory data on theinfluence of age on the association between IGF-I levels andprostate cancer risk. In our study, most (80%) of the case sub-jects diagnosed with prostate cancer were aged 59–60 years orolder at the time of the health survey, while 20% were men aged58 years or younger. There were two clusters in age distributionbecause of the recruitment at even decades: 74% of the menwere aged 59–60 years, and 15% of the men were aged 49–50years. Men younger than 59 years had significantly higher IGF-Ilevels than did men older than 59 years (231 ng/mL ± 63 versus209 ng/mL ± 60; P<.005). The difference in mean IGF-I levelsbetween case and control subjects was larger in younger men(269 ng/mL ± 76 versus 228 ng/mL ± 54; P � .007) than inolder men (219 ± 65 versus 211 ± 58 ng/mL; P � .25). Whenwe restricted our logistic regression analyses to men below age59 years at study entry, a strong positive trend in risk was foundfor increasing tertiles of IGF-I but not for increasing tertiles ofIGFBP-3. Furthermore, adjustment for IGFBP-3 levels did notdiminish the association of prostate cancer risk with IGF-I levelsin these younger men. In contrast, adjustment for IGF-I levelscaused IGFBP-3 levels to be negatively associated with prostatecancer risk, although the trend was not statistically significant(Table 4).

DISCUSSION

In this prospective cohort study, we observed an increase inprostate cancer risk with rising levels of circulating IGF-I andIGFBP-3. In subgroup analyses, we observed a particularlystrong association between prostate cancer risk and IGF-I levelsin men who were younger than 59 years at recruitment. Plasmalevels of insulin and IGFBPs-1 and -2, which are themselvesnegatively regulated by insulin, were not associated with anyincrease or decrease in prostate cancer risk, either in the wholecohort or in subgroups.

Methodologic Considerations

A methodologic strength of our study is that blood sampleswere obtained before cancer diagnosis, which makes it unlikelythat levels of IGF-I, IGFBPs, or insulin had been affected byalterations in metabolism that might be caused by the presenceof an advanced prostate tumor. This conclusion is corroboratedby the fact that OR estimates remained unchanged after we





ownloaded from


Table 3. Odds ratios of prostate cancer for quartiles of plasma peptide measurements*

Quartile level†

Pfor trend‡1 2 3 4

IGF-INo. of case subjects 33 31 42 43 .02No. of control subjects 77 81 69 68Mean§ 143.0 200.6 235.7 297.3OR (95% CI) 1.00 (referent) 0.89 (0.50–1.57) 1.48 (0.85–2.59) 1.57 (0.88–2.81)

IGF-I adjusted for BMI and smokingNo. of case subjects 32 33 35 40 .006No. of control subjects 70 71 69 63Mean§ 143.0 200.7 234.5 297.1OR (95% CI) 1.00 (referent) 1.00 (0.56–1.78) 1.21 (0.66–2.20) 1.72 (0.93–3.19)

IGFBP-1No. of case subjects 39 41 35 34 .19No. of control subjects 71 70 75 77Mean§ 14.1 28.7 45.0 79.0OR (95% CI) 1.00 (referent) 1.07 (0.61–1.87) 0.85 (0.48–1.51) 0.80 (0.44–1.43)

IGFBP-1 adjusted for BMI and smokingNo. of case subjects 39 35 30 35 .23No. of control subjects 63 68 73 68Mean§ 17.7 30.2 42.4 76.4OR (95% CI) 1.00 (referent) 0.73 (0.40–1.34) 0.60 (0.32–1.13) 0.74 (0.39–1.40)

IGFBP-2No. of case subjects 33 38 40 32 .64No. of control subjects 72 68 66 74Mean§ 217.0 476.4 733.2 1231.6OR (95% CI) 1.00 (referent) 1.17 (0.66–2.08) 1.29 (0.74–2.23) 0.92 (0.51–1.67)

IGFBP-2 adjusted for BMI and smokingNo. of case subjects 33 35 35 32 .49No. of control subjects 65 64 64 67Mean§ 264.7 475.8 704.0 1216.5OR (95% CI) 1.00 (referent) 1.18 (0.64–2.15) 1.07 (0.58–2.00) 0.83 (0.44–1.57)

IGFBP-3No. of case subjects 31 37 39 41 .03No. of control subjects 79 74 72 70Mean§ 1856 2365 2667 3250OR (95% CI) 1.00 (referent) 1.31 (0.74–2.31) 1.42 (0.80–2.52) 1.56 (0.86–2.83)

IGFBP-3 adjusted for BMI and smokingNo. of case subjects 28 34 38 39 .007No. of control subjects 74 69 65 65Mean§ 1861 2363 2655 3233OR (95% CI) 1.00 (referent) 1.49 (0.82–2.71) 1.70 (0.93–3.11) 1.83 (0.98–3.24)

InsulinNo. of case subjects 37 35 45 32 .72No. of control subjects 74 73 69 79Mean§ 2.9 5.0 6.8 17.6OR (95% CI) 1.00 (referent) 1.01 (0.57–1.78) 1.36 (0.78–2.36) 0.83 (0.48–1.42)

Insulin adjusted for BMI and smokingNo. of case subjects 32 36 42 30 .23No. of control subjects 70 68 61 74Mean§ 4.3 5.2 6.6 16.4OR (95% CI) 1.00 (referent) 1.20 (0.61–2.34) 1.62 (0.86–3.04) 0.98 (0.53–1.81)

IGF-I adjusted for IGFBP-3No. of case subjects 37 31 36 44 .15No. of control subjects 72 80 75 67Mean§ 160.2 207.0 227.7 283.0OR (95% CI) 1.00 (referent) 0.73 (0.41–1.3) 0.98 (0.56–1.72) 1.32 (0.73—2.39)

IGFBP-3 adjusted for IGF-INo. of case subjects 36 37 34 41 .48No. of control subjects 74 73 77 70Mean§ 2021 2364 2633 3133OR (95% CI) 1.00 (referent) 1.07 (0.61–1.89) 0.96 (0.52–1.74) 1.19 (0.66–2.15)

*All odds ratio (OR) estimates are adjusted for age and fasting time. Numbers of case and control subjects do not always add up to the total of 149 and 298,respectively, because of missing values for the measurements of IGF-I (three control subjects), IGFBP-1 (five control subjects), IGFBP-2 (six case subjects and 18control subjects), IGFBP-3 (one case subject and three control subjects), and insulin (three control subjects). In the analyses adjusted for BMI and smoking status,up to nine additional case subjects and 19 additional control subjects were excluded because of missing values in the latter adjustment variables. BMI � body massindex; IGF-I � insulin-like growth factor-I; IGFBP-1 � insulin-like growth factor-binding protein-1; IGFBP-2 � insulin-like growth factor-binding protein-2;IGFBP-3 � insulin-like growth factor-binding protein-3; PSA � prostate-specific antigen; CI � confidence interval.

†Quartile cut points determined on case subjects plus control subjects.‡Tests for trend were performed on the original measurements, as a continuous variable.§Mean levels of peptides (unadjusted) within the exposure category; units are ng/mL for IGF-I and the three IGFBPs and pmol/mL for insulin.





ownloaded from


excluded either those case subjects whose cancer was diagnosedless than 2 years after blood donation or those with large, me-tastasized, or poorly differentiated tumors. Furthermore, the pro-spective design of our study, which included a complete follow-up for cancer incidence, minimized the possibility of any biasthat might have been caused by systematic case–control differ-ences in blood sample collection, processing, and storage or byselection of case and control subjects from different sourcepopulations. Several studies (7,30,31) have shown that bloodlevels of IGF-I and IGFBP-3 have relatively low within-subjectvariation over time, so that measurements in a single bloodsample should provide a reliable indicator of a subject’s habitualblood concentrations of these peptides over time periods up to atleast 1 year.

Influence of IGF-I and IGFBP-3 on ProstateCancer Development

The positive association that we have observed betweenplasma IGF-I levels and prostate cancer risk is consistent withsimilar findings from a prospective cohort study of U.S. malephysicians (7) and two case-control studies (8,9). It is interestingthat recent prospective cohort studies have also shown that in-creases in the risks of colon (32) and breast (33) cancers areassociated with elevated plasma IGF-I levels. Together, thesefindings suggest that an elevation of plasma IGF-I levels mightbe a common risk factor for different forms of cancer. Theobserved increase in total plasma IGF-I concentrations in mendeveloping prostate cancer may reflect a relative elevation ofpituitary secretion of GH, which provides the main stimulus forsynthesis of IGF-I and IGFBP-3 in the liver as well as in othertissues (1,11).

Our findings of increased prostate cancer risk in men withelevated IGFBP-3 levels are inconsistent with results obtained in

other studies. Neither the Physicians’ Health Study (7) nor onecase–control study (8) found an association between prostatecancer risk and total IGFBP-3 levels. Furthermore, in the Phy-sicians’ Health Study, the association of risk with IGF-I levelswas considerably stronger when IGF-I levels were adjusted forlevels of IGFBP-3, whereas levels of IGFBP-3 itself were foundto be significantly inversely related to risk when adjusted forIGF-I (7). In our study, adjustment for IGFBP-3 levels reducedthe association of risk with IGF-I levels in a full cohort analysisbut did not affect this association when the analysis was re-stricted to younger men. We observed an inverse, but statisti-cally nonsignificant, association of risk with IGFBP-3 levelsonly in younger men when IGFBP-3 levels were adjusted forIGF-I levels. In addition to revealing that younger men have ahigher risk of prostate cancer as IGF-I levels increase, our datashow that younger men had higher mean IGF-I levels than oldermen. Pituitary GH secretion, plasma IGF-I levels, and the IGF-I/IGFBP-3 ratio all peak during adolescence and then graduallydecline with age (34–37). Furthermore, many epidemiologicstudies (25,26,38–40), including ours, have shown positive as-sociations of prostate cancer risk with adult height, which maybe a possible marker of bioavailable IGF-I levels during theprepubertal and adolescent growth spurt. Therefore, it seemslikely that the OR estimates of prostate cancer for quartiles ofIGF-I corrected for height in our analysis were most likely over-adjusted. Taken together, these various observations fit a modelin which prostate cancer risk is specifically increased in menwho have elevated IGF-I levels at a young age and whose IGF-Ilevels, therefore, show a stronger age-related decline than dothose in men with lower initial IGF-I levels. Elevated IGF-Ilevels during adolescence and early adulthood may favor thedevelopment of early (pre) neoplastic lesions, which may prog-ress into clinically manifest cancers much later in life.

Table 4. Odds ratios of prostate cancer for tertiles of plasma IGF-I and IGFBP-3 in men below age 59 years at study entry*

Tertile level†

Pfor trend§1 2 3

IGF-INo. of case subjects 5 11 14 .01No. of control subjects 24 19 17Mean‡ 175.5 239.3 305.4OR (95% CI) 1.00 (referent) 2.97 (0.84–10.49) 4.30 (1.19–15.50)

IGF-I adjusted for IGFBP-3No. of case subjects 8 5 17 .01No. of control subjects 21 24 15Mean‡ 193.3 237.8 294.6OR (95% CI) 1.00 (referent) 0.61 (0.16–2.31) 4.75 (1.15–19.61)

IGFBP-3No. of case subjects 9 12 9 .44No. of control subjects 20 18 22Mean‡ 2121 2673 3345OR (95% CI) 1.00 (referent) 1.55 (0.48–4.95) 0.86 (0.25–2.99)

IGFBP-3 adjusted for IGF-INo. of case subjects 12 10 8 .20No. of control subjects 18 20 22Mean‡ 2214 2746 3257OR (95% CI) 1.00 (referent) 0.58 (0.14–2.40) 0.31 (0.068–1.38)

*All odds ratio (OR) estimates are adjusted for age. IGF-I � insulin-like growth factor-I; IGFBP-3 � insulin-like growth factor-binding protein-3; CI �

confidence interval.†Tertile cut points determined on case subjects plus control subjects.‡Mean levels of peptides (unadjusted) within the exposure category; units are ng/mL for IGF-I and the three IGFBPs and pmol/mL for insulin.§Tests for trend were performed on continuous variables.





ownloaded from


Insulin, IGFBP-1, and IGFBP-2

Circulating levels of IGF-I and IGFBPs-1, -2, and -3 areregulated by both GH and insulin. GH provides the main stimu-lus for the synthesis of IGF-I and IGFBP-3, whereas insulincan increase IGF-I bioactivity by inhibiting the synthesis ofIGFBPs-1 and -2. It is interesting that we found that plasmalevels of insulin and the insulin-dependent IGFBPs -1 and -2were not related to prostate cancer risk. It is unlikely that an ORof any importance would have been missed because of randomerrors in the measurements of peptide levels. Most (96%) bloodsamples were collected at least 4 hours after last consumptionof any food or drink, an interval that is usually sufficient forlevels of insulin and IGFBP-1 to return to fasting levels. Previ-ous studies (31,41) have shown that measurements of insulin,IGFBP-1, and IGFBP-2 levels, especially in blood from fastingindividuals, generally show relatively little intraindividual varia-tion over time, suggesting that a single blood sample is adequateto test for associations with cancer risk in epidemiologic studies.Furthermore, we found no association between prostate cancerrisk and high BMI, which is a major determinant of hyperinsu-linemia and decreased IGFBP-1 and IGFBP-2 levels. This ab-sence of association fits with results from many (27,39,42,43),although not all (25,26), epidemiologic studies on BMI andprostate cancer risk.

It is difficult to explain why only the absolute level of circu-lating IGF-I is related to prostate cancer risk, while an increasein IGF-I bioactivity that may result from insulin-induced reduc-tions in IGFBP-1 and IGFBP-2 levels apparently is not. First, itis only partially understood how IGF-I bioactivity within theprostate is quantitatively determined by the concentrations ofIGF-I and IGFBPs. For example, although reductions in levelsof IGFBPs-1, -2, or -3 are generally thought to increase IGF-Ibioactivity, some in vitro studies (10,44) have shown that theseIGFBPs can sometimes actually potentiate IGF-I receptor bind-ing and action, depending on their concentrations relative tothose of IGF-I and on the cell type studied. Second, the relativeconcentrations of IGF-I and IGFBPs in the circulation may dif-fer from those in the prostate gland because some of the circu-lating IGFBPs can diffuse more easily through the endothelialbarrier than others and because IGF-I and IGFBPs may also besynthesized in the prostate. Furthermore, circulating levels ofIGF-I and IGFBPs may also reflect the levels of synthesis ofthese peptides within the prostate itself because the same factors(e.g., GH and insulin) that regulate the synthesis of these pep-tides within the liver also regulate their synthesis within theprostate.

Implications for Prevention

Prostate cancer incidence rates vary widely between high-riskareas, such as the United States and Scandinavia, and low-risk areas, such as Northern Africa or Southeast Asia. In high-risk areas, average body height and prostate cancer incidencerates have increased in parallel since the early 1900s, and growthrates during puberty and adolescence are related to levels ofIGF-I. These observations, together with our results, would sug-gest that elevated IGF-I levels might provide a physiologic linkbetween a Western lifestyle that is characterized by an energy-dense diet and an increased risk of prostate cancer. However,although chronic energy restriction has been shown to decreasecirculating IGF-I levels, obesity, which is a reflection of long-

term positive energy balance, is not related to an increase inIGF-I levels compared with the normally nourished but non-obese state (11). In addition to (or possibly in interaction with)nutritional lifestyle, genetic predisposition has been shown todetermine a large part (30%–60%) of variation in circulatingIGF-I levels (45–48). Identification of the nutritional and geneticfactors that may cause the interindividual variations in IGF-Ilevels is an important area for future research.

In conclusion, our results indicate that elevated plasma IGF-Ilevels may be an important factor in the etiology of prostatecancer. In contrast, high levels of plasma insulin and insulin-induced reductions in IGFBP levels do not appear to be relatedto prostate cancer development. Studies on factors influencingplasma levels of IGF-I may provide insights that could be usedin the design of novel preventive and therapeutic strategiesagainst prostate cancer.

REFERENCES

(1) Jones JI, Clemmons DR. Insulin-like growth factors and their bindingproteins: biological actions. Endocr Rev 1995;16:3–34.

(2) Dunn SE, Kari FW, French J, Leininger JR, Travlos G, Wilson R, et al.Dietary restriction reduces insulin-like growth factor I levels, which modu-lates apoptosis, cell proliferation, and tumor progression in p53-deficientmice. Cancer Res 1997;57:4667–72.

(3) Cohen P, Peehl DM, Lamson G, Rosenfeld RG. Insulin-like growth factors(IGFs), IGF receptors, and IGF-binding proteins in primary cultures ofprostate epithelial cells. J Clin Endocrinol Metab 1991;73:401–7.

(4) Iwamura M, Sluss PM, Casamento JB, Cockett AT. Insulin-like growthfactor I: action and receptor characterization in human prostate cancer celllines. Prostate 1993;22:243–52.

(5) Pollak M, Beamer W, Zhang JC. Insulin-like growth factors and prostatecancer. Cancer Metastasis Rev 1999;17:383–90.

(6) Prager D, Li HL, Asa S, Melmed S. Dominant negative inhibition oftumorigenesis in vivo by human insulin-like growth factor I receptor mu-tant. Proc Natl Acad Sci U S A 1994;91:2181–5.

(7) Chan JM, Stampfer MJ, Giovannucci E, Gann PH, Ma J, Wilkinson P, etal. Plasma insulin-like growth factor-I and prostate cancer risk: a prospec-tive study. Science 1998;279:563–6.

(8) Wolk A, Mantzoros CS, Andersson SO, Bergstrom R, Signorello LB, La-giou P, et al. Insulin-like growth factor 1 and prostate cancer risk: a popu-lation-based, case-control study. J Natl Cancer Inst 1998;90:911–5.

(9) Mantzoros CS, Tzonou A, Signorello LB, Stampfer M, Trichopoulos D,Adami HO. Insulin-like growth factor 1 in relation to prostate cancer andbenign prostatic hyperplasia. Br J Cancer 1997;76:1115–8.

(10) Rajaram S, Baylink DJ, Mohan S. Insulin-like growth factor-binding pro-teins in serum and other biological fluids: regulation and functions. EndocrRev 1997;18:801–31.

(11) Thissen JP, Ketelslegers JM, Underwood LE. Nutritional regulation of theinsulin-like growth factors. Endocr Rev 1994;15:80–101.

(12) Pollard M, Luckert PH, Snyder D. Prevention of prostate cancer and livertumors in L-W rats by moderate dietary restriction. Cancer 1989;64:686–90.

(13) Donaghy AJ, Baxter RC. Insulin-like growth factor bioactivity and itsmodification in growth hormone resistant states. Baillieres Clin EndocrinolMetab 1996;10:421–46.

(14) Nam SY, Lee EJ, Kim KR, Cha BS, Song YD, Lim SK, et al. Effect ofobesity on total and free insulin-like growth factor (IGF)-1, and their re-lationship to IGF-binding protein (BP)-1, IGFBP-2, IGFBP-3, insulin, andgrowth hormone. Int J Obes Relat Metab Disord 1997;21:355–9.

(15) Nyomba BL, Berard L, Murphy LJ. Free insulin-like growth factor I (IGF-I) in healthy subjects: relationship with IGF-binding proteins and insulinsensitivity. J Clin Endocrinol Metab 1997;82:2177–81.

(16) Scacchi M, Pincelli AI, Cavagnini F. Growth hormone in obesity. IntJ Obes Relat Metab Disord 1999;23:260–71.

(17) Mukherjee P, Sotnikov AV, Mangian HJ, Zhou JR, Visek WJ, Clinton SK.Energy intake and prostate tumor growth, angiogenesis, and vascular en-dothelial growth factor expression. J Natl Cancer Inst 1999;91:512–23.





ownloaded from


(18) Rosen CJ, Pollak M. Circulating IGF-I: new perspectives for a new cen-tury. Trends Endocrinol Metab 1999;10:136–41.

(19) Soderberg S, Ahren B, Jansson JH, Johnson O, Hallmans G, Asplund K, etal. Leptin is associated with increased risk of myocardial infarction. J InternMed 1999;246:409–18.

(20) Mattsson B, Wallgren A. Completeness of the Swedish Cancer Register.Non-notified cancer cases recorded on death certificates in 1978. ActaRadiol Oncol 1984;23:305–13.

(21) Oncological Centre, Umeå University Hospital, Sweden. Prostate cancer;primary registration in Northern Sweden 1992–1997.

(22) Union International Contre le Cancer (UICC). In: TNM classification ofmalignant tumours. 4th ed. Berlin (Germany): Springer; 1992.

(23) Mostofi FK, Sesterhenn IA, Sobin LH. Histological typing of tumours. No.22. In: International histological classification of tumours. Geneva (Swit-zerland): World Health Organization; 1980.

(24) Vessella RL, Noteboom J, Lange PH. Evaluation of the Abbott IMx Au-tomated Immunoassay of Prostate-Specific Antigen. Clin Chem 1992;38:2044–54.

(25) Andersson SO, Wolk A, Bergstrom R, Adami HO, Engholm G, Englund A,et al. Body size and prostate cancer: a 20-year follow-up study among135,006 Swedish construction workers. J Natl Cancer Inst 1997;89:385–9.

(26) Gronberg H, Damber L, Damber JE. Total food consumption and bodymass index in relation to prostate cancer risk: a case-control study inSweden with prospectively collected exposure data. J Urol 1996;155:969–74.

(27) Giovannucci E, Rimm EB, Ascherio A, Colditz GA, Spiegelman D,Stampfer MJ, et al. Smoking and risk of total and fatal prostate cancer inUnited States health professionals. Cancer Epidemiol Biomarkers Prev1999;8(4 Pt 1):277–82.

(28) Coughlin SS, Neaton JD, Sengupta A. Cigarette smoking as a predictor ofdeath from prostate cancer in 348,874 men screened for the Multiple RiskFactor Intervention Trial. Am J Epidemiol 1996;143:1002–6.

(29) Lumey LH. Prostate cancer and smoking: a review of case–control andcohort studies. Prostate 1996;29:249–60.

(30) Goodman-Gruen D, Barrett-Connor E. Epidemiology of insulin-likegrowth factor-I in elderly men and women. The Rancho Bernardo Study.Am J Epidemiol 1997;145:970–6.

(31) Kaaks R, Toniolo P, Akhmedkhanov A, Lukanova A, Biessy C, DechaudH, et al. Serum C-peptide, insulin-like growth factor (IGF)-I, IGF-bindingproteins, and colorectal cancer risk in women. J Natl Cancer Inst 2000;92:1592–600.

(32) Ma J, Pollak MN, Giovannucci E, Chan JM, Tao Y, Hennekens CH, et al.Prospective study of colorectal cancer risk in men and plasma levels ofinsulin-like growth factor (IGF)-I and IGF-binding protein-3. J Natl CancerInst 1999;91:620–5.

(33) Hankinson SE, Willett WC, Colditz GA, Hunter DJ, Michaud DS, Deroo B,et al. Circulating concentrations of insulin-like growth factor-I and risk ofbreast cancer. Lancet 1998;351:1393–6.

(34) Veldhuis JD, Iranmanesh A. Physiological regulation of the human growthhormone (GH)-insulin-like growth factor type I (IGF-I) axis: predominantimpact of age, obesity, gonadal function, and sleep. Sleep 1996;19:S221–4.

(35) Juul A, Main K, Blum WF, Lindholm J, Ranke MB, Skakkebaek NE. Theratio between serum levels of insulin-like growth factor (IGF)-I and the IGFbinding proteins (IGFBP-1, 2 and 3) decreases with age in healthy adultsand is increased in acromegalic patients. Clin Endocrinol (Oxf) 1994;41:85–93.

(36) Juul A, Bang P, Hertel NT, Main K, Dalgaard P, Jorgensen K, Muller J, etal. Serum insulin-like growth factor-I in 1030 healthy children, adolescents,and adults: relation to age, sex, stage of puberty, testicular size, and bodymass index. J Clin Endocrinol Metab 1994;78:744–52.

(37) Hesse V, Jahreis G, Schambach H, Vogel H, Vilser C, Seewald HJ, et al.Insulin-like growth factor I correlations to changes of the hormonal statusin puberty and age. Exp Clin Endocrinol 1994;102:289–98.

(38) Giovannucci E, Rimm EB, Stampfer MJ, Colditz GA, Willett WC. Height,body weight, and risk of prostate cancer. Cancer Epidemiol BiomarkersPrev 1997;6:557–63.

(39) Nilsen TI, Vatten LJ. Anthropometry and prostate cancer risk: a prospec-tive study of 22,248 Norwegian men. Cancer Causes Control 1999;10:269–75.

(40) Tulinius H, Sigfusson N, Sigvaldason H, Bjarnadottir K, Tryggvadottir L.Risk factors for malignant diseases: a cohort study on a population of22,946 Icelanders. Cancer Epidemiol Biomarkers Prev 1997;6:863–73.

(41) Muti P, Trevisan M, Micheli A, Krogh V, Bolelli G, Sciajno R, et al.Reliability of serum hormones in premenopausal and postmenopausalwomen over a one-year period. Cancer Epidemiol Biomarkers Prev 1996;5:917–22.

(42) Villeneuve PJ, Johnson KC, Kreiger N, Mao Y. Risk factors for prostatecancer: results from the Canadian National Enhanced Cancer SurveillanceSystem. The Canadian Cancer Registries Epidemiology Research Group.Cancer Causes Control 1999;10:355–67.

(43) Whittemore AS, Kolonel LN, Wu AH, John EM, Gallagher RP, Howe GR,et al. Prostate cancer in relation to diet, physical activity, and body size inblacks, whites, and Asians in the United States and Canada. J Natl CancerInst 1995;87:652–61.

(44) Kelley KM, Oh Y, Gargosky SE, Gucev Z, Matsumoto T, Hwa V, et al.Insulin-like growth factor-binding proteins (IGFBPs) and their regulatorydynamics. Int J Biochem Cell Biol 1996;28:619–37.

(45) Hall K, Hilding A, Thoren M. Determinants of circulating insulin-likegrowth factor-I. J Endocrinol Invest 1999;22(5 Suppl):48–57.

(46) Hong Y, Pedersen NL, Brismar K, Hall K, de Faire U. Quantitative geneticanalyses of insulin-like growth factor I (IGF-I), IGF-binding protein-1, andinsulin levels in middle-aged and elderly twins. J Clin Endocrinol Metab1996;81:1791–7.

(47) Harrela M, Koistinen H, Kaprio J, Lehtovirta M, Tuomilehto J, Eriksson J,et al. Genetic and environmental components of interindividual variation incirculating levels of IGF-I, IGF-II, IGFBP-1, and IGFBP-3. J Clin Invest1996;98:2612–5.

(48) Verhaeghe J, Loos R, Vlietinck R, Herck EV, van Bree R, Schutter AM.C-peptide, insulin-like growth factors I and II, and insulin-like growthfactor binding protein-1 in cord serum of twins: genetic versus environ-mental regulation. Am J Obstet Gynecol 1996;175:1180–8.

NOTES

Supported by grants from The Swedish Cancer Society, The Lions ResearchFoundation, Umeå, Sweden, and The Medical Faculty, Umeå University.

We thank all of the participants in the VIP and MONICA projects; Åsa Ågrenand Charlotte Ingri for their assistance in project coordination and data handling;and David Achaintre, Beatrice Vozar, and Francine Claustrat for their help withlaboratory assays.

Manuscript received March 29, 2000; revised September 18, 2000; acceptedSeptember 25, 2000.





ownloaded from


plasma insulin-like growth factor-i, insulin-like growth factor-binding proteins, and prostate...

Documents