Selenium, Genetic Variation, and Prostate CancerRisk: Epidemiology Reflects Back on Selenium andVitamin E Cancer Prevention TrialElizabeth A. Platz, Department of Epidemiology, Johns Hopkins Bloomberg School of Public Health; James Buchanan BradyUrological Institute; and Sidney Kimmel Comprehensive Cancer Center, Johns Hopkins Medical Institutions, Baltimore, MD

Scott M. Lippman, Divisions of Cancer Medicine and Cancer Prevention and Population Sciences, The University of TexasM. D. Anderson Cancer Center, Houston, TX

The randomized, controlled Selenium and Vitamin E CancerPrevention Trial (SELECT)1 found that selenium supplementationdid not decrease the risk of prostate cancer, which contradicted asecondary finding from the Nutritional Prevention of Cancer (NPC)Study.2 In the wake of SELECT, we expect that many groups willreport results fromepidemiologic studies on seleniumandprostatecancer, some of which may illuminate reasons for SELECTsnull results.

One such effort is reported in this issue of Journal of ClinicalOncology by Chan et al,3 who conducted a cross-sectional epidemio-logic study of plasma selenium, an SOD2 variant, and aggressive pros-tate cancer in men diagnosed with clinically localized or locallyadvanced prostate cancer and who provided a blood specimen. Theyobserved an unexpected positive association between plasma sele-nium and aggressive disease. As expected, an SOD2 variant, a substi-tutionof alanine for valine at aminoacid16 in the antioxidant enzymemanganese superoxide dismutase, was not statistically significantlyassociatedwith aggressive disease, although the relative risks (RRs) foroneor twoalanine alleleswere above 1.0. Twopatterns emerged in theanalysis of the joint association of plasma selenium and SOD2: menwith low selenium and two alanine alleles had a higher risk of aggres-sive disease (compared with low selenium and none or one alaninealleles); high selenium seemingly protected men with two alaninealleles; and men with high selenium and none or one alanine allelehad a higher risk of aggressive disease (compared with low seleniumand none or one alanine allele). In light of these findings, Chan et al3

commented that their data indicate caution against broad use ofselenium supplementation for men with prostate cancer and thatcomplete interpretation of results from SELECT may depend onassessment of SOD2 genotype in trial participants.

Before discussing the broader context (related observationalstudies, the NPC study, and SELECT) and implications of this study,we will discuss a design feature that complicates drawing etiologicinferencesand thus implications for selenium supplementationfromtheChanetal study.3Theyestimated theprevalenceofaggressivedisease in a cohort of men with prostate cancer, making the denomi-nator for risk calculationmenwith prostate cancer. Etiologic researchon aggressive prostate cancer typically estimates risk of aggressive

disease in a cohort of men who do not have a diagnosis of prostatecancer at baseline, making at-risk men without prostate cancer thedenominator. Table 1 illustrates several hypothetical scenarios (a sub-set of all possible scenarios) that could explain the higher risk ofaggressive prostate cancer found by Chan et al in men with higherplasma selenium. Columns 4 to 6 involve a typical cohort of at-riskmen. The cohort of cancer patients (aggressive and nonaggressive)involved incolumn7 isderived fromtheat-risk cohort andreflects thedesign of Chan et al. Although all are hypothetical, the data andassumptions in Table 1 are consistent with the epidemiologic litera-ture on nutrients and genetic variants and are not out of line with thedata of Chan et al (the same can be said of the hypothetical data andassumptions in Table 2, which is discussed below). Scenario D is aplausible scenario in which the increased RR of aggressive disease(1.25) in a design similar to that of the study by Chan et al would bemisleading because the true RR of aggressive disease is 0.75, as illus-tratedby themore informative typical cohort design. ScenariosA toD are all plausible but unknowable because of this studys design.Differences in implication make it absolutely critical to discern thecorrect scenario behind the Chan et al results. The implication ofhypothetical scenarios A and Ccaution against taking a seleniumsupplementis diametrically opposed to that of B and Da recom-mendation for taking a selenium supplement. The typical cohortdesign for etiologic research, using at-risk individuals, enables correctimplications that are not derivable fromadesignusing cancer patientsat baseline.

The design of the study byChan et al is frequently used in clinicalstudies of potential biomarker associationsor correlationswith canceraggressiveness, where inferences are unambiguous (barring othersources of error). This design also sometimes is used for etiologicresearch, but investigators should be alert to the etiologic ambiguity itraises for interpreting a positive or inverse association between anexposure and aggressive disease.

Previously published studies in at-riskmenmay help in drawinginferences from the Chan et al3 findings on the main effects of sele-niumand SOD2 and their joint effects, which are evenmore challeng-ing to interpret with a cohort of only cancer patients. Regardingassociations between circulating seleniumandoverall prostate cancer,

JOURNAL OF CLINICAL ONCOLOGY E D I T O R I A L

VOLUME 27 NUMBER 22 AUGUST 1 2009

2009 by American Society of Clinical Oncology 3569Journal of Clinical Oncology, Vol 27, No 22 (August 1), 2009: pp 3569-3572DOI: 10.1200/JCO.2009.22.2117; published online ahead of print at www.jco.org on June 15, 2009

most large studies foundnoassociation,4-7whether the study involvedpopulations with low or high selenium exposure or with or withoutroutineprostate-specific antigen (PSA) screening (one small studydidfind an inverse association8). Regarding associations between sele-nium levels and aggressive disease, a study of selenium content intoenails9 and some other studies4,6 reported inverse, albeit nonlinear,associations and other studies reported no association.5,6 Regardingassociations between selenium and nonaggressive disease,4-6 only thePhysicians Health Study (PHS)4 found an association, which was aninverse association with nonaggressive disease diagnosed before butnot since the PSA era began. If selenium is not associatedwith nonag-gressive disease in thePSA era, then theChan et al finding of a positiveassociation for aggressive disease likely is not explained by scenario B,C, or D in Table 1 (suggesting prevention of nonaggressive disease).The worrisome scenario A showing an increased risk of aggressivediseasewith selenium,however, contradicts thefindings of some largestudies.4,6,9 Again, we cannot be certain which scenario is correct

because of the etiologic ambiguity resulting from the Chan et alstudy design.

Large studies with a risk-estimation denominator of at-riskmensuggest that the SOD2 alanine allele is positively associated with pros-tate cancer,10-12 although so weakly in some studies that the authorsconcluded no association.13,14 In studies finding a moderate associa-tion, it was present for aggressive and nonaggressive disease10,12 butpossibly was stronger for the former. The Chan et al3 results areconsistent with these results10,12; if the alanine allele is more stronglypositively associated with aggressive than nonaggressive disease in atypical, at-risk cohort design, then the RR of aggressive disease wouldbe above 1.0 when using the Chan et al design (Table 2, in which thehypothetical scenario is constructed as are those in Table 1).

The PHS and another cohort study with an at-risk denomina-tor for risk estimation11,13 found a joint association for two alaninealleles plus low seleniumwith ahigherprostate cancer risk (comparedwith none or one alanine alleles and low selenium), which was more

Table 1. Hypothetical Scenarios Involving an At-Risk Cohort of Men That May Explain the Apparent Positive Association Made by Chan et al3 Between HigherCirculating Selenium Level and Aggressive Prostate Cancer

Scenario

Hypothetical True Direction ofAssociation Between

Selenium and ProstateCancer

Typical Cohort Design for Etiologic Research: Denominator Is At-Risk Men Without ProstateCancer at Baseline

Chan et al3 Design:Denominator Is Men

With Prostate Cancer

Aggressive NonaggressiveRR of Total Prostate

CancerRR of Aggressive Prostate

CancerRR of Nonaggressive

Prostate CancerRR of AggressiveProstate Cancer

A 1 Null (53.3 60/1,000)/(40 60/1,000) 1.13

(53.3/1,000)/(40/1,000) 1.33 (60/1,000)/(60/1,000) 1.00 (53.3/53.3 60)/(40/40 60) 1.18

B Null 2 (40 45/1,000)/(40 60/1,000) 0.85

(40/1,000)/(40/1,000) 1.00 (45/1,000)/(60/1,000) 0.75 (40/40 45)/(40/40 60) 1.18

C 1 2 (53.3 45/1,000)/(40 60/1,000) 1.00

(53.3/1,000)/(40/1,000) 1.33 (45/1,000)/(60/1,000) 0.75 (53.3/53.3 45)/(40/40 60) 1.36

D 2 22 (30 30/1,000)/(40 60/1,000) 0.60

(30/1,000)/(40/1,000) 0.75 (30/1,000)/(60/1,000) 0.50 (30/30 30)/(40/40 60) 1.25

Abbreviation: RR, relative risk.1 and 2 signify an increase or decrease in risk of 1.33 or 0.75 (these RRs are equidistant from the null on the natural logarithm scale), respectively in the

high-selenium group. 22 signifies a decrease in risk of 0.5 in the high-selenium group.The cohort in these columns consists of 2,000 at-risk men, half of whom are in the high-selenium group. In the low-selenium group, 100 men develop prostate

cancer, 40% of which is aggressive disease.The cohort in this column consists of the prostate cancer cases that are diagnosed in the cohort in .Denominator is all prostate cancer cases in the high-selenium group.Denominator is all prostate cancer cases in the low-selenium group.

Table 2. Hypothetical Scenario Involving an At-Risk Cohort of Men That May Explain the Apparent Positive Association Made by Chan et al3 Between theAlanine Allele of SOD2 and Aggressive Prostate Cancer

Hypothetical True Direction ofAssociation Between the Alanine

Allele and Prostate CancerTypical Cohort Design: Denominator Is Men Without Prostate Cancer at Baseline

Chan et al3 Design:Denominator Is Men

With Prostate Cancer

Aggressive NonaggressiveRR of Total Prostate

CancerRR of Aggressive Prostate

CancerRR of Nonaggressive

Prostate CancerRR of AggressiveProstate Cancer

11 1 (80 79.8/1,000)/(40 60/1,000) 1.60

(80/1,000)/(40/1,000) 2.00 (79.8/1,000)/(60/1,000) 1.33 (80/80 79.8)/(40/40 60) 1.25

Abbreviation: RR, relative risk.1signifies an increase in risk of 1.33 in the group with 2 alanine alleles. 11 signifies an increase in risk of 2.0 in the group with 2 alanine alleles.The cohort in these columns consists of 2,000 at-risk men, sampled so that half have two alanine alleles and half have zero alanine alleles. In the group with zero

alanine alleles, 100 men develop prostate cancer, 40% of which is aggressive disease.The cohort in this column consists of the prostate cancer cases that are diagnosed in the cohort in .Denominator is all prostate cancer cases in the two alanine-allele group.Denominator is all prostate cancer cases in the zero alanine-allele group.

Platz and Lippman

3570 2009 by American Society of Clinical Oncology JOURNAL OF CLINICAL ONCOLOGY

pronounced for aggressive disease in the PHS.13 In contrast,menwithhigh selenium had a lower risk of prostate cancer irrespective of thenumber of alanine alleles they carried.13 This pattern is not unique toselenium, however. An increased risk of aggressive disease was associ-ated with two alanine alleles and low lycopene in the Health Profes-sionals Follow-up Study (HPFS)14 and with two alanine alleles andlow lycopene and low total antioxidants in the PHS.13 Regardingeffects on nonaggressive disease, there was no clear interaction be-tween antioxidant status and SOD2 genotype in the PHS,13 and it canbe inferred that there was no interaction between lycopene and SOD2genotypeor that itpossiblywas inverselyassociated in theHPFS.14Theincreased risk of aggressive prostate cancer in men with two alaninealleles and low selenium in the study byChan et al3 is compatiblewiththe results from these at-riskdenominator studies (paralleling theargument depicted in Table 2). But again, the interaction with SOD2may not be selenium specific.

The only finding in Chan et al3 that is not compatible with theprevious studies of the association between selenium and prostatecancer reviewed here1,2,4-14 is the overall increased risk of aggressivedisease associated with higher circulating selenium, which was drivenby men with none or one alanine allele. Possible contributors to thisincompatible result were the Chan et al study design and method-ologic limitationsdiscussedby the authors (eg, lackof temporality anda small sample size for evaluating joint effects). The study population,however, had a selenium exposure (eg, distribution comparable tothat inUS adultmen15) and outcome (eg, PSA era) similar to those insome of the other observational studies. Serum selenium in the Chanet al study was higher than in the NPC study16 but lower than inSELECT,1 and the distribution of stage/grade was likely intermediatebetween the NPC study (which straddled the pre-PSA and PSA era)and SELECT (PSA era, and virtually all patients had T1/T2 and Glea-son 3 4 disease). In the NPC study, the inverse association forselenium supplementation was present for both local and advanceddisease andwas strongest inmenwith low circulating selenium.16 Thenull SELECT result implies no association of selenium supplementa-tion with nonaggressive disease. Baseline selenium level associationswith outcome have not been evaluated in SELECT,1 so it is unknownwhether supplementation in this trial may have decreased prostatecancer risk in men with low baseline selenium, as in the NPC study,and concurrently increased the risk of overall or aggressive disease inmen with high baseline selenium, as in the Chan et al study. Thatselenium may have opposing effects at high versus low exposures isplausible. In theNPCstudy,baselineplasmaseleniumandtotal cancer(all sites) riskweredirectly associated in the seleniumsupplement armand possibly inversely associated in the placebo arm.17 Therefore,what seems to be an anomalous prostate cancer result of Chan et alwarrants additional investigation.

The findings of Chan et al3 and others1,2,4-14 raise questions thatmay lead to an increasedunderstanding of selenium supplementationin men at risk of prostate cancer. Did selenium supplementation inSELECT reduce the risk of total or high-grade prostate cancer inmenwith baseline circulating selenium levels as low as in the bottom tertileof theNPCstudy?Did seleniumsupplementation inSELECT increasethe risk of total or high-grade disease inmenwith baseline circulatingselenium levels as high as in the top quintile of the Chan et al study? Ifone combined across large observational studies to achieve adequatesample size for phenotype and a wider range between low and highselenium levels, would higher circulating selenium be associated with

the risk of prostate cancer that is aggressive or nonaggressive as deter-mined by stage alone, grade alone, both, and both plus PSA level? InSELECT or combined large observational studies, is the seleniumeffect on (or association with) prostate cancer risk beneficial, detri-mental, ornull in thosegroupsofmenwithhigher and lower exposureto other antioxidants; higher and lower production or activity ofantioxidant enzymes including manganese superoxide dismutase,perhaps as measured by genotype; higher and lower exposure to oxi-dant sources; and with variously combined exposures to selenium,other antioxidants, antioxidant enzymes, and oxidant sources? Read-ers shouldnote that theSOD2 interactionwith circulating selenium inrelation to prostate cancer risk was present for low lycopene and totalantioxidants, not just selenium. Furthermore, manganese superoxidedismutase encoded by SOD2 is only one of a large number of antiox-idant enzymes. All of these interactions deserve in-depth study. A lastquestion, which probably cannot be resolved by data from epidemio-logic studies, is whether lifelong, distant-past, or recent selenium levelis critical to supplement effects on prostate cancer risk.

Noneof the largeobservational studiesor trials addressed the roleof selenium in preventing prostate cancer recurrence, nor didChan etal3 despite their caution against broad use of selenium supplementa-tion for men with prostate cancer. Long-term follow-up of prostatecancer patients characterized for selenium levels before, at, and afterdiagnosis and treatment will be required to address recurrence. Ad-justing for stage and grade at diagnosis will be necessary to determineif selenium predicts recurrence independently of these two prognos-tic factors.

Placing the findings of Chan et al3 and SELECT in the contextof the existing literature does not clarify the role of selenium inprostate cancer etiology. Nevertheless, we offer a couple of workinghypotheses. Selenium supplementation does not decrease risk exceptpossibly in selenium-deficientpopulations. Supplementationpossiblyincreases risk of prostate cancer, especially aggressive disease, inselenium-replete men or men with a particular genotype for antioxi-dant enzymes. These hypotheses and the questions posed above sug-gest the need for personalized risk prediction. At present, we do notknow enough to determine howmuch selenium any man or womanshould receive from the diet or a supplement.

This lack of knowledge supports the common public health rec-ommendation of moderation with respect to supplements for menand women. Furthermore, we should encourage men and women toeat awide arrayof foods,maintainnormalweight, bephysically active,not smoke, and drink in moderation if at all to prevent chronicdiseases in general.Unlike theprospect of personalized chemopreven-tion, this is not new or exciting advice, but it is common sense.

AUTHORS DISCLOSURES OF POTENTIAL CONFLICTS OF INTERESTThe author(s) indicated no potential conflicts of interest.

AUTHOR CONTRIBUTIONSConception and design: Elizabeth A. Platz, Scott M. LippmanData analysis and interpretation: Elizabeth A. Platz, Scott M. LippmanManuscript writing: Elizabeth A. Platz, Scott M. LippmanFinal approval of manuscript: Elizabeth A. Platz, Scott M. Lippman

REFERENCES1. Lippman SM, Klein EA, Goodman PJ, et al: Effect of selenium and vitamin

E on risk of prostate cancer and other cancers: The Selenium and Vitamin ECancer Prevention Trial (SELECT). JAMA 301:39-51, 2009

2. Clark LC, Combs GF Jr, Turnbull BW, et al: Effects of selenium supple-mentation for cancer prevention in patients with carcinoma of the skin: A

Editorial

www.jco.org 2009 by American Society of Clinical Oncology 3571

randomized controlled trialNutritional Prevention of Cancer Study Group.JAMA 276:1957-1963, 1996

3. Chan JM, Oh WK, Xie W, et al: Plasma selenium, manganese superoxidedismutase, and intermediate- or high-risk prostate cancer. J Clin Oncol 27:3577-3583, 2009

4. Li H, Stampfer MJ, Giovannucci EL, et al: A prospective study of plasmaselenium levels and prostate cancer risk. J Natl Cancer Inst 96:696-703, 2004

5. Peters U, Foster CB, Chatterjee N, et al: Serum selenium and risk ofprostate cancer: A nested case-control study. Am J Clin Nutr 85:209-217, 2007

6. Allen NE, Appleby PN, Roddam AW, et al: Plasma selenium concentrationand prostate cancer risk: Results from the European Prospective Investigationinto Cancer and Nutrition (EPIC). Am J Clin Nutr 88:1567-1575, 2008

7. Gill JK, Franke AA, Steven Morris J, et al: Association of selenium,tocopherols, carotenoids, retinol, and 15-isoprostane F(2t) in serum or urine withprostate cancer risk: The multiethnic cohort. Cancer Causes Control [epub aheadof print on February 11, 2009]

8. Brooks JD, Metter EJ, Chan DW, et al: Plasma selenium level beforediagnosis and the risk of prostate cancer development. J Urol 166:2034-2038,2001

9. Yoshizawa K, Willett WC, Morris SJ, et al: A study of prediagnosticselenium level in toenails and the risk of advanced prostate cancer. J Natl CancerInst 90:1219-1224, 1998

10. Woodson K, Tangrea JA, Lehman TA, et al: Manganese superoxidedismutase (MnSOD) polymorphism, alpha-tocopherol supplementation and pros-

tate cancer risk in the alpha-tocopherol, beta-carotene cancer prevention study(Finland). Cancer Causes Control 14:513-518, 2003

11. Kang D, Lee KM, Park SK, et al: Functional variant of manganesesuperoxide dismutase (SOD2 V16A) polymorphism is associated with prostatecancer risk in the prostate, lung, colorectal, and ovarian cancer study. CancerEpidemiol Biomarkers Prev 16:1581-1586, 2007

12. Cooper ML, Adami HO, Gronberg H, et al: Interaction between singlenucleotide polymorphisms in selenoprotein P and mitochondrial superoxidedismutase determines prostate cancer risk. Cancer Res 68:10171-10177, 2008

13. Li H, Kantoff PW, Giovannucci E, et al: Manganese superoxide dismutasepolymorphism, prediagnostic antioxidant status, and risk of clinical significantprostate cancer. Cancer Res 65:2498-2504, 2005

14. Mikhak B, Hunter DJ, Spiegelman D, et al: Manganese superoxidedismutase (MnSOD) gene polymorphism, interactions with carotenoid levels andprostate cancer risk. Carcinogenesis 29:2335-2340, 2008

15. Niskar AS, Paschal DC, Kieszak SM, et al: Serum selenium levels in the USpopulation: Third National Health and Nutrition Examination Survey, 1988-1994.Biol Trace Elem Res 91:1-10, 2003

16. Clark LC, Dalkin B, Krongrad A, et al: Decreased incidence of prostatecancer with selenium supplementation: Results of a double-blind cancer preven-tion trial. Br J Urol 81:730-734, 1998

17. Duffield-Lillico AJ, Dalkin BL, Reid ME, et al: Selenium supplementation,baseline plasma selenium status and incidence of prostate cancer: An analysis ofthe complete treatment period of the Nutritional Prevention of Cancer Trial. BJUInt 91:608-612, 2003

Platz and Lippman

3572 2009 by American Society of Clinical Oncology JOURNAL OF CLINICAL ONCOLOGY