Serum levels of selenium and smoking habits at age 50 influence long term prostate cancer risk; a 34 year ULSAM follow-up

Oxidative DNA base modification likely plays an important role in prostate carcinogenesis [1]. Both endogenous and exogenous antioxidants act interdependently in preventing clinically significant prostate cancer [2]. The selenoproteins active in oxidative DNA damage repair need the antioxidant selenium to be fully effective [3–8]. High serum selenium levels correlate with decreased incidence of cancer in general [9, 10], and prostate cancer in particular [11–16], an association by some found to be mainly present in smokers [17]. Low serum selenium levels inversely relate to overall accumulated DNA damage [18] and selenoproteins are highly expressed in the prostate [19].

Despite that selenium thus theoretically may be a candidate anti-oxidant to use for prevention, the recent SELECT trial [20] - a large interventional study with selenium supplementation and/or vitamin E supplementation to reduce the risk of prostate cancer - was prematurely ended due to lack of effect. However, selenium may still be protective in a subset of men with specific genetic polymorphisms of the MnSOD gene [2] or in men with low baseline levels of serum selenium [21–23] which was not studied in SELECT but could be relevant to study further in light of recent findings[24].

The role of tobacco smoking in prostate cancer etiology is still unclear [25, 26] but as smoking causes oxidative damage [27, 28] and also is associated with lower serum selenium levels [29, 30] a relation between smoking and prostate cancer seems biologically plausible.

Genes of interest to a relation between selenium and prostate cancer risk include those involved in DNA repair and oxidative stress. Oxidative stress triggers the base mutation 8-oxoguanine (OG) in DNA [31]. Single nucleotide polymorphisms (SNPs) in the gene coding for 8-oxoguanine glucosylase (OGG1), a selective base excision repair enzyme for OG-lesions [32], are associated with cancer progression [33, 34] with evidence also for an association with prostate cancer [35–38]. Manganese superoxide dismutase (MnSOD) is the primary antioxidant enzyme in mitochondria. A functional SNP in the MnSOD gene, rs4880, modifies risk of prostate cancer associated with serum levels of pre-diagnostic selenium [2] although it not in itself was associated with prostate cancer risk [39]. MnSOD enzyme activity has a 56-fold range of values between polymorphic variants [40].

The aim of our study was to investigate the association between serum selenium levels at age 50 and later non-screening detected prostate cancer in a population based study with up to 34 years of individual follow up, and to assess if smoking status at baseline modified this association. In this cohort we also had information on factors with a potential to decrease serum selenium levels: Erythrocyte sedimentation rate (ESR), total serum cholesterol [41], and also of body mass index (BMI). In addition, we explored as a hypotheses generating analysis in a subcohort if polymorphism in the genes OGG1 and MnSOD influenced any effect of serum selenium.

The study was approved by the Ethics Committee of the Faculty of Medicine at Uppsala University and written informed consent to participate was obtained from all participants.

The source population of the study is a cohort in the Uppsala Longitudinal Study of Adult Men (ULSAM). In 1970-1973, all male residents in Uppsala county in January 1970, born in 1920-24 (n = 2841) were invited to take part in a prospective health survey aimed at identifying risk factors for diabetes and cardiovascular disease. Eighty two per cent (n = 2322) of the invited men participated at baseline, at age 50, forming ULSAM [42].

The serum selenium level at age 50 was determined for 2045 men in ULSAM and they constitute our study base (figure 1). After 21 years of follow-up, at age 71, the cohort was invited to a third cycle of investigation, performed from 1991 to 1995, including 1 221 of 1681 men still alive in January 1991. At this third cycle of investigation, blood samples for DNA extraction and later genotyping were collected from 1052 men giving consent for such sampling. Genotypes of OGG1 and MnSOD were available for1005 of the 1221 participants. The baseline mean serum levels of selenium, and two factors negatively correlated to selenium concentration: ESR and total serum cholesterol [41] did not differ between the 1005 genotyped and the full cohort of 2045 men who at baseline had their serum selenium levels determined. Therefore, the genotyped men were considered a representative sub-sample of the full cohort. Use of the drug clozapine may decrease serum selenium levels [43] but none of the participants reported use of this drug at age 50, i.e. concurrently with the blood sampling for selenium.

In a study population where little PSA screening occurred, we found a previously not reported relation between serum selenium levels, smoking and prostate cancer risk: Smokers with serum selenium levels in the two lower tertiles (≤80 μg/L) experienced a higher cumulative incidence of prostate cancer than smokers in the high serum selenium tertile. In hypothesis generating explorative analyses, we found indications that the association between serum selenium levels and prostate cancer risk was modified by genetic polymorphisms, in the OGG1 gene (rs125701) and MnSOD gene (rs2758331- linked to rs4880).

The ULSAM cohort is population based and homogenous as regards the ethnic background and the age of the participants [42]. The participant rate was high and the follow-up up to 34 years almost complete through using linkage to national registers with high coverage. ULSAM has a thorough characterization of factors influencing serum selenium levels including information on smoking habits[42]. Only few prostate cancers were detected by PSA-screening and our results thus mainly pertain to clinically relevant prostate cancer with a high risk of progress.

For cancer studies, ULSAM is comparatively small and modest or weak associations may go undetected: our data indicate an overall modest inverse relation of serum selenium level with prostate cancer risk, which we could not substantiate. The serum selenium levels and co- variates were measured at a median of 23 years before diagnosis, a disadvantage, which however is shared with the other studies in the field [10, 11, 16]. In a Swedish population the range of selenium exposure is limited, with few high serum selenium level values. The size of the study implies that our analyses of the interaction of gene polymorphism and serum selenium levels are of an exploratory character. Another disadvantage is that a proxy SNP was used for the analysis of the main MnSOD polymorphism which made a detailed analysis on a molecular level unattainable.

Our finding of an increased risk for prostate cancer in smokers with low serum selenium is a new finding. As smoking increases oxidative damage, it is not biologically unreasonable. This relation may in other studies have been masked by the strong competing risk of concomitant early death from other causes, mainly cardiovascular disease, predominantly in smokers. If smoking is associated with both early death and risk of prostate cancer later in life, conventional time to event analyses may give a false impression that smoking protects from prostate cancer since methods traditionally employed assumes that censoring is non-informative. We noted and analysed a similar pattern for obesity in the ULSAM cohort [59]. This type of competing risk problem may be particularly evident in this study representing a pre-screening era where the median age at diagnosis of prostate cancer was 73 years of age. Thus, a screening program for prostate cancer introducing long lead times may alter this pattern.

Our hypothesis generating results that serum selenium levels and the polymorphic genes OGG1 and MnSOD involved in the protection from oxidative stress, act concurrently in the defence of prostate cancer development are in accord with previous knowledge [2, 11, 12, 15, 16, 60]. A recent nested case-control study within the Physicians Health Study [24] analyzing polymorphisms within the selenoprotein gene, SEP15 found genetic variants associated with prostate cancer mortality and also modifying the association of serum selenium with prostate cancer survival. If, as the recent report suggest [24] and our study may be interpreted, only subgroups of men would benefit from high serum selenium levels, then beneficial effects of supplementation in interventional studies may pertain only to subgroups and therefore be difficult to detect. There was an overall lack of effect of selenium supplementation in the SELECT trial [20]. The trial inclusion criteria did neither enrich a study population with low baseline selenium levels nor patients defined by pre-specified genetic polymorphisms. The differences in results from the interventional study, ours and another recent observational study [24] are not contradictory but rather emphasize the possibility for interventional studies in predefined groups of men with low selenium concentrations and certain genetic profiles. Due to the limited size of our study, the impact of genetic variation in relation to smoking being a risk factor for prostate cancer was not possible to analyse further.

Our results indicate that when competing risks of death are taken into account, middle age levels of serum selenium and smoking habits give long term information regarding the risk of prostate cancer presenting in a setting with a low prevalence of screening. We thus here identify a need for further development of statistical methods to take competing risk into account in prospective studies of risk factors for diseases with the prevalence peak late in life. The finding that smoking may be a risk factor for prostate cancer in men with low serum selenium levels is relevant to explore further. We found indications that patterns of exposure to selenium and smoking combined with data on genetic variation in genes involved in DNA repair for oxidative damage can be valuable to explore. Our findings are of relevance for the analysis of interventional studies with selenium supplementation.