Selenium for preventing cancer

Marco Vinceti; Gabriele Dennert; Catherine M Crespi; Marcel Zwahlen; Maree Brinkman; Maurice PA Zeegers; Markus Horneber; Roberto D'Amico; Cinzia Del Giovane

doi:10.1002/14651858.CD005195.pub3

. 2014 Mar 30;2014(3):CD005195. doi: 10.1002/14651858.CD005195.pub3

Selenium for preventing cancer

Marco Vinceti ^1,^✉, Gabriele Dennert ², Catherine M Crespi ³, Marcel Zwahlen ⁴, Maree Brinkman ⁵, Maurice PA Zeegers ⁶, Markus Horneber ⁷, Roberto D'Amico ⁸, Cinzia Del Giovane ⁸

PMCID: PMC4441528 NIHMSID: NIHMS684156 PMID: 24683040

Abstract

Background

This review is an update of the first Cochrane publication on selenium for preventing cancer (Dennert 2011).

Selenium is a metalloid with both nutritional and toxicological properties. Higher selenium exposure and selenium supplements have been suggested to protect against several types of cancers.

Objectives

Two research questions were addressed in this review: What is the evidence for:

1. an aetiological relation between selenium exposure and cancer risk in humans? and 2. the efficacy of selenium supplementation for cancer prevention in humans?

Search methods

We conducted electronic searches of the Cochrane Central Register of Controlled Trials (CENTRAL, 2013, Issue 1), MEDLINE (Ovid, 1966 to February 2013 week 1), EMBASE (1980 to 2013 week 6), CancerLit (February 2004) and CCMed (February 2011). As MEDLINE now includes the journals indexed in CancerLit, no further searches were conducted in this database after 2004.

Selection criteria

We included prospective observational studies (cohort studies including sub‐cohort controlled studies and nested case‐control studies) and randomised controlled trials (RCTs) with healthy adult participants (18 years of age and older).

Data collection and analysis

For observational studies, we conducted random effects meta‐analyses when five or more studies were retrieved for a specific outcome. For RCTs, we performed random effects meta‐analyses when two or more studies were available. The risk of bias in observational studies was assessed using forms adapted from the Newcastle‐Ottawa Quality Assessment Scale for cohort and case‐control studies; the criteria specified in the Cochrane Handbook for Systematic Reviews of Interventions were used to evaluate the risk of bias in RCTs.

Main results

We included 55 prospective observational studies (including more than 1,100,000 participants) and eight RCTs (with a total of 44,743 participants). For the observational studies, we found lower cancer incidence (summary odds ratio (OR) 0.69, 95% confidence interval (CI) 0.53 to 0.91, N = 8) and cancer mortality (OR 0.60, 95% CI 0.39 to 0.93, N = 6) associated with higher selenium exposure. Gender‐specific subgroup analysis provided no clear evidence of different effects in men and women (P value 0.47), although cancer incidence was lower in men (OR 0.66, 95% CI 0.42 to 1.05, N = 6) than in women (OR 0.90, 95% CI 0.45 to 1.77, N = 2). The most pronounced decreases in risk of site‐specific cancers were seen for stomach, bladder and prostate cancers. However, these findings have limitations due to study design, quality and heterogeneity that complicate interpretation of the summary statistics. Some studies suggested that genetic factors may modify the relation between selenium and cancer risk—a hypothesis that deserves further investigation.

In RCTs, we found no clear evidence that selenium supplementation reduced the risk of any cancer (risk ratio (RR) 0.90, 95% CI 0.70 to 1.17, two studies, N = 4765) or cancer‐related mortality (RR 0.81, 95% CI 0.49 to 1.32, two studies, N = 18,698), and this finding was confirmed when the analysis was restricted to studies with low risk of bias. The effect on prostate cancer was imprecise (RR 0.90, 95% CI 0.71 to 1.14, four studies, N = 19,110), and when the analysis was limited to trials with low risk of bias, the interventions showed no effect (RR 1.02, 95% CI 0.90 to 1.14, three studies, N = 18,183). The risk of non‐melanoma skin cancer was increased (RR 1.44, 95% CI 0.95 to 1.17, three studies, N = 1900). Results of two trials—the Nutritional Prevention of Cancer Trial (NPCT) and the Selenium and Vitamin E Cancer Trial (SELECT)—also raised concerns about possible increased risk of type 2 diabetes, alopecia and dermatitis due to selenium supplements. An early hypothesis generated by NPCT that individuals with the lowest blood selenium levels at baseline could reduce their risk of cancer, particularly of prostate cancer, by increasing selenium intake has not been confirmed by subsequent trials. As the RCT participants were overwhelmingly male (94%), gender differences could not be systematically assessed.

Authors' conclusions

Although an inverse association between selenium exposure and the risk of some types of cancer was found in some observational studies, this cannot be taken as evidence of a causal relation, and these results should be interpreted with caution. These studies have many limitations, including issues with assessment of exposure to selenium and to its various chemical forms, heterogeneity, confounding and other biases. Conflicting results including inverse, null and direct associations have been reported for some cancer types.

RCTs assessing the effects of selenium supplementation on cancer risk have yielded inconsistent results, although the most recent studies, characterised by a low risk of bias, found no beneficial effect on cancer risk, more specifically on risk of prostate cancer, as well as little evidence of any influence of baseline selenium status. Rather, some trials suggest harmful effects of selenium exposure. To date, no convincing evidence suggests that selenium supplements can prevent cancer in humans.

Keywords: Female, Humans, Male, Case‐Control Studies, Neoplasms, Neoplasms/prevention & control, Observational Studies as Topic, Odds Ratio, Randomized Controlled Trials as Topic, Selenium, Selenium/administration & dosage, Selenium/adverse effects, Sex Factors, Trace Elements, Trace Elements/administration & dosage, Trace Elements/adverse effects

Selenium for preventing cancer

Review question

We reviewed the evidence suggesting that selenium can help to prevent cancer. This review updates the first Cochrane review on this topic (Dennert 2011).

Background

Selenium is a naturally occurring element found in crops, animal products and water. Small amounts of selenium are needed for proper human nutrition. Starting in the 1960s, numerous studies reported that people with high levels of selenium in their diet or in their body tissues had lower rates of cancer. Some laboratory studies also suggested that selenium could inhibit the growth of cancer cells. This led to widespread interest and claims that taking selenium supplements could prevent cancer. Over the next decades, many more studies were conducted to compare cancer rates among individuals with high and low selenium levels, and several trials were conducted in which individuals were randomly assigned to receive selenium supplements or placebo and then were followed so their cancer rates could be determined. Particular interest focused on whether selenium could prevent prostate, skin or other specific types of cancer.

Study characteristics

This review includes 55 studies in which adults observed to have high or low selenium levels were followed over time to determine whether they developed cancer, along with eight trials in which adults were randomly assigned to receive selenium supplements or placebo. The evidence is current to February 2013.

Key results

We found limited evidence suggesting that individuals observed to have higher selenium levels have a lower incidence of cancer. However, it is not possible to conclude from these studies that selenium was the reason for the lower cancer risk, because a high selenium level might be associated with other factors that reduce cancer risk, such as a healthier diet or lifestyle. Also, selenium comes in many different chemical forms that have different biological activity, and these studies did not identify which chemical forms were being measured. Selenium levels in body tissues in which people might develop cancer (e.g. the prostate) also were not examined.

The randomised controlled trials that assessed whether taking selenium supplements might prevent cancer differed considerably in methodological quality and are not equally reliable. Several studies reported that individuals receiving selenium supplements decreased their liver cancer risk, but these studies reported insufficient details about their randomisation process and participant follow‐up to be convincing. Recent trials that were judged to be well conducted and reliable have found no effects of selenium on reducing the overall risk of cancer or on reducing the risk of particular cancers, including prostate cancer. In contrast, some trials suggest that selenium may increase the risk of non‐melanoma skin cancer, as well as of type 2 diabetes, raising concern about the safety of selenium supplements.

Overall, no convincing evidence suggests that selenium supplements can prevent cancer. However, for a full understanding of the role of this metalloid in cancer development, more research is needed on how selenium may act differently in individuals with different genetic backgrounds or nutritional status, and on the different biological activities of the various selenium compounds, which are still largely unknown.

Background

This review is an update of the first Cochrane publication on selenium for preventing cancer (Dennert 2011).

Description of the condition

Cancer is a leading cause of death worldwide (WHO 2008). According to World Health Organization (WHO) estimates, 14.1 million people developed and 8.2 million died of cancer in 2012, with more than half of all new cases occurring in less developed regions of the world (IARC 2012).

The role of diet and nutrition in carcinogenesis and cancer prevention has been an area of active research for decades. A holy grail has been the identification of nutritional supplements with cancer preventive properties. Such dietary factors would clearly have major public health implications, but unfortunately, investigations into supplementation of various vitamins, trace elements and other dietary constituents have generally yielded disappointing and even troubling results (Ashar 2010; Bjelakovic 2012; Driscoll 2010; Fortmann 2013; Guallar 2013; Jerome‐Morais 2011; Marik 2012; Martinez 2012; Mayne 2012; Rocourt 2013).

Description of the intervention

The metalloid selenium is one of the dietary elements that has received considerable attention as a potential cancer preventive agent. Selenium is nutritionally essential for humans but is toxic at higher levels, with a narrow safe range of intake (Rayman 2012; Vinceti 2013a; Vinceti 2013b). Whether selenium provides various health benefits (including a cancer preventive effect) beyond its essential nutritional role is a matter of ongoing debate (Bodnar 2012; Fortmann 2013; Karp 2013; Lippman 2009, in: SELECT 2009; Rayman 2012; Stranges 2010;Vinceti 2013a; Vinceti 2013b; Vinceti 2013d). Humans usually ingest this trace element with crop and animal products and sometimes in functional foods or supplements (Hurst 2013; Vinceti 2000a). Chemical forms and concentrations of selenium in environmental matrices, foods, drinking water and other sources of exposure vary considerably, depending on factors such as plant and animal metabolism and growth conditions or animal nutrition (Rayman 2008a; Rayman 2008b).

Selenium species can be classified into organically bound selenium forms (e.g. selenomethionine, selenocysteine) and inorganic forms (e.g. selenate, selenite) (Gammelgaard 2011; Weekley 2013). Selenium yeast refers to a selenium‐enriched yeast medium that usually contains nearly entirely organically bound selenium with a high proportion of selenomethionine (Block 2004; Rayman 2004).

The recommended intake of selenium differs between regulatory agencies (Hurst 2013; Vinceti 2009; Vinceti 2013a). For example, the US Institute of Medicine recommends a daily intake of 55 µg/d for adults (Institute of Medicine 2009), whereas the WHO recommends values ranging from 25 to 34 µg/d, depending on age and sex (WHO 2004). These various standards do not take into account the chemical forms of selenium, despite growing evidence of the importance of selenium speciation (Vinceti 2013a; Vinceti 2013c; Weekley 2013).

To prevent adverse effects due to excessive selenium intake, the US Institute of Medicine has set the tolerable upper intake level to 400 µg/d for adults (Office of Dietary Supplements 2009); however, recent epidemiological studies suggest toxicity at lower intake levels (Lippman 2009, in: SELECT 2009; Stranges 2007; Vinceti 2013a). In addition to the acute and chronic toxicity of high selenium exposure, possible harmful effects of long‐term intake of lower dosages have been a matter of concern. However, such effects are still inadequately investigated (Vinceti 2001; Vinceti 2009). Furthermore, strong evidence shows different biological activities of the various organic and inorganic forms of selenium (Hazane‐Puch 2013; Rayman 2008a; Vinceti 2009; Vinceti 2013c;Weekley 2013), suggesting the opportunity to better characterise the specific toxicological and nutritional properties of each selenium species in humans, in animals and in the environment. Recent publications have questioned the adequacy of the current upper 'safe' limit of intake (Jerome‐Morais 2011;Morris 2013; Moyad 2012;Rocourt 2013;Sacco 2013;Vinceti 2009; Vinceti 2013b) and have espoused the need to set different limits for the many different sources of organic and inorganic selenium.

Accurate estimation of selenium exposure in epidemiological studies presents several challenges. Individual exposure is typically assessed by using peripheral biomarkers of exposure, such as blood (generally plasma or serum) or nail concentrations, or by estimating dietary intake (Ashton 2009). All of these methods have strengths and limitations, and their validity has been questioned (Ashton 2009;Haldimann 1996;Vinceti 2013b). However, levels of selenium in peripheral biomarkers such as blood, toenail and hair have been found to correlate to a moderate degree with dietary intake as assessed through self reported consumption of supplements, food frequency questionnaires and dietary records (Hurst 2013; Longnecker 1996; Ovaskainen 1993; Pestitschek 2013; van den Brandt 1993a). )Stronger correlation has been seen at high intake levels (Morris 2013), although results of other studies were not consistent (Hunter 1990; Karita 2003; Satia 2006; Vinceti 2012). Assessment of selenium levels in highly specific body tissues, is extremely complex, as these levels are not necessarily homogeneously reflected by all biomarkers because overall selenium exposure, as well as its chemical forms and other factors, influences distribution of the metalloid into various body compartments (Behne 1996; Behne 2010; Panter 1996; Vinceti 2000a; Vinceti 2013c). For example, circulating levels of some selenium species and of total selenium did not correlate with selenium content in the central nervous system as assessed by cerebrospinal fluid concentrations (Solovyev 2013; Vinceti 2013c), indicating not only the tissue‐specific significance of biomarkers but also the importance of selenium speciation when the distribution of selenium in different body compartments is assessed, representing target organs for different diseases.

Selenium levels found in human specimens (Rayman 2008b), as well as the estimated intake of selenium (Fairweather‐Tait 2011; Haldimann 1996; Jablonska 2013), show high global variability due to factors such as dietary habits, ethnicity, gender, age, individual metabolism, occupational exposure, exposure to coal and other sources of combustion and smoking. It is interesting to note that smoking tends to lower selenium biomarker concentrations, although it is a source of selenium exposure (Jossa 1991;Kafai 2003)—a phenomenon that might be related to altered metabolism of the metalloid due to an interaction with cadmium. Globally, inconsistencies have been noted as to how these factors are associated with selenium levels (Haldimann 1996; Vinceti 2000a). For example, selenium levels increased with age in women, but not in men, in the French SU.VI.M.AX cohort study (Arnaud 2007) and decreased with age in a female population in Ohio (Smith 2000); however, two studies in Switzerland and Austria could not find an association between age and selenium status in either gender (Burri 2008; Gundacker 2006). Gender‐specific nutritional and health behaviours, as well as gender‐specific differences in selenium metabolism, may contribute to observed discrepancies in selenium levels between males and females (Combs 2012; Rodriguez 1995). Gender might more generally influence the ability of selenium to induce adverse metabolic effects, as suggested by the recent observation of a direct association between metabolic syndrome and selenium in females but not in males in a European case‐control study (Arnaud 2012).

How the intervention might work

The ability of selenium to counteract cancer cell growth, as has been observed in a large number of laboratory studies, may be due to its effects on DNA stability, cell proliferation, necrotic and apoptotic cell death in healthy and malignant cells, regulation of oxidative stress and the immune system (for reviews, see: Davis 2012; Jackson 2008; Steinbrenner 2013; Weekley 2013). These features have also suggested the possibility of using selenium compounds in cancer therapy—a hypothesis that has been under investigation (Chintala 2012; Fan 2013; Kim 2012; Sonaa 2013). Selenium may be involved in these processes through several mechanisms as a source of selenometabolites and as a component of selenium‐containing enzymes (Davis 2012; Hatfield 2009; Jackson 2008; Steinbrenner 2013; Weekley 2013). The optimum level for the retardation of carcinogenesis in human cells has been debated and is thought to be higher than the level commonly achieved through dietary changes (Whanger 2004). However, in laboratory studies, selenium has been shown to promote malignant cell transformation and progression (Chen 2000; Kandas 2009; National Toxicology Program 2011; Novoselov 2005; Rose 2014; Su 2005), thus confirming a ‘dual personality’ of this Janus‐faced element and of selenoproteins in both preventing and promoting cancer (Hatfield 2014).

Numerous epidemiological studies have reported an inverse association between selenium exposure and cancer risk. The first such studies had ecological study designs (Schrauzer 1977; Shamberger 1969). These were followed by case‐control and cohort observational studies and randomised trials, some of which received substantial attention from both the general population and the scientific community (Brinkman 2006; Fortmann 2013; Steinbrenner 2013; Vinceti 2013b). Gender‐related differences regarding the effects of selenium on cancer risk have also been suggested by some observational and experimental human studies, and differences in selenium tissue distribution, tumour biology and other factors have been suggested to explain a possible greater beneficial effect in males than in females (NPCT 2002; Waters 2004).

Why it is important to do this review

Selenium has been suggested to be involved in central anticarcinogenic processes. This has led to wide marketing of selenium supplements with associated health claims, particularly the prevention of both cancer (Dennert 2011; Vinceti 2013b) and cardiovascular disease (Rees 2013). In recent decades, worldwide debate has continued about the association between selenium exposure and cancer risk, including whether selenium supplements are effective in decreasing the incidence of or mortality from cancer. Epidemiological and other data have yielded conflicting results, sometimes suggesting different effects in men and women, and it has been suggested that selenium supplements might even have harmful effects. This review is timely and important, as several meta‐analyses and systematic reviews have been published, but an updated comprehensive summary synthesising evidence from both observational studies and intervention trials that include all types of cancer and look for gender‐related differences has not been conducted since the

time of the first Cochrane publication on the use of selenium for preventing cancer (Dennert 2011).

Objectives

Two research questions were addressed in this review: What is the evidence for:

an aetiological relation between selenium exposure and cancer risk in humans? and
the efficacy of selenium supplementation for cancer prevention in humans?

Methods

Criteria for considering studies for this review

Types of studies

Randomised controlled trials (RCTs) and prospective observational studies (cohort studies and nested case‐control studies) were included, irrespective of publication year, publication status or language, provided they were published in extenso. We did not include conference abstracts in this review.

Types of participants

All adult participants (18 years of age and older).

Types of interventions

We considered prospective observational studies (cohort studies and cohort‐nested and nested case‐control studies) for inclusion if they assessed baseline exposure to selenium in apparently cancer‐free individuals either as biochemical selenium status or as estimated selenium intake at study entry.

We considered RCTs for inclusion if they used selenium supplementation at any dose or route of administration for a minimum of four weeks versus placebo or no intervention. We excluded trials using selenium supplementation as part of a multi‐component preparation without a study arm using selenium monotherapy supplementation.

Types of outcome measures

We analysed primary and secondary outcomes.

Primary outcomes

Incidence of any cancer and of site‐specific cancers, assessed as the proportion of participants developing cancers during the study period.
Mortality from any cancer and from site‐specific cancer, assessed as the proportion of participants dying from cancers during the study period.

Secondary outcomes

Incidence of selected adverse effects, assessed as the proportion of participants developing adverse health conditions. These outcomes were assessed in RCTs only.

Search methods for identification of studies

We conducted electronic searches of the Cochrane Central Register of Controlled Trials (CENTRAL, 2013, Issue 1), MEDLINE (Ovid, 1966 to February 2013 week 1), EMBASE (1980 to 2013 week 6), CancerLit (February 2004) and CCMed (February 2011). We conducted the initial search in 2004 and updates in July 2007, January 2009, October 2009, February 2011 and February 2013. As MEDLINE now includes the journals indexed in CancerLit, no further searches were conducted in this database after 2004.

We also searched the following online clinical trials databases in the previous review (Dennert 2011).

Clinical Trials of the American Cancer Society (http://www.cancer.gov, February 2011).
The metaRegister of Controlled Trials (mRCT, http://www.controlled‐trials.com, February 2011).
The German Cancer Study Register (http://www.studien.de, February 2011).
The System for Information on Grey Literature in Europe (SIGLE) (February 2004, discontinued in 2005).

The search strategies are provided in Appendix 1.

Data collection and analysis

Selection of studies

Two review authors independently checked all electronic search results for eligibility. When search results could not be rejected with certainty on the basis of title, abstract or both, we obtained full‐text material.

We scanned bibliographies of papers retrieved using the described search strategy to identify additional studies. If additional information was needed, we contacted the correspondent authors of the included studies; we also asked investigators for information about unpublished trials.

Two review authors (MV and MH) independently applied the inclusion and exclusion criteria, if necessary with the assistance of a translator. We resolved disagreements by discussion and with the involvement of a third review author.

Data extraction and management

We used piloted extraction forms for epidemiological studies and RCTs to document data from the original material and to assess the quality of studies. One review author (CDG) extracted data, and a second review author (MV) checked extracted data for discrepancies, which were discussed between the two review authors (CDG and MV). In a small number of cases, we sought the opinion of a third review author (GD or CMC) to reach a consensus. If several reports from the same study were available, we considered as primary publications studies reporting the entire period of follow‐up with active selenium supplementation, when available, but study details available from other publications were also extracted if not reported in the primary study reference.

For comparison of selenium exposure measured in serum and plasma specimens, we converted all data into the unit µg/L. Results provided as ppm (parts per million) or µg/g were converted using the factor 1.026 g/mL (density of blood plasma), and data provided as µmol/L were converted using the factor 78.96 (molecular weight of selenium).

To be included, prospective observational studies had to report estimates of risk ratio (RR), for example, odds ratio (OR), for various selenium exposure levels. Studies reporting only the RR for a one‐unit increase in selenium exposure were not included in the analysis.

Assessment of risk of bias in included studies

Observational studies

The risk of bias in observational studies was assessed using assessment forms adapted from the Newcastle‐Ottawa Quality Assessment Scale (NOS) for cohort and case‐control studies (Wells 2004). The NOS form for cohort studies was used for all included observational studies, and the NOS case‐control form was used for nested case‐control studies. Both forms must be adapted a priori for use in a systematic review according to the research question and the review topic. The NOS uses a star system in which studies are judged on key domains pertaining to the selection and comparability of study groups, the ascertainment of exposure and outcome, and the duration of follow‐up. For each domain, either a 'star' or 'no star' is assigned, with a 'star' indicating that that study design element was considered adequate and less likely to introduce bias. A study could receive a maximum of nine stars in the cohort assessment (Appendix 2) and nine stars in the assessment of the case‐control portion (Appendix 3).

The risk of bias assessment was based on data provided in the included publications. We did not check other publications for details if they were not included in the review. If an included study encompassed more than one publication with divergent ratings in the NOS, we used the publication with the highest score.

Randomised controlled trials

We categorised generation of allocation sequence, allocation concealment, blinding and completeness of outcome data as adequate (low risk of bias), inadequate (high risk of bias) or unclear, according to the criteria specified in the Cochrane Handbook for Systematic Reviews of Interventions (Higgins 2011a) and suggested by Higgins et al. (Higgins 2011b). We considered these four items to be key domains for risk of bias assessment. Studies that were categorised as "adequate" in all four domains were considered to have a low risk of bias; studies with inadequate procedures in one or more key domains were considered to have a high risk of bias. Studies with unclear procedures in one or more key domains were considered to have an unclear risk of bias.

We assessed the fulfilment of ethical standards as follows.

Was informed consent obtained from participants? (yes/no/unclear).
Was approval obtained from an ethics board? (yes/no/unclear).

Measures of treatment effect

This review includes only the binary outcome of cancer diagnosis (i.e. cancer incidence) or death from cancer (i.e. cancer mortality), or a combination of both. The term 'cancer risk' is used in this paper as a generic term and refers generally to cancer incidence, cancer mortality and combined incidence/mortality data.

For observational studies, we used the odds ratio (OR) or the risk ratio (RR) and its 95% confidence interval (95% CI) as measures of the association between cancer risk and selenium exposure. When adjusted ORs were reported, we used the OR with the most extensive covariate adjustment reported in the publication.

For RCTs, we used RRs and their 95% CIs. When hazard ratios (HRs) rather than RRs were reported in the original study, we reported the individual study results as HRs with their 95% CIs; however, when data from such studies were included in meta‐analyses, we entered the RRs, and only RRs were pooled.

Dealing with missing data

When data were missing or when discrepancies in study publications were found, we tried to contact the study investigators to request further information. Contacting study authors helped to clarify discrepancies in several publications (e.g. differing data in text and tables within the same report); however, we retrieved no missing data or study details.

Assessment of heterogeneity

We performed a Chi² test for heterogeneity of study results. Additionally, we used I² statistics (Higgins 2003) to quantify inconsistency.

Assessment of reporting biases

The possibility of reporting bias was evaluated by using funnel plots.

Data synthesis

We performed data synthesis and analysis separately for RCTs and observational studies.

For observational studies, we conducted random effects meta‐analyses for all cancers or for site‐specific cancers for which at least five studies were available. We applied this restriction for two reasons. The first was practical: to limit the number of analyses to be performed. The second was that we expected results to be heterogeneous, but heterogeneity cannot be described and quantified well if too few studies are available (Higgins 2009). Although the cutoff at five studies is somewhat arbitrary, this decision was made very early in the review process; it was declared in the protocol and confirmed in its update. RCTs were less numerous, but given their fundamental importance in epidemiological research, we decided in the current review update to perform meta‐analyses for all cancers or site‐specific cancers when data from two or more trials were available.

Observational studies

We conducted random effects meta‐analyses of summary statistics from observational studies if data were available from at least five studies for all cancers or specific types of cancer. We used the OR or RR comparing the highest and lowest selenium exposure categories. Effect estimates were entered as the natural logarithm of the OR or RR, and the squared standard error of the natural logarithm of the OR or RR was used as a weight. The latter was calculated from the reported upper and lower boundaries of the 95% CI of the OR or RR. If a 95% CI was not reported, we used the total number of cases and the total number of controls, as well as the number of categories of selenium exposure, to estimate the numbers of cases and controls per exposure category. We then used the standard normal approximation formula to calculate the standard error of the OR (comparing the highest versus the lowest exposure category (lnOR = (1/a + 1/b + 1/c + 1/d), where a, b, c and d are the four counts needed to calculate the OR via (a*d)/(b*c)).

Meta‐analyses were conducted by using STATA (version 10 to 12) statistical software. We repeated meta‐analyses that were included in this review publication using the Review Manager 5 statistical tool; for this, logarithmic data for the OR and the standard error were copied from STATA into Review Manager 5, and results were double‐checked for errors.

Randomised controlled trials

We performed random effects meta‐analyses of summary statistics using RCT data if data were available from at least two studies for all cancers or specific types of cancer. When more than one publication from the same trial was available and reported different periods of follow‐up for the same cancer site, we included in the meta‐analysis only the longest period of follow‐up, provided that the experimental protocol was still ongoing at the time of follow‐up (i.e. that selenium supplementation was still actively supplied).

RRs and 95% CIs were calculated on the basis of the numbers of participants and cases when these were provided in the publication, using the meta‐analysis tool provided by Review Manager 5; otherwise, we used the RRs reported in the original publication. When an adjusted measure was also reported, we reported both the crude RR and the adjusted RR. We also calculated the RR of adverse outcomes and 95% CIs if sufficient data were available.

Subgroup analysis and investigation of heterogeneity

For observational studies, we used gender‐disaggregated data from mixed‐gender studies, together with data from single‐gender cohorts, for subgroup analyses by gender. We conducted the latter subgroup analyses to account for potential gender differences in selenium health effects (see Background).

Sensitivity analysis

For RCTs, we repeated analyses confining the included studies versus those with low risk of bias. For observational studies, we conducted sensitivity analyses to assess the effects of the different methods used to assess selenium status/intake.

Results

Description of studies

Citation style: Please note that we reference the sources of relevant information in a certain way to enhance traceability of our results for interested readers. When the source of information is not the primary publication of an included study, the specific publication of interest is also referenced. For example "Hakama 1990, in: Knekt 1990" indicates that the cited paper is "Hakama 1990" as part of the mentioned study.

Three full‐text theses published in the US could not be accessed (Coates 1987, in: Coates 1988; Menkes 1986a, in: Menkes 1986; Schober 1986, in: Menkes 1986). However, later journal publications were available and were included in this review as main study publications (Coates 1988, in: Coates 1988; Menkes 1986b, in: Menkes 1986; Schober 1987, in: Menkes 1986). Thus retrieval of the full‐text theses was considered to be unnecessary.

Results of the search

In the previous Cochrane review, of 4082 hits of potential relevance, 268 publications were retrieved in full text. Of these, 137 papers were considered as relevant (see the flow chart of the literature search in Dennert 2011).

In our updated search, after internal duplicates and duplicates against the database of the literature search conducted in January 2011 were excluded, 766 hits were retrieved. Of these, we excluded 744 references as being clearly irrelevant on the basis of title and abstract (flow chart of literature search: Figure 1). The reasons for exclusion were as follows.

Type of study: no prospective observational study or no randomised controlled trial (n = 213).
Type of outcome measure: no cancer epidemiology/prevention (n = 86).
Types of participants: no healthy adults (n = 397).
Type of exposure/intervention: no selenium exposure or no selenium supplements (n = 48).

The remaining 22 publications were considered of possible relevance and were reevaluated and retrieved in full text from this updated search (268 were retrieved in full text from the previous review). Upon further review, 11 of these publications were deemed relevant.

Included studies

In total, from the previous Cochrane review and from our update, 148 papers were identified for inclusion in this review: 89 papers referred to one ongoing and 55 completed observational studies, and 59 papers referred to four ongoing and eight completed RCTs.

A detailed description of the studies included is given in the table Characteristics of included studies.

1. Observational studies

Fifty‐five completed observational studies were included in this review. Forty‐one studies were nested case‐control studies, the others were subcohort controlled or cohort studies, and one study used a cohort together with a nested case‐control design. Subcohort controlled studies used (random) samples of the cohort as controls. The original papers were published between 1983 and 2013. Six studies were conducted in Asia (China, Japan and Taiwan), one in Australia, 22 in Europe (including data from Belgium, Denmark, Germany, Greece, Italy, Netherlands, Norway, Spain, Sweden, Channel Islands, Finland, France and UK) and 25 in the US. Overall, the studies included more than 1,100,000 participants. European study populations made up 45%, US 45%, Asia 9.4% and Australia 0.2% of all study participants. The median size of the study populations was 8801. Twenty‐eight studies included men and women, one did not report gender, 21 included only men and five only women. For a substantial proportion of the study populations (38%), gender was not reported. Forty‐three percent of participants were men, and 23% were women. Six studies with mixed‐gender populations reported results stratified by gender. The study populations were derived from 48 different cohorts. Twenty‐three cohorts were non‐randomly recruited (e.g. included volunteers), and 20 cohorts consisted of a random (or total) sample of the population of interest, which was either a specifically exposed population such as male tin‐miners in China or the general population.

Forty‐three studies specified the age range of their included participants; most included adults older than 40 years of age.

Seven studies investigated nutritional and/or supplemental selenium intake by using food frequency questionnaires or interviews. Forty‐eight studies assessed biochemical selenium status where:

8 used toenail specimens,
12 plasma specimens,
27 serum specimens,
and one used both serum and plasma specimens.

One study measured both serum selenium levels and intake.

The mean follow‐up period was up to three years in five studies and longer than three years in the remaining studies. Generally, study authors grouped the cases according to the International Classification of Diseases (ICD) classification that was up‐to‐date at the inception of the cohort observation. The level of disaggregation of data varied markedly between the studies. Although some studies reported cancer risk according to organ system (e.g. urinary tract, respiratory tract), others reported cancer risk for one or two organs (e.g. female breast, urinary bladder). Only in the case of skin cancer did studies also differentiate according to histological type (e.g. melanoma, basal cell carcinoma).

For the following outcomes, five or more studies were included in the review, and observational data were meta‐analysed.

Any cancer (16 studies).
Female breast cancer (7 studies).
Urinary bladder cancer (6 studies).
Lung cancer (14 studies).
Prostate cancer (17 studies).
Stomach cancer (5 studies).
Colon/colorectal cancer (5 studies).

Bates 2011 was not included in the meta‐analysis for any cancer, as it provided only the HR associated with an increase of one standard deviation of selenium exposure.

Table 3 provides an overview of the studies for each outcome. Five studies gave data for the group of “other” cancers, which encompassed any type of cancer not reported separately in the study publications. The definition of “other” cancers varied between studies, including predominantly rare cancers but also cancers of unknown origin. The results of the studies within the category "other cancers" are mentioned for the sake of completeness; however, because of the diversity of outcomes, the results were not included in further analysis or discussion of this review.

Table 1.

Included observational studies by outcome

Organ system	Outcome	Number of studies/case definitions	Meta‐ analysis	Countries	Number of participants	Number of cases	Selenium assessment	Reporting study
Any cancer	Any cancer	total: 16 incidence: 8 mortality: 6 incidence and mortality combined: 1	✓ yes	US Finland Netherlands Sweden Norway Belgium France Japan	total: ˜ 152,000	total: 3010 male: 1700 female: 736	serum: 12 plasma: 2 serum + plasma: 1 plasma selenium P: 1	Knekt 1990 Coates 1988 Kok 1987a Salonen 1984 Nomura 1987 Virtamo 1987 Willett 1983 Fex 1987 Ringstad 1988 Persson 2000 Salonen 1985 Peleg 1985,Kornitzer 2004 Akbaraly 2005 Bleys 2008 Fujishima 2011
Gynaecological cancer	Female breast cancer	total: 7 incidence: 7 incidence and mortality combined: 0	✓ yes	US Finland Netherlands Channel Islands	total/female: > 155,000 (one study did not report cohort size)	total/female: 1078	serum: 2 plasma: 1 serum + plasma: 1 toenail: 3	Dorgan 1998 van den Brandt 1993a Coates 1988 Overvad 1991 Knekt 1990 Garland 1995 van Noord 1987
	Cervical cancer	total: 2 incidence: 2 mortality: 0 incidence and mortality combined: 0	✗ no	US	total/female: > 15,161 (one study did not report cohort size)	total/female: 62	serum: 2	Menkes 1986 Coates 1988
	Uterine cancer	total: 1 incidence: 1 mortality: 0 incidence and mortality combined: 0	✗ no	US	total/female: 62,641	total/female: 91	toenail: 1	Garland 1995
	Ovarian cancer	total: 4 incidence: 4 mortality: 0 incidence and mortality combined: 0	✗ no	US Finland	total/female: ˜ 214,000	total/female: 568	serum: 2 toenail: 1 supplemental intake: 1	Knekt 1990 Garland 1995 Menkes 1986 Thomson 2008
	Gynaecological cancer (without breast cancer)	total: 1 incidence: 1 mortality: 0 incidence and mortality combined: 0	✗ no	Finland	total/female: ˜ 18,000	total/female: 86	serum: 1	Knekt 1990
Urological cancers	Urinary bladder cancer	total: 6 incidence: 6 mortality: 0 incidence & mortality combined: 0	✓ yes	US/Hawaii Finland Netherlands	total: 356,150 female: 130,786 male: 128,009	total: 1295 female: 175 male 755	serum: 3 toenail: 3	Menkes 1986 Nomura 1987 Michaud 2002 van den Brandt 1993a Michaud 2005 Hotaling 2011
Urological cancers	Urinary tract cancer	total: 2 incidence: 2 mortality: 0 incidence & mortality combined: 0	✗ no	Netherlands Finland	total: 48,000	total: 104 male: 91 female: 13	serum: 1 plasma: 1	Knekt 1990 Persson 2000
Respiratory tract cancers	Lung cancer	total: 14 incidence: 12 mortality: 2 incidence and mortality combined: 0	✓ yes	China Japan US Finland Netherlands Denmark	total: ˜ 336,000 male: 125,341 female: 181,895	total: 2002 male: 1256 female: 333	serum: 9 serum + plasma: 2 toenail: 2 dietary intake: 1 (one study reported both serum levels and food intake)	Knekt 1990 Knekt 1998 Garland 1995 Coates 1988 Nomura 1987 van den Brandt 1993a Kabuto 1994 Menkes 1986 Goodman 2001 Comstock 1997 Kromhout 1987 Ratnasinghe 2000 Epplein 2009 Suadicani 2012
	Oral/pharyngeal cancer	total: 1 incidence: 1 mortality: 0 incidence and mortality combined: 0	✗ no	US	total: 25,804	total: 28	serum: 1	Menkes 1986
	Any cancer of the respiratory tract	total: 1 incidence: 1 mortality: 0 incidence and mortality combined: 0	✗ no	Sweden	total/male: ˜ 9500	total/male: 69	plasma selenium P: 1	Persson 2000
Andrological cancers	Prostate cancer	total: 17 incidence: 17 mortality: 0 incidence and mortality combined: 0	✓ yes	US Europe	total/male: > 421,000 (one study did not report cohort size)	total/male: 6366	serum: 8 plasma: 3 toenail: 3 dietary intake: 3	Hartman 1998,Helzlsouer 2000 Coates 1988 Brooks 2001 van den Brandt 1993a Nomura 2000 Goodman 2001 Yoshizawa 1998 Li 2004a Peters 2007 Peters 2008 Allen 2008 Epplein 2009
Gastrointestinal cancers	Oesophageal cancer	total: 2 incidence: 2 mortality: 1 incidence and mortality combined: 0	✗ no	China US	total: 29,923	total: > 959	serum: 1 supplemental intake: 1	Wei 2004 Dong 2008
	Oesophageal squamous cell carcinoma	total:1 incidence: 1 mortality:0 incidence and mortality combined: 0	✗ no	Netherlands	total: 120,852	total: 64	toenail: 1	Steinbrecher 2010
	Oesophageal adenocarcinoma	total:1 incidence:1 mortality:0 incidence and mortality combined: 0	✗ no	Netherlands	total: 120,852	total: 112	toenail: 1	Steinbrecher 2010
	Oesophageal/stomach cancer	total: 1 incidence: 1 mortality: 0 incidence and mortality combined: 0	✗ no	Netherlands	total: 36,265	total: 86 male: 51 female: 35	serum: 1	Knekt 1998
	Gastric cardia adenocarcinoma	total:1 incidence:1 mortality:0 incidence and mortality combined: 0	✗ no	Netherlands	total: 120,852	total:114	toenail: 1	Steinbrecher 2010
	Stomach cancer	total: 5 incidence: 5 mortality: 1 incidence and mortality combined: 0	✓ yes	China Japan US/Hawaii Finland Netherlands	total: ˜ 197,000 male: 86,311 female: 80,669	total: 955 male: 626 female: 329	serum: 4 toenail: 1	Knekt 1990 van den Brandt 1993a Nomura 1987 Kabuto 1994 Wei 2004
	Primary liver cancer	total: 2 incidence: 1 mortality: 1 incidence and mortality combined: 0	✗ no	Taiwan	total: 46,404	total: 235 male: 223 female: 12	plasma: 1 toenail: 1	Yu 1999 Sakoda 2005
	Pancreatic cancer	total: 2 incidence: 2 mortality: 0 incidence and mortality combined: 0	✗ no	US Finland	total: 65,072	total: 67 male: 31 female: 36	serum: 2	Menkes 1986 Knekt 1990).
	Colon/colorectal cancer	total: colon 2, colorectum 3 incidence: 5 mortality: 0 incidence and mortality combined: 0	✓ yes	US Netherlands Finland	total: 255,425 male: 86,311 female: 143,310	total: 617 male: 285 female: 332	serum: 3 toenail: 2	van den Brandt 1993a Nomura 1987 Menkes 1986 Garland 1995 Knekt 1990
	Rectal cancer	total: 2 incidence: 2 mortality: 0 incidence and mortality combined: 0	✗ no	US/Hawaii Netherlands	total: 127,712	total: 145 male: 109 female: 36	serum: 1 toenail: 1	van den Brandt 1993a Nomura 1987
	All gastrointestinal cancers	total: 2 incidence: 2 mortality: 0 incidence and mortality combined: 0	✗ no	US Sweden	total: > 9500 (one study did not report cohort size)	total: 143	plasma + serum: 1 plasma selenium P: 1	Coates 1988 Persson 2000
Skin cancer	Melanoma	total: 3 incidence: 3 mortality: 0 incidence and mortality combined: 0	✗ no	US	total: ˜ 158,000	total: 547	serum: 1 toenail: 1 supplemental intake: 1	Garland 1995,Menkes 1986,Peters 2008
	Basal cell carcinoma	total: 3 incidence: 3 mortality: 0 incidence and mortality combined: 0	✗ no	Australia US Finland	total: > 66,000	total: 292	serum: 3 dietary intake: 1	Knekt 1990 Menkes 1986 McNaughton 2005
	Squamous cell carcinoma	total: 4 incidence: 4 mortality: 0 incidence and mortality combined: 0	✗ no	Australia US	total: ˜ 30,000	total: 488	serum: 2 plasma: 1 dietary intake: 1	Combs 1993 Karagas 1997 Menkes 1986 McNaughton 2005
	Total non‐melanoma skin cancer	total: 1 incidence: 1 mortality: 0 incidence and mortality combined: 0	✗ no	US	total: 117	total: 19	plasma: 1	Clark 1985
Rare and other cancers	Haematological cancers	total: 1 incidence: 1 mortality: 0 incidence and mortality combined: 0	✗ no	US	total: ˜ 6200	total: 12	serum + plasma: 12	Coates 1988
	Thyroid cancer	total: 1 incidence: 1 mortality: 0 incidence and mortality combined: 0	✗ no	Norway	total: 100,000	total: 43 male: 12 female: 31	serum: 1	Glattre 1989
	Other cancers	total: 5 incidence:4 mortality:1 incidence and mortality combined: 0	✗ no	China US Finland Sweden		total: 573 male: 230 female: 285		Garland 1995 Coates 1988 Knekt 1990 Wei 2004 Persson 2000

Study	Publication	Newcastle Ottawa Scale (cohort)				Newcastle Ottawa Scale (case‐control)
Study	Publication	Selection	Comparability	Outcome	Total	Selection	Comparability	Exposure	Total
Kabuto 1994	Kabuto 1994	0‐1‐1‐1	2	1‐1‐0	7	0‐1‐1‐1	2	1‐1‐1	8
Ratnasinghe 2000	Ratnasinghe 2000	1‐1‐1‐1	2	1‐0‐0	7	0‐0‐1‐1	2	1‐1‐1	7
Sakoda 2005	Sakoda 2005	0‐1‐1‐0	1	1‐1‐0	5	1‐1‐1‐1	1	1‐1‐1	8
Wei 2004	Wei 2004	1‐1‐1‐1	1	1‐1‐1	8	.‐.‐.‐.	.	.‐.‐.	.
Wei 2004	Mark 2000	1‐1‐1‐1	1	1‐1‐1	8	.‐.‐.‐.	.	.‐.‐.	.
Yu 1999	Yu 1999	0‐1‐1‐1	2	1‐1‐0	7	1‐1‐1‐1	2	1‐1‐1	9
McNaughton 2005	McNaughton 2005	1‐1‐1‐1	1	1‐1‐0	7	1‐1‐1‐1	1	1‐1‐1	8
	Heinen 2007	1‐1‐1‐1	2	1‐1‐1	9	.‐.‐.‐.	.	.‐.‐.	.
	van der Pols 2009	1‐1‐1‐1	2	1‐1‐0	8	.‐.‐.‐.	.	.‐.‐.	.
Akbaraly 2005	Akbaraly 2005	0‐1‐1‐1	2	0‐1‐0	6	.‐.‐.‐.	.	.‐.‐.	.
Allen 2008	Allen 2008	1‐1‐1‐1	2	1‐1‐0	8	1‐1‐1‐1	2	1‐1‐1	9
Fex 1987	Fex 1987	1‐1‐1‐0	2	1‐1‐1	8	1‐0‐1‐1	2	1‐1‐1	8
Glattre 1989	Glattre 1989	0‐1‐1‐0	1	1‐1‐1	6	1‐1‐1‐1	1	1‐1‐1	8
Hartman 1998	Hartman 1998	1‐1‐0‐1	2	1‐1‐0	7	.‐.‐.‐.	.	.‐.‐.	.
Knekt 1990	Knekt 1990	1‐1‐1‐1	2	1‐1‐1	9	0‐1‐1‐1	2	1‐1‐1	8
	Hakama 1990	1‐1‐1‐1	2	1‐1‐1	9	0‐1‐1‐1	2	1‐1‐1	8
	Knekt 1988	1‐1‐1‐1	2	1‐1‐1	9	0‐0‐1‐1	2	1‐1‐1	7
	Knekt 1996	1‐1‐1‐1	1	1‐1‐1	8	0‐1‐1‐1	1	1‐1‐1	7
	Knekt 1991	1‐1‐1‐1	2	1‐1‐1	9	0‐1‐1‐1	2	1‐1‐1	8
Knekt 1998	Knekt 1998	1‐1‐1‐1	2	1‐1‐1	9	0‐1‐1‐1	2	1‐1‐1	8
Kok 1987a	Kok 1987b	1‐1‐1‐1	2	1‐1‐1	9	1‐0‐1‐1	2	1‐1‐1	8
Kok 1987a	Kok 1987a	.‐.‐.‐.	.	.‐.‐.	.	.‐.‐.‐.	.	.‐.‐.	.
Kornitzer 2004	Kornitzer 2004	1‐1‐1‐0	1	1‐1‐1	7	1‐1‐1‐1	1	1‐1‐1	8
Kromhout 1987	Kromhout 1987	1‐1‐1‐0	2	1‐1‐1	8	.‐.‐.‐.	.	.‐.‐.	.
Michaud 2002	Michaud 2002	1‐1‐1‐1	2	1‐1‐0	8	0‐1‐1‐1	2	1‐1‐1	8
Overvad 1991	Overvad 1991	1‐1‐1‐0	1	1‐1‐0	6	.‐.‐.‐.	.	.‐.‐.	.
Persson 2000	Persson‐Moschos 2000	1‐1‐1‐0	2	1‐1‐1	8	1‐0‐1‐1	2	1‐1‐1	8
Ringstad 1988	Ringstad 1988	1‐1‐1‐1	2	1‐1‐0	8	1‐1‐1‐1	2	1‐1‐1	9
Salonen 1984	Salonen 1984	1‐1‐1‐1	2	1‐1‐1	9	0‐1‐1‐1	2	1‐1‐1	8
Salonen 1985	Salonen 1985	1‐1‐1‐1	2	1‐1‐1	9	1‐1‐1‐1	2	1‐1‐1	9
van Noord 1987	van Noord 1987	1‐1‐1‐0	1	1‐0‐1	6	1‐1‐1‐0	1	1‐1‐1	7
van den Brandt 1993a	van den Brandt 1993	1‐1‐1‐1	2	1‐1‐1	9	.‐.‐.‐.	.	.‐.‐.	.
	van den Brandt 1994	1‐1‐1‐1	2	1‐1‐1	9	.‐.‐.‐.	.	.‐.‐.	.
	van den Brandt 1993	1‐1‐1‐1	2	1‐1‐1	9	.‐.‐.‐.	.	.‐.‐.	.
	van den Brandt 2003	1‐1‐1‐1	2	1‐1‐1	9	.‐.‐.‐.	.	.‐.‐.	.
	Zeegers 2002	1‐1‐1‐1	2	1‐1‐1	9	.‐.‐.‐.	.	.‐.‐.	.
	Steevens 2010	1‐1‐1‐1	2	1‐1‐1	9	0‐1‐1‐1	2	1‐0	6
Virtamo 1987	Virtamo 1987	0‐1‐1‐1	2	1‐1‐1	8	.‐.‐.‐.	.	.‐.‐.	.
Bleys 2008	Bleys 2008	1‐1‐1‐1	2	1‐1‐1	9	.‐.‐.‐.	.	.‐.‐.	.
Brooks 2001	Brooks 2001	0‐1‐1‐0	2	1‐0‐0	5	1‐0‐1‐1	2	1‐1‐0	7
Clark 1985	Clark 1985	0‐1‐1‐0	0	0‐0‐0	2	.‐.‐.‐.	.	.‐.‐.	.
Coates 1988	Coates 1988	0‐1‐1‐0	1	1‐1‐0	5	1‐0‐1‐0	1	1‐1‐1	6
Coates 1988	Coates 1987	.‐.‐.‐.	.	.‐.‐.	.	.‐.‐.‐.	.	.‐.‐.	.
Combs 1993	Combs Jr 1993	0‐1‐1‐0	2	1‐0‐0	5	.‐.‐.‐.	.	.‐.‐.	.
Comstock 1997	Comstock 1997	0‐1‐1‐0	2	1‐1‐0	6	1‐1‐1‐1	2	1‐1‐1	9
Dong 2008	Dong 2008	1‐1‐1‐1	2	1‐1‐1	9	.‐.‐.‐.	.	.‐.‐.	.
Dorgan 1998	Dorgan 1998	0‐1‐1‐1	2	0‐1‐0	6	1‐1‐1‐1	2	1‐1‐1	9
Epplein 2009	Epplein 2009	0‐1‐1‐1	2	1‐1‐0	7	0‐1‐1‐1	2	1‐1‐1	8
Epplein 2009	Gill 2009	0‐1‐1‐1	1	1‐1‐0	6	0‐1‐1‐1	1	1‐1‐1	7
Garland 1995	Garland 1995	0‐1‐1‐1	2	1‐1‐1	8	1‐1‐1‐1	2	1‐1‐1	9
Garland 1995	Hunter 1990	0‐1‐1‐1	2	1‐1‐1	8	1‐1‐1‐1	2	1‐1‐1	9
Goodman 2001	Goodman 2001	0‐1‐1‐0	2	1‐1‐0	6	1‐1‐1‐1	2	1‐1‐1	9
Helzlsouer 2000	Helzlsouer 2000	0‐1‐1‐1	1	1‐1‐0	6	1‐1‐1‐1	1	1‐1‐1	8
Karagas 1997	Karagas 1997	0‐1‐1‐1	2	1‐1‐1	8	1‐1‐1‐1	2	1‐1‐1	9
Li 2004a	Li 2004	0‐1‐1‐1	2	0‐1‐1	7	1‐1‐1‐1	2	1‐1‐1	9
Menkes 1986	Menkes 1986	0‐1‐1‐1	2	1‐1‐0	7	1‐1‐1‐1	2	1‐1‐1	9
	Batieha 1993	0‐1‐1‐1	2	1‐1‐0	7	1‐1‐1‐1	2	1‐1‐1	9
	Breslow 1995	0‐1‐1‐1	2	1‐1‐0	7	1‐0‐1‐1	2	1‐1‐1	8
	Burney 1989	0‐1‐1‐1	2	1‐1‐0	7	0‐1‐1‐1	2	1‐1‐1	8
	Helzlsouer 1996	0‐1‐1‐1	2	1‐1‐0	7	0‐1‐1‐1	2	1‐1‐1	8
	Helzlsouer 1989	0‐1‐1‐1	2	1‐1‐0	7	1‐1‐1‐1	2	1‐1‐1	9
	Ko 1994	0‐1‐1‐0	2	1‐1‐0	6	1‐1‐1‐1	2	1‐1‐1	9
	Menkes 1986	.‐.‐.‐.	.	.‐.‐.	.	.‐.‐.‐.	.	.‐.‐.	.
	Schober 1987	0‐1‐1‐1	1	1‐1‐0	6	0‐1‐1‐1	1	1‐1‐1	7
	Schober 1986	.‐.‐.‐.	.	.‐.‐.	.	.‐.‐.‐.	.	.‐.‐.	.
	Zheng 1993	0‐1‐1‐1	2	1‐1‐0	7	0‐1‐1‐1	2	1‐1‐1	8
Michaud 2005	Michaud 2005	0‐1‐1‐1	2	0‐1‐0	6	1‐1‐1‐1	2	1‐1‐1	9
Nomura 1987	Nomura 1987	1‐1‐1‐1	2	1‐1‐1	9	1‐1‐1‐1	2	1‐1‐1	9
Nomura 2000	Nomura 2000	1‐1‐1‐1	2	1‐1‐1	9	1‐1‐1‐1	2	1‐1‐1	9
Peleg 1985	Peleg 1985	1‐1‐1‐1	1	1‐1‐0	7	1‐1‐1‐1	1	1‐1‐1	8
Peters 2007	Peters 2007	0‐1‐1‐1	2	1‐1‐0	7	1‐1‐1‐1	2	1‐1‐1	9
Peters 2008	Peters 2008	0‐1‐1‐1	1	1‐1‐1	7	.‐.‐.‐.	.	.‐.‐.	.
	Asgari 2009	0‐1‐1‐1	1	1‐1‐0	6	.‐.‐.‐.	.	.‐.‐.	.
	Hotaling 2011	0‐1‐0‐1	0	1‐1‐1	5	.‐.‐.‐.	.	.‐.‐.	.
	Walter 2011	0‐1‐0‐1	2	1‐1‐1	7	.‐.‐.‐.	.	.‐.‐.	.
Thomson 2008	Thomson 2008	0‐1‐1‐1	2	0‐1‐0	6	.‐.‐.‐.	.	.‐.‐.	.
Willett 1983	Willett 1983	1‐1‐1‐0	2	1‐1‐0	7	1‐1‐1‐1	2	1‐1‐1	9
Yoshizawa 1998	Yoshizawa 1998	0‐1‐1‐1	2	1‐1‐1	8	1‐0‐1‐1	2	1‐1‐1	8
Fujishima 2011	Fujishima 2011	1‐1‐1‐1	2	1‐1‐1	9	.‐.‐.‐.	.	.‐.‐.	.
Grundmark 2011	Grundmark 2011	1‐1‐1‐1	2	1‐1‐1	9	.‐.‐.‐.	.	.‐.‐.	.
Bates 2011	Bates 2011	1‐1‐1‐1	1	1‐1‐1	8	.‐.‐.‐.	.	.‐.‐.	.
Suadicani 2012	Suadicani 2012	0‐1‐1‐1	2	1‐1‐1	8	.‐.‐.‐.	.	.‐.‐.	.
Agalliu 2011	Agalliu 2011	0‐1‐0‐1	1	1‐1‐0	5	0‐1‐0‐1	1	1‐1‐0	5
Steinbrecher 2010	Steinbrecher 2010	1‐1‐1‐1	2	0‐1‐0	7	1‐1‐1‐1	2	0‐1‐1	8

Study	Sequence generation	Allocation concealment	Blinding	Completeness of outcome data	Risk of bias
NPCT 2002	adequate	adequate	unclear	adequate	unclear
Li 2000	unclear	unclear	adequate	adequate	unclear
Yu 1997	unclear	unclear	adequate	unclear	unclear
Yu 1991	unclear	unclear	adequate	unclear	unclear
SELECT 2009	adequate	adequate	adequate	adequate	low
Algotar 2013	adequate	adequate	adequate	adequate	low
Marshall 2011	adequate	adequate	adequate	adequate	low
Dreno 2007	adequate	adequate	unclear	adequate	unclear

Organ system	Cancer	Case definition	Relative risk estimate (highest vs lowest exposure category)	95% CI	Selenium marker	Gender	Study
Gynaecological	Cervix	incidence	0.89	0.40 to 2.00	serum	women	Menkes 1986 (Batieha 1993)
	Cervix	incidence	1.10	n.r.	serum		Coates 1988
	Gynaecological (without breast)	incidence	0.96	n.r.	serum		Knekt 1990
	Ovary	incidence	0.87	0.25 to 5.26	serum		Knekt 1990 (Knekt 1996)
			1.22	0.44 to 3.38	toenail		Garland 1995
			0.58	0.2 to 1.7	serum		Menkes 1986 (Helzlsour 1996)
			1.00 (HR)	0.73 to 1.37	suppl. intake		Thomson 2008
	Uterus	incidence	1.38	0.62 to 3.08	toenail		Garland 1995
Gastrointestinal	Gastrointestinal tract (all)	incidence	1.00	n.r.	serum/plasma	both	Coates 1988
	Gastrointestinal tract (all)	incidence	0.29	0.10 to 0.91	plasma	men	Persson 2000
	Oesophageal squamous cell carcinoma	incidence	0.37	0.16 to 0.86	toenail	both	Steevens 2010
	Oesophageal adenocarcinoma	incidence	0.76	0.41 to 1.40	toenail	both	Steevens 2010
	Oesophagus	incidence	0.56	0.44 to 0.71	serum	both	Wei 2004 (Mark 2000)
		mortality	0.62	0.44 to 0.89	serum	both	Wei 2004 (Mark 2000)
		mortality	0.35	0.16 to 0.81	serum	both	Wei 2004 (Wei 2004)
		incidence	0.27	0.03 to 2.21	suppl. intake	both	Dong 2008
	Gastric cardio adenocarcinoma	incidence	0.52	0.27 to 1.02	toenail	both	Steevens 2010
	Oesophagus and stomach	incidence	0.45	n.r.	serum	men	Knekt 1990 (Knekt 1988)
	Oesophagus and stomach	incidence	0.67	n.r.	serum	women	Knekt 1990 (Knekt 1988)
	Liver	incidence	0.62	0.21 to 1.86	plasma	men	Yu 1999
		incidence	0.50	0.28 to 0.90	toenail	both	Sakoda 2005
		mortality	0.50	0.28 to 0.90		both
			0.57	0.31 to 1.05		men
			0.18	0.03 to 1.13		women
	Pancreas	incidence	0.08	0.01 to 0.56)	serum	men	Menkes 1986 (Burney 1989)
			0.83	0.4 to 1.67	serum	women	Menkes 1986 (Burney 1989)
			0.58	n.r.	serum	men	Knekt 1990
			3.49	n.r.	serum	women	Knekt 1990
	Rectum	incidence	0.625	n.r.	serum	men	Nomura 1987
			1.05	0.54 to 2.03	toenail	both	van den Brandt 1993a
			0.91	0.41 to 2.00		men
			1.58	0.59 to 4.22		women
Urinary tract	Urinary tract (all)	incidence	0.97	0.72 to 1.31	serum	both	Hotaling 2011
			5.0	0.71 to ∞	plasma	men	Persson 2000
			0.81	n.r.	serum	men	Knekt 1990
			4.12	n.r.	serum	women	Knekt 1990
Respiratory tract	Cavum oris/pharynx	incidence	5.43	n.r.	serum	both	Menkes 1986 (Zheng 1993)
Respiratory tract	Respiratory tract (all)	incidence	6.0	1.5 to 24.2	plasma	men	Persson 2000
Skin	Melanoma	incidence	1.66	0.71 to 3.85	toenail	women	Garland 1995
			0.90	0.30 to 2.50	serum	both	Menkes 1986 (Breslow 1995)
			0.98	0.69 to 1.41	suppl. intake	both	Peters 2008 (Asgari 2009)
	Any non‐melanoma cancer	incidence	0.77	n.r.	plasma	both	Clark 1985
	Basal cell carcinoma	incidence	0.54	n.r.	serum	men	Knekt 1990
			1.55	n.r.	serum	women	Knekt 1990
			0.80	0.10 to 4.5	serum	both	Menkes 1986 (Breslow 1995)
			0.86	0.38 to 1.96	serum	both	McNaughton 2005
			0.95	0.59 to 1.50	nutritional intake	both	McNaughton 2005
	Squamous cell carcinoma	incidence	0.69	0.51 to 0.92	plasma	both	Combs 1993
			0.60	0.20 to 1.50	serum	both	Menkes 1986 (Breslow 1995)
			0.86	0.47 to 1.58	plasma	both	Karagas 1997
			1.30	0.77 to 2.3	nutritional intake	both	McNaughton 2005
			0.49	0.24 to 0.99	serum	both	McNaughton 2005
Other	Haematological	incidence	0.60	n.r.	serum/plasma	both	Coates 1988
	Haematological	incidence	0.95	0.75 to 1.20	suppl. intake	both	Walter 2011
	Thyroid	incidence	0.15	0.0 to 5.0	serum	men	Glattre 1989
			0.12	0.01 to 1.11		women
			0.13	0.02 to 0.77		both
Not defined		mortality	0.72 (HR)	0.58 to 0.89	plasma	both	Bates 2011

Database	Date of most recent literature search	Search strategy	Comment
www.cancer.gov	4 Feb 2011	medication: selenium indication: prevention
Cancerlit	Oct 2004	1 selen* OR organoselen* OR natriumselen* 2 random* OR placebo* OR clinical trial* OR controlled trial* OR controlled clinical trial* OR double blind* OR single blind* 3 epidemiologic stud* OR cohort OR case‐control stud* OR nested case‐control* OR case‐control design* OR prospectiv* 4 2 OR 3 5 1 AND 4	Now included in MEDLINE database
Clinical Contents in Medicine (CCMed)	4 Feb 2011	selen* OR organoselen* OR natriumselen*
CENTRAL	2013, Issue 1	#1 MeSH descriptor: [Selenium] this term only #2 MeSH descriptor: [Selenium Compounds] explode all trees #3 MeSH descriptor: [Organoselenium Compounds] explode all trees #4 selen* #5 #1 or #2 or #3 or #4 #6 MeSH descriptor: [Neoplasms] explode all trees #7 (neoplasm* or cancer* or tumor* or tumour* or carcino* or malignan* or adenocarcinoma* or sarcoma* or adenoma* or chondrosarcoma* or fibrosarcoma* or dermatofibrosarcoma* or neurofibrosarcoma* or hemangiosarcoma* or leiomyosarcoma* or liposarcoma* or myosarcoma* or rhabdomyosarcoma* or myxosarcoma* or osteosarcoma* or lymphoma*) #8 #6 or #7 #9 #5 and #8
metaRegister of Controlled Trials (mRCT, www.controlled‐trials.com)	4 Feb 2011	selen AND cancer
EMBASE Ovid	2013 week 6	1 selenium/ 2 selen.mp. 3 selenium derivative/ 4 methylseleninic acid/ 5 methylselenium.mp. 6 exp organoselenium derivative/ 7 1 or 2 or 3 or 4 or 5 or 6 8 exp neoplasm/ 9 (neoplasm or cancer* or tumor* or tumour* or carcino* or malignan* or adenocarcinoma* or sarcoma* or adenoma* or chondrosarcoma* or fibrosarcoma* or dermatofibrosarcoma* or neurofibrosarcoma* or hemangiosarcoma* or leiomyosarcoma* or liposarcoma* or myosarcoma* or rhabdomyosarcoma* or myxosarcoma* or osteosarcoma* or lymphoma).mp. 10 8 or 9 11 7 and 10 12 exp clinical study/ 13 crossover procedure/ 14 double‐blind procedure/ 15 single‐blind procedure/ 16 cohort analysis/ 17 observational study/ 18 (random or factorial* or crossover* or cross‐over* or cross over* or placebo* or (double adj blind) or (singl adj blind) or assign or allocat* or volunteer* or observ* or cohort* or prospectiv* or (case* and control*)).mp. 19 12 or 13 or 14 or 15 or 16 or 17 or 18 20 11 and 19 21 (exp animal/ or nonhuman/ or exp animal experiment/) not human/ 22 20 not 21 key: [mp=title, abstract, subject headings, heading word, drug trade name, original title, device manufacturer, drug manufacturer, device trade name, keyword]
German Cancer Study Register: www.studien.de	4 Feb 2011	selen
MEDLINE (via Ovid)	Feb 2013 week 1	1 Selenium/ 2 exp Selenium Compounds/ 3 exp Organoselenium Compounds/ 4 selen.mp. 5 1 or 2 or 3 or 4 6 exp Neoplasms/ 7 (neoplasm or cancer* or tumor* or tumour* or carcino* or malignan* or adenocarcinoma* or sarcoma* or adenoma* or chondrosarcoma* or fibrosarcoma* or dermatofibrosarcoma* or neurofibrosarcoma* or hemangiosarcoma* or leiomyosarcoma* or liposarcoma* or myosarcoma* or rhabdomyosarcoma* or myxosarcoma* or osteosarcoma* or lymphoma).mp. 8 6 or 7 9 5 and 8 10 randomized controlled trial.pt. 11 controlled clinical trial.pt. 12 randomized.ab. 13 placebo.ab. 14 drug therapy.fs. 15 randomly.ab. 16 trial.ab. 17 groups.ab. 18 exp case‐control studies/ 19 exp Cohort Studies/ 20 (cohort or observ* or prospectiv* or (case* and control*)).mp. 21 10 or 11 or 12 or 13 or 14 or 15 or 16 or 17 or 18 or 19 or 20 22 9 and 21 23 exp animals/ not humans.sh. 24 22 not 23 key: mp=title, abstract, original title, name of substance word, subject heading word, keyword heading word, protocol supplementary concept, rare disease supplementary concept, unique identifier pt=publication type ab=abstract fs=floating subheading
SIGLE	Oct 2004	?selen?	database discontinued in 2005

Outcome or subgroup title	No. of studies	No. of participants	Statistical method	Effect size
1 Total cancer incidence and mortality	14		Odds Ratio (Random, 95% CI)	Subtotals only
1.1 Incidence	8		Odds Ratio (Random, 95% CI)	0.69 [0.53, 0.91]
1.2 Mortality	6		Odds Ratio (Random, 95% CI)	0.60 [0.39, 0.93]
2 Total cancer incidence and mortality (men)	8		Odds Ratio (Random, 95% CI)	Subtotals only
2.1 Incidence	5		Odds Ratio (Random, 95% CI)	0.66 [0.42, 1.05]
2.2 Mortality	3		Odds Ratio (Random, 95% CI)	0.56 [0.38, 0.81]
3 Total cancer incidence and mortality (women)	5		Odds Ratio (Random, 95% CI)	Subtotals only
3.1 Incidence	2		Odds Ratio (Random, 95% CI)	0.90 [0.45, 1.77]
3.2 Mortality	3		Odds Ratio (Random, 95% CI)	0.92 [0.79, 1.07]
4 Total cancer incidence and mortality (ascending order of selenium levels)	12		Odds Ratio (Random, 95% CI)	Subtotals only
4.1 Incidence	6	1297	Odds Ratio (Random, 95% CI)	0.69 [0.52, 0.91]
4.2 Mortality	6	1041	Odds Ratio (Random, 95% CI)	0.60 [0.39, 0.93]
5 Breast cancer risk (women)	7		Odds Ratio (Random, 95% CI)	1.00 [0.77, 1.29]
5.1 Breast cancer (all)	6		Odds Ratio (Random, 95% CI)	1.01 [0.74, 1.36]
5.2 Breast cancer (premenopausal)	1		Odds Ratio (Random, 95% CI)	1.10 [0.46, 2.65]
6 Bladder cancer risk	5		Odds Ratio (Random, 95% CI)	0.67 [0.46, 0.97]
6.1 All (male + female)	2		Odds Ratio (Random, 95% CI)	0.65 [0.46, 0.92]
6.2 Male	3		Odds Ratio (Random, 95% CI)	0.82 [0.41, 1.62]
6.3 Female	1		Odds Ratio (Random, 95% CI)	0.36 [0.14, 0.92]
7 Lung cancer risk (gender‐aggregated data)	12		Odds Ratio (Random, 95% CI)	0.83 [0.61, 1.13]
7.1 Incidence	10		Odds Ratio (Random, 95% CI)	0.75 [0.54, 1.03]
7.2 Mortality	2		Odds Ratio (Random, 95% CI)	1.34 [0.93, 1.93]
8 Lung cancer risk (gender‐disaggregated data)	12		Odds Ratio (Random, 95% CI)	0.84 [0.65, 1.09]
8.1 All (female + male)	4		Odds Ratio (Random, 95% CI)	0.58 [0.39, 0.86]
8.2 Female	4		Odds Ratio (Random, 95% CI)	0.83 [0.43, 1.61]
8.3 Male	7		Odds Ratio (Random, 95% CI)	0.98 [0.68, 1.39]
9 Lung cancer risk	12		Odds Ratio (Random, 95% CI)	0.83 [0.61, 1.13]
9.1 Intake	1		Odds Ratio (Random, 95% CI)	0.98 [0.41, 2.35]
9.2 Serum or plasma	9		Odds Ratio (Random, 95% CI)	0.91 [0.70, 1.18]
9.3 Toenail	2		Odds Ratio (Random, 95% CI)	1.05 [0.11, 10.36]
10 Lung cancer risk (ascending order of selenium levels)	8	1867	Odds Ratio (Random, 95% CI)	0.97 [0.74, 1.27]
11 Prostate cancer risk	17		Odds Ratio (Random, 95% CI)	0.79 [0.69, 0.90]
12 Prostate cancer risk (by selenium measurement)	17		Odds Ratio (Random, 95% CI)	0.79 [0.69, 0.90]
12.1 Biochemical selenium level	15		Odds Ratio (Random, 95% CI)	0.76 [0.67, 0.88]
12.2 Estimated selenium intake	2		Odds Ratio (Random, 95% CI)	1.00 [0.73, 1.36]
13 Prostate cancer risk (by exposure assessment)	17		Odds Ratio (Random, 95% CI)	0.79 [0.69, 0.90]
13.1 Intake	2		Odds Ratio (Random, 95% CI)	1.00 [0.73, 1.36]
13.2 Serum or plasma	12		Odds Ratio (Random, 95% CI)	0.82 [0.72, 0.93]
13.3 Toenail	3		Odds Ratio (Random, 95% CI)	0.53 [0.35, 0.81]
14 Prostate cancer risk (by continent)	17		Odds Ratio (Random, 95% CI)	0.79 [0.69, 0.90]
14.1 Europe	6		Odds Ratio (Random, 95% CI)	0.86 [0.73, 1.02]
14.2 North America	11		Odds Ratio (Random, 95% CI)	0.73 [0.60, 0.88]
15 Prostate cancer risk (by country)	17		Odds Ratio (Random, 95% CI)	0.79 [0.69, 0.90]
15.1 Several European countries	3		Odds Ratio (Random, 95% CI)	0.87 [0.71, 1.07]
15.2 Finland	2		Odds Ratio (Random, 95% CI)	1.24 [0.75, 2.05]
15.3 The Netherlands	1		Odds Ratio (Random, 95% CI)	0.69 [0.48, 0.99]
15.4 US	11		Odds Ratio (Random, 95% CI)	0.73 [0.60, 0.88]
16 Prostate cancer risk (ascending order of selenium levels)	12	2982	Odds Ratio (Random, 95% CI)	0.82 [0.72, 0.93]
17 Stomach cancer risk	5		Odds Ratio (Random, 95% CI)	0.66 [0.43, 1.01]
17.1 Stomach	4		Odds Ratio (Random, 95% CI)	0.65 [0.35, 1.19]
17.2 Stomach: cardia cancer	1		Odds Ratio (Random, 95% CI)	0.47 [0.33, 0.66]
17.3 Stomach: non‐cardia cancer	1		Odds Ratio (Random, 95% CI)	1.07 [0.55, 2.08]
18 Stomach cancer risk (by gender)	5		Odds Ratio (Random, 95% CI)	0.66 [0.42, 1.04]
18.1 All (female + male)	2		Odds Ratio (Random, 95% CI)	0.75 [0.41, 1.36]
18.2 Female	2		Odds Ratio (Random, 95% CI)	0.73 [0.12, 4.35]
18.3 Male	3		Odds Ratio (Random, 95% CI)	0.43 [0.14, 1.32]
19 Colorectal cancer risk	5		Odds Ratio (Random, 95% CI)	0.89 [0.65, 1.23]
19.1 Colon and rectal cancer	2		Odds Ratio (Random, 95% CI)	1.11 [0.50, 2.46]
19.2 Colon cancer	3		Odds Ratio (Random, 95% CI)	0.80 [0.56, 1.15]
20 Colorectal cancer risk (by gender)	5		Odds Ratio (Random, 95% CI)	0.89 [0.65, 1.23]
20.1 All (female + male)	1		Odds Ratio (Random, 95% CI)	1.22 [0.52, 2.86]
20.2 Female	3		Odds Ratio (Random, 95% CI)	1.06 [0.57, 2.00]
20.3 Male	3		Odds Ratio (Random, 95% CI)	0.69 [0.42, 1.12]

Outcome or subgroup title	No. of studies	No. of participants	Statistical method	Effect size
1 Any cancer risk	2	18698	Risk Ratio (IV, Random, 95% CI)	0.90 [0.70, 1.17]
2 Cancer mortality	2	18698	Risk Ratio (IV, Random, 95% CI)	0.81 [0.49, 1.32]
3 Liver cancer risk	3	4765	Risk Ratio (IV, Random, 95% CI)	0.50 [0.35, 0.71]
4 Non‐melanoma skin cancer risk	3		Risk Ratio (Random, 95% CI)	1.44 [0.95, 2.17]
5 Prostate cancer risk	4	19110	Risk Ratio (IV, Random, 95% CI)	0.90 [0.71, 1.14]
6 Prostate cancer risk for studies with low RoB	3	18183	Risk Ratio (IV, Random, 95% CI)	1.02 [0.90, 1.14]
7 Lung cancer risk	2	18698	Risk Ratio (IV, Random, 95% CI)	0.94 [0.62, 1.42]
8 Bladder cancer risk	2	18698	Risk Ratio (IV, Random, 95% CI)	1.14 [0.81, 1.61]
9 Colorectal cancer risk	2	18698	Risk Ratio (IV, Random, 95% CI)	0.77 [0.37, 1.62]

Date	Event	Description
10 February 2015	Amended	Minor edit made to feedback response
3 February 2015	Feedback has been incorporated	Feedback and author's response added
18 March 2014	New citation required but conclusions have not changed	New trials added. Meta‐analysis of data from RCTs was applied when at least two studies were available for each outcome
15 February 2013	New search has been performed	Search strategy updated
9 January 2013	Amended	Authors' list changed
14 August 2012	Feedback has been incorporated	Additional feedback and author response incorporated.
8 March 2012	Feedback has been incorporated	Feedback submitted and author's reply added.
6 December 2011	Amended	Sources of support amended.
6 July 2011	Amended	Search dates added to abstract.

Methods	Nested case‐cohort study Country: Canada
Participants	Name of parent cohort: Canadian Study of Diet, Lifestyle and Health (CSDLH) Participants: 22.975 participants (alumni associations of the University of Western Ontario, 67% of 34.291) Recruitment: between 1995 and 1998 Outcome assessment: December 2003 Number of cases: Prostate cancer: 661 Case definition: incidence Years of follow‐up: 4.3 to 7.7 years mean Type of selenium marker: supplementation
Interventions	d.n.a.
Outcomes	Statistical methods: Cox proportional hazard model Variables controlled in analysis: age at baseline, race, BMI, exercise activity, and education
Risk estimates [95% CI]	Reference category: zero Results: Prostate cancer highest quartile: HR 0.76 (95% CI 0.43 to 1.33)
Selenium levels in exposure categories	lowest quartile (median value): 15.7 µg highest quartile (median value): 105.0 µg
Notes

Methods	Cohort/sub‐cohort controlled cohort study Country: France
Participants	Name of parent cohort: Etude du Vieillissement Antériel Study (EVA study) Participants: 1389 participants (41% male, 59% female) Inclusion criteria: 59 to 71 years of age; residents of Nantes; able to undergo examination at study centre Recruitment: 1991 to 1993 Outcome assessment: December 2001 Number of cases: Any cancer: 45 (male/female: n.r.) Case definition: mortality Years of follow‐up: 9.0 years Type of selenium marker: plasma
Interventions	d.n.a.
Outcomes	Statistical methods: Cox proportional hazard model Variables controlled in analysis: gender, smoking, alcohol intake, medication use, obesity, diabetes mellitus, hypertension, CVD, age, education, dyslipidaemia, low cognitive function
Risk estimates [95% CI]	Reference category: highest quartile Results: Any cancer both genders: lowest quartile: RR 4.06 (95% CI 1.51 to 10.92)
Selenium levels in exposure categories	lowest quartile: 0.18 to 0.95 µmol/l highest quartile: 1.22 to 1.97 µmol/l
Notes

Methods	Randomised controlled trial Allocation: random Sequence generation: unclear Concealment: The study agent (two doses) and matched placebo caplets were coated with titanium oxide to ensure identical appearance, weight, taste, and smell. Blinding: only described as double‐blinded Dropouts/withdrawals: study dropouts percentage was 34.1%, 41.9%, and 40.8% for placebo, 200 mg/ day selenium group and 400 mg/day selenium group respectively (P=0.173). Intention‐to‐treat‐analysis: yes Recruitment period: not specified Treatment duration: not specified Observation period/dermatologic follow‐up: Subjects were followed every 6 months for up to 5 years Detection of cases: Tissue samples from the subject’s qualifying biopsy were requested from the subject’s physician and compiled in a biospecimen repository Informed consent: An external Data and Safety Monitoring Committee (DSMC) was established before study initiation.This committee was responsible for reviewing protocol amendments, consent forms, accrual and retention rates, adverse events, and data analysis reports
Participants	699 male participants with a negative prostate biopsy Country: US and NZ Number of patients: 699 (randomised to selenium 200 ug/day group: 234, to selenium 400 ug/day group: 233; to placebo group: 233) Condition: male patients at high risk for prostate cancer (prostate specific antigen (PSA) >4 ng/ml and/or suspicious digital rectal examination and/or PSA velocity >0.75 ng/ml/year), but with a negative prostate biopsy Demographics: mean age 65.2± SD 8 years (selenium 200ug/day), 65.5±7.7 years (selenium 400ug/day), 65.5±7.4 years (placebo); Recruitment and setting: from urology offices at 20 sites in the United States and New Zealand
Interventions	Intervention: 200 µg/day selenium supplied as selenium yeast 400 µg/day selenium supplied as selenium yeast Control: placebo Recruitment: not reported End of the blinded treatment period: For subjects in the US, participation was complete at 5 years, whereas subjects in New Zealand received intervention for no more than 3 years.
Outcomes	Primary outcome measure: the incidence of biopsyproven prostate cancer over the course of the study. Other reported outcomes: The secondary endpoint was the rate of change of PSA over time (i.e., PSA velocity) using biannual PSA measurements.
Risk estimates [95% CI]	Primary outcomes: The hazard ratios [95% confidence intervals] for risk of developing prostate cancer in the selenium 200 mg/day or the selenium 400 mg/day group were 0.94 [0.52, 1.7] and 0.90 [0.48, 1.7], respectively. Other reported outcomes: PSA velocity in the selenium arms was not significantly different from that observed in the placebo group (P= 0.18 and P = 0.17, respectively)
Selenium levels in exposure categories	d.n.a.
Notes	The DSMC recommended that the trial be stopped before all participants completed the full intervention duration Adverse effects: No significant differences were seen in the incidences of cataract/glaucoma or in hair/nail changes in the three treatment groups HR: adjusted for: age at baseline, baseline PSA, baseline selenium concentrations.

Methods	Matched, nested case‐control study Countries: Denmark, Germany, Greece, Italy, the Netherlands, Spain, Sweden, the UK
Participants	Participants: approximately 130,000 men Inclusion criteria: male participants of the EPIC study Name of parent cohort: European Prospective Investigation into Cancer and Nutrition (EPIC) Recruitment: 1992 to 2000 Outcome assessment: at each country's study closure date (between June 1999 and January 2003) Number of cases: Prostate cancer: 959 (male/female: 959/0) Case definition: incidence Years of follow‐up: median 2.6 years (Greece) to 9.2 years (Sweden) Type of selenium marker: plasma
Interventions	d.n.a.
Outcomes	Statistical methods: conditional logistic regression Variables controlled in analysis: BMI, smoking, alcohol consumption, physical activity, marital status, education Variables controlled by matching: age, study centre, time of day of blood collection, time between blood collection and last meal, sex
Risk estimates [95% CI]	Reference category: lowest quintile Results: Prostate cancer highest quintile: OR 0.96 (95% CI 0.70 to 1.31)
Selenium levels in exposure categories	lowest quintile < 62.0 µg/l highest quintile ≥ 84.1 µg/l
Notes

Methods	Cohort Study Country: UK
Participants	Participants: 1,054 men and women Inclusion criteria: people aged 65 years and over Name of parent cohort: British National Diet and Nutrition Survey Recruitment: 1994 to 1995 Outcome assessment: September 2008 Number of cases: Cancer deaths: 140 Case definition: mortality Type of selenium marker: plasma concentration
Interventions	d.n.a.
Outcomes	Statistical methods: Cox proportional hazard regression Variables controlled in analysis: age, sex, a1‐antichymotrypsin (an acute‐phase indicator), plasma creatinine (a renal status indicator), plasma total and HDL‐cholesterol concentrations and plasma albumin concentration.
Risk estimates [95% CI]	Cancer deaths HR 0.72; 95 % CI 0.58 to 0.89
Selenium levels in exposure categories	plasma concentrations
Notes

Methods	Cohort Study Country: US
Participants	Participants: 13,887 men and women Inclusion criteria: male and female adults, aged 20 to 90 years, participating in the NHANES III: "stratified, multistage probability cluster to provide data representing the noninstitutionalized US population" (Bleys 2008, p. 404) Name of parent cohort: Third National Health and Nutrition Examination Survey (NHANES III) Recruitment: 1988 to 1994 Outcome assessment: 15 December 2000 Number of cases: Cancer deaths: 457 (male/female: n.r.) Case definition: mortality Years of follow‐up: 6 to 12 years Type of selenium marker: serum
Interventions	d.n.a.
Outcomes	Statistical methods: Cox proportional hazard regression Variables controlled in analysis: age, sex, race, education, annual family income, post‐menopausal status (women), cigarette smoking, serum cotinine level, alcohol consumption
Risk estimates [95% CI]	Reference category: lowest tertile Results: Cancer deaths both genders: highest tertile: HR 0.69 (95% CI 0.53 to 0.90) both genders: highest tertile: HR 0.68 (95% CI 0.48 to 0.97); cases at baseline were excluded
Selenium levels in exposure categories	lowest tertile < 117.31 ng/ml highest tertile ≥ 130.39 ng/ml
Notes

Methods	Cohort/sub‐cohort‐controlled cohort study Country: US
Participants	Participants: 177 participants; no information on gender Inclusion criteria: persons at high risk of non‐melanoma skin cancer Recruitment: n.r. Outcome assessment: n.r. Number of cases: skin (non‐melanoma): 19 (male/female: n.r.) Case definition: incidence Years of follow‐up: mean: 3.0 years Type of selenium marker: plasma
Interventions	d.n.a.
Outcomes	Statistical methods: Cox proportional hazard model
Risk estimates [95% CI]	Reference category: lower half Results: Skin (non‐melanoma) gender n.r.: higher half: RR 0.77 (CI not reported)
Selenium levels in exposure categories	n.r.
Notes

Methods	Multicentre, randomised, placebo‐controlled, parallel‐group trial Allocation: random Sequence generation: unclear Concealment: unclear Blinding: only described as double‐blinded Dropouts/withdrawals: During the treatment phase, 38 in the selenium group and 37 in the placebo group withdrew from the study. This distribution was similar in both treatment groups. Intention‐to‐treat‐analysis: unclear Recruitment period: not specified Treatment duration: 3 years of treatment Observation period/dermatologic follow‐up: Subjects were followed for 2 years more after treatment Detection of cases: Patients were seen by a dermatologist before grafting; and any patients presenting with a non‐malignant or malignant skin keratosis or viral warts that had been present for less than 3 months were not selected. Within 10 weeks following the graft, a second visit was performed by a dermatologist to check that no new cutaneous lesion had appeared. Informed consent:The protocol and consent form had been approved by a National Ethics Committee prior to starting the study. Written informed consent was mandatory.
Participants	184 participants Number of patients: 184 (randomised to selenium 200 ug/day group: 91, to placebo group: 93) Condition: organ transplant recipient population Demographics: mean age 44.3± SD 13 years (selenium 200ug/day), 44.4± 10.7 years (placebo);
Interventions	Intervention: 200 µg/day selenium supplied as selenium yeast Control: placebo
Outcomes	Primary outcome measure: Occurrence rates of warts and various keratoses Other reported outcomes: skin cancers
Risk estimates [95% CI]	Primary outcome: events in selenium group=33 (36.3%), events in placebo group=31 (33.3%); odds‐ratio 1.09, P = 0.72 Secondary outcome: events in selenium group=6 (6.6%), events in placebo group=2 (2.2%); odds‐ratio 3.08, P = 0.15
Selenium levels in exposure categories
Notes

Methods	Matched, nested case‐control study (Epplein 2009, Gill 2009) Country: US
Participants	Inclusion criteria: participants of the Multiethnic Cohort, aged 45 to 75 years (native Hawaiians: aged 42 years and older), blood sample provided before cancer diagnosis between 1997 and 2006 Name of parent cohort: Multiethnic Cohort Recruitment: 1993 to 1996 Case definition: incidence Type of selenium marker: serum Epplein 2009: Participants: 67,594 (male: 29,009 / female: 38,585) men and women Outcome assessment: 2006 Number of cases: Lung cancer: 207 (male/female: 136/71) Years of follow‐up: 0 to 10 years Gill 2009: Participants: 29,009 men Outcome assessment: n.r. Number of cases: Prostate cancer: 467 (male/female: 467/0) Years of follow‐up: n.r.
Interventions	d.n.a.
Outcomes	Statistical methods: conditional logistic regression Epplein 2009: Variables controlled in analysis: age, fasting hours, pack‐years, pack‐years squared, years of schooling, family history of lung cancer Variables controlled by matching: age, sex, race/ethnicity, date of sample collection, time of day of sample collection, fasting status, smoking Gill 2009: Analysed cases: 450 of 467 cases analysed Variables controlled in analysis: age, fasting hours, BMI, family history of prostate cancer, education Variables controlled by matching: age, race/ethnicity, date of sample collection, geographic site (California, Hawaii), time of day of sample collection, fasting status
Risk estimates [95% CI]	Epplein 2009: Reference category: lowest tertile Results: Lung cancer male: highest tertile: OR 0.70 (95% CI 0.37 to 1.33) female: highest tertile: OR 0.98 (95% CI 0.42 to 2.29) Gill 2009: Reference category: lowest quartile Results: Prostate cancer highest quartile: OR 0.82 (95% CI 0.59 to 1.14)
Selenium levels in exposure categories	Epplein 2009: lowest tertile: median 0.12 µg/g of sodium highest tertile: median 0.15 µg/g of sodium Gill 2009: lowest quartile: median 0.12 µg/g highest quartile: median 0.16 µg/g
Notes	Primary publication: Epplein 2009 Other publications: Gill 2009

Methods	Prospective cohort study Country: northern part of Japan
Participants	Participants: 1,041 men and women Inclusion criteria: adult haemodialysis patients Name of parent cohort: ‘Kaleidoscopic Approaches to patients with end‐stage RENal disease Study’ (the KAREN Study) Recruitment: June 2003 to March 2004 Number of cases: malignant disease‐related death: 17 Case definition: mortality Years of follow‐up: 5‐year Type of selenium marker: serum
Interventions	d.n.a.
Outcomes	Statistical methods: Cox logistic regression Variables controlled by matching: age, male gender, BMI, hypertension, dyslipidaemia, diabetes mellitus, serum albumin levels, high‐sensitivity CRPlevels, history of myocardial infarction, history of stroke, history of malignant disease, smoking status and regular drinking habit
Risk estimates [95% CI]	Reference category: lowest quartile Results: malignant disease‐related death highest quartile: HR 2.98 (95% CI 0.62 to 14.35)
Selenium levels in exposure categories	lowest quartile: 18.4‐85.3 pg/L highest quartile: 114.2‐226.2 pg/L
Notes

Methods	Cohort study Country: Swedish
Participants	Participants: 2322 males Inclusion criteria: male residents in Uppsala county in January 1970, born in 1920‐24 Name of parent cohort: Uppsala Longitudinal Study of Adult Men (ULSAM). Recruitment: 1991 to 1995 Outcome assessment: 31/12/2003 Number of cases: Prostate cancer: 208 Case definition: incidence Years of follow‐up: 26.5‐years (median) Type of selenium marker: serum
Interventions	d.n.a.
Outcomes	Statistical methods: proportional hazard model
Risk estimates [95% CI]	Reference category: lowest level Results: Prostate cancer: highest level: RR 0.83 (95% CI 0.60 to 1.16)
Selenium levels in exposure categories	lowest level: <=70 µg/l highest level: >81 µg/l
Notes

Methods	Matched, nested case‐control study (Knekt 1990, Hakama 1990, Knekt 1988, Knekt 1996) Cohort study (Knekt 1991) Country: Finland
Participants	Inclusion criteria: no history of cancer at baseline Name of parent cohort: Social Insurance Institution's Mobile Clinic Health Examination Survey Recruitment: 1968 to 1972 Knekt 1990: Participants: 39,268: 21,172 men and 18,096 women Outcome assessment: 31 December 1980 Number of cases: Any cancer: 1096 (male/female: 597/499) Stomach cancer: 95 (male/female: 58/37) Colon and rectal cancer: 91 (male/female: 32/59) Lung cancer: 198 (male/female: 189/9) Prostate cancer: 51 (male/female: 51/0) Urinary tract cancer: 47 (male/female: 34/13) Pancreatic cancer: 45 (male/female: 22/23) Breast cancer: 90 (male/female: 0/90) Gynaecological cancer (without breast): 86 (male/female: 0/86) Basal cell carcinoma (skin): 126 (male/female: 64/62) Other: 267 (male/female: 147/120) Hakama 1990: Participants: number of participants n.r.; both genders Inclusion criteria: aged 15 years and older Outcome assessment: 1977 Number of cases: Any cancer: 766 (male/female: n.r.) Lung cancer: 151 (male/female: 151/0) Breast cancer: 67 (male/female: 0/67) Stomach cancer: 76 (male/female: n.r.) Prostate cancer: 37 (male/female: 37/0) Knekt 1988: Participants: 36,265: 21,172 men and 15,093 women Outcome assessment: 31 December 1977 Number of cases: Oesophageal and stomach cancer: 86 (male/female: 51/35) Colon and rectal cancer: 57 (male/female: 21/36) Knekt 1991: Participants: 4538 men Inclusion criteria: aged 20 to 69 years, with dietary history taken Outcome assessment: 1986 Number of cases: Lung cancer: 117 (male/female: 117/0) Knekt 1996: Participants: 1896 women Outcome assessment: 1980 Number of cases: Ovarian cancer: 24 (male/female: 0/24) Case definition: incidence Years of follow‐up: 9 to 20 years Type of selenium marker: serum (Knekt 1990, Hakama 1990, Knekt 1988, Knekt 1996), intake (Knekt 1991: dietary history)
Interventions	d.n.a.
Outcomes	Knekt 1990: Statistical methods: conditional logistic regression Variables controlled in analysis: smoking Variables additionally controlled in analysis of highest four quintiles versus lowest quintile: occupation, BMI, parity, cholesterol, haematocrit Variables controlled by matching: age, gender, municipality, time of baseline examination, duration of storage of sample Hakama 1990: Analysed cases: 766 of 864 cases analysed (reason for non‐inclusion: no serum sample) Statistical methods: conditional logistic regression Variables controlled in analysis: smoking Variables additionally controlled in analysis of highest four quintiles versus lowest quintile: retinol level, alpha‐tocopherol level Variables controlled by matching: age, gender, municipality, time of baseline examination, duration of storage of sample Knekt 1988: Statistical methods: n.r. Variables controlled in analysis: smoking, serum cholesterol Variables controlled by matching: age, gender, municipality, time of baseline examination, duration of storage of sample Knekt 1991: Statistical methods: Cox‐proportional hazards model Variables controlled in analysis: age, smoking (data stratified according to smoking status) Knekt 1996: Statistical methods: conditional logistic regression Variables controlled by matching: age, gender, municipality, time of baseline examination, duration of storage of sample
Risk estimates [95% CI]	Knekt 1990: Reference category: lowest quintile Results: Any cancer male: highest quintile: OR 0.41 (CI not reported) above 20th percentile: OR 0.67 (CI not reported); cases during first 2 years of follow‐up excluded: 476 cases: OR 0.65 (95% CI 0.48 to 0.89) female: highest quintile: OR 0.86 (CI not reported) above 20th percentile: OR 0.93 (CI not reported); cases during first 2 years of follow‐up excluded: 423 cases: OR 0.97 (95% CI 0.68 to 1.39) Stomach cancer male: highest quintile: OR 0.09 (CI not reported) above 20th percentile: OR 0.26 (CI not reported); cases during first 2 years of follow‐up excluded: 43 cases: OR 0.24 (95% CI 0.09 to 0.69) female: highest quintile: OR 0.27 (CI not reported) above 20th percentile: OR 0.59 (CI not reported); cases during first 2 years of follow‐up excluded: 30 cases: OR 0.48 (95% CI 0.14 to 1.66) Colon and rectal cancer male: highest quintile: OR 0.53 (CI not reported) above 20th percentile: OR 0.69 (CI not reported); cases during first 2 years of follow‐up excluded: 29 cases: OR 1.01 (95% CI 0.18 to 5.65) female: highest quintile: OR 0.80 (CI not reported) above 20th percentile: OR 1.26 (CI not reported); cases during first 2 years of follow‐up excluded: 48 cases: OR 1.10 (95% CI 0.42 to 2.92) Lung cancer male: highest quintile: OR 0.30 (CI not reported) above 20th percentile: OR 0.60 (CI not reported); cases during first 2 years of follow‐up excluded: 153 cases: OR 0.66 (95% CI 0.37 to 1.19) female: third highest quintile: OR 4.62 (CI not reported) (quintile 4 and 5 did not contain any cases) Prostate cancer highest quintile: OR 1.15 (CI not reported) above 20th percentile: OR 1.13 (CI not reported); cases during first 2 years of follow‐up excluded: 46 cases: OR 1.00 (95% CI 0.42 to 2.40) Urinary tract cancer male: highest quintile: OR 0.81 (CI not reported) above 20th percentile: OR 0.89 (CI not reported); cases during first 2 years of follow‐up excluded: 26 cases: OR 0.34 (95% CI 0.06 to 2.06) female: highest quintile: OR 4.12 (CI not reported) above 20th percentile: not reported; cases during first 2 years of follow‐up excluded: 9 cases: OR 2.51 (95% CI 0.13 to 47.9) Pancreatic cancer male: fourth quintile versus lowest: OR 0.58 (CI not reported) (highest quintile did not contain any cases) above 20th percentile: OR 0.11 (CI not reported); cases during first 2 years of follow‐up excluded: not reported female: highest quintile: OR 3.49 (CI not reported) above 20th percentile: not reported; cases during first 2 years of follow‐up excluded: 22 cases: OR 0.86 (95% CI 0.21 to 3.52) Breast cancer highest quintile: OR 0.64 (CI not reported) above 20th percentile: OR 0.52 (CI not reported); cases during first 2 years of follow‐up excluded: 74 cases: OR 0.57 (95% CI 0.18 to 1.81) Gynaecological cancer (without breast) highest quintile: OR 0.96 (CI not reported) above 20th percentile: OR 0.91 (CI not reported); cases during first 2 years of follow‐up excluded: 70 cases: OR 1.03 (95% CI 0.43 to 2.50) Basal cell carcinoma (skin) male: highest quintile: OR 0.54 (CI not reported) above 20th percentile: OR 0.65 (CI not reported); cases during first 2 years of follow‐up excluded: 54 cases: OR 0.86 (95% CI 0.35 to 2.12) female: highest quintile: OR 1.55 (CI not reported) above 20th percentile: OR 1.73 (CI not reported); cases during first 2 years of follow‐up excluded: 52 cases: OR 1.54 (95% CI 0.64 to 3.73) Other or unspecified cancer: male: highest quintile: OR 0.42 (CI not reported) above 20th percentile: OR 0.72 (CI not reported); cases during first 2 years of follow‐up excluded: 110 cases: OR 0.70 (95% CI 0.36 to 1.36) female: highest quintile: OR 0.71 (CI not reported) above 20th percentile: OR 0.87 (CI not reported); cases during first 2 years of follow‐up excluded: 111 cases: OR 0.92 (95% CI 0.44 to 1.92) Hakama 1990: Reference category: highest quintile Results: Any cancer male: lowest quintile: OR 2.40 (CI not reported) lowest quintile vs. four highest quintiles: OR 1.60 (CI not reported) female: lowest quintile: OR 1.20 (CI not reported) lowest quintile vs. four highest quintiles:0.90 (CI not reported) Lung cancer male: lowest quintile vs. four highest quintiles: OR 1.80 (CI not reported) Breast cancer lowest quintile vs. four highest quintiles: OR 3.10 (CI not reported) Stomach cancer male: lowest quintile vs. four highest quintiles: OR 6.70 (CI not reported) female: lowest quintile vs. four highest quintiles: OR 2.00 (CI not reported) Prostate cancer lowest quintile vs. four highest quintiles: OR 0.80 (CI not reported) Knekt 1988: Reference category: highest quintile Results: Oesophageal and stomach cancer male: lowest tertile: OR 2.20 (CI not reported) lowest quintile vs. four highest quintiles: OR 3.3 (95% CI 1.3 to 9.1) female: lowest tertile: OR 1.50 (CI not reported) lowest quintile vs. four highest quintiles: OR 2.4 (95% CI 0.7 to 8.3) Colon and rectal cancer male: lowest tertile: OR 0.90 (CI not reported) lowest quintile vs. four highest quintiles: OR 1.7 (95% CI 0.4 to 7.7) female: lowest tertile: OR 0.60 (CI not reported) lowest quintile vs. four highest quintiles: OR 0.8 (95% CI 0.2 to 2.4) Knekt 1991: Reference category: highest tertile Results: Lung cancer male non‐smokers: lowest tertile: OR 1.03 (CI not reported) male smokers: lowest tertile: OR 0.83 (CI not reported) Knekt 1996: Reference category: highest tertile Results: Ovarian cancer lowest tertile: OR 1.15 (95% CI 0.19 to 4.06)
Selenium levels in exposure categories	Knekt 1990: lowest quintile: ≤ 48.90 µg/l; highest quintile ≥ 78.00 µg/l Hakama 1990: quintiles: not specified Knekt 1988: both genders: lowest tertile: ≤ 56.90 µg/l; highest tertile ≥ 70.10 µg/l lowest quintile: ≤ 50 µg/l; highest four quintiles > 50 µg/l Knekt 1991: tertiles: n.r. Knekt 1996: lowest tertile: ≤ 56.90 µg/l; highest tertile: ≥ 68.10 µg/l
Notes	Primary publication: Knekt 1990 Other publications: Hakama 1990, Knekt 1988, Knekt 1991, Knekt 1996

Methods	Matched, nested case‐control study (McNaughton 2005b) Cohort study (Heinen 2007, van der Pols 2009) Country: Australia
Participants	Name of parent cohort: Nambour Skin Cancer Study Recruitment: 1992 to 1996 Case definition: incidence McNaughton 2005b: Participants: approximately 1000 men and women Inclusion criteria: randomly selected adults, aged 20 to 69 years; recruited for participation in a randomised controlled trial for skin cancer prevention with beta‐carotene supplements and sunscreen application in 1992; living in the Nambour community; free of SCC at baseline; with blood sample and FFQ provided in 1996; participants with extreme energy intakes in FFQ excluded Outcome assessment: December 2001 Number of cases: Basal cell carcinoma of the skin: 90 (male/female: 39/51) Years of follow‐up: 5.5 years Type of selenium marker: serum and nutritional intake (FFQ) Heinen 2007: Participants: 1001 men and women Inclusion criteria: randomly selected adults, aged 20 to 69 years; recruited for participation in randomised controlled trial for skin cancer prevention with beta‐carotene supplements and sunscreen application in 1992; living in the Nambour community; with blood sample and FFQ provided in 1996; participants with extreme energy intakes in FFQ and missing consumption frequencies for more than 10% of food items excluded Outcome assessment: 31 December 2004 Number of cases: Basal cell carcinoma of the skin: 149 (male/female: 87/62) participants with 321 BCC tumours Squamous cell carcinoma of the skin: 116 (male/female: 70/46) participants with 221 SCC tumours, Case definition: incidence (tumour‐based incidence and person‐based incidence) Years of follow‐up: 8 years Type of selenium marker: nutritional intake (FFQ) van der Pols 2009: Participants: 485 (male/female: 223/262) men and women Inclusion criteria: randomly selected adults, aged 20 to 69 years; recruited for participation in randomised controlled trial for skin cancer prevention with beta‐carotene supplements and sunscreen application in 1992; randomised to placebo in the intervention trial; living in the Nambour community; free of SCC at baseline; with blood sample and FFQ provided in 1996; participants with extreme energy intakes in FFQ excluded Outcome assessment: 31 December 2004 Number of cases: Basal cell carcinoma of the skin: 77 (male/female: 46/31) participants with 173 BCC tumours Squamous cell carcinoma of the skin: 59 (male/female: 38/21) participants with 124 SCC tumours, Years of follow‐up: 8 years Type of selenium marker: serum
Interventions	d.n.a.
Outcomes	McNaughton 2005b: Statistical methods: conditional logistic regression Variables controlled in analysis: age, gender Variables controlled by matching: age, gender Heinen 2007: Statistical methods: generalised linear models Variables controlled in analysis: age, sex, intervention arm in RCT, energy intake, skin colour, elastosis of the neck, smoking, use of dietary supplements, history of skin cancer van der Pols 2009: Statistical methods: generalised linear models Variables controlled in analysis: age, sex, pack‐years of smoking, alcohol intake, time spent outdoors on weekdays, history of skin cancer before 1996
Risk estimates [95% CI]	McNaughton 2005b: Reference category: lowest quartile Results: Basal cell carcinoma (skin) both genders: highest quartile: OR 0.86 (95% CI 0.38 to 1.96) biochemical selenium level both genders: highest quartile: OR 1.13 (95% CI 0.47 to 2.74) selenium intake Heinen 2007: Reference category: lowest tertile Results: Basal cell carcinoma (skin) both genders: highest tertile: RR 0.95 (95% CI 0.59 to 1.50) Squamous cell carcinoma (skin) both genders: highest tertile: RR 1.3 (95% CI 0.77 to 2.3) van der Pols 2009: Reference category: lowest exposure category Results: Basal cell carcinoma (skin) both genders: highest exposure category: RR 0.58 (95% CI 0.32 to 1.07) Squamous cell carcinoma (skin) both genders: highest exposure category: RR 0.49 (95% CI 0.24 to 0.99)
Selenium levels in exposure categories	McNaughton 2005b: n.r. Heinen 2007: lowest tertile ≤ 76.20 µg/day highest tertile ≥ 89.31 µg/day van der Pols 2009: lowest exposure category ≤ 1.0 µmol/l highest exposure category ≥ 1.3 µmol/l
Notes	Primary publication:McNaughton 2005b Other publication: Heinen 2007, van der Pols 2009 tumour‐based incidence: number of newly developed histologically confirmed BCC or SCC divided by the person‐years of follow‐up accumulated over follow‐up period person‐based incidence: number of persons newly affected by BCC or SCC during the same person‐years of follow‐up time as calculated for the tumour‐based analysis

PERMALINK

Selenium for preventing cancer

Marco Vinceti

Gabriele Dennert

Catherine M Crespi

Marcel Zwahlen

Maree Brinkman

Maurice PA Zeegers

Markus Horneber

Roberto D'Amico

Cinzia Del Giovane

Abstract

Background

Objectives

Search methods

Selection criteria

Data collection and analysis

Main results

Authors' conclusions

Selenium for preventing cancer

Background

Description of the condition

Description of the intervention

How the intervention might work

Why it is important to do this review

Objectives

Methods

Criteria for considering studies for this review

Types of studies

Types of participants

Types of interventions

Types of outcome measures

Primary outcomes

Secondary outcomes

Search methods for identification of studies

Data collection and analysis

Selection of studies

Data extraction and management

Assessment of risk of bias in included studies

Observational studies

Randomised controlled trials

Measures of treatment effect

Dealing with missing data

Assessment of heterogeneity

Assessment of reporting biases

Data synthesis

Observational studies

Randomised controlled trials

Subgroup analysis and investigation of heterogeneity

Sensitivity analysis

Results

Description of studies

Results of the search

Figure 1.

Included studies

1. Observational studies

Table 1.

2. Randomised controlled trials

Excluded studies

Risk of bias in included studies

Observational studies

Table 2.

Figure 2.

Figure 3.

Figure 4.

Figure 5.

Randomised controlled trials

Table 3.

Ethical criteria

Effects of interventions

1. Observational studies

1.1. Aetiological association: results from meta‐analyses

1.1.1. Any cancer

Analysis 1.1.

Analysis 1.2.

Analysis 1.3.

Analysis 1.4.

1.1.2. Female breast cancer

Analysis 1.5.

1.1.3. Bladder cancer

Methods	Unmatched, nested case‐control study Country: US
Participants	Participants: 6860 men Inclusion criteria: born 1900 to 1919; Japanese ancestry; inhabitants of Oahu/Hawaii; participants in the Honolulu Heart Program (1965 to 68) Name of parent cohort: Honolulu Heart Program Recruitment: 1971 to 1975 Outcome assessment: n.r. Number of cases: Any cancer: 280 (male/female: 280/0) Stomach cancer: 66 (male/female: 66/0) Rectal cancer: 32 (male/female: 32/0) Lung cancer: 71 (male/female: 71/0) Colon cancer: 82 (male/female: 82/0) Bladder cancer: 29 (male/female: 29/0) Case definition: incidence Years of follow‐up: 11.0 years Type of selenium marker: serum
Interventions	d.n.a.
Outcomes	Statistical methods: proportional hazards regression/Cox regression Variables controlled in analysis: age at examination, cigarettes/day (any cancer, lung cancer, bladder cancer) age at examination (stomach, rectum, colon)
Risk estimates [95% CI]	Reference category: highest quintile Results: Stomach cancer: male: lowest quintile: OR 0.9 (CI not reported) Rectal cancer male: lowest quintile: OR 1.6 (CI not reported) Lung cancer male: lowest quintile: OR 1.1 (CI not reported) Colon cancer male: lowest quintile: OR 1.8 (CI not reported) Bladder cancer male: lowest quintile: OR 3.1 (CI not reported) All five types of cancer male: lowest quintile: OR 1.3 (CI not reported)
Selenium levels in exposure categories	lowest quintile: ≤ 10.30 µg/dl highest quintile: ≥ 13.31 µg/dl
Notes	N.B.: "Any cancer" in this study comprises all cancer cases for stomach, rectal, lung, colon and bladder cancer.

Methods	Randomised controlled trial Nutritional Prevention of Cancer Trial (NPCT) Allocation: random, block/stratified by clinic Sequence generation: computer generated random numbers Concealment: central assignment (sealed pill bottles) Blinding: participant blinded, doctor blinded, outcome assessor/pathologist unclear, review/coding of medical records blinded Dropouts/withdrawals: “9 patients (5 in the selenium group and 4 in the placebo group) declined to provide additional illness information” (Clark 1996, p. 1959) ‐ 0 participants lost to vital follow‐up Intention‐to‐treat‐analysis: yes Recruitment period: 1983 to 1991 End of predefined study period: 31 December 1993 Blinded intervention continued until the end of the blinded period: 31 January 1996 Intervention duration: 31 December 1993 (end of study period): mean = 4.5 years 31 January 1996 (end of blinded period): mean = 7.9 years Observation period/dermatologic follow‐up: 31 December 1993 (end of study period): mean = 6.4 years 31 January 1996 (end of blinded period): mean = 7.4 years Detection of cases: dermatologic examination and interview every 6 months during follow‐up; incident BCC and SCC were diagnosed by biopsy and confirmed by another dermatopathologist Informed consent: written informed consent forms, approval by institutional review board of participating institutions
Participants	Country: US Number of participants: 1312 (randomised to selenium group: 653, to placebo group: 659) Condition: male and female participants with history of 2 or more squamous cell or basal cell skin cancers Demographics: mean age 63.4 years (selenium)/63.0 years (placebo); 73.8% men (selenium). 75.6% men (placebo) Recruitment and setting: seven dermatological clinics (three academic units, four private practices) in the US
Interventions	Intervention: 200 µg selenium supplied as 500 mg selenium yeast tablets p.o./daily. Control: placebo
Outcomes	Primary outcome measure: incidence of basal and squamous cell carcinoma of the skin: all analyses were based on 1250 participants with initial blood collection within four days after randomisation (621 in the selenium group and 629 in the placebo group) Other reported outcomes and secondary outcome measures: Reported in Clark 1996: Incidence of lung cancer, prostate cancer, colorectal cancer, any cancer, head and neck cancer, bladder cancer, oesophageal cancer, breast cancer, melanoma, haematologic cancer, Reported in Duffield‐Lillico 2002: Overall cancer mortality
Risk estimates [95% CI]	Primary outcomes: 1) at the end of study period (31 December 1993) (Clark 1996): BCC: RR 1.10 (95% CI 0.95 to 1.28); cases: selenium group: 377, placebo group: 350; incidence per person‐year under follow‐up: selenium group 0.16, placebo group 0.15 SCC: RR 1.14 (95% CI 0.93 to 1.39); cases: selenium group 218, placebo group: 190; incidence per person‐year under follow‐up: selenium group 0.07, placebo group 0.06 2) at the end of blinded period (31 January 1996) (Duffield‐Lillico 2003): BCC: RR 1.17 (95% CI 1.02 to 1.35), HR 1.09 (95% CI 0.94 to 1.26); number of cases not reported; incidence per person‐year under follow‐up: selenium group: 0.16, placebo group 0.13 SCC: RR 1.32 (95% CI 1.09 to 1.60), HR 1.25 (95% CI 1.03 to 1.51); number of cases not reported; incidence per person‐year under follow‐up: selenium group: 0.05, placebo group 0.07 NMSC: RR 1.27 (95% CI 1.11 to 1.45) HR 1.17 (95% CI 1.02 to 1.34); number of cases not reported; incidence per person‐year under follow‐up: selenium group: 0.20, placebo group 0.16 Other reported outcomes and secondary outcomes: 1) at the end of study period (31 December 1993) (Clark 1996): lung cancer RR 0.54 (95% CI 0.30 to 0.98), adjusted HR 0.56 (95% CI 0.31 to 1.01) cases selenium: 17, placebo: 31 prostate cancer RR 0.37 (95% CI 0.18 to 0.71), adjusted HR 0.35 (95% CI 0.18 to 0.65) cases selenium: 13, placebo: 35 colorectal cancer RR 0.42 (95% CI 0.18 to 0.95), adjusted HR 0.39 (95% CI 0.17 to 0.90) cases selenium: 8, placebo: 19 any cancer RR 0.63 (95% CI 0.47 to 0.85), adjusted HR 0.61 (95% CI 0.46 to 0.82) cases selenium: 77, placebo: 119 head and neck cancer RR 0.74 (95% CI 0.21 to 2.43), adjusted HR 0.77 (95% CI 0.27 to 2.24) cases selenium: 6, placebo: 8 bladder cancer RR 1.32 (95% CI 0.40 to 4.61), adjusted HR 1.27 (95% CI 0.44 to 3.67) cases selenium: 8, placebo: 6 oesophageal cancer RR 0.33 (95% CI 0.03 to 1.84), adjusted HR 0.30 (95% CI 0.06 to 1.49) cases selenium: 2, placebo: 6 breast cancer RR 2.88 (95% CI 0.72 to 16.5), adjusted HR 2.95 (95% CI 0.80 to 10.9) cases selenium: 9, placebo:3 melanoma RR 0.97 (95% CI 0.32 to 2.96), adjusted HR 0.92 (95% CI 0.34 to 2.45) cases selenium: 8, placebo: 8 haematological cancer RR 1.58 (95% CI 0.46 to 6.14), adjusted HR 1.50 (95% CI 0.49 to 4.60) cases selenium: 8, placebo: 5 other specific carcinomas RR 0.55 (95% CI 0.14 to 1.82), adjusted HR 0.54 (95% CI 0.18 to 1.62), cases selenium: 5, placebo: 9 total carcinoma RR 0.55 (95% CI 0.40 to 0.77), adjusted HR 0.54 (95% CI 0.39 to 0.75), cases selenium: 59; placebo: 104 leukaemia /lymphomas RR 1.58 (95% CI 0.46 to 6.14), adjusted HR 1.50 (95% CI 0.49 to 4.60), cases selenium: 8, placebo 5 other specific non‐carcinomas RR 0.99 (95% CI 0.13 to 7.37), HR 0.99 (95% CI 0.20 to 4.94), cases selenium: 3, placebo: 3 total non‐carcinomas RR 1.17 (95% CI 0.57 to 2.44), adjusted HR 1.16 (95% CI 0.60 to 2.27), cases selenium: 19; placebo: 16 2) at the end of the blinded period (31 January 1996) (Duffield‐Lillico 2002): lung cancer RR 0.70 (95% CI 0.40 to 1.21), adjusted HR 0.74 (95% CI 0.44 to 1.24), cases selenium: 25, placebo: 35 prostate cancer RR 0.51 (95% CI 0.29 to 0.87), adjusted HR 0.48 (95% CI 0.28 to 0.80), cases selenium: 22, placebo: 42 colorectal cancer RR 0.46 (95% CI 0.19 to 1.08), adjusted HR 0.46 (95% CI 0.21 to 1.02), cases selenium: 9, placebo: 19 any cancer RR 0.75 (95% CI 0.58 to 0.98), adjusted HR 0.75 (95% CI 0.58 to 0.97), cases selenium: 105, placebo: 137 head and neck cancer RR 1.27 (95% CI 0.42 to 4.01), adjusted HR 1.27 (95% CI 0.47 to 3.42), cases selenium: 9, placebo: 7 bladder cancer RR 1.24 (95% CI 0.44 to 3.61), adjusted HR 1.28 (95% CI 0.50 to 3.25), cases selenium: 10, placebo: 8 oesophageal cancer RR 0.39 (95% CI 0.04 to 2.41), adjusted HR 0.40 (95% CI 0.08 to 2.07), cases selenium: 2, placebo: 5 breast cancer RR 1.82 (95% CI 0.62 to 6.01), adjusted HR 1.89 (95% CI 0.69 to 5.14), cases selenium: 11, placebo: 6 melanoma RR 1.21 (95% CI 0.46 to 3.30), adjusted HR 1.18 (95% CI 0.49 to 2.85), cases selenium: 11, placebo: 9 haematological cancer (lymphoma and leukaemia) RR 1.32 (95% CI 0.40 to 4.61), adjusted HR 1.25 (95% CI 0.43 to 3.61), cases selenium: 8, placebo: 6 cancer mortality, all sites RR 0.59 (95% CI 0.39 to 0.89), adjusted HR 0.59 (95% CI 0.39 to 0.87), cases selenium: 40, placebo: 66 other carcinomas RR 0.66 (95% CI 0.19 to 2.07), adjusted HR 0.67 (95% CI 0.24 to 1.88), cases selenium: 6, placebo:9 other non‐carcinomas RR 0.59 (95% CI 0.09 to 3.04), adjusted HR 0.59 (95% CI 0.14 to 2.47), cases selenium: 3, placebo: 5
Selenium levels in exposure categories	d.n.a.
Notes	Adverse effects: Clark 1996: 35 participants (21 in selenium and 14 in control group) complained of adverse effects, mostly involving gastrointestinal upset, and withdrew treatment. Post‐hoc introduced secondary outcomes were: all‐cause mortality, total cancer mortality, total cancer incidence and incidence of lung / prostate / colorectal cancers HR: adjusted for sex, age, smoking status, clinic site, plasma selenium concentration, clinical sun damage, sunscreen use at baseline and number of BCCs/SCCs/NMSCs in the 12 months before randomisation

Methods	Randomised controlled trial Sub‐study of the Nutritional Prevention of Cancer Trial (NPCT 2002) Allocation: random Sequence generation: computer generated random numbers Concealment: central assignment (sealed pill bottles) Blinding: participant blinded, doctor blinded, outcome assessor/pathologist unclear, review/coding of medical records blinded Dropouts/withdrawals: two participants declined to provide additional illness information, no participant lost to vital follow ‐up Intention‐to‐treat‐analysis: yes Recruitment period: 1989‐1992 Treatment duration: Blinded intervention continued until the end of the blinded period; 1 February 1996. Observation period/dermatologic follow‐up: 1 February 1996 Detection of cases: dermatological examination and interview every 6 months during follow‐up; incident BCC and SCC were diagnosed by biopsy and confirmed by another dermatopathologist Informed consent: written informed consent forms, approval by institutional review board of participating institutions
Participants	423 male and female participants with prior non‐melanoma skin cancer Country: US Number of patients: 423 (randomised to selenium group: 210, to placebo group: 213) Condition: male and female patients with history of 2 or more squamous cell or basal cell skin cancers Demographics: mean age 63.8 years (selenium)/63.8 years (placebo); 66.2% men (selenium). 68.2% men (placebo) Recruitment and setting: dermatologic clinic in Macon, Georgia
Interventions	Intervention: 400 µg selenium supplied as selenium yeast tablets p.o./daily. Control: placebo 400 µg/day of selenium yeast or identical‐appearing low selenium yeast placebo Recruitment: 12 September 1989 to 3 April 1992 End of the blinded treatment period: 2 February 1996
Outcomes	Primary outcome measure: incidence of basal and squamous cell carcinoma of the skin: all analyses were based on n = 423 participants with initial blood collection within 4 days after randomizations Other reported outcomes: total internal cancer incidence
Risk estimates [95% CI]	Primary outcomes: BCC: RR 0.90 (95% CI 0.65 to 1.24); cases: selenium group: 76, placebo group: 83; adjusted HR: 0.95 (95% CI 0.69 to 1.29) SCC: RR 1.05 (95% CI 0.71 to 1.56); cases: selenium group: 56, placebo group: 53; adjusted HR: 1.05 (95% CI 0.72 to 1.53) NMSC: RR 0.88 (95% CI 0.66 to 1.16); cases: selenium group: 98, placebo group: 108; adjusted HR: 0.91 (95% CI 0.69 to 1.20) NMSC in women: RR 0.40 (95% CI 0.20 to 0.80) Other reported outcomes: total internal cancer incidence: RR 1.10 (95% CI 0.57 to 2.17); cases: selenium group: 21, placebo group: 19
Selenium levels in exposure categories	d.n.a.
Notes	Information on study design, which was not reported in Reid 2008, was taken from the information available on the Nutritional Prevention of Cancer Trial. Adverse effects: not reported HR: adjusted for: age (continuous), smoking status (never, former, current), gender

Methods	Randomised controlled trial SELECT (Selenium and Vitamin E Cancer Prevention Trial) Allocation: random, block/stratified by clinic Sequence generation: computer‐generated random numbers Concealment: central assignment (pill bottles) Blinding: participant blinded, doctor blinded, outcome assessor/pathologist blinded, review/coding of medical records blinded Dropouts/withdrawals: of 35,533 randomised participants, 645 were excluded from analysis because they had prior prostate cancer, did not give informed consent or participated at two study sites, which were excluded due to management and regulatory issues Intention‐to‐treat‐analysis: yes Recruitment period: 22 August 2001 to 24 June 2004 End of study period: 1 August 2009 Blinded intervention was discontinued on 23 October 2008 following the recommendation of the data safety and monitoring committee after the second formal interim analysis in September 2008 Detection of cases: Participants had clinic visits once every 6 months and reported prostate cancers to the study staff. Study staff obtained medical records to verify the diagnosis. Tissue and the corresponding pathology report were sent to the central pathology laboratory for confirmation. Informed consent: yes
Participants	Countries: US, Canada, Puerto Rico Number of participants: 34,888 men, randomised to four groups: placebo (8696), vitamin E (8737), selenium (8752), selenium + vitamin E (8703) Condition: healthy men, aged 50 years or older (African American) or 55 years or older (all other), no prior diagnosis of prostate cancer, 4 ng/ml or less of PSA in serum, a digital rectal examination not suspicious for cancer, no current use of anticoagulant therapy other than 175 mg/day or less of acetylsalicylic acid or 81 mg/day or less of acetylsalicylic acid with clopidogrel bisulphate, no history of haemorrhagic stroke, normal blood pressure. Demographics: median age: 62.3‐62.6 years in all four intervention groups, 79% white in all four intervention groups Recruitment and setting: 427 participating sites
Interventions	Group 1: placebo + placebo Group 2: 400 IU/day all rac‐alpha‐tocopheryl acetate + placebo Group 3: 200 µg/day L‐selenomethionine + placebo Group 4: 400 IU/day all rac‐alpha‐tocopheryl acetate + 200 µg/day L‐selenomethionine
Outcomes	Primary outcome: incidence of prostate cancer as determined by routine clinical management Secondary outcomes: incidence of any cancer / lung cancer / colorectal cancer, diabetes mellitus, cardiovascular events, death from any cause
Risk estimates [95% CI]	Results are presented for the comparison of selenium alone (group 3) versus placebo (group 1) Primary outcome: prostate cancer HR 1.04, (95% CI 0.90 to 1.18), (99% CI 0.87 to 1.24), cases: selenium 432 (5‐year rate: 4.56%), placebo 416 (5‐year rate 4.43%) Secondary outcomes: any cancer HR 1.01, (95% CI 0.89 to 1.15) lung cancer HR 1.12, (99% CI 0.73 to 1.72) colorectal cancer 1.05, (99% CI 0.66 to 1.67) other primary cancer (excluding prostate cancer, basal cell and squamous cell skin cancer) 0.95, (99% CI 0.77 to 1.17) diabetes mellitus 1.07, (99% CI 0.94 to 1.22) cardiovascular events 1.02, (99% CI 0.92 to 1.13) deaths 0.99, (99% CI 0.82 to 1.19) deaths from cancer 1.02, (99% CI 0.74 to 1.41)
Selenium levels in exposure categories	d.n.a.
Notes	Adverse effects: alopecia RR 1.28, (99% CI 1.01 to 1.62) dermatitis grade 1‐2 RR 1.17, (99% CI 1.00 to 1.35) dermatitis grade 3‐4 RR 1.74, (99% CI 0.56 to 5.44) halitosis RR 1.17, (99% CI 0.99 to 1.38) nail changes RR 1.04, (99% CI 0.94 to 1.16) fatigue grade 1‐2 RR 1.09, (99% CI 0.95 to 1.26) fatigue grade 3‐4 RR 0.87, (99% CI 0.40 to 1.88) nausea grade 1‐2 RR 1.19, (99% CI 0.94 to 1.52) nausea grade 3 RR 0.99, (99% CI 0.30 to 3.34)

Methods	Nested case‐control study Country: Germany
Participants	Participants: 11928 men (from the total cohort of 25540 men and women) Name of parent cohort: EPIC‐Heidelberg cohort Recruitment: 1994‐1998. Outcome assessment: 2/2007 Number of cases: prostate cancer: 248 Case definition: incidence Years of follow‐up: mean: 3 years Type of selenium marker: serum selenium concentration
Interventions	d.n.a.
Outcomes	Statistical methods: conditional logistic regression Variables controlled in analysis: family history of prostate cancer, participation in PSA testing, smoking status, and vigorous physical activity. variables controlled in matching: age group and time of recruitment
Risk estimates [95% CI]	Prostate cancer Reference category: lowest quartile highest quartile OR 1.10 (95% CI, 0.58‐2.09)
Selenium levels in exposure categories	lowest quartile: ≤ 78.9 µg/L highest quartile: ≥ 95 µg/L
Notes

Methods	Cohort Study Country: Denmark
Participants	Participants: 3,333 males, male participants were derived from 14 workplaces in Copenhagen: the air force, army, navy, emergency management agency, postal service, customs service, a railroad company, national bank, a telephone company, three municipal service centres (for electricity and engineering and a fire brigade), a pharmaceutical company, and a building contractor company. Name of parent cohort: Copenhagen male study Recruitment: from 1970‐1971/1985‐1986 Outcome assessment: 1985‐1986/2001 Number of cases: deaths for lung cancer: 167 Case definition: death for lung cancer Years of follow‐up: 16 years Type of selenium marker: serum selenium concentration
Interventions	d.n.a.
Outcomes	Statistical methods: Cox logistic regression Variables controlled in analysis: age, pack‐years of smoking, spirits intake and dietary markers
Risk estimates [95% CI]	Reference category: lowest exposure category: 0.4‐1.0 µmol.L^‐1 Results: Deaths for lung cancer highest exposure category: HR 1.43 (95% CI 0.96 to 2.14)
Selenium levels in exposure categories	lowest category: 0.4‐1.0 µmol.L^‐1 highest category: 1.3‐3.0 µmol.L^‐1
Notes

Methods	Frequency‐matched cohort controlled study Country: China
Participants	Participants:Mark 2000: 29,584 men and women; Wei 2004: 1103 people who were originally selected as disease‐free controls in Mark 2000 Inclusion criteria: 40 to 69 years of age; healthy inhabitants of 4 Linxian communities; participants of a randomised controlled trial Name of parent cohort: General Population Trial Linxian Recruitment: 1985 Outcome assessment: May 1991 (Mark 2000); n.r. (Wei 2004) Number of cases: Wei 2004: oesophageal cancer: 75 (male/female: 49/26) mortality stomach, cardia cancer: 36 (male/female: 22/14) mortality stomach, non‐cardia cancer: 24 (male/female: 20/4) mortality other: 32 (male/female: 22/10) mortality Mark 2000: oesophageal cancer: 590 (male/female: 286/304) incidence oesophageal cancer: 332 (male/female: n.r.) mortality stomach, cardia cancer: 402 (male/female: 239/163) incidence stomach, cardia cancer: 232 (male/female: n.r.) mortality stomach, non‐cardia cancer: 87 (male/female: 66/21) incidence stomach, non‐cardia cancer: 68 (male/female: n.r.) mortality Case definition: mortality, incidence Years of follow‐up: unclear/approximately 9 years (Wei 2004), 6 years (Mark 2000) Type of selenium marker: serum
Interventions	d.n.a.
Outcomes	Statistical methods: cox‐proportional hazard model Variables controlled in analysis:Wei 2004: age, cholesterol, smoking, alcohol intake, BMI; Mark 2000: age Variables controlled by matching: age category, gender
Risk estimates [95% CI]	Wei 2004: Reference category: lowest quartile Results: Oesophageal cancer both genders: highest quartile: RR 0.35 (95% CI 0.16 to 0.81) Stomach, cardia cancer both genders: highest quartile: RR 0.31 (95% CI 0.11 to 0.87) Stomach, non‐cardia cancer both genders: highest quartile: RR 1.64 (95% CI 0.49 to 5.48) Other cancers both genders: highest quartile: RR 1.95 (95% CI 0.66 to 5.81) Mark 2000: Reference category: lowest quartile Results: Oesophageal cancer both genders / incidence: highest quartile: RR 0.56 (95% CI 0.44 to 0.71) both genders / mortality: highest quartile: RR 0.62 (95% CI 0.44 to 0.89) Stomach, cardia cancer both genders / incidence: highest quartile: RR 0.47 (95% CI 0.33 to 0.65) both genders / mortality: highest quartile: RR 0.59 (95% CI 0.39 to 0.90) Stomach, non‐cardia cancer both genders / incidence: highest quartile: OR 1.07 (95% CI 0.55 to 2.08) both genders / mortality: highest quartile: OR 1.03 (95% CI 0.85 to 2.02)
Selenium levels in exposure categories	Wei 2004: lowest quartile: 0.00 to 0.76 µmol/l highest quartile ≥ 1.07 µmol/l Mark 2000: lowest quartile: 0.00 to 59.70 µg/l highest quartile ≥ 82.20 µg/l
Notes	Primary publication: Wei 2004 Other publication: Mark 2000 Remark: Wei 2004 measured serum selenium in a sub‐cohort derived from 29,584 male and female participants of the Linxian Population Trial. The earlier publication of this study, Mark 2000 reported 332 fatal cases and 590 incident cases. The later publication, Wei 2004 reported deaths from oesophageal cancer in the disease‐free controls of Mark 2000 and analysed 75 fatal cases.

Study	Reason for exclusion
Bostick 1993	Cohort study: Iowa Women's Health Study cohort Selenium exposure not assessed according to eligibility: only intake of selenium supplements yes/no in questionnaire assessed
Brock 1991	Case‐control study with precancerous condition (carcinoma in situ of the cervix)
Chen 1988	Case‐control study
Chen 2003	Case‐control study
Connelly‐Frost 2009	Case‐control study
Costello 2001	APPOSE (Australian Prostate Cancer Prevention Trial Using Selenium): publication describes study design, trial was not started
Criqui 1991	Population‐based prospective case‐control study: Lipid Research Clinic Prevalence and Follow‐Up study Results not reported according to inclusion criteria: differences in mean selenium levels reported
Cui 2007	Nested case‐control study Selenium exposure not assessed according to eligibility: selenium measurement conducted in tissue of benign breast disease
Davies 2002	Nested case‐control study: EPIC Norfolk study cohort Results not reported according to inclusion criteria: RR estimate per unit increase in selenium level reported
Fleshner 2003	Randomised Study of Vitamin E, Selenium, and Soy Protein Isolate in Patients with High‐Grade Prostatic Intraepithelial Neoplasia: Multicomponent Intervention
Hagmar 1992	Historical cohort study
Harris 2012	Cancer was not a study endpoint
Hartman 2002	Nested case‐control study: ATBC cohort Results not reported according to inclusion criteria: differences in mean selenium levels reported; OR reported as graph and could not be calculated from reported data
Huzarski 2006	Interventional study without control group with 1489 female participants with BRCA1 mutation who received a selenium‐containing nutritional supplement
Joniau 2007	Intervention study without control group with male participants with high‐grade intraepithelial neoplasia of the prostate who received a selenium‐containing nutritional supplement
Kellen 2008	Case‐control study
Kilander 2001	Cohort study in Uppsala/Sweden Results not reported according to inclusion criteria: RR estimate per unit increase in selenium level reported
Knekt 1988a	Nested case‐control study: Mobile Health Clinic cohort Results not reported according to inclusion criteria: differences in mean selenium levels reported
Knekt 1988b	Nested case‐control study: Mobile Health Clinic cohort Results not reported according to inclusion criteria: differences in mean selenium levels reported
Knekt 1991	Nested case‐control study: Mobile Health Clinic cohort Results not reported according to inclusion criteria: differences in mean selenium levels reported
Kok 1987b	Nested case‐control study: Zoetermeer cohort Results not reported according to inclusion criteria: differences in mean selenium levels reported
Kune 2006	Case‐control study
Kuroda 1988	Case‐control study
Lawson 2007	Cohort study on multivitamin use and risk of prostate cancer
Le Marchand 2006	Case‐control study
Li 2004b	RCT for gastric cancer prevention with multicomponent intervention (200 mg synthetic allitridum and 100 μg selenium per day)
Limburg 2005	Randomised controlled trial: primary endpoint in this two‐by‐two factorial design trial with selenomethionine 200 µg daily and/or celecoxib 200 mg twice daily was the per‐participant change (regression, stable, progression) of preexisting oesophageal dysplasia—cancer incidence and mortality were not endpoints in this study
Linxian Pilot 2000	Randomised controlled trial with selenium supplements and celecoxib in participants with oesophageal squamous dysplasia in Linxian, China Endpoint was "regression of disease", cancer was not an endpoint in this investigation
Neuhouser 2009	Cohort study (Women's Health Initiative) on multivitamin use and risk of cancer and cardiovascular disease No data for selenium and cancer risk reported
Ray 2006	Cohort study (Women's Health and Aging Studies I and II) on selenium and carotenoid serum levels and mortality No data for selenium and cancer mortality reported
Rayman 2001	PRECISE trial (Prevention of Cancer by Intervention with Selenium): trial has been stopped
Rendon	Randomised controlled trial: Vitamin E, Selenium, and Soy Protein in Preventing Cancer in Patients with High‐Grade Prostate Neoplasia: Multicomponent Intervention
Steevens 2010b	Cancer was not a study endpoint
Thompson 2009	Cohort study: Iowa Women's Health Study cohort Selenium exposure not assessed according to eligibility: only intake of selenium supplements yes/no in questionnaire assessed
Tsugane 1996	Case‐control and cross‐sectional studies
Ujiie 2002	A part of this study is a prospective cohort study in Miyagi/Japan Results not reported according to inclusion criteria: differences in mean selenium levels reported
van Noord 1992	Nested case‐control study: DOM cohort Results not reported according to inclusion criteria: differences in mean selenium levels reported
van Noord 1993	Nested case‐control study: DOM II cohort Results not reported according to inclusion criteria: RR estimate per unit increase of selenium level reported
van't Veer 1996	Case‐control study
Wallace 2009	Case‐control study
Watters 2009	Cohort study on smoking and prostate cancer risk. Selenium not reported as independent variable
Wright 2004	Cohort study: ATBC cohort Exposure to antioxidants was assessed using a self‐developed index
You 2005	Randomised controlled trial to test retardation of the progression of precancerous gastric lesions among 3400 adults in Shandong, China. Intervention: vitamin C, vitamin E, selenium, garlic preparation Multicomponent intervention
Yuan 2006	Nested case‐control study: Shanghai cohort study No data on selenium and cancer risk reported
Zeegers 2009	Cohort study on factors influencing recurrence or progression of bladder cancer: West Midlands Bladder Cancer Prognosis Programme

Trial name or title	Cancer Sero Epidemiology Among the Japanese in Hawaii
Methods	This is a sero‐epidemiological prospective study to identify biochemical markers related to common cancers Among the aims are (a) to see whether low serum selenium levels increase prostate cancer risk, and (b) to determine whether low serum selenium levels increase urinary bladder cancer risk in men
Participants	9345 male American Japanese subjects, examined in Hawaii
Interventions	d.n.a.
Outcomes	Risk of prostate and urinary bladder cancer
Starting date	Project start: 15 September 1983, Project end planned for 30 June 2004
Contact information	Abraham M. Nomura Kuakini Medical Center 347 N Kuakini St Honolulu, HI 96817
Notes

Trial name or title	Selenium Supplementation for the Prevention of Hepatocellular Carcinomas in HBsAg Positive Patients (pilot study)
Methods	Randomised controlled trial
Participants	Men 45 to 64 years of age with positive HBsAg test, negative AFP test and normal ALT values
Interventions	Placebo or 200 mg (sic!)/d selenium as selenised yeast
Outcomes	Primary liver cancer
Starting date	2003
Contact information	Prof Kar Keung Cheng, University of Birmingham, UK
Notes	Study author contacted for further information, but no reply received Should probably say 200 µg/d selenium yeast as intervention in the publication