Birgit Debrabant

Does selenium supplementation affect thyroid function? Results from a randomized, controlled, double-blinded trial in a Danish population.

OBJECTIVE Selenium is present in the active site of proteins important for thyroid hormone synthesis and metabolism. The objective of this study is to investigate the effect of selenium supplementation in different doses on thyroid function, under conditions of suboptimal dietary selenium intake. DESIGN The Danish PREvention of Cancer by Intervention with SElenium pilot study (DK-PRECISE) is a randomized, double-blinded, placebo-controlled trial. A total of 491 males and females aged 60-74 years were randomized to 100 μg (n=124), 200 μg (n=122), or 300 μg (n=119) selenium-enriched yeast or matching yeast-based placebo tablets (n=126). A total of 361 participants, equally distributed across treatment groups, completed the 5-year intervention period. METHODS Plasma samples were analyzed for selenium and serum samples for TSH, free triiodothyronine (FT3), and free thyroxine (FT4) at baseline, and after 6 months, and 5 years of supplementation. RESULTS Plasma selenium concentrations increased significantly and dose-dependently in treatment groups receiving selenium (P<0.001). Serum TSH and FT4 concentrations decreased significantly and dose-dependently by 0.066 mIU/l (P=0.010) and 0.11 pmol/l (P=0.015), respectively, per 100 μg/day increase, with insignificant differences between 6 months and 5 years. No significant effects were found for FT3 and FT3:FT4 ratio. CONCLUSIONS In euthyroid subjects, selenium supplementation minutely and dose-dependently affects thyroid function, when compared with placebo, by decreasing serum TSH and FT4 concentrations. Based on these findings, selenium supplementation is not warranted under conditions of marginal selenium deficiency. However, a role for selenium supplementation in the treatment of autoimmune thyroid diseases is still unresolved.

Human longevity and variation in DNA damage response and repair: study of the contribution of sub-processes using competitive gene-set analysis.

DNA-damage response and repair are crucial to maintain genetic stability, and are consequently considered central to aging and longevity. Here, we investigate whether this pathway overall associates to longevity, and whether specific sub-processes are more strongly associated with longevity than others. Data were applied on 592 SNPs from 77 genes involved in nine sub-processes: DNA-damage response, base excision repair (BER), nucleotide excision repair, mismatch repair, non-homologous end-joining, homologous recombinational repair (HRR), RecQ helicase activities (RECQ), telomere functioning and mitochondrial DNA processes. The study population was 1089 long-lived and 736 middle-aged Danes. A self-contained set-based test of all SNPs displayed association with longevity (P-value=9.9 × 10−5), supporting that the overall pathway could affect longevity. Investigation of the nine sub-processes using the competitive gene-set analysis by Wang et al indicated that BER, HRR and RECQ associated stronger with longevity than the respective remaining genes of the pathway (P-values=0.004–0.048). For HRR and RECQ, only one gene contributed to the significance, whereas for BER several genes contributed. These associations did, however, generally not pass correction for multiple testing. Still, these findings indicate that, of the entire pathway, variation in BER might influence longevity the most. These modest sized P-values were not replicated in a German sample. This might, though, be due to differences in genotyping procedures and investigated SNPs, potentially inducing differences in the coverage of gene regions. Specifically, five genes were not covered at all in the German data. Therefore, investigations in additional study populations are needed before final conclusion can be drawn.

Early Dynamics of P-selectin and Interleukin 6 Predicts Outcomes in Ischemic Stroke

BACKGROUND Thromboinflammatory molecules connect the prothrombotic state, endothelial dysfunction, and systemic/local inflammation in the acute phase of ischemic stroke. METHODS We prospectively investigated (1) serial changes in the levels of thromboinflammatory biomarkers in 76 patients with acute ischemic stroke (6, 24, and 72 hours after onset); (2) compared with 44 patients with asymptomatic severe (≥70%) carotid stenosis and 66 patients with Parkinson disease; and (3) we applied multiple regression methods, relating biological biomarkers combined with demographic data and comorbidities to poststroke infection, death, and functional outcome, and assessed the ability of the models to predict each outcome. RESULTS Interleukin 6 (IL-6) levels and change of IL-6 concentrations by 72 hours correlated with the size of tissue damage indicated by S100B titers. Levels of IL-6 and P-selectin at 72 hours were higher in patients with large-artery versus lacunar stroke. High concentration of IL-6, monocyte chemotactic protein 1, and S100B at 6 hours were associated with poststroke infections; high concentration of IL-6, S100B, and high-sensitivity C-reactive protein (hsCRP) correlated with death. Change of P-selectin from 6 to 72 hours by 1 unit increased the incidence of poststroke infections with an odds ratio of 22.7; each 100 units of IL-6 at baseline increased the odds of death by 9‰, and at 72 hours, the odds of poststroke infections by 4‰. Each unit of baseline hsCRP elevated the odds of death by 7%. CONCLUSIONS In regression models, in which biological, demographic, and comorbid factors were combined, those biological biomarkers predicted poor outcome with high accuracy, which were characterized by an increasing concentration by 72 hours. Two particular biomarkers emerged to predict outcomes besides hsCRP: early dynamic changes in the systemic levels of P-selectin and IL-6.

Copy number variation associates with mortality in long-lived individuals: a genome-wide assessment

Copy number variants (CNVs) represent a significant source of genetic variation in the human genome and have been implicated in numerous diseases and complex traits. To date, only a few studies have investigated the role of CNVs in human lifespan. To investigate the impact of CNVs on prospective mortality at the extreme end of life, where the genetic component of lifespan appears most profound, we analyzed genomewide CNV data in 603 Danish nonagenarians and centenarians (mean age 96.9 years, range 90.0–102.5 years). Replication was performed in 500 long‐lived individuals from the Leiden Longevity Study (mean age 93.2 years, range 88.9–103.4 years). First, we assessed the association between the CNV burden of each individual (the number of CNVs, the average CNV length, and the total CNV length) and mortality and found a significant increase in mortality per 10 kb increase in the average CNV length, both for all CNVs (hazard ratio (HR) = 1.024, P = 0.002) and for duplications (HR = 1.011, P = 0.005), as well as per 100 kb increase in the total length of deletions (HR = 1.009, P = 0.0005). Next, we assessed the relation between specific deletions and duplications and mortality. Although no genome–wide significant associations were discovered, we identified six deletions and one duplication that showed consistent association with mortality in both or either of the sexes across both study populations. These results indicate that the genome–wide CNV burden, specifically the average CNV length and the total CNV length, associates with higher mortality in long‐lived individuals.

No Association between Variation in Longevity Candidate Genes and Aging-related Phenotypes in Oldest-old Danes

In this study we explored the association between aging-related phenotypes previously reported to predict survival in old age and variation in 77 genes from the DNA repair pathway, 32 genes from the growth hormone 1/ insulin-like growth factor 1/insulin (GH/IGF-1/INS) signalling pathway and 16 additional genes repeatedly considered as candidates for human longevity: APOE, APOA4, APOC3, ACE, CETP, HFE, IL6, IL6R, MTHFR, TGFB1, SIRTs 1, 3, 6; and HSPAs 1A, 1L, 14. Altogether, 1,049 single nucleotide polymorphisms (SNPs) were genotyped in 1,088 oldest-old (age 92-93 years) Danes and analysed with phenotype data on physical functioning (hand grip strength), cognitive functioning (mini mental state examination and a cognitive composite score), activity of daily living and self-rated health. Five SNPs showed association to one of the phenotypes; however, none of these SNPs were associated with a change in the relevant phenotype over time (7 years of follow-up) and none of the SNPs could be confirmed in a replication sample of 1,281 oldest-old Danes (age 94-100). Hence, our study does not support association between common variation in the investigated longevity candidate genes and aging-related phenotypes consistently shown to predict survival. It is possible that larger sample sizes are needed to robustly reveal associations with small effect sizes.

Point Processes with a Generalized Order Statistic Property

Mixed Poisson processes are characterized by a well-known order statistic property: their occurrence times are distributed like ordinary uniform order statistics, given the state of the process at a certain time. We study a generalization of this property using a generalized model of ordered random variables, including a sequence of real parameters.

DNA methylation and all-cause mortality in middle-aged and elderly Danish twins

Several studies have linked DNA methylation at individual CpG sites to aging and various diseases. Recent studies have also identified single CpGs whose methylation levels are associated with all-cause mortality. In this study, we perform an epigenome-wide study of the association between CpG methylation and mortality in a population of 435 monozygotic twin pairs from three Danish twin studies. The participants were aged 55–90 at the time of blood sampling and were followed for up to 20 years. We validated our results by comparison with results from a British and a Swedish cohort, as well as results from the literature. We identified 2806 CpG sites associated with mortality (false discovery rate (FDR)<0.05), of which 24 had an association p-value below 10−7. This was confirmed by intra-pair comparison controlling for confounding effects. Eight of the 24 top sites could be validated in independent datasets or confirmed by previous studies. For all these eight sites, hypomethylation was associated with poor survival prognosis, and seven showed monozygotic correlations above 35%, indicating a potential moderate to strong heritability, but leaving room for substantial shared or unique environmental effects. We also set up a predictor for mortality using least absolute shrinkage and selection operator (LASSO) regression. The predictor showed good performance on the Danish data under cross-validation, but did not perform very well in independent samples.

Identifying traffic accident black spots with Poisson-Tweedie models

This paper aims at the identification of black spots for traffic accidents, i.e. locations with accident counts beyond what is usual for similar locations, using spatially and temporally aggregated hospital records from Funen, Denmark. Specifically, we apply an autoregressive Poisson-Tweedie model, which covers a wide range of discrete distributions and handles zero-inflation as well as overdispersion. The estimated power parameter of the model was 1.6 (SE=0.06) suggesting a distribution close to the Pólya-Aeppli distribution. We identified nine black spots consistently standing out in all six considered calendar years and calculated by simulations a probability of p=0.03 for these to be chance findings. Altogether, our results recommend these sites for further investigation and suggest that our simple approach could play a role in future area based traffic accident prevention planning.

DNA methylation age and perceived age in elderly Danish twins

Perceived age is an easily accessible biomarker of aging. Here, we studied its relation to DNA methylation age (DNAm age) as introduced in (Horvath, 2013) in 180 elderly Danish twins. We found perceived age and DNAm age to be associated with chronological age (P=0.04 resp. P=2.2e-10) when correcting for gender, but did not see an association between perceived age and DNAm age (P=0.44). Intrapair-analysis showed that the proportion of pairs where the twin with the highest perceived age also had the highest DNAm age was not different from 0.5 (P=1), and we did not see a trend when dividing pairs according to their difference in perceived age (P=0.36). Hence, intrapair analysis did not reveal links between perceived age and DNAm age. Moreover, none of the 353 CpGs underlying DNAm age was individually associated with perceived age after correction for multiple-testing (P>6e-4, FDR>0.21). Finally, when constructing an epigenetic signature based on these CpGs to predict perceived age, we only found a correlation of 0.18 (95%CI: -0.06 to 0.40) and a mean square error of 13.6 years2 between observed and predicted values in the test dataset, indicating poor predictive strength. Altogether, our results suggest that perceived age and DNAm age capture different aging aspects.

The null hypothesis of GSEA, and a novel statistical model for competitive gene set analysis

Motivation: Competitive gene set analysis intends to assess whether a specific set of genes is more associated with a trait than the remaining genes. However, the statistical models assumed to date to underly these methods do not enable a clear cut formulation of the competitive null hypothesis. This is a major handicap to the interpretation of results obtained from a gene set analysis. Results: This work presents a hierarchical statistical model based on the notion of dependence measures, which overcomes this problem. The two levels of the model naturally reflect the modular structure of many gene set analysis methods. We apply the model to show that the popular GSEA method, which recently has been claimed to test the self‐contained null hypothesis, actually tests the competitive null if the weight parameter is zero. However, for this result to hold strictly, the choice of the dependence measures underlying GSEA and the estimators used for it is crucial. Contact: bdebrabant@health.sdu.dk Supplementary information: Supplementary material is available at Bioinformatics online.MOTIVATION Competitive gene set analysis intends to assess whether a specific set of genes is more associated with a trait than the remaining genes. However, the statistical models assumed to date to underly these methods do not enable a clear cut formulation of the competitive null hypothesis. This is a major handicap to the interpretation of results obtained from a gene set analysis. RESULTS This work presents a hierarchical statistical model based on the notion of dependence measures, which overcomes this problem. The two levels of the model naturally reflect the modular structure of many gene set analysis methods. We apply the model to show that the popular GSEA method, which recently has been claimed to test the self-contained null hypothesis, actually tests the competitive null if the weight parameter is zero. However, for this result to hold strictly, the choice of the dependence measures underlying GSEA and the estimators used for it is crucial. CONTACT bdebrabant@health.sdu.dk.