Synthesizing subject-matter expertise for variable selection in causal effect estimation: A case study.

BACKGROUND: Causal graphs are an important tool for covariate selection but there is limited applied research on how best to create them. Here, we used data from the Coronary Drug Project (CDP) trial to assess a range of approaches to directed acyclic graph (DAG) creation. We focused on the effect of adherence on mortality in the placebo arm, since the true causal effect is believed with a high degree of certainty. METHODS: We created DAGs for the effect of placebo adherence on mortality using different approaches for identifying variables and links to include or exclude. For each DAG, we identified minimal adjustment sets of covariates for estimating our causal effect of interest, and applied these to analyses of the CDP data. RESULTS: When we used only baseline covariate values to estimate the cumulative effect of placebo adherence on mortality, all adjustment sets performed similarly. The specific choice of covariates had minimal effect on these (biased) point estimates, but including non-confounding prognostic factors resulted in smaller variance estimates. When we additionally adjusted for time-varying covariates of adherence using inverse probability weighting, covariates identified from the DAG created by focusing on prognostic factors performed best. CONCLUSION: Theoretical advice on covariate selection suggests including prognostic factors that are not exposure predictors can reduce variance without increasing bias. In contrast, for exposure predictors that are not prognostic factors, inclusion may result in less bias control. Our results empirically confirm this advice. We recommend that hand-creating DAGs begin with identification of all potential outcome-prognostic factors.

Associations between metals and metabolomic profiles related to diabetes among adults in a rural region.

INTRODUCTION: Exposure to metals is associated with increased risk of type 2 diabetes (T2D). Potential mechanisms for metals-T2D associations involve biological processes including oxidative stress and disruption of insulin-regulated glucose uptake. In this study, we assessed whether associations between metal exposure and metabolite profiles relate to biological pathways linked to T2D. MATERIALS AND METHODS: We used data from 29 adults rural Colorado residents enrolled in the San Luis Valley Diabetes Study. Urinary concentrations of arsenic, cadmium, cobalt, lead, manganese, and tungsten were measured. Metabolic effects were evaluated using untargeted metabolic profiling, which included 61,851 metabolite signals detected in serum. We evaluated cross-sectional associations between metals and metabolites present in at least 50% of samples. Primary analyses adjusted urinary heavy metal concentrations for creatinine. Metabolite outcomes associated with each metal exposure were evaluated using pathway enrichment to investigate potential mechanisms underlying the relationship between metals and T2D. RESULTS: Participants had a mean age of 58.5 years (standard deviation = 9.2), 48.3% were female, 48.3% identified as Hispanic/Latino, 13.8% were current smokers, and 65.5% had T2D. Of the detected metabolites, 455 were associated with at least one metal, including 42 associated with arsenic, 22 with cadmium, 10 with cobalt, 313 with lead, 66 with manganese, and two with tungsten. The metabolic features were linked to 24 pathways including linoleate metabolism, butanoate metabolism, and arginine and proline metabolism. Several of these pathways have been previously associated with T2D, and our results were similar when including only participants with T2D. CONCLUSIONS: Our results support the hypothesis that metals exposure may be associated with biological processes related to T2D, including amino acid, co-enzyme, and sugar and fatty acid metabolism. Insight into biological pathways could influence interventions to prevent adverse health outcomes due to metal exposure.

Antibodies to oral pathobionts and colon cancer risk in the CLUE I cohort study.

Periodontitis has been associated with an increased risk for gastrointestinal cancers. The objective of our study was to investigate the association of antibodies to oral bacteria and the risk of colon cancer in a cohort setting. Using the CLUE I cohort, a prospective cohort initiated in 1974 in Washington County, Maryland, we conducted a nested case-control study to examine the association of levels of IgG antibodies to 11 oral bacterial species (13 total strains) with risk of colon cancer diagnosed a median of 16 years later (range: 1-26 years). Antibody response was measured using checkerboard immunoblotting assays. We included 200 colon cancer cases and 200 controls matched on age, sex, cigarette smoking status, time of blood draw and pipe or cigar smoking status. Controls were selected using incidence density sampling. Conditional logistic regression models were used to assess the association between antibody levels and colon cancer risk. In the overall analysis, we observed significant inverse associations for 6 of the 13 antibodies measured (P-trends <.05) and one positive association for antibody levels to Aggregatibacter actinomycetemcomitans (ATCC 29523; P-trend = .04). While we cannot rule out a role for periodontal disease in colon cancer risk, findings from our study suggest that a strong adaptive immune response may be associated with a lower risk of colon cancer. More studies will need to examine whether the positive associations we observed with antibodies to A. actinomycetemcomitans reflect a true causal association for this bacterium.