Severity scoring of manganese health effects for categorical regression

•Development of Manganese toxicity database for categorical regression outlined.•Severity scoring matrix for health outcomes developed and applied to database.•Database is applicable to broad range of modeling approaches.•Severity scoring matrix can be adopted as general template for any essential substances.

Characterizing the U-shaped exposure response relationship for manganese (Mn) is necessary for estimating the risk of adverse health from Mn toxicity due to excess or deficiency. Categorical regression has emerged as a powerful tool for exposure-response analysis because of its ability to synthesize relevant information across multiple studies and species into a single integrated analysis of all relevant data. This paper documents the development of a database on Mn toxicity designed to support the application of categorical regression techniques. Specifically, we describe (i) the conduct of a systematic search of the literature on Mn toxicity to gather data appropriate for dose-response assessment; (ii) the establishment of inclusion/exclusion criteria for data to be included in the categorical regression modeling database; (iii) the development of a categorical severity scoring matrix for Mn health effects to permit the inclusion of diverse health outcomes in a single categorical regression analysis using the severity score as the outcome variable; and (iv) the convening of an international expert panel to both review the severity scoring matrix and assign severity scores to health outcomes observed in studies (including case reports, epidemiological investigations, and in vivo experimental studies) selected for inclusion in the categorical regression database. Exposure information including route, concentration, duration, health endpoint(s), and characteristics of the exposed population was abstracted from included studies and stored in a computerized manganese database (MnDB), providing a comprehensive repository of exposure-response information with the ability to support categorical regression modeling of oral exposure data.

Graphical abstractDownload high-res image (173KB)Download full-size image