Se Status Prediction by Food Intake as Compared to Circulating Biomarkers in a West Algerian Population

Abstract

Algeria is the largest country in Africa, located close to the Mediterranean coastal area, where nutrients consumption varies widely. Local data on selenium composition of foods are not available. We postulated a close correlation between selenium status predictions from food consumption analysis with a quantitative analysis of circulating biomarkers of selenium status. Population characteristics were recorded from 158 participants and dietary selenium intake was calculated by 24-h recall. The average total plasma selenium was 92.4 ± 18.5 µg/L and the mean of selenium intake was 62.7 µg/day. The selenoprotein P concentration was 5.5 ± 2.0 mg/L and glutathione peroxidase 3 activity was 247.3 ± 41.5 U/L. A direct comparison of the dietary-derived selenium status to the circulating selenium biomarkers showed no significant interrelation. Based on absolute intakes of meat, potato and eggs, a model was deduced that outperforms the intake composition-based prediction from all food components significantly (DeLong's test, p = 0.029), yielding an area under the curve of 82%. Selenium status prediction from food intake remains a challenge. Imprecision of survey method or information on nutrient composition makes extrapolating selenium intake from food data providing incorrect insights into the nutritional status of a given population, and laboratory analyses are needed for reliable information.

Keywords: food intake; glutathione peroxidase 3; monitoring; predictive model; selenium; selenoprotein P.

Figure 1

Comparison of calculated Se intakes per day as determined by the food recall method in combination with the food composition data in relation to the laboratory analysis of total Se and SELENOP concentrations measured in the plasma samples. No significant differences (“ns”) were detected between the two groups of different Se intakes (less or more than 55.0 µg Se/day) with respect to (A) total plasma Se, of (B) plasma SELENOP concentrations. Significance calculated by the Mann-Whitney U test, ns; p > 0.05.

Figure 2

Correlation analysis of plasma Se with SELENOP concentrations. All of the available plasma samples (n = 134) of the patients enrolled were analysed for total plasma Se and SELENOP concentrations. The samples were separated into two groups based on total plasma Se deficiency into Se-deficient (<70 µg/L, green) and Se-replete (>70 µg/L, red). The biomarkers showed a significant and positive linear correlation (Spearman, R = 0.79, p = 0.0036) in the Se-deficient samples, whereas Se-replete subjects revealed a non-significant, positive correlation (Spearman, R = 0.06, p = 0.47).

Figure 3

ROC-analysis of different food categories to differentiate between Se-deficient and Se-replete subjects. (A) Absolute intakes of several food categories (g/day) yielded similar results and poor predictive information. (B) The multiple regression model based on eggs, meat and potatoes intakes outperformed any other combination of variables via stepwise AIC selection. The final model (blue) based on these three parameters yielded a high AUC of 82%, and performed significantly better compared to the model based on the calculated Se intake (grey) from all categories deduced via the available composition data (DeLong’s test, p = 0.029). (C) The estimates of the final model are given alongside with the corresponding confidence intervals.