Evaluation of inflammation adjustment methods to assess iron deficiency using longitudinal data from norovirus human challenge trials

Abstract

Accounting for inflammation is necessary to assess iron deficiency using ferritin. A limitation of existing inflammation-adjustment methods is reliance on cross-sectional data to evaluate method performance. The study objective was to evaluate three inflammation-adjustment methods using longitudinal data from two controlled trials where apparently healthy adults (n = 52) were exposed to norovirus. Correction factors (CF), the Biomarkers Reflecting Inflammation and Nutritional Determinants of Anemia (BRINDA) regression correction (BRC), and restricted cubic splines (RCS) were used to adjust the influence of inflammation on ferritin using alpha-1-acid glycoprotein (AGP) and/or C-reactive protein (CRP). Blood was collected at baseline (day 0, pre-exposure to norovirus) and at 9 time points post-exposure (days 1, 2, 3, 4, 7, 14, 21, 28, and 35). Inflammation-adjusted ferritin concentrations were compared with 1) baseline, 2) endline, 3) the average of baseline and endline, and 4) predicted ferritin concentrations among subjects with infection, expressed as percent difference. The predicted ferritin concentrations were modeled using data from 26 subjects without infection in a linear mixed model. Adjusting for CRP or AGP, the median differences between adjusted ferritin using CF, BRC, and RCS were respectively [0.2%, 2.5%], [-22.2%, -20.8%], [-16.7%, -7.1%] compared with the average of baseline and endline values and were 0%, [-16.8%, -18.5%], [-8.9%, -2.8%] compared with predicted ferritin concentrations. For BRC, adjusting for both CRP and AGP tended to result in more over-adjustment of ferritin compared to using a single inflammatory protein. The BRC appeared to overcorrect ferritin in this study setting, while the CF yielded adjusted ferritin concentrations closer to the average baseline and endline concentrations and the predicted concentrations. Longitudinal studies with larger sample sizes exposed to other infectious agents are needed to further evaluate inflammation-adjustment methods and the need for including multiple inflammation biomarkers.

Fig 1. Distributions of percent difference in ferritin comparing inflammation-adjusted values with the baseline, endline, average of baseline and endline, and predicted (for those with infection) and unadjusted (for those without infection) values among adults exposed to norovirus, using Correction Factors (CF), BRINDA Regression Correction (BRC), and restricted cubic spines (RCS).

Inflammatory biomarkers were α-1-acid glycoprotein (AGP) and C-reactive protein (CRP). The midline indicates the median difference in percentage, with the upper and lower limits of the box being the third and first quartile (75th and 25th percentile) respectively. The whiskers extend up to 1.5 times the interquartile range from the top (bottom) of the box to the furthest data within that distance. The reference values (or deciles) used for CRP and AGP were 0.18 mg/L and 0.38 g/L.

Fig 2

Percentage of data points categorized as iron deficient (serum ferritin <15 μg/L) based on inflammation-adjusted ferritin concentration for (A) all observations (n = 445 time-person observations), (B) among subjects with infection excluding baseline measurements (n = 200), and (C) among those without infection including baseline measurements (n = 219). Inflammation adjustment methods are Correction Factors (CF), BRINDA Regression Correction (BRC), and restricted cubic spines (RCS). Each method was adjusted for α-1-acid glycoprotein (AGP) and/or C-reactive protein (CRP). The horizontal line indicates the percentage of iron deficiency using raw, unadjusted ferritin concentrations.