Inclusion of Vitamin A Intake Data Provides Improved Compartmental Model-Derived Estimates of Vitamin A Total Body Stores and Disposal Rate in Older Adults

Abstract

Background Sampling times and study duration impact estimates of kinetic parameters and variables including total body stores (TBS) and disposal rate (DR) when compartmental analysis is used to analyze vitamin A kinetic data. Objective We hypothesized that inclusion of dietary intake (DI) of vitamin A as an additional input would improve confidence in predictions of TBS and DR when modeling results appear to indicate that studies are not long enough to accurately define the terminal slope of the plasma retinol isotope response curve. Methods We reanalyzed previously published data on vitamin A kinetics monitored over 52 d in 7 US and 6 Chinese adults (means: 56 y, BMI 26.6 kg/m2, 38% males), adding an estimate for vitamin A intake [2.8 µmol/d (mean RDA)] as an input during application of the Simulation, Analysis and Modeling software. Results Use of a model with 1 extravascular compartment (1 EV), as in the original analysis, resulted in predictions of vitamin A intake that were higher than physiologically reasonable; inclusion of intake data in a model with 2 extravascular compartments (2 EV DI) resulted in more realistic estimates of intake and DR. Specifically, predictions of DR by the 2 EV DI (versus 1 EV) model were 2.10 compared with 12.2 µmol/d (US) and 2.21 compared with 5.13 µmol/d (Chinese). Predictions of both TBS [2056 compared with 783 µmol (US) and 594 compared with 219 µmol (Chinese)] and days of vitamin A stores [981 compared with 64 d (US) and 269 compared with 43 d (Chinese)] were higher using the new approach. Conclusions Inclusion of vitamin A intake as additional data input when modeling vitamin A kinetics can compensate for less-than-optimal study duration, providing more realistic predictions of vitamin A TBS and DR. This approach advances the application of compartmental analysis to the study of vitamin A and, potentially, other nutrients.

Addition of Vitamin A Intake Data during Compartmental Modeling of Retinol Kinetics in Theoretical Humans Leads to Accurate Prediction of Vitamin A Total Body Stores and Kinetic Parameters in Studies of Reasonable Duration

“…In related work, we recently showed (7) that by adding an estimate of vitamin A intake into the data input for modeling, we obtained physiologically reasonable predictions of intake and DR, and thus TBS, in older US and Chinese adults previously studied in 52-d kinetic experiments (8, 9). In contrast to values for DR calculated as in the original analysis (10) (12 µmol/d for 7 US subjects and 5.1 µmol/d for 6 Chinese subjects), which were reflected in unrealistic predictions of dietary vitamin A intake (16 and 7 µmol/d, respectively), the addition of an estimate of intake equivalent to the mean RDA for adults (2.8 µmol/d) during modeling resulted in DR predictions of 2.1 (US) and 2.2 µmol/d (Chinese) and 2.8 and 2.9 µmol/d, respectively, for intake.…”

“…Our current results suggest that TBS reported for that group (823 µmol) in Lopez-Teros et al (17) is likely accurate. More recently, as discussed previously, we applied the approach of including intake in the model to a different problem (7). Specifically, we realized in retrospect that originally published (10) model predictions for vitamin A intake and DR in older adults were unrealistically high (and thus model-predicted TBS was lower than if the predictions for those parameters had been more physiologically reasonable), presumably because the study length (52 d) was not long enough to fully define the terminal slope of the plasma isotope response curve for all subjects.…”

“…Using methods similar to those described by Ford et al (14), we postulated 12 theoretical subjects ( Supplemental Table 1 ), 4 each of children, younger adults, and older adults. For each subject, we assigned physiologically reasonable values for relevant demographics and state variables (including vitamin A intake, plasma retinol pool size, and extravascular vitamin A) based on results adapted from previously published vitamin A studies in children (15–17), younger adults (5, 18–21), and older adults (7–9, 18, 22), as well as unpublished data (Supplemental Table 1). Then, parameters specified in the model described in the next section were assigned by adapting published values or choosing physiologically reasonable estimates.…”

“…Initially, model simulations included loss from both compartment 6 and compartment 7, as shown in Figure 1 and as presumably occurs in vivo. However, because only plasma is typically sampled in human retinol kinetic studies (and this does not allow identification of the site or sites of loss), and because current results were only minimally affected if output was modeled from 1 compared with both extravascular compartments, we used a model with output only from compartment 6, as in a number of previous studies (3, 7, 17, 23).…”

“…TBS results were compared to the reference value (discussed previously) for each subject; we specified that a value for TBS within 15% of the reference value would be considered accurate. Before testing, we fixed the values for the 2 model parameters that describe the initial absorptive phase [L(2, 1) and DT(3); Figure 1] at the reference values for each individual because these parameters are difficult to identify with precision and they have little influence on model predictions of postabsorptive kinetics and state variables; previous studies have taken the same approach (5, 7, 28). After fitting, iterations were performed using weighted nonlinear regression analysis, with an FSD of 0.05 as the weighting factor for plasma and, when applicable, for intake data, to estimate final values for the model parameters and their statistical uncertainty (25, 26).…”