Objective: To evaluate whether a compartmental model could estimate iron absorption as accurately as the well-validated technique of plasma area under the curve using labelled test meals.

Design: The study is a randomised cross-sectional intervention.

Setting: The study was carried out at the Human Nutrition Unit at the Institute of Food Research, Norwich, UK.

Subjects: A total of nine female volunteers, aged 33±8 y.

Interventions: Volunteers were given an oral dose (approximately 5 mg) of Fe-57 as iron sulphate in an orange juice test drink and simultaneously infused Fe-58 (approximately 200 μg) as iron citrate over 90 min. Multiple blood samples were taken for the following 6 h. The samples were analysed by mass spectrometry and iron absorption was estimated using a mathematical model based on the appearance of Fe isotopes in plasma and the area under the curve technique.

Results: The geometric mean (−1 s.d., +1 s.d.) absorption of the model estimate is 16% (9, 31) and the area under curve estimate is 18% (8, 29).

Conclusions: Results indicate that a compartmental model can be used to estimate labelled iron absorption although it is unlikely that this new method will be used in favour of an existing one. Further studies are now needed with unlabelled iron to assess whether the technique could have application in the assessment of total (haem+nonhaem) iron absorption from food.

Sponsorship: European Union Marie Curie Fellowship and Biotechnology and Biological Sciences Research Council (BBSRC), UK.

The aim of several previous studies investigating iron absorption methodologies in human volunteers has been to correlate the peak increase in serum iron concentration with the absorption of iron from a test meal (Hallberg & Solvell, 1960; Ekenved et al, 1976). This has failed because the underlying assumption behind it that the rate of elimination of iron from serum does not vary markedly between individuals is false. Where this has been recognised, an intravenous (i.v.) labelled iron has been given in conjunction with an oral test dose to calculate iron absorption (Gonzalez et al, 2001). The drawback to the latter method is that the oral and i.v. doses must behave in the same (kinetic) way once in the systemic circulation, but there is some evidence that this does not always happen (Winchell, 1968).

Compartmental modelling can overcome these problems because it can be used to estimate the quantity of iron absorbed from an oral dose without the need to give a simultaneous i.v. dose of iron. It is a well-established technique for estimating fluxes and pool sizes, and does not require invasive sampling of body tissues (Carson et al, 1983; Green & Green, 1990; Jacquez, 1996). Its use in iron metabolism is based on radioisotope work, published mainly in the 1960s, which resulted in the development of complex multicompartmental models (Pollycove & Mortimer, 1961; Sharney et al, 1965; Hosain et al, 1967; Najean et al, 1967). A pilot study (Dainty et al, 2003) has indicated that simple, single compartmental modelling shows promise in estimating unlabelled iron absorption; however, bias due to modelling assumptions needs further investigation. The current paper attempts to address this issue by comparing the estimate of iron absorption made from the plasma area under the curve (AUC) method (Barrett et al, 1994) with that from the compartmental model using labelled test doses.

Volunteers were recruited through advertisements placed around the Norwich Research Park. Nine healthy women between the ages of 20 and 45 y volunteered for the study. Their health was assessed through prestudy screening, which included a blood test and a medical questionnaire. None of the subjects were taking dietary supplements, was pregnant or a smoker. The study was approved by the Norwich and Norfolk District Ethics Committee and all recruits signed informed consent forms before entering the study. After an overnight fast, the subjects had an i.v. cannula inserted into each arm. The experimental protocol and dose preparation are similar to that described by Dainty et al (2003). In all, 12 serial blood samples were taken at t=0, 20, 40, 60, 80, 120, 145, 175, 200, 260, 320 and 380 min after the simultaneous oral and intravenous dose. A single focussing multicollector inductively coupled plasma mass spectrometer (Isoprobe, Micromass, UK) with a desolvating sample introduction system and microconcentric nebulizer (Aridus and T1 H, both from Cetac) was used for isotope analysis. All samples were run in duplicate and calibrated against IRMM014 (Rosman & Taylor, 1998).

Model analysis

The majority of the analysis is contained in a recently published paper (Dainty et al, 2003) and is not reproduced here. Briefly, the model estimates the quantity of iron that has been absorbed from a test meal by assuming that it crosses the gut wall as a constant infusion, appears in the plasma and is cleared. By fitting equations to the labelled plasma iron concentration data, it is possible to estimate the quantity of iron absorbed from the test meal (M_oral). By using the same model for the i.v. infusion, it is possible to validate some of its assumptions by estimating the ‘i.v. recovery’, which is defined as the calculated mass coming from the i.v. dose, M_i.v., divided by the actual mass of labelled iron in the i.v. dose (dose_i.v.). The ratio of these two masses should be an independent validation of the model's suitability for predicting absorption from the oral dose.

Statistical analysis

All data are presented as arithmetic mean±s.d. and are based on nine volunteers (n=9) unless stated otherwise. A paired, two-tail, Student's t-test was carried out to assess the significance of numerical differences arising from different routes of isotope administration. Differences were considered significant if P<0.05. A graphical technique was used to assess the agreement between estimation of iron absorption from the AUC method and the model method (Bland & Altman, 1986). The difference in iron absorption between the two methods is plotted against the average value of iron absorption found using both techniques. Two methods for measuring the same quantity are said to have good agreement provided (1) variations within the mean of the difference in iron absorption±2 s.d. are not significant and (2) the average bias of one method relative to the other is small.

Absorption data, plasma half-lives and i.v. recovery data are shown in Table 1. The geometric mean (−1 s.d.,+1 s.d.) absorption for the AUC method is 18% (10, 32) and that for the model is 16% (8, 29). When the model is applied to the labelled i.v. plasma concentration data, it can be seen that there is a small underestimation of the true i.v. dose that was infused into the subjects (i.v. recovery=95±11%). The half-life (t_1/2) of the labelled iron in the plasma from the oral dose is 2.67±0.36 h (k=0.26±0.04 h⁻¹) and is not significantly different (P=0.43) from that for the i.v. dose (t_1/2=2.60±0.36 h, k=0.27±0.04 h⁻¹). Figure 1 is a graphical depiction of the agreement between the two methods. The mean difference between the estimation of %Fe absorption is calculated at −1.8% (ie the model method gives a lower estimate) and the 95% confidence interval is estimated as being between −7.5 and 3.9 units of difference (% iron absorption).

Bland and Altman plot showing mean difference between methods (—–) and 95% confidence interval (····).

Full size image

In the present paper, we have attempted to show that a single compartment model can be used to estimate the absorption of extrinsically labelled nonhaem iron from a test drink. For a new measurement technique to be accepted, it must be shown to perform as well as existing methods and to offer some advantages. Inspection of the Bland and Altman analysis (Figure 1) suggests that there is a small bias (mean difference=−1.8%) between the two methods for calculating iron absorption. This finding is reinforced by the geometric mean iron absorption values, 18% for the AUC method and 16% for the model. The underestimation in i.v. recovery (Table 1) also supports these observations and the reasons for this have already been discussed in an earlier publication (Dainty et al, 2003) and can be summarised as an oversimplified model structure that does not take account of labelled iron that disappears and then refluxes back into the plasma compartment during the experimental period. Despite these concerns, the model method appears to compare well with the AUC method although the width of the 95% confidence interval needs further discussion.

In their original paper (Bland & Altman, 1986), a key question was posed as to how much disagreement can be tolerated between ‘old’ and ‘new’ measurement techniques before a new method is considered to be acceptable. Using the model method in the present study, we may expect an estimate of iron absorption to be somewhere between 22 and 10% instead of the ‘true’ value of 18% as estimated by the AUC technique. In most nutritional studies, this would clearly be unacceptable and the model method would not be used. However, agreement between the two techniques would almost certainly have been improved with a larger sample size. For this kind of comparative study it is recommended that 50 data points be collected (Altman, 1991) and this would lead to a much more definitive statement regarding the agreement of the two methods. The promising nature of the results from this pilot study suggests that such an exercise should be undertaken because, while it is clear that the model method for predicting labelled iron absorption is not perfect, the method could be useful for unlabelled iron studies, where there is presently no method for quantifying the absorption of total iron from single test meals. Further work is needed to assess the quantity of unlabelled iron that would be needed in a test meal to ensure adequate characterisation of the plasma response curve, and to establish whether the method is applicable to a range of test meals with differing iron release properties.

We wish to thank John Eagles and Jurian Hoogewerff for processing the samples on the mass spectrometer and Treasa Nic Suibhne, Fiona O'Neill and Gosia Newman for help in the laboratory. Sources of support: Beatriz Sarria was funded by a European Union Marie Curie Fellowship. The other authors are funded by the Biotechnology and Biological Sciences Research Council (BBSRC), UK.

Affiliations

1. Instituto del Frío, Consejo Superior de Investigaciones Científicas, Madrid, Spain; and

   • B Sarria

2. Institute of Food Research, Norwich Research Park, Colney, Norwich, UK

   • J R Dainty
   • , T E Fox
   •  & S J Fairweather-Tait

Corresponding author

Correspondence to J R Dainty.

To obtain permission to re-use content from this article visit RightsLink.

Publication history

DOI

https://doi.org/10.1038/sj.ejcn.1602030