![[Feedback]](/icons/banner/feedbackACT.gif){kind=icon}
![[For Subscribers]](/icons/banner/subscriberACT.gif){kind=icon}
![[Archive]](/icons/banner/archiveACT.gif){kind=icon}
![[Search]](/icons/banner/searchACT.gif){kind=icon}
![[Contents]](/icons/banner/tocACT.gif){kind=icon}
|
|
|
|
|
||
U.S. Department of Agriculture/ARS Children’s Nutrition Research Center, Department of Pediatrics, Baylor College of Medicine, Houston, TX 77030
3To whom correspondence should be addressed. E-mail: zchen1{at}bcm.edu.
 |
ABSTRACT |
|---|
 TOP ABSTRACT SUBJECTS AND METHODSRESULTSDISCUSSIONLITERATURE CITED |
|---|
KEY WORDS: • inductively coupled plasma MS • iron absorption • isobaric interference • mass bias • stable isotopes
The WHO has identified iron deficiency anemia as one of the top 10 global health problems (1–5). Studies of iron absorption, retention, and utilization by the human body have been carried out using stable isotopes for more than 3 decades (6–17), but have been limited by the relatively large amount of blood required for analysis and by the high costs and limited availability of isotope ratio analysis.
The current standard methodology for analyzing blood samples is thermal ionization MS (TIMS), which requires 0.5–1.0 mL of RBC, and extensive sample preparation time and expertise (18). TIMS equipment is rarely available for nutritional research; it is extremely expensive and only a few groups in the world have experience with biological sample preparation and analysis. Additionally, TIMS analysis, although not requiring a large blood volume, necessitates venipuncture to collect an adequate sample volume, which can be difficult for babies and small children, especially in a field setting in developing countries. Reducing the amount of blood required would therefore be very advantageous, possibly allowing the use of capillary blood samples. With the development and improvement of inductively coupled plasma MS (ICP-MS) instrumentation, high-resolution, single- and multicollector ICP-MS may play important roles in human nutrition research because of their ability to resolve interferences from argides, polyatoms, and doubly charged interference isotopes, and their high throughput and easy sample pretreatment (19–22). We previously described ICP-MS methods for the measurement of calcium isotope ratios that allowed much more rapid, cost-efficient, sample analysis and reduced sample volumes compared with existing TIMS methods (23). In this paper, we describe new techniques for interference correction and factors that affect the determination of iron isotope ratios by ICP-MS and the potential use of ICP-MS in human nutrition research.
|
SUBJECTS AND METHODS |
|---|
|
TOPABSTRACTSUBJECTS AND METHODSRESULTSDISCUSSIONLITERATURE CITED |
|---|
Stable iron isotopes of 58Fe and 57Fe were purchased from Trace Sciences and prepared as sterile, aqueous solutions for iron bioavailability studies. High-quality ultrapure concentrated nitric acid (Fisher Scientific) was used for sample digestion. Teflon containers and glass tubes (Fisher Scientific) were treated with 2.2 mol/L HNO3 overnight, rinsed with deionized water from Milli-Q Water System (Millipore), and air-dried under cover before use. Iron and nickel solutions of natural abundance were prepared from standard stock solutions (SCP Science).
Sample pretreatment
The majority of samples were obtained from volunteers involved in iron bioavailability studies in South America. Subjects were between 5 and 10 mo of age and approximately half were iron deficient. Subjects were administered 2 mg 57Fe isotope as an aqueous solution of ferrous sulfate with 50 mg vitamin C added, and 0.15–0.2 mg of 58Fe (as ferrous sulfate) that had been added to 60 mL maternal breast milk and allowed to equilibrate for 24 h. After the 58Fe labeled breast milk was administered, an additional 60 mL of unlabeled breast milk was given. A smaller number of samples were obtained from a study of 12- to 48-mo-old children in North America. Subjects consumed a breakfast meal extrinsically labeled with 1 mg 58Fe, as ferrous sulfate, in the morning. The same evening they were given a reference dose of 3 mg 57Fe as an aqueous solution of ferrous sulfate with 50 mg vitamin C. In both cases, blood samples were collected 14–15 d after the administration of stable iron isotopes and were centrifuged at 530 x g for 10 min to obtain RBC for iron incorporation measurement (11.12). Written informed consent was obtained from a parent or legal guardian for each subject; written or verbal assent was obtained based on age from the study subject. The Institutional Review Board of Baylor College of Medicine and Affiliated Hospitals as well as the local Institutional Review Board approved the clinical protocol. The samples analyzed in this study are a subset of the entire dataset that had sufficient blood left for ICP-MS analysis, after TIMS analysis was completed.
Pure concentrated nitric acid (30 µL) was first measured into a clean Teflon container; 2 µL of the RBC sample was placed next to the acid and immediately mixed with the acid. The mixed sample was placed on a hot plate for digestion at 
110°C for 
5 min. The clear, digested solution was then diluted to a volume of 3 mL to make a final solution with a nitric acid concentration of 0.1 mol/L. The solution was mixed in a Vortex-mix Genie (Fisher Scientific) before being transferred into a clean glass tube for ICP-MS iron isotope analysis. This sample pretreatment technique is much simpler than that for TIMS. It also eliminates any potential contaminants and mass fractionation during the complicated sample preparation process for TIMS analysis. Typically, TIMS analysis requires 500 µL of RBCs for digestion followed by an ion exchange procedure before being loaded onto a degassed rhenium filament (18).
ICP-MS iron isotope analysis
Among the 4 stable iron isotopes, 58Fe and 57Fe are the most commonly used as tracers due to their low natural abundances (6–14). During ICP-MS analysis using argon as the carrier gas, a relatively high mass resolution (M/
M) of several thousands is required to resolve polyatomic and doubly charged interferences on iron isotopes (Table 1). Currently, high-resolution ICP-MS instrumentation is not capable of resolving isobaric interferences of 54Cr and 58Ni on iron isotopes; therefore, interference correction is inevitably required for ICP-MS iron isotope ratio analysis. An analytical methodology was developed to scan iron and isobaric interference isotopes at medium mass resolution (M/ M 4000) using a double-focusing, single-collector Thermo Element 2 ICP-MS (Thermo) with software v2.354. Sample solutions were introduced into a spray chamber by a low-flow nebulizer using argon as a carrier gas. Dead time was determined and applied to the detector, as recommended by the manufacturer. The signal intensity was obtained by integration of the counting signal of the scanning mass over a 4-min acquisition period. Sample blanks were taken into account and the major interference of spectral overlaps from 58Ni on 58Fe was corrected mathematically by a concurrently measured 60Ni isotope signal. A natural abundance iron standard solution was analyzed in between sample runs to monitor changes of systematic bias throughout the day. All argide, oxide, and doubly charged spectral interferences were eliminated at a resolution of 4000. Tests of 30 human RBCs revealed negligible chromium interference on 54Fe in human blood samples. 54Cr contributed to <0.03% of the mass-54 peak, which is less than the counting statistic error for the mass-54 signal. 54Cr correction, however, may be necessary for other types of human samples, such as fecal samples (20). Hydride interferences in human RBC samples were also negligible at this resolution. The 57Fe/54Fe and 58Fe/54Fe ratios were obtained using the corrected counting intensities and normalized for mass bias.
|
Iron isotope ratios derived by TIMS and ICP-MS were used to calculate iron absorption, as previously described (18). Red blood incorporation of 57Fe and 58Fe was measured 14–15 d after isotope administration. Iron absorption was calculated based on the assumption that 90% of the absorbed iron had been incorporated into RBC.
Statistical analysis
Linear regression analysis was used to examine the correlations of iron isotope ratios measured by TIMS and ICP-MS. Agreement between TIMS and ICP-MS data was assessed using the Bland-Altman approach (25). Student’s paired t tests were used to examine differences between means using the 2 MS methods and Levene’s test to examine differences in variance. Data for iron absorption were analyzed using both data log-transformation (to base 10) to normalize the distribution and the raw (untransformed) data. Analysis was carried out using SSPS version 11 for Macintosh. Differences were considered significant if P < 0.05.
Correction of iron isotope counting signal
Dead time. Detector dead time was determined using a series of uranium solutions and was applied to the detector for automatic correction of the integrated counting signal for each scanned mass. Dead time varied up to 7 ns during the day in our ICP-MS. Similar observations on dead time uncertainty were reported previously (26–27).
Mass bias correction. External standardization was applied for mass bias correction and normalization of stable iron isotope ratios. Iron isotope ratios of standards (natural abundance) with iron concentrations (150–200 µg/L) similar to the diluted RBC samples were prepared and analyzed frequently using the sample-standard bracketing method. Minor changes of mass bias occurred during the day. Small day-to-day variations also occurred. Tests results from 7 human RBC samples of natural abundance suggested that the matrix effect was negligible, presumably due to the fact that all RBC samples were diluted 1500 times by volume.
Isobaric interference correction. Nickel is virtually always present in samples undergoing iron isotope analysis, and the isobaric interference of nickel on the 58Fe isotope requires correction. 58Fe has the lowest abundance (0.28%) among iron isotopes, whereas 58Ni is the most abundant (68.3%) nickel isotope; thus, nickel contamination during sample transfer or preparation may present a major spectral interference for 58Fe. Inaccurate measurement of 60Ni intensity may introduce a 1.5 times greater error for the 58Ni signal determination. This may not be a problem for most diluted RBC samples with relatively high iron concentration but very low nickel content. However, nickel contamination is likely to occur during sample preparation unless an extremely well-filtered clean room environment is utilized. Contamination during sample handling and transfer in the field setting due to the utilization of dirty containers and pipette tips is also possible. When nickel contamination is severe, even the ion exchange chromatography may not adequately separate nickel from iron. Assays of these contaminated samples may not proceed successfully with TIMS and the samples usually end up being discarded, although it is theoretically possible to correct for these difficulties on newer, but less widely available TIMS equipment. The ICP-MS analytical technique may overcome nickel contamination problems by mathematical correction of nickel interference. It is possible to obtain useful iron isotope data even if the intensity of the interference is greater than the analyte signal.
Correction of the 58Ni isotope signal was carried out in 2 ways. Standard nickel solutions of natural abundance with relatively high nickel concentrations (>50 µg/L) were utilized to define systematic bias-dependent 58Ni/60Ni values. They were analyzed in between samples and standards. The 58Ni/60Ni value changed from day to day (2.54–2.68) and fluctuated slightly (<0.04) during the day. The nickel isotope ratio obtained was applied to other samples for mathematical corrections of 58Ni signals overlapping 58Fe. This technique works well for most RBC samples with nickel concentrations < 0.7 mg/L (<0.5 µg/L in sample solutions) because the small variation of daily 58Ni/60Ni values will have little affect on 58Fe correction when 58Ni is much lower than 58Fe.
We used another technique to correct nickel interference using a nickel calibration curve. This curve was developed from a series of iron and nickel mixing standard solutions with the same iron concentration but a wide range of nickel concentrations. A consistent 58Fe/54Fe ratio after correction is expected regardless of nickel content, provided that the correction is accurate. In this case, a best-estimated 58Ni/60Ni value of 2.655 used for nickel interference correction yielded relatively consistent 58Fe/54Fe ratios throughout the entire nickel range (the horizontal solid curve in Fig. 1). This technique has a major advantage over that using single standard nickel solution for nickel correction. It may be used to obtain a more accurate 58Ni/60Ni ratio, which is essential for nickel interference correction when samples have relatively high nickel concentrations (>0.7 mg/L) and the error of 58Fe correction may propagate, as illustrated in Figure 1. It compensates for the fluctuation and precision limit (0.5% relative SD) our ICP-MS may produce. The machine virtually cannot differentiate 58Ni/60Ni ratios between 2.65 and 2.66 with analysis of a single nickel solution.
|
|
RESULTS |
|---|
|
TOPABSTRACTSUBJECTS AND METHODSRESULTSDISCUSSIONLITERATURE CITED |
|---|
|
4-min assay. Reproducibility of iron isotope ratio analysis was excellent during analysis of a single day, with an external relative SD < 0.4% in most cases. In general, analytical precision for TIMS analysis was <0.3%. The ratios from ICP-MS were converted to 57Fe/56Fe and 58Fe/56Fe using the expected constant 54Fe/56Fe value from TIMS and the results were compared with 57Fe/56Fe and 58Fe/56Fe from TIMS analysis (Fig. 2and 3). There was a good correlation between 57Fe/56Fe pairs and 58Fe/56Fe pairs from both machines using linear regression analysis, with R2 = 0.92 and P < 0.0001 for the 57Fe/56Fe pairs (Fig. 2) and R2 = 0.99 and P < 0.0001 for the 58Fe/56Fe pairs (Fig. 3). 57Fe/54Fe and 58Fe/54Fe ratios from both machines also showed good correlations, with R2 = 0.92 and 0.99, and P < 0.0001 for 57Fe/54Fe pairs and 58Fe/54Fe pairs, respectively. The 2 methods did not differ. No significant bias occurred in the Bland-Altman analysis of 57Fe/56Fe ratio pairs from the 2 machines (Fig. 4); all data are within a narrow range of 2% difference except for one outlier, which had a somewhat lower ratio from TIMS. There was no significant bias throughout the entire data range in the Bland-Altman analysis of 58Fe/56Fe ratio pairs; all paired data were within a narrow range of 1.5% difference (Fig. 5). These results suggest that iron isotope ratios from ICP-MS are consistent with those from TIMS, and the techniques we used for mass bias and nickel interference correction worked well.
|
|
|
|
|
|
DISCUSSION |
|---|
|
TOPABSTRACTSUBJECTS AND METHODSRESULTSDISCUSSIONLITERATURE CITED |
|---|
2 µL RBCs for iron isotope analysis with good precision. No lengthy anion exchange chromatography is required for sample pretreatment. Sample throughput with ICP-MS is much higher than with TIMS analysis. The advantages of using different types of MS for bioavailability studies using stable isotopes have always been a concern (28). Although TIMS has better accuracy and precision than most currently available ICP-MS, it is a specialized device available in very few nutrition laboratories and a challenge to sharing analysis with others, such as geologists who operate them. ICP-MS, on the other hand, has analytical advantages over TIMS in ionization efficiency, sample volume, sample throughput, sample preparation, and overall cost (28–32). Because ICP-MS suffices for most applications in precision and accuracy, it may easily become the method of choice (29). In the application of human nutrition research, some reduction in precision compared with TIMS is acceptable because most biological samples are relatively highly enriched. Furthermore, ICP-MS is able to account for nickel contamination, which is not currently readily available for TIMS. This may offset small differences in precision between the 2 methods. ICP-MS may have up to 5 times higher throughput than TIMS (29), which is important in human nutrition studies due to the large number of samples involved. ICP-MS is also among the fastest growing instruments available in the analytical industry, with several major manufactures competing in the market. For example, 325 high-resolution single-collector and 40+ multiple-collector ICP-MS instruments manufactured by a single large manufacturer are active in different institutions worldwide (Chuck Douthitt, ThermoElectron, personal communication). ICP-MS is used in many toxicology laboratories and the new generation of magnetic sector ICP-MS is rapidly become standard in many laboratories. Therefore, using the much more widely available ICP-MS for analysis in life sciences, at a precision similar if not identical to that of TIMS, would allow for substantial benefits in terms of sample throughput, time, and cost.
Our results show excellent agreement between TIMS and ICP-MS in terms of both iron isotope ratios and calculated iron absorption. Not only did the mean iron absorption not differ significantly, but the variance of the measurement was similar as well. The nonsignificant difference in means and variances of iron absorption results from both machines is of importance. Small errors in isotope ratio may be multiplied during the calculation of iron absorption. If using ICP-MS data had significantly increased the variability of measured iron absorption values, then studies using this methodology would require larger sample sizes or larger isotope doses than those in which samples were analyzed by TIMS. Such changes would have undermined the time and cost savings associated with ICP-MS analysis. It is reassuring, therefore, that the ICP-MS method did not lead to more variability in measured iron absorption and is as compatible with similar study designs (in terms of sample size and isotope dosing) as the TIMS method.
Two samples, Samples 15 and 23 (Table 3), gave TIMS 57Fe absorption values < 7%, much lower than those from ICP-MS (12–14%). Among them, one produced the outlier 57Fe/56Fe ratio as showed in Figure 4. The higher absorption rates from ICP-MS data were probably more reliable and more in keeping with the rest of the results. The TIMS result may have been erroneously low due to contamination during the iron exchange process, or mass fractionation during the lengthy and extreme heating in TIMS analysis.
Improvement of the standardization technique, data quality, and data comparability become a more intriguing issue for stable isotope nutritional studies using ICP-MS (28). In particular, problems such as systematic bias, matrix effect, standardization of data, and interference correction should be addressed and resolved (28–30). Tremendous efforts also have been made to improve the sample introduction system, including nebulizer, spray chamber, and more efficient solution desolvation devices. Many choices of sample introduction devices are commercially available for different analytical purposes. They have been designed for better ionization efficiency and signal stability, and less polyatomic interference. We are optimistic that high-resolution ICP-MS instrument sensitivity and analytical precision will reach a new level, which may greatly benefit elemental bioavailability studies in the future.
|
ACKNOWLEDGMENTS |
|---|
|
FOOTNOTES |
|---|
2 Financed in part with federal funds from the USDA/ARS under Cooperative Agreement number 58-6250-6-001. The contents of this publication do not necessarily reflect the views or policies of the USDA, nor does mention of trade names, commercial products, or organizations imply endorsement by the U.S. Government.
Manuscript received 21 December 2004. Initial review completed 7 February 2005. Revision accepted 11 April 2005.
|
LITERATURE CITED |
|---|
|
TOPABSTRACTSUBJECTS AND METHODSRESULTSDISCUSSIONLITERATURE CITED |
|---|
1. Herman S., Griffin I. J., Suwarti S., Ernawati F., Permaesih D., Pambudi D., Abrams S. A. Cofortification of iron-fortified flour with zinc sulfate, but not zinc oxide, decreases iron absorption in Indonesian children. Am. J. Clin. Nutr. 2002;76:813-817.
2. Zlotkin S. H. Another small step in the path to controlling micronutrient deficiencies...but we still have a long way to go. J. Pediatr. 2004;145:4-6.[Medline]
3. Mishaan A.M.A., Zavaleta N., Griffin I. J., Hilmers D. C., Hawthorne K. M., Abrams S. A. Bioavailability of iron and zinc from a multiple micronutrient-fortified beverage. J. Pediatr. 2004;145:26-31.[Medline]
4. Abrams S. A. New approaches to iron fortification: role of bioavailability studies. Am. J. Clin. Nutr. 2004;80:1104-1105.
5. Griffin I. J., Abrams S. A. Iron and breastfeeding. Pediatr. Clin. North Am. 2001;48:401-413.[Medline]
6. King J. C., Raynolds W. L., Margen S. Absorption of stable isotopes of iron, copper, and zinc during oral contraceptives use. Am. J. Clin. Nutr. 1978;31:1198-1203.
7. Janghorbani M., Young V. R. Use of stable isotopes to determine bioavailability of minerals in human diets using the method of fecal monitoring. Am. J. Clin. Nutr. 1980;33:2021-2030.
8. Turnlund J. R., Michel M. C., Keyes W. R., King J. C., Margen S. Use of enriched stable isotopes to determine zinc and iron absorption in elderly men. Am. J. Clin. Nutr. 1982;35:1033-1040.
9. Woodhead J. C., Drulis J. M., Rogers R. R., Ziegler E. E., Stumbo P. J., Janghorbani M., Ting B. T., Fomon S. J. Use of the stable isotope, 58Fe, for determining availability of nonheme iron in meals. Pediatr. Res. 1988;23:495-499.[Medline]
10. Whittaker P. G., Lind T., Williams J. G. Iron absorption during normal human pregnancy: a study using stable isotopes. Br. J. Nutr. 1991;65:457-463.[Medline]
11. Abrams S. A., O’Brien K. O., Wen J., Liang L. K., Stuff J. E. Absorption by 1-year-old children of an iron supplement given with cow milk or juice. Pediatr. Res. 1996;39:171-175.[Medline]
12. Abrams S. A., Wen J., Stuff J. E. Absorption of calcium, zinc, and iron from breast milk by five- to seven-month-old infants. Pediatr. Res. 1997;41:384-390.[Medline]
13. McDonald M. C., Abrams S. A., Schanler R. J. Iron absorption and red blood cell incorporation in premature infants fed an iron-fortified infant formula. Pediatr. Res. 1998;44:507-511.[Medline]
14. Domellof M., Lonnerdal B., Abrams S. A., Hernell O. Iron absorption in breast-fed infants: effects of age, iron status, iron supplements, and complementary foods. Am. J. Clin. Nutr. 2002;76:198-204.
15. Turnlund J. R. The use of stable isotopes in mineral nutrition research. J. Nutr. 1989;119:7-14.
16. Turnlund J. R. Bioavailability of dietary minerals to humans: the stable isotope approach. Crit. Rev. Food. Sci. Nutr. 1991;30:387-396.[Medline]
17. Griffin I. J., Abrams S. A. Methodological issues in stable isotope-based kinetic studies in children. Adv. Exp. Med. Biol. 2003;537:117-130.[Medline]
18. Abrams S. A. Using stable isotopes to assess mineral absorption and utilization by children. Am. J. Clin. Nutr. 1999;70:955-964.
19. Hamester M., Wiederin D., Wills J., Kerl W., Douthitt C. B. Strategies for isotope ratio measurements with a double focusing sector field ICP-MS. J. Anal. Chem. 1999;364:495-497.
20. Ingle C., Langford N., Harvey L., Dainty J. R., Armah C., Fairweather-Tait S., Sharp B., Crews H., Rose M., et al. Development of a high-resolution ICP-MS method, suitable for the measurement of iron and iron isotope ratios in acid digests of faecal samples from a human nutrition study. J. Anal. At. Spectrom. 2002;17:1498-1501.
21. Weyer S., Schwieters J. B. High precision Fe isotope measurements with high mass resolution MC-ICPMS. Int. J. Mass Spectrom. 2003;226:355-368.
22. Vanhaecke F., Moens L. Overcoming spectral overlap in isotopic analysis via single- and multi-collector ICP-mass spectrometry. Anal. Bioanal. Chem. 2004;378:232-240.[Medline]
23. Chen Z., Griffin I. J., Kriseman Y. L., Liang L. K., Abrams S. A. Inductively coupled plasma mass spectrometric analysis of calcium isotopes in human serum: a low-sample-volume acid-equilibration method. Clin. Chem. 2003;49:2050-2055.
24. Coplen T. B., Böhlke J. K., De Bièvre P., Ding T., Holden N. E., Hopple J. A., Krouse H. R., Lamberty A., Peiser H. S., et al. Isotope-abundance variations of selected elements (IUPAC Technical Report). Pure Appl. Chem. 2002;74:1987-2017.
25. Bland J. M., Altman D. G. Statistical methods for assessing agreement between two methods of clinical measurement. Lancet. 1986;1:307-310.[Medline]
26. Vanhaecke F., de Wannemacker G., Moens L., Dams R., Latkoczy C., Prohaska T., Stingeder D. Dependence of detector dead time on analyte mass number in inductively coupled plasma mass spectrometry. J. Anal. At. Spectrom. 1998;13:567-571.
27. Stürup S. Application of HR-ICP-MS for the simultaneous measurement of zinc isotope ratios and total zinc content in human samples. J. Anal. At. Spectrom. 2000;15:315-321.
28. Walczyk T., Coward A., Schoeller D. A., Preston T., Dainty J., Turnlund J. R., Iyengar V. Stable isotope techniques in human nutrition research: concerted action is needed. Food Nutr. Bull. 2002;23(suppl.):69-75.[Medline]
29. Turnlund J. R., Keyes W. R. Isotope ratios of trace elements in samples from human nutrition studies determined by TIMS and ICP-MS: precision and accuracy compared. Food Nutr. Bull. 2002;23(suppl.):129-132.[Medline]
30. Ingle C. P., Sharp B. L., Horstwood M.S.A., Parrish R. R., Lewis D. J. Instrument response functions, mass bias and matrix effects in isotope ratio measurements and semi-quantitative analysis by single and multi-collector ICP-MS. J. Anal. At. Spectrom. 2003;18:219-229.
31. Ingle C. P., Langford N., Harvey L. J., Dainty J. R., Turner P. J., Sharp B. L., Lewis D. J. Comparison of three different instrumental approaches to the determination of iron and zinc isotope ratios in clinical samples. J. Anal. At. Spectrom. 2004;19:404-406.
32. Walczyk T. TIMS versus multicollector-ICP-MS: coexistence or struggle for survival?. Anal. Bioanal. Chem. 2004;378:229-331.[Medline]
This article has been cited by other articles:
|
|
 |
|
  M. F. Lynch, I. J. Griffin, K. M. Hawthorne, Z. Chen, M. G. Hamzo, and S. A. Abrams Iron Absorption Is More Closely Related to Iron Status Than to Daily Iron Intake in 12- to 48-Mo-Old Children J. Nutr., January 1, 2007; 137(1): 88 - 92. [Abstract] [Full Text] [PDF]
|
|
| |||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
