criteria that included the following: subjects had a wide range of calcium intakes, as variability in retention increases at higher intakes; the balance studies were initiated at least 7 days after starting the diet in order for subjects to approach a steady state, as observed by Dawson-Hughes et al. (1988); and, where possible, the adult balance studies included were only for subjects who were consuming calcium at their usual intakes, unless otherwise indicated. By selecting studies conducted on such subjects, the 1997 committee concluded that it obviated the concern about whether the bone remodeling transient (i.e., the temporary alteration in the balance between bone formation and bone resorption) might introduce bias in the calcium retentions observed (IOM, 1997). Such selection was not possible in studies in children who were randomized to one of two calcium intakes. However, in children, the impact of the bone remodeling transient related to changing intake is overshadowed by their rapid and constantly changing rates of calcium accretion (i.e., their modeling and remodeling rates are not in steady state, even without an intake change).
For the 1997 DRI development (IOM, 1997), the non-linear regression model describing the relationship between calcium intake and retention was solved to obtain a predetermined desirable calcium retention that was specific for each age group. According to the report, the major limitation of the data available was that bone mineral accretion during growth had not yet been studied over a wide range of calcium intakes. Overall, the committee expressed concern about the uncertainties in the methods inherent in balance studies.
Specifics about calcium balance studies that relate to DRI development are provided in Chapter 4, but, as background the recent work of Hunt and Johnson (2007) offers some remedy for the uncertainties surrounding the precision of balance studies. Hunt and Johnson (2007) examined data from 155 subjects—men and women between the ages of 20 and 75 years—who took part in 19 feeding studies conducted at one site (Grand Forks Human Nutrition Research Unit) between 1976 and 1995 in a metabolic unit under carefully controlled conditions.
In their overall analysis, the relationship between intake and output was examined by fitting random coefficient models. Rather than model calcium retention compared with calcium intake by using the Jackman et al. (1997) model, as was done in the 1997 DRI report (IOM, 1997), Hunt and Johnson (2007) modeled output rather than retention to avoid confounding in the precision of estimates that would be caused by including intake as a component of the dependent variable. In the Hunt and Johnson (2007) analysis, the data summary did not show non-linearity and therefore did not justify the use of a more complex non-linear model. The authors noted that the coefficients of the 1997 approach appeared to be greatly influenced by data points above the 99th percentile of daily