Vitamin D status is known to be poor in obese individuals; there is no consensus as to the reason. Cross-sectional study of the relation between serum 25-hydroxyvitamin D (25(OH)D) concentration and body size in the baseline data from unsupplemented adults entering two study cohorts in our research unit, N = 686. Regression analyses of body size variables against serum 25(OH)D concentration, using both linear and hyperbolic models. The fit to a hyperbolic model of 25(OH)D against body weight completely removed the obesity-related component of inter-individual variability in serum 25(OH)D concentration. The hyperbolic fit using total body weight was significantly better than any linear model, and specifically better than any using BMI. Dilution of ingested or cutaneously synthesized vitamin D in the large fat mass of obese patients fully explains their typically low vitamin D status. There is no evidence for sequestration of supplemental or endogenous cholecalciferol. Vitamin D replacement therapy needs to be adjusted for body size if desired serum 25(OH)D concentrations are to be achieved.
Obesity has been long associated with low vitamin D status (1,2,3,4,5,6,7,8,9,10,11,12,13,14,15). Many epidemiological studies, including National Health and Nutrition Examination Survey (NHANES) (12) and Framingham (14), have shown increasing prevalence of hypovitaminosis D with increasing BMI. A negative correlation of vitamin D status with anthropometric variables is observed across all ages—from adolescents to postmenopausal women, and blacks as well as whites. The cause of this tendency to low 25-hydroxyvitamin D (25(OH)D) concentrations in obese individuals is poorly understood. Various mechanisms have been proposed. Several are preeminently plausible, such as: (i) limited skin exposure to sun because of decreased mobility and the social stigma of obesity; (ii) impaired ability of skin to convert 7-dehydrocholesterol to vitamin D3; and (iii) negative feedback from elevated 1,25(OH)2D and parathyroid hormone level on hepatic synthesis of 25(OH)D (16).
However the mechanism most often invoked is sequestration of cholecalciferol in body tissues, particularly fat. Although never precisely defined, sequestration implies that vitamin D, whether entering the body by mouth or via cutaneous synthesis, is relatively tightly bound in tissue depots and thus not appropriately released into the general circulation to support serum 25(OH)D concentrations. Sequestration was originally invoked by Wortsman et al. (7)., who found that obese individuals, despite greater body surface area, exhibited a much smaller response to a standard dose of UV-B radiation than did normal weight individuals. How vitamin D might be held tightly in fat (i.e., “sequestered”) is not stated.
Unraveling the mechanism beyond low vitamin D status in obesity has important therapeutic implications—both in deciding on appropriate vitamin D replacement doses for obese individuals and in evaluating potential effects of treatment of vitamin D inadequacy. In this paper, we approach the issue of low vitamin D status in obesity (and especially sequestration) by testing a dilutional model on serum 25(OH)D concentration values in a cohort spanning a range of body weights running from 41 to 166 kg and of BMI, from 16.5 to 61.2 kg/m2.
Methods and Procedures
The individuals whose baseline data are the subject of this report were derived from (i) a population-based cohort of small-town, Midwestern women, described in detail in an earlier publication (17), and (ii) a previously undescribed adult cohort entered into an obesity study. The first group excluded women, with morbid obesity, while the second group excluded normal weight individuals. All participants were recruited from Omaha and surrounding communities. For both groups, those ingesting, by history, vitamin D supplements of 400 IU/day or higher or receiving bone active agents such as bisphosphonates were excluded from this analysis. The total sample size was 686, and pertinent personal data are set forth in Table 1. For the population-based cohort, sample size was 651; mean (s.d.) age was 66.8 (7.3) years, and BMI 26.1 (7.3) kg/m2. The included members of the obesity cohort consisted of 32 women and 3 men, all with BMI values above 30 kg/m2. Their mean age was 45 (10.2), and mean BMI, 36.9 (5.8). Serum 25(OH)D values were seasonally adjusted to the empirical null point of the annual sine wave curve. All were in good general health (except for the obesity in the second cohort), and none was taking medications known to influence vitamin D status or metabolism. All gave written consent to the investigation and the respective projects had been approved by the Creighton University institutional review board.
Table 1. Pertinent subject characteristics
The design of this investigation was a cross-sectional analysis of data on body size and serum 25(OH)D obtained on entry into their respective studies.
Serum 25(OH)D in the first cohort was measured by the Immunodiagnostic Systems, Ltd. (IDS) method (radioimmunoassay; Nichols/Quest Diagnostics, San Clemente, CA), and the second by the DiaSorin Liaison system (chemiluminescence; DiaSorin, Stillwater, MN). All assays were performed in the investigators' laboratory and the two methods had been calibrated against one another. For this analysis, the DiaSorin results were adjusted to match the IDS results. Coefficient of variation within-assay for both was ∼7%. In addition, our laboratory participates in the International Quality Assessment Scheme for Vitamin D metabolites (DEQAS) system (18), which provides international, inter-laboratory calibration. Body weight was measured on a beam balance scale and body fat was measured by dual energy X-ray absorptiometry on a Hologic QDR 4500 instrument (Hologic, Waltham MA), calibrated weekly against whole body standards.
The hypothesis to be tested in this study was that serum 25(OH)D was an inverse function of body size. We approached this testing in two ways: (i) stepwise multiple linear regression using SPSS (SPSS, Chicago, IL), treating 25(OH)D values as the dependent variable, and various body size variables (height, weight, BMI, and body fat mass) as predictor variables; and (ii) fitting the most strongly related size variable to 25(OH)D with a hyperbolic model, using the curve-fitting function of SigmaPlot 11.1 (Systat Software, San Jose, CA). The specific equation fitted was
In this model y = serum 25(OH)D in nmol/l; x = one of the several body size variables; a = the calculated zero size value of y in nmol/l; and b = a constant that localizes the point of maximum curvature of the fitted hyperbola relative to the X-Y coordinates.
A hyperbolic model was selected for this analysis since it is the mathematical representation of the relationship between concentration and volume. Even so, such a model assumes a constant content; thus the hyperbola applied to real data can only be an approximation since differences in body size are likely to be associated with different, though partially offsetting, effects on daily cutaneous inputs of cholecalciferol.
In addition to the curve fitting by SigmaPlot, standard descriptive statistics and various linear regression models were implemented using SPSS for Windows (SPSS). In the fitting of serum 25(OH)D to various models, we specifically evaluated the residuals for each fit to see if variability could be further reduced by other body size indexes.
Subject characteristics related to this analysis are set forth in Table 1. Note especially that the s.d. for height was less than 4% of the mean, while for weight, it was nearly 22%. Thus by far the largest component of variation in body size was weight, not stature.
In multiple linear modeling, while all the body size variables (except height) were significantly inversely correlated with 25(OH)D, the best fit was obtained by a combination of body weight and body fat mass, which together explained 10.4% of the total variance. Weight entered the equation first (R2 = 0.088) and fat mass was the only other size variable that played a significant role (addition to R2 = 0.015). BMI, by itself, was also significantly inversely related to serum 25(OH)D, but the fit was not as good as with body weight or body weight plus fat mass.
The fact that fat mass was significantly correlated with the residuals of the regression of 25(OH)D on body weight (a variable that itself already includes fat mass) indicates that a linear model was not optimal in explaining the body size-related variation in 25(OH)D. This defect directly supports use of a hyperbolic model, which more precisely describes volumetric dilution.
Because weight was the strongest univariate predictor, it was used as the body size variable in fitting the data to a hyperbola. Figure 1 shows the resulting plot. Three features of the graph stand out: (i) there is a visually evident inverse relationship, with lower 25(OH)D values at high body weights and vice versa; (ii) there is a small upward-directed concavity, as would be expected for data exhibiting the predicted hyperbolic relationship; and (iii) there is a very large spread of values at any given body size value.
The R2 value indicated that the hyperbolic fit explained 13.0 % of the total variance. Moreover, it was significantly greater than for the linear regression models using either weight alone or weight plus fat. In order to determine whether fat mass explained some of the remaining spread around the regression line in the hyperbolic model, the residuals from that regression were further regressed against measured fat mass. However, there was no hint of a correlation between the 25(OH)D residuals and body fat mass, indicating that fitting to a hyperbola captured essentially all the body size-related variance in the data.
In this study we show that a volumetric dilutional model accounts for essentially all the variability in serum 25(OH)D concentration attributable to obesity. The residual s.d. around the regression line in Figure 1 does not differ from the s.d. of serum 25(OH)D concentration in normal weight individuals. To the extent that several of the suggested mechanisms for low vitamin D status in obesity (e.g., reduced surface-to-volume ratio, less outdoor time, etc.) may be operative, the applicable factors are all effectively captured by body weight. The one remaining “explanation ” (sequestration and inappropriate storage of vitamin D in adipose tissue) (7) is not necessary to explain the low vitamin D status of obesity, as once the values in obesity are adjusted for body size, there is no longer a difference in 25(OH)D concentrations between normal and obese individuals. Further, direct measurement of fat content of cholecalciferol (19,20) lead to the same conclusion.
Despite a great paucity of human data, there can be little doubt that vitamin D is stored in body fat. This point has been more than amply demonstrated in animal studies (21,22); and in a few tracer-labeled studies of human fat, the isotopic label has been found concentrated in fat, relative to lean tissue (23). The issue, vis-à-vis sequestration, is not whether cholecalciferol is stored in fat, but whether it is somehow bound there so that, despite an ostensibly “adequate ” input, the obese person cannot readily access his or her own reserves of the vitamin.
The fact that weight is a slightly better predictor of serum 25(OH)D concentration than is fat mass is not surprising, since 25(OH)D, the dependent variable in this relationship, is more widely distributed in nonadipose tissue than is cholecalciferol (24,25,26).
Although the precise equilibrium between the cholecalciferol contents of serum and fat has not been explored, the most likely basis for a relationship is a simple passive diffusion model involving cholecalciferol bound to D-binding protein (DBP) in serum on the one hand, and cholecalciferol dissolved in the fat globules of adipose tissue on the other. This arrangement is illustrated in Figure 2. Cholecalciferol normally enters the body either from cutaneous synthesis or from oral intake and is bound to DBP in blood. Figure 2 postulates an equilibrium between cholecalciferol bound to DBP and cholecalciferol in body fat depots, with the relative affinities for vitamin D being ∼1:12, respectively (20). The 25-hydroxylation of the compound to 25(OH)D represents, in a sense, a competitive pathway, pulling cholecalciferol out of fat by hepatic uptake of a fraction of the cholecalciferol bound to DBP.
In a dilutional model, such as we propose here, any given increment of cholecalciferol would be distributed not simply in the serum, bound to DBP, but in the totality of the body fat. If that fat mass is, for example, twice as great in an obese individual as in a normal weight individual, then the induced rise in serum cholecalciferol (and ultimately in 25(OH)D) of any given input, would be predicted to be roughly half as great. If, on the other hand, there were some active process whereby cholecalciferol was aggressively taken up by body fat and tenaciously held onto (as suggested by the notion “sequestration”), then one would expect to find disproportionately high levels of cholecalciferol (and by some reckoning potentially toxic levels) in samples of fat from obese individuals.
In the one study of which we are aware, Blum et al. (20), evaluating fat content as a function of serum cholecalciferol concentration in patients undergoing bariatric surgery, found a mean content of vitamin D in adipose tissue amounting to ∼1,500 IU/kg fat. This value was, in turn, associated with a mean serum cholecalciferol level of 7.8 nmol/l, with the fat content values positively correlated with the serum levels at an approximate slope of 12:0 (fat on serum). These data do not suggest that vitamin D is being stored in fat in concentrations greater than would be predicted from the serum level. This conclusion is buttressed by a recent study using 50,000 IU cholecalciferol per week (19). At the end of 12 weeks' treatment, while serum 25(OH)D concentration had risen appreciably, subcutaneous fat biopsies revealed that only a minor fraction of the 600,000 IU administered had been stored in fat (the bulk being hydroxylated and consumed metabolically).
It has been shown for normal weight adults that total body utilization of vitamin D needed to sustain a serum level of 25(OH)D of 32 ng/ml (80 nmol/l) is on the order of 4,000 IU/day (27), and for 40 ng/ml (100 nmol/l), 5,000 IU/day. It is clear from the data of Blum et al. (20) that a kilogram of fat in obese patients contains appreciably less than one day's supply of vitamin D for a healthy individual. It can be calculated from these data that a daily input in the range of 70–80 IU/kg body weight will produce mean serum 25(OH)D concentrations of 30–40 ng/ml (75–100 nmol/l). (If a lower 25(OH)D concentration is targeted (e.g., 20 ng/ml), 30–40 IU/kg body weight will suffice.)
As already noted, most of the reports of lower vitamin D status in obesity have used BMI or percent body fat as the index variables for obesity (e.g., ref. (3,5,6,7,8,9,10,11,15,28)). For mechanical reasons this approach often produces a statistically significant linear relationship (see below), despite the fact that it is conceptually incorrect. Both BMI and percent body fat, as is well understood, measure fatness, not fat mass. A simple thought experiment illustrates this point: consider two women, one 1.5 m, weighing 67.5 kg and the other 1.8 m, weighing 97.2 kg. Both have a BMI of exactly 30 kg/m2. But clearly the taller woman has better than 40% more tissue (fat and lean) than the shorter woman. As a result the identical oral dose of vitamin D would be expected to produce a substantially smaller rise in serum 25(OH)D in the larger woman. The reason BMI often correlates inversely with vitamin D status is that, at a population level, weight is a much larger source of inter-individual variation in body size than is stature, and thus BMI inevitably reflects some (but not all) of that size variation. This is illustrated dramatically in the body size variables in this study cohort, in whom the coefficient of variation for height was less than 4%, while the coefficient of variation for weight was more than five times as great (Table 1).
This point is not just a trivial academic distinction. Studies involving multivariate analyses and attempting to adjust for the obesity factor by using BMI or percent fat, will leave an unrecognized body size component in the residual variability. Thus, unless the hypothesis concerned relates specifically to fatness, rather than to fat mass, BMI or percent body fat should not be used as the body size variable when making such adjustments.
We have confirmed an inverse association between vitamin D levels and anthropometric measures of body size. Body weight and body fat are inversely correlated with 25(OH)D levels across the spectrum of body weight ranging from normal to obese. This inverse association is related to the greater volume of distribution for both vitamin D and 25(OH)D in tissue mass. We suggest that simple volumetric dilution is the most parsimonious explanation for the low vitamin D status in obesity. A hyperbolic model (which is the mathematical expression of dilution), best explains the lower 25(OH)D values in obesity. When serum 25(OH)D values are suitably adjusted for body weight, the difference between obese and normal individuals disappears, and there is no longer an effect to be “explained ” by hypothetical mechanisms such as sequestration. In obese individuals, vitamin D dosing for treatment of deficiency should be based on body weight, i.e., “one size does not fit all”.
Work reported in this paper was supported by DHHS Grant AG14683 and by Health Future Foundation funds.