Dose-Response Relationships Between Energy Availability and Bone Turnover in Young Exercising Women^†

Abstract

To help refine nutritional guidelines for military servicewomen, we assessed bone turnover after manipulating the energy availability of 29 young women. Bone formation was impaired by less severe restrictions than that which increased bone resorption. Military servicewomen and others may need to improve their nutrition to avoid these effects.

Introduction: We determined the dose-response relationship between energy availability (defined as dietary energy intake minus exercise energy expenditure) and selected markers of bone turnover in 29 regularly menstruating, habitually sedentary, young women of normal body composition.

Materials and Methods: For 5 days in the early follicular phase of two menstrual cycles separated by at least 2 months, subjects expended 15 kcal/kgLBM/day in supervised exercise at 70% of aerobic capacity and consumed controlled amounts of a clinical dietary product in balanced (45 kcal/kgLBM/day) and one of three restricted (either 10, 20, or 30 kcal/kgLBM/day) energy availability treatments in random order. Blood was sampled at 10-minute intervals, and urine was collected for 24 h. Samples were assayed for plasma osteocalcin (OC), serum type I procollagen carboxy-terminal propeptide (PICP), and urinary N-telopeptide (NTX).

Results: NTX concentrations (p < 0.01) and indices of bone resorption/formation uncoupling (Z_NTX-OC and Z_NTX-PICP; both p < 10⁻⁴) were increased by the 10 kcal/kgLBM/day treatment. OC and PICP concentrations were suppressed by all restricted energy availability treatments (all p < 0.05). PICP declined linearly (p < 10⁻⁶) with energy availability, whereas most of the suppression of OC occurred abruptly between 20 and 30 kcal/kgLBM/day (p < 0.05).

Conclusions: These dose-response relationships closely resembled those of particular reproductive and metabolic hormones found in the same experiment and reported previously: similar relationships were observed for NTX and estradiol; for PICP and insulin; and for OC, triiodothyronine (T₃), and insulin-like growth factor (IGF)-I. The uncoupling of bone resorption and formation by severely restricted energy availability, if left to continue, may lead to irreversible reductions in BMD, and the suppression of bone formation by less severe restrictions may prevent young women from achieving their genetic potential for peak bone mass. More prolonged experiments are needed to determine the dose-response relationships between chronic restrictions of energy availability and bone turnover.

INTRODUCTION

THE U.S. DEPARTMENT of Defense has an immediate need to reduce the prevalence of stress fractures in military servicewomen. Over a period of 20 years, many studies have found the prevalence of stress fractures in U.S. military trainees to be as much as 11 times higher in servicewomen than in servicemen,¹ and 16% of active duty Army servicewomen report having been diagnosed with a stress fracture.² Participation in weight control programs, running exercise, underweight, and amenorrhea are all associated with the reporting of stress fractures by servicewomen.² It is well known that malnutrition and sex steroid deficiency interrupt bone mineral accrual and contribute to early bone loss,^{3, 4} and that young civilian women with anorexia nervosa and exercise-associated amenorrhea also display deficits in bone mineral and an increased prevalence of fractures.^{5, 6} Because of the established relationship between nutrition, exercise, and fractures in anorexia and amenorrhea, the U.S. military wanted to investigate how nutritional guidelines for military servicewomen might be refined to better protect their skeletal health while improving their fitness for physically demanding military missions.

In mammals, reproductive function depends on the cellular availability of oxidizable metabolic fuels, which can be impaired by dietary restriction, pharmacologic inhibitors of oxidative metabolism, insulin administration, thermogenesis during cold exposure, and physical activity.⁷ Ovarian steroidogenesis critically depends on the pulsatile release of luteinizing hormone (LH) from the pituitary. Through short-term, randomized, controlled experiments, we have shown that LH pulsatility is disrupted in young women, not by the stress of exercise, but rather by low energy availability (operationally defined as dietary energy intake minus exercise energy expenditure), regardless of whether the low energy availability is caused by dietary restriction or exercise energy expenditure.⁸ Specifically, we have shown that LH pulsatility is disrupted below a threshold of energy availability at ∼30 kcal/kgLBM/day.⁹ Not only does dietary supplementation prevent the disruption of LH pulsatility in exercising women, but dietary supplementation alone, without any moderation of the exercise regimen, restores menstrual cycles in monkeys whose amenorrhea was induced by exercise training.¹⁰

Bone tissue is continuously turned over at millions of local remodeling sites through processes of bone resorption by osteoclasts followed by bone formation by osteoblasts. In adults, the rates of these processes are usually coupled so that little net change in BMD accumulates. For many years, the low BMDs observed in anorexia nervosa patients and amenorrheic athletes had been attributed to their chronic hypoestrogenism, because the principal role of estrogen in bone turnover is to suppress osteoclast activity. Unfortunately, however, the low BMDs in these women have not been fully reversed by either estrogen therapy or the restoration of menses.⁶,^11-13 Therefore, investigators have begun to suspect that chronic undernutrition may also act through an estrogen-independent mechanism to impair bone formation. This speculation has drawn attention to a broad spectrum of metabolic hormones also disrupted by low energy availability⁹ that influence bone formation.¹⁴ Such an uncoupling of bone turnover that decreases formation while increasing resorption may cause irreversible reductions in BMD.¹⁵ Therefore, it is important to better understand the influence of energy availability on bone turnover.

Others have shown that a few days of complete fasting reduces concentrations of bone formation markers and increases mineral dissolution in young women.^{16, 17} The two purposes of this experiment were to determine the physiological dose-response relationships between energy availability and certain markers of bone turnover in young women and to gain insight into these relationships by comparing them to dose-response relationships that we previously reported between energy availability and certain reproductive and metabolic hormones in the same experiment.⁹

In a randomized, repeated-measures, prospective cohort experiment, we controlled the diet and exercise of regularly menstruating, habitually sedentary, young women for 5 days in the early follicular phase of two menstrual cycles.⁹ We selected plasma osteocalcin (OC) and serum type I procollagen carboxy-terminal propeptide (PICP) as markers of bone formation and urinary N-terminal telopeptide (NTX) as a marker of bone resorption. In the formation of bone matrix, PICP and amino-terminal (PINP) ends are cleaved off of type-I procollagen molecules. These find their way into the bloodstream and are considered to be quantitative measures of newly formed type I collagen. In constant molar ratios with collagen and hypoxyapatite in bone, OC appears in bone with the onset of matrix mineralization and functions in calcium binding. A small proportion finds its way instead into the bloodstream. During bone resorption, N-terminal (NTX) and C-terminal (CTX) telopeptides of type I collagen are released into the circulation and are specific for bone resorption. In regularly menstruating, habitually sedentary, young women, an energy availability of 45 kcal/kgLBM/day causes 24-h energy intake to equal (i.e., balance) 24-h energy expenditure.^{8, 9} This report describes the incremental effects of balanced (45 kcal/kgLBM/day) and three restricted (10, 20, and 30 kcal/kgLBM/day) energy availability treatments on OC, PICP, and NTX.

MATERIALS AND METHODS

Subject selection

Healthy, young, regularly menstruating women were recruited. All volunteers signed consent forms and received a full verbal and written description of the nature of the experiment, of its associated risks and benefits, and of their ability to withdraw from the experiment at any time. The qualification criteria for screening these volunteers for inclusion in the experiment have been published previously.⁹ The protocol for this experiment was approved by the Institutional Review Boards of Ohio University, The Ohio State University, and the Surgeon General of the Department of the Army.

Summary information about the demographic characteristics of the subjects is shown in Table 1. Specific demographic information about the subjects assigned to the three restricted energy availability treatment groups has been published previously.⁹ One-way ANOVA detected no physiologically or statistically significant differences in age, age of menarche, gynecological age, menstrual cycle length, luteal phase length, any parameter of body size or composition, maximal aerobic capacity, or habitual dietary intake between the women assigned to the three restricted energy availability treatments (all p > 0.08). The normalized habitual dietary energy intake of the subjects, 44.2 ± 1.1 kcal/kgLBM/day, was also indistinguishable from the balanced energy availability treatment (p > 0.4).

Table Table 1.. Demographic Characteristics of the Subjects

Experiment

Design:

In this experimental design (Fig. 1), subjects participated twice: once being administered a balanced and once being administered one of three different restricted energy availability (EA) treatments in random order. All subjects performed 15 kcal/kgLBM (63 kJ/kgLBM) of 70% VO_2max exercise energy expenditure (EEE) each day for 5 days while their daily dietary energy intake (CDI) was controlled. In one group, dietary intake was controlled at CDI = 60 and 45 kcal/kgLBM/day (251 and 188 kJ/kgLBM/day) in separate treatments to provide balanced and restricted energy availabilities (EA = CDI − EEE) of 45 and 30 kcal/kgLBM/day (188 and 125 kJ/kgLBM/day), respectively. In a second group, dietary energy intake was controlled at CDI = 60 and 35 kcal/kgLBM/day (251 and 146 kJ/kgLBM/day) for energy availabilities of 45 and 20 kcal/kgLBM/day (188 and 84 kJ/kgLBM/day). In the third group, dietary energy intake was controlled at CDI = 60 and 25 kcal/kgLBM (251 and 105 kJ/kgLBM/day) for energy availabilities of 45 and 10 kcal/kgLBM/day (188 and 42 kJ/kgLBM/day). Treatments were administered at intervals of at least 2 months to allow time for full recovery from blood sampling.

Figure FIG. 1.

Experimental design.

Protocol:

The 9-day experiment began on the second to fifth day of the menstrual cycle. Subjects provided a urine and blood sample between 7:30 a.m. and 8:30 a.m. on the 3 pretreatment days and on the 5 treatment days. Immediately after treatments had been completed on the eighth day, subjects were driven to the General Clinical Research Center (GCRC) at The Ohio State University Hospital where blood samples were drawn through a venous catheter at 10-minute intervals for 24 h, extending into the ninth day. All urine voided during the 24 h was also collected.

Energy expenditure:

During the pretreatment and treatment days, each subject wore an accelerometric physical activity monitor (Tritrac; Hemokinetics) during all waking hours, except while bathing, to estimate 24-h energy expenditure (24EE). The total energy expenditure for each 24EE was calculated on each treatment day.

Controlled energy expenditure (CEE) was defined as the total energy expended during exercise as measured by indirect calorimetry, which was administered as the sum of the intended exercise energy expenditure (EEE = 15 kcal/kgLBM/day) plus the portion of the each subject's total 24EE occurring during the exercise time period as estimated by the subject's physical activity monitor records on the pretreatment days. Twenty-four-hour energy balance (EB) was calculated as the difference between the 24-h CDI and total 24EE. All these quantities were normalized by lean body mass to control the energy available to actively metabolizing tissue, regardless of individual differences in body composition.

Control of treatments

Diet:

A commercially available clinical dietary product (Ross Laboratories' Ensure Plus) was used to set energy intake for the selected levels of energy availability. This product is composed of 28% fat, 15% protein, and 57% carbohydrate. Subjects were also provided with a daily multivitamin and mineral tablet containing both vitamin K and calcium. A daily qualitative urinary dipstick assay for the ketone aceto-acetate was used as an indicator of noncompliance. Meals were administered at standardized times each day.

Exercise:

For the exercise treatment, subjects walked up a grade on a motorized treadmill ergometer under continuous supervision in 30- to 40-minute sessions separated by 10-minute rest periods. Exercise intensity was controlled by setting treadmill speed and grade to elicit 70% of each individual's maximal oxygen consumption. Because the energy cost of exercise per liter of oxygen consumed depends on substrate use, the total duration of each individual's daily exercise was adjusted according to the individual's rate of exercise energy expenditure as indicated by the individual's oxygen uptake and respiratory exchange ratio during each exercise session.

Effectiveness of treatment administration:

The balanced and three restricted energy availability treatments actually administered to the subjects are described in Table 2. By design, there were no differences between the exercise regimens (%VO_2max, CEE, and EEE) in any of the energy availability treatments. In contrast and also by design, the controlled dietary energy intakes, and thereby, the energy availabilities and 24-h EBs of the subjects during the three restricted energy availability treatments were extremely different from those in the balanced energy availability treatment (all p < 10⁻⁷) and from one other, whereas the 24-h EB was indistinguishable from zero (p = 0.8) during the balanced energy availability treatment.

Table Table 2.. The Paired Balanced and Three Restricted Energy Availability Treatments

Experimental data collection

Plasma OC and serum PICP were measured in a baseline blood sample (pooled from days 1-3) and in a sample collected on the morning of the ninth day. Urinary NTX and creatinine were measured in the 24-h urine sample.

Assays:

All assays were done in duplicate, and all samples from each individual subject were run in a single assay. Plasma OC was measured by RIA (Diagnostic Systems Laboratories) with intra- and interassay CVs of 5% and 9%, respectively, at 8 ng/ml. Serum PICP was measured by RIA (Orion Diagnostica) with intra- and interassay CVs of 4% and 7%, respectively, at 183 ng/ml. Urinary NTX was measured by ELISA (Ostex International) with intra- and interassay CVs of 13% and 2%, respectively, at 223 nM BCE. For normalization of urinary NTX, urinary creatinine was measured by an enzymatic assay (Sigma Diagnostics) with intra- and interassay CVs of 3% and 2%, respectively, at 4.5 mM. Plasma and serum concentrations were corrected for differences in plasma volume between the baseline days and ninth day of the protocol.¹⁸

Data analysis

Bone marker analysis:

For each subject, baseline concentrations of PICP and OC were subtracted from the same subject's concentrations on the morning of the ninth day of treatment as an estimate of the subject's individual response to treatment. The difference between her responses to the restricted and balanced energy availability treatments was calculated as the effect of energy availability. In the urine collected in the GCRC, the 24-h pooled average of each subject's urinary NTX concentration was normalized by the 24-h pooled average of her urinary creatinine concentration. The difference between the normalized urinary NTX concentrations between the restricted and balanced energy availability treatments was calculated as the effect of energy availability. Z scores for bone marker measurements in each subject after restricted energy availability treatments (Z_Ri = [X_Ri − Xbar_B]/SD_B) were determined by their location relative to the Z distribution of measurements made in all the subjects after the balanced energy availability treatment (Z_Bi = [X_Bi − Xbar_B]/SD_B). Two indices of the uncoupling of bone formation and resorption were calculated as the differences between the Z scores of (1) NTX and OC and (2) NTX and PICP after the balanced and restricted energy availability treatments.

Statistical analysis:

In this laboratory, all data sets are routinely tested for non-normality, heteroscedasticity, and outliers before statistical hypothesis tests are performed. Outliers detected are rejected, and non-normal data sets are transformed as necessary. In this experiment, one subject administered 10 kcal/kgLBM/day displayed PICP and OC responses that were not only 3.5 and 2.5 SD, respectively, higher than the cluster of responses of the other subjects, but they were also of the opposite sign. Her baseline PICP concentration was also 4.5 SD higher than those of all other subjects. Therefore, all her data were excluded from further statistical analysis. Another subject administered the same treatment displayed an OC response 2.5 SD higher and also of the opposite sign from the OC responses of the other subjects receiving the same treatment. Because her other data were similar to those of the other subjects, only her OC response data were excluded from further analysis. No data sets indicated a degree of non-normality warranting transformation before analysis.

One-way ANOVA was used to compare the demographic characteristics of the subjects who were administered the three restricted energy availability treatments. Single-sample Student's t-tests were used to quantify dose-dependent effects of low energy availability at 10, 20, and 30 kcal/kgLBM/day on the bone markers and uncoupling indices. Repeated measures, one-way ANOVA with two-sample posthoc least significant difference (LSD) tests were performed to compare these restricted energy availability treatment effects. All single-sample and two-sample tests were single-sided, because the direction of interest in the outcome variables was known in advance. Paired data sets from N = 10, 10, and 9 women at a balanced energy availability of 45 kcal/kgLBM/day and at restricted energy availability treatments of 10, 20, or 30 kcal/kgLBM/day, respectively, provided sufficient statistical power to detect and quantify effects of 1.1, 1.1, and 1.2 SD, respectively, at 100α = 5% and 100β = 10% probabilities of type I and type II errors in single-sample tests, and differences of 1.4, 1.5. and 1.5 SD in two-sample comparisons between effects at 10 and 20, 20 and 30, and 10 and 30 kcal/kgLBM/day, respectively, at the same error rates.

Table Table 3.. Incremental Effects of Restricted Energy Availability on Bone Turnover Markers

RESULTS

The incremental effects of restricted energy availability on the three markers of bone turnover are quantified in Table 3. Figure 2 shows the dose-response relationships as percentage changes from bone marker concentrations after the balanced energy availability treatment. Restricted energy availability treatments at 10, 20, and 30 kcal/kgLBM/day reduced plasma OC concentrations by 28% (p = 0.0001), 32% (p = 0.002), and 11% (p = 0.02), respectively. The similar (p = 0.7) effects of 10 and 20 kcal/kgLBM/day were 166% larger (p = 0.03) than the effect of 30 kcal/kgLBM/day. In contrast, regression analysis showed that PICP declined linearly (p < 10⁻⁶) with energy availability. The restricted energy availability treatments at 10, 20, and 30 kcal/kgLBM/day reduced serum PICP concentrations by 26% (p = 0.001), 19% (p = 0.01), and 12% (p = 0.03), respectively.

Figure FIG. 2.

Incremental effects of low energy availability on NTX (•), PICP (▪), and OC (▴). Significance of treatment effects:^ap < 0.05,^bp < 0.01,^cp < 0.001, and^dp < 10⁻³. Significance of difference between treatment effects on OC at 20 vs. 30 kcal/kgLBM/day:^ep < 0.01. Significance of linear dependence of PICP on energy availability:^fp < 10⁻⁶.

The restricted energy availability treatment at 10 kcal/kgLBM/day raised 24-h urinary NTX concentrations by 34.0% (p < 0.001), whereas the treatments at 20 and 30 kcal/kgLBM/day had no effect (p > 0.4). Correspondingly, the two indices of resorption/formation uncoupling were also increased by the 10 kcal/kgLBM/day treatment (Z_NTX-OC = 1.7 ± 0.2; Z_NTX-PICP = 1.9 ± 0.2; both p < 10⁻⁴) but not by 20 (Z_NTX-OC = 0.7 ± 0.6; Z_NTX-PICP = 0.3 ± 0.5; both p > 0.2) or 30 kcal/kgLBM/day (Z_NTX-OC = 0.3 ± 0.3; Z_NTX-PICP = 0.4 ± 0.2; both p > 0.1).

DISCUSSION

This experiment is the first to quantify physiological dose-response relationships between energy availability and selected markers of bone turnover in healthy, young, regularly menstruating, habitually sedentary women, extending previous reports of the effects of complete fasting on bone markers.^{16, 17} The clinical significance of such effects on bone turnover is indicated by experience with postmenopausal women receiving pulsed estrogen therapy. Changes in CTX (∼43%) and OC (∼25%) in postmenopausal women similar to those in NTX (∼34%) and OC (∼30%) in this experiment led to a ∼6% change in BMD after 2 years.¹⁹

We found that NTX, PICP, and OC depend on energy availability in three distinctly different ways. Restricting energy availability only affected NTX, and by inference, bone resorption, when energy restriction was extreme, and NTX concentrations increased greatly. Calculated indices of uncoupling indicated that, in this extreme degree of energy restriction, increased bone resorption became uncoupled from decreased bone formation. By contrast, PICP and OC were significantly suppressed at all levels of energy restriction, showing for the first time that bone formation is impaired at much higher levels of energy availability than is bone resorption. Furthermore, the incremental responses of PICP and OC to energy restriction were distinctly different from one another. Whereas PICP, and by inference type I collagen formation, declined linearly with energy availability, the decline in OC occurred predominantly between 20 and 30 kcal/kgLBM/day, showing, again for the first time, a threshold effect of energy availability on OC secretion, and by inference, on matrix mineralization. The difference between these incremental effects on bone formation suggests that different mechanisms may mediate the influence of energy availability on collagen formation and matrix mineralization.

The observed changes in these markers of bone turnover were undoubtedly caused by our restriction of energy availability. The diet and exercise regimens of the subjects were precisely controlled according to the experimental design, and strict subject screening criteria ensured that results were not confounded by pre-existing medical conditions, unusual dietary habits, or reproductive disorders. Moreover, although markers of bone turnover do fluctuate rhythmically during the menstrual cycle, with formation markers maximal and resorption markers minimal when estrogen peaks at the time of ovulation,²⁰ the directions of those changes are opposite to the effects induced during the follicular phase in this experiment. Any such confounding of our results by menstrual cycle phase was nullified by our calculation of energy availability effects from repeated measures at the same phase of two menstrual cycles. Therefore, our findings can be reliably attributed to the energy availability treatments that we applied.

The energy availability treatments administered in this experiment span the range of energy restriction habitually self-administered by physically active women. Amenorrheic athletes reportedly practice diet and exercise regimens providing energy availabilities of ∼16 kcal/kgLBM/day,²¹ a level between the two most severely restricted energy availability treatments administered in this experiment. In comparison, regularly menstruating athletes self-administer energy availabilities of ∼30 kcal/kgLBM/day,²¹ similar to our mildest treatment. Increased markers of bone resorption and/or reduced markers of bone formation have been found in some but not all observational comparisons of amenorrheic athletes^22-26 and anorexia nervosa patients¹⁴,^27-30 to regularly menstruating controls. Such inconsistencies may be attributable to high variabilities in small samples of cases and controls. Alternatively, systemic markers of bone turnover may not reliably reflect localized demineralization in the lumbar spine, where reductions in BMD are most consistently found in these women, especially if their BMD is also increased in the heel and other load-bearing bones, as is found in some physically active women. Abnormalities in bone markers may also gradually change over time if BMD or turnover approaches a lower equilibrium in prolonged amenorrhea. Prospective experiments such as ours are inherently more sensitive for detecting the effects of administered treatments than are observational studies for finding differences of the same magnitude between groups of cases and controls. Previous prospective experiments have found that bone turnover was improved in anorexia nervosa patients by refeeding^30-32 and in amenorrheic athletes by vitamin K supplementation.³³

Unlike our 5-day prospective experiment, a 6-year prospective experiment that restricted the dietary intake of female rhesus monkeys by 30% had no effect on OC.³⁴ Considering the reduced lean body mass of the monkeys, however, their normalized dietary intake (kcal/kgLBM/day) had been reduced by only 20%, a restriction substantially milder than the mildest restriction administered in this experiment and self-administered by amenorrheic athletes.²¹

Previously, we reported dose-response relationships between energy availability and various reproductive and metabolic hormones in this same experiment.⁹ Some of these other dose-response relationships contributed insights into the present findings (Fig. 3). Like NTX, estradiol was unaffected by energy restriction until that restriction became severe. Like PICP and OC, a wide spectrum of metabolic hormones was disrupted at all levels of energy restriction. All of these metabolic hormones may participate to some degree in mediating the influence of energy availability on bone formation, but the dose-response relationships of insulin, triiodothyronine (T₃), and insulin-like growth factor (IGF)-I closely resembled those of PICP and OC. These similarities suggest that these particular hormones may play specific predominant roles in mediating the influence of energy availability on collagen formation and matrix mineralization.

Figure FIG. 3.

Comparison of effects of low energy availability on (A) NTX (•) and estradiol (E2; ○); (B) PICP (▪) and insulin (INS/2; □); and (C) OC (▴), T₃ (2*T3; ▵), and IGF-I (open diamond). For graphical clarity, values for insulin and T₃ have been divided and multiplied, respectively, by 2. Significance of treatment effects:^ap < 0.05,^cp < 0.001,^dp < 10⁻³,^ep < 10⁻⁴, and^fp < 10⁻⁵. Significance of difference between treatment effects on OC, T₃, and IGF-I at 20 vs. 30 kcal/kgLBM/day:^hp < 0.05. Significance of linear dependence of PICP and insulin on energy availability: ^gp < 10⁻⁶.

Dose-dependent effects on bone resorption and reproductive hormones

Figure 3A compares the energy availability dose-response relationships of the bone resorption marker NTX with that of estradiol, which we reported earlier.⁹ The 34% increase in NTX and the uncoupling of bone resorption and formation occurred when the restriction of energy availability was sufficiently severe to suppress estradiol by 18%. Finding associated opposite effects of energy availability on estradiol and NTX is not surprising considering the well-known role of estrogen in suppressing osteoclast activity.¹⁵ The ovarian suppression induced by severe energy restriction in this experiment was a secondary effect of a ∼40% reduction in LH pulse frequency,⁹ on which ovarian function critically depends. The disruption of LH pulsatility began at a threshold of energy availability at ∼30 kcal/kgLBM/day and became progressively more extreme as energy availability declined further.⁹ The fall in estradiol concentrations at 10 kcal/kgLBM/day indicates that the disruption of LH pulsatility at this energy availability was severe enough to suppress ovarian function.

In amenorrheic athletes, ovarian steroid production is unchanging at early follicular phase levels, indicating a complete absence of ovarian follicular development, ovulation, and luteal function.³⁵ Regularly menstruating athletes do display monthly rhythms in ovarian steroids, but their rate of follicular development is slower than regularly menstruating sedentary women, and their luteal function after a later ovulation is blunted and abbreviated.³⁵ Whether this luteal suppression also impairs bone metabolism is unclear.

Among premenopausal women, BMD is negatively correlated with the number of menstrual cycles missed since menarche.³⁶ Among young, amenorrheic women, NTX is negatively correlated with BMD.²⁷ Skeletal demineralization, osteopenia, osteoporosis, and fractures are all observed in young women who are hypoestrogenic because of anorexia nervosa²⁷,^37-39 and athletic amenorrhea.^{4, 6, 40} For these reasons, correcting chronic hypoestrogenism has, until recently, been the aim of treatment to restore the skeletal health of amenorrheic women who are physically active and/or restrained eaters. Because estrogen supplementation has not fully reversed their bone loss in clinical trials,^{11, 13} however, the focus on hypoestrogenism is being reconsidered. While the failure of estrogen therapy may be because of irreversible consequences of uncoupling bone formation and resorption,¹⁵ it may also be caused by the continued action of non-estrogen-dependent mechanisms.

Such mechanisms might be indicated by the hypometabolic status of amenorrheic athletes.^{24, 41} Female athletes eat less than would be expected for their level of physical activity, and amenorrheic athletes display low levels of plasma glucose, insulin, IGF-I, leptin, and T₃, as well as elevated levels of growth hormone (GH), IGF-I binding protein-1, and cortisol. These abnormalities are evidence of chronic energy deficiency with a compensatory slowing of metabolic rate, GH resistance, mobilization of fat stores, reduction in glucose use, and protein sparing.

Dose-dependent effects on bone formation and metabolic hormones

In this experiment, bone formation declined with energy availability although GH was increased,⁹ as it is in amenorrheic and regularly menstruating athletes.⁴² Thus, the stimulative effects of GH on bone formation seem to have been overridden by other factors.

Effects on PICP:

Figure 3B compares the energy availability dose-response relationships of the bone formation marker PICP and insulin, which we reported earlier.⁹ PICP and insulin both declined linearly with energy availability and were significantly reduced even at an energy availability as high as 30 kcal/kgLBM/day. If this energy availability were imposed on our ∼45 kgLBM women by dietary restriction alone, it would correspond to a dietary energy intake of ∼1350 kcal/day, a restriction not at all unusual in weight control programs.

In humans, chronic hypoinsulinemia in type I diabetes is associated with reduced skeletal mass and delayed healing of fractures.⁴³ In diabetic rats, active cuboidal osteoblasts are virtually absent from the endocortical surface and are replaced by bone-lining cells with no detectable uptake of proline for collagen synthesis,⁴⁴ indicating that the basic defect is in the number of active osteoblasts. Streptozotocin-induced diabetic mice fail to adequately express genes that regulate osteoblast differentiation, Cbfa1, Runx-2, and Dix-5, which leads to decreased bone formation.⁴⁵ Their diminished expression of type I collagen and OC is reversed by insulin treatment,⁴⁵ showing a specific causal relationship between inadequate insulin production and abnormal bone formation. Thus, the suppression of insulin by ∼50% in amenorrheic athletes⁴² may impair their bone formation and contribute to their persistent low BMD.

Type I diabetes may not be a good model for low energy availability, however, because the reduced plasma glucose levels in amenorrheic athletes and anorexia nervosa patients contrast sharply with the hyperglycemia in untreated type I diabetes. In amenorrheic athletes, lower plasma glucose levels derive not only from the fact that female athletes consume ∼30% less energy and carbohydrate per kilogram of body weight than male athletes,⁴⁶ but also from the fact that working muscle derives most of its energy from carbohydrate during prolonged intense exercise. Similar effects of energy and carbohydrate deficiency on bone can also occur in men: PINP was suppressed by as much as 30% within 3 days when trained male runners performed intensive exercise while their dietary energy intake was restricted by 50% in an experimental protocol designed to deplete their muscle glycogen stores.⁴⁷

Effects on OC:

Figure 3C compares the energy availability dose-response relationships of the bone formation marker OC with those of T₃ and circulating IGF-I, which were reported earlier.⁹ OC, T₃, and IGF-I all declined nonlinearly with energy availability, and most of this decline occurred abruptly between 20 and 30 kcal/kgLBM/day.

The different forms of the dose-response relationships of PICP and OC in this experiment are provocative because they suggest that collagen formation and matrix mineralization may have different sensitivities to energy availability. The linear decline in collagen formation is, perhaps, what one would expect, because the availability of amino acids declined with energy availability because we reduced the quantity without changing the quality of the diet, and because insulin, which stimulates amino acid uptake and protein synthesis, also declined linearly with energy availability. The unexpected abrupt decline in OC secretion below 30 kcal/kgLBM/day is what requires additional explanation.

Experiments with osteosarcoma cells in culture seem to provide this explanation, because they show that T₃ has a much stronger influence, mediated by IGF-I, on OC secretion than on collagen formation. Rat osteosarcoma ROS 17/2.8 cells express a mature osteoblast phenotype similar to the osteoblasts that would be found in adult humans.⁴⁸ In ROS 17/2.8 cells, T₃ increases OC expression in a dose-dependent fashion,⁴⁸ but it does so selectively without inducing collagen expression.⁴⁹ In mature mouse osteoblasts, a normal physiological 1 nM dose of T₃ induced a 5000% increase in OC expression, whereas the lowest effective 10 nM dose of T₃ induced only a 40% increase in proline uptake, which predominantly reflects collagen formation.⁵⁰

The dose-response relationship of IGF-I in this experiment probably tracked that of T₃ because ∼75% of circulating IGF-I is produced by the liver in response to GH stimulation, and T₃ modulates GH sensitivity in the liver and other tissues.⁵¹ At a normal physiological dose of 1 nM, T₃ stimulates IGF-I production in cultured rat calvarial osteoblasts,⁵² and interference with IGF-I action decreases the anabolic effects of T₃ on both OC and collagen synthesis.⁵⁰ IGF-I increases collagen synthesis in bones of hypophysectomized rats⁵³ and OC synthesis in human MG-63 osteosarcoma cells,⁵⁴ but has no effect without co-factors, such as calcitriol,⁵⁴ that are also induced by T₃.⁵⁰

Short-term administration of recombinant human (rh)IGF-I alone has selectively increased bone formation without increasing bone resorption in young, fasting, healthy women¹⁷ and in osteopenic anorexia nervosa patients,¹⁴ but chronic rhIGF-I administration to anorexia nervosa patients improved bone turnover and BMD only when used in combination with concurrent antiresorptive therapy.⁵⁵ However, because a nonresponding control group receiving antiresorptive therapy alone unaccountably reduced their ad libitum dietary intake of energy and calcium by 40% during the experiment, resulting in a 15% decline in IGF-I, the small (∼1.8%) increase in BMD resulting from the combined treatment may have been caused by the antiresorptive therapy and not the rhIGF-I.

Serum OC levels are elevated in patients with hyperthyroidism and reduced in patients with hypothyroidism.⁵⁶ In overt hypothyroidism, the time required to complete resorption/formation remodeling of bone is prolonged from 6 months to 2 years.⁵⁷ Low T₃ syndrome is found in anorexia nervosa,⁵⁸ dietary amenorrhea,⁵⁹ and athletic amenorrhea⁶⁰ as one facet of a multifaceted disruption of metabolic hormones and substrates.⁴² In reproducing the full spectrum of these metabolic aberrations experimentally, we have shown that T₃ concentrations depend on energy availability (or more specifically on the carbohydrate availability associated with that energy availability) so that T₃ can be suppressed either by reducing dietary intake or by increasing exercise energy expenditure without reducing dietary intake.⁶¹ This suppression occurs abruptly below a threshold of energy availability⁶² at ∼30 kcal/kgLBM/day.⁹ Thus, we suspect that the ∼25% reduction in T₃ levels in amenorrheic athletes⁶⁰ may also impair their bone formation and especially their bone mineralization.

In summary, this short-term experiment has contributed new qualitative and quantitative information about the dependence of bone turnover on energy availability in regularly menstruating young women. When energy availability was restricted severely enough to suppress estradiol, bone resorption increased and became uncoupled from suppressed bone formation within 5 days, a condition that, if left to continue, may cause irreversible reductions in BMD. Bone formation was impaired by much less severe restrictions of energy availability that also disrupt a wide spectrum of metabolic hormones and substrates, a condition that may prevent young women from achieving their genetic potential for peak bone mass. Military servicewomen and others in weight control and physical training programs may need to maintain their energy availability above 30 kcal/kgLBM/day to avoid these effects. More prolonged clinical experiments are needed to determine whether this short-term experiment predicts the influence of chronic restrictions of energy availability on bone turnover.

Acknowledgements

We thank JR Thuma, B Baumer, P Cadamagnani, AL Cornelius, S Demarchi, T Grindstaff, ER Jopperi, JK Lavery, WB Malarkey, D Murray, K Ragg, JM Slade, K Swain, TM Vallecorsa, KA Varmus, T Wiese, K Zaylor, and the nursing staff of the General Clinical Research Center at The Ohio State University Hospital for important contributions to this research. We also appreciate the extraordinary cooperation of the subjects. This research was supported in part by U.S. Army Medical Research and Material Command (Defense Women's Health & Military Medical Readiness Research Program) Grant DAMD 17-95-1-5053, the General Clinical Research Branch, Division of Research Resources, National Institutes of Health Grant M01 RR00034, Ohio University Research Enhancement Fund, and Ross Laboratories. The content of the information reported in this paper does not necessarily reflect the position or the policy of the Government, and no official endorsement should be inferred.