Introduction

Osteoporosis currently affects 10–12 million people in the United States. Although the majority of those affected are postmenopausal women, osteoporosis also affects middle-aged men [1,2,3] and premenopausal women [4,5,6,7]. In these men and premenopausal women, an underlying etiology can often be determined [4,5]. The underlying disorder may be due to interference with the acquisition of peak bone mass or to excessive bone loss after peak bone mass was attained. However, there remains a cohort of men and women with low bone mass and/or low trauma fractures, in whom no contributing etiology can be identified. Osteoporosis that affects young and otherwise healthy individuals is operationally defined as “idiopathic” osteoporosis (IOP).

In the review by Heshmati and Khosla [8], IOP was diagnosed primarily in Caucasians and affected men and women equally. Fractures, a requirement for inclusion in their review, were usually multiple, occurred over a 5- to 10-year period, and predominantly involved sites rich in cancellous bone, such as the vertebrae. Khosla et al. found histomorphometric evidence of osteoblast dysfunction in 18 men and women with IOP [4], an observation confirmed by other investigators in studies of men with IOP [3,9,10]. In our prospective study of men with IOP [3], we found a reduction in bone formation rate that ranged from 28% to 54% compared with normal controls. Thus, evidence is accumulating in men for a defect in osteoblast function [11]. While the cause of the osteoblast dysfunction is not clear, reduced serum levels of insulin-like growth factor-I (IGF-I), a major determinant of bone growth and remodeling, have been implicated. We documented low serum IGF-I levels in men with IOP [3], as did Ljunghall et al. in men [12], and Reed et al. [13] in both men and women with IOP. Moreover, significant correlations between histomorphometric parameters of bone formation and serum IGF-I levels have been reported [3,13]. The reasons for low IGF-I levels in patients with IOP are unclear, but in men may be associated with a specific polymorphism of the IGF-I gene [14].

Most studies of IOP have focused upon men [3,9,10,12] or have not distinguished between men and women [4]. Whether premenopausal women with IOP have similar decreases in IGF-I levels and bone formation is unknown. Postulating that serum IGF-1 levels would also be low in premenopausal women with IOP, we prospectively evaluated a group of premenopausal women with unexplained low bone mass and/or fragility fractures and compared them with normal premenopausal women with respect to serum concentrations of IGF-I and IGFBP3.

Materials and methods

This study was conducted in the Irving Center for Clinical Research and the Metabolic Bone Diseases Unit at Columbia-Presbyterian Medical Center (CPMC) with the approval of the Institutional Review Board of Columbia University. All participants gave written informed consent.

Study design

The study was a cross-sectional, case-control design. Subjects were identified from our practice database, by physician referral and advertisement. Controls were recruited by advertisement and flyers posted at the medical center. Bone mineral density (BMD) measurements were performed by dual-energy X-ray absorptiometry (DXA) at CPMC on a Hologic QDR-4500 bone densitometer (Hologic, Inc., Waltham, Mass., USA). Because these were younger women, in whom Z- and T-scores are similar, our inclusion criteria were based upon Z-scores that compare BMD to normal individuals matched for age, gender and race. To be eligible, subjects had to have Z-scores less than −2.0 at the lumbar spine, total hip or femoral neck, with or without a history of fracture. Controls were women with normal BMD (Z-scores less than −0.5 at all three sites). Since smaller bones have a lower areal BMD than larger bones and the definition of case or control was based on BMD (an areal measurement of bone density; g/cm2), we confirmed that cases and controls would still have that designation if a calculated measurement that converts areal to volumetric BMD, bone mineral apparent density (BMAD; g/cm3), were used. BMAD was calculated for the lumbar spine and femoral neck sites [Spine BMAD=BMC/(area)1.5; FN BMAD=BMC/(area)2], as an approach to controlling for errors in the measurement caused by size difference [15,16].

To be included, participants had to have regular monthly menses, with no history of oligomenorrhea or amenorrhea, excepting pregnancy. Use of oral contraceptives for purposes of birth control was permitted only if the menstrual history prior to their initiation was completely normal. All were evaluated during the follicular phase of the menstrual cycle (days 1–5). Participants had a complete history and physical examination, with particular focus on secondary causes of osteoporosis, and completed a questionnaire that provided information on medical, surgical, and reproductive history, risk factors for osteoporosis (alcohol, tobacco and caffeine), personal and family history of fractures and current and past medication use. All completed an Eating Attitudes Test to identify subclinical eating disorders [17]. Dietary calcium intake and use of calcium and vitamin D supplements were assessed. A detailed laboratory investigation excluded disorders of calcium metabolism and known causes of osteoporosis. Subjects were considered to have IOP if no cause of osteoporosis (specifically glucocorticoid excess, type I diabetes, primary hyperparathyroidism, vitamin D deficiency, eating disorders, hypogonadism, malabsorption, celiac disease, hypercalciuria, renal dysfunction) could be identified on the basis of the clinical and biochemical investigations. Participants who had used glucocorticoids, calcineurin inhibitors, loop diuretics, anticonvulsants, heparin, l-thyroxine, and gonadotrophin-releasing hormone agonists were excluded.

Fracture ascertainment was based upon history. Only fractures that occurred after age 18 were included. Fractures were considered fragility fractures if they were associated with minimal or no trauma (such as a fall from a standing height or less). Those associated with motor vehicle accidents, major falls, or sports injuries were excluded. A family history of osteoporosis was considered positive if a grandparent, parent or sibling had a history of a typical fragility fracture (hip, vertebral, or wrist fracture) or had osteoporosis based on a BMD measurement.

Measurements

All blood specimens were collected in the morning, after an overnight fast during the first 5 days of the menstrual cycle. Serum electrolytes, calcium, phosphorus, albumin, total alkaline phosphatase activity, creatinine and blood urea nitrogen, were measured in the clinical laboratories of CPMC by standard autoanalyzer techniques (Technicon Instruments, Tarrytown, N.Y., USA). Serum estradiol and FSH measurements were also performed in the clinical laboratory of CPMC. The estradiol assay was a commercial solid-phase, chemiluminescent immunoassay (Immulite; Diagnostic Products Co, DPC, Los Angeles, Calif., USA). The intra-assay and inter-assay variability was 9.3% and 10.5%, respectively. Serum for indices of mineral metabolism was frozen immediately and stored in aliquots at –70o for analysis in the Core Laboratory of the General Clinical Research Center (GCRC) of Columbia University. Kits for parathyroid hormone (PTH) and vitamin D metabolites were purchased from Corning-Nichols (San Juan Capistrano, Calif., USA). PTH was measured by immunoradiometric (IRMA) assay [18]. Intra-assay and inter-assay variability was 3.4% and 5.6%, respectively. Serum 25-hydroxyvitamin D (25-OHD) and 1,25-dihydroxyvitamin D [1,25(OH)2D] were measured after extraction by radiobinding and radioreceptor assays, respectively. The intra-assay and inter-assay variability of 25-OHD was 7.5% and 9.6%, respectively. The intra-assay and inter-assay variability of 1,25-OH2D was 7.6% and 9.8%, respectively. Bone-specific alkaline phosphatase (BAP) was measured by a solid phase, two-site immunoradiometric assay (Hybritech Inc., San Diego, Calif., USA). The intra-assay and inter-assay variability was 4.2% and 7.2%, respectively. Serum osteocalcin was measured by radioimmunoassay (RIA) with intra-assay and inter-assay variability of 4.3% and 5.7%, respectively. SHBG was measured with a coated-tube immunoradiometric assay kit (DSL, Webster, Tex., USA). The inter-assay variability was 5.6% and the intra-assay variability was 2.4%. Samples for IGF-I were first prepared by acid-ethanol cryoprecipitation according to the method of Breier et al. [19] and then analyzed in duplicate by RIA using a polyclonal antibody supplied by Nichols Institute (San Juan Capistrano, Calif., USA). This extraction technique removes almost all IGF binding proteins from serum, and values for IGF-I by RIA after acid-ethanol cryoprecipitation performed in our laboratory correlate closely (r=0.97) with samples extracted by acid-gel chromatography [20]. Reference values for serum IGF-I over the 3 decades of ages included in our cohort were generated by serum measurements performed on 195 women who participated in other investigations. The interassay and intra-assay coefficients of variation were 8.8% and 2.73%, respectively. The lower detection limit of the assay was 10 ng/ml. Serum IGF binding protein 3 (IGFBP3) concentration was measured using an IRMA assay (DSL, Webster, Tex., USA) with interassay and intra-assay coefficients of variation of 7.8% and 3.02%, respectively. The lower level of detection in this assay is 250 ng/ml. A 24-h urine collection was analyzed for creatinine by standard autoanalyzer techniques (Technicon Instruments Corporation). Urinary calcium was measured by atomic absorption spectrophotometry. Urinary N-telopeptide (NTX) excretion was measured by a competitive-inhibition enzyme-linked immunosorbent assay (Ostex, Seattle, Wash., USA) [21]. The intra- and interassay variability were 7.6% and 4.0%, respectively. Urinary free cortisol was determined by a well-established protocol employing an initial extraction step followed by RIA [22]. Free estradiol levels were calculated using the method of Sodergard [23].

Statistical analysis

All continuous data are presented as mean value±SEM, and all categorical data are reported as percentage or absolute number. Distributions of continuous variables were tested for normality, and descriptive statistics were calculated for raw data. Student’s t-tests and Fisher’s exact tests were used to assess differences between groups. For all inferential tests, a 5% type 1 error rate was used. A regression analysis was performed to assess the influence of low BMD on estradiol and urinary NTX while controlling for oral contraceptive use. Z-scores for IGF-I were developed from laboratory reference ranges by decade of age.

Results

Patient characteristics

Thirteen women with IOP were compared with 13 healthy control subjects. Most of the subjects (n=9) had been found to have low bone density after a fragility fracture; the remaining subjects had undergone bone mineral density screening because of a family history of osteoporosis. Subjects and controls did not differ with respect to age, height, ethnic composition (11 Caucasian patients in both groups; the two non-Caucasian subjects were Hispanic and Asian Indian, while the two non-Caucasian controls were Hispanic and Asian) or proportion using oral contraceptives (Table 1), but body mass index (BMI) was lower in subjects than controls. Dietary calcium and vitamin D intake were higher in subjects than in controls. Subjects reported both a family history of osteoporosis and fragility fractures more frequently than controls (P≤0.05). The majority of the fractures involved the extremities (Table 2). According to the selection criteria, BMD, measured in g/cm2, T- and Z-scores were lower in subjects than controls at the lumbar spine and femoral neck (Table 2). Within the subjects, BMD was lower among those who had fractured (lumbar spine: 0.810±0.03 versus 0.897±0.07 g/cm2; femoral neck: 0.607±0.02 versus 0.622±0.05 g/cm2), although not statistically significant. When size differences were taken into account by measurement of BMAD, subjects still had significantly lower values (Table 2).

Table 1 Patient characteristics. Data are expressed as the mean±SEM
Full size table
Table 2 Fracture history and bone mineral density. Data are expressed as the mean±SEM
Full size table

Serum IGF-I levels

As expected, serum IGF-I declined with age (r=−0.59, P<0.01; Fig. 1). However, in contrast to our observations in men with IOP, mean serum IGF-I and IGFBP3 concentrations did not differ between subjects and controls, even after controlling for group differences in BMI. To appreciate better the age-specific reduction in IGF-I levels, each patient’s IGF-I was expressed by an age-based norm, or Z-score (see Methods). However, IGF-I Z-scores were still not different between groups (subjects: −0.28±0.25 versus controls: −0.41±0.16; P=0.64). Moreover, serum IGF-I correlated neither with serum estradiol nor with BMD or BMAD at either the lumbar spine or femoral neck.

Fig. 1
figure 1

IGF-I levels in premenopausal women with IOP (■) and controls (□). Serum IGF-I levels decline significantly with age. The data points are shown in relationship to the age specific mean±1 SD for IGF-I, calculated from 1031 normal subject samples assayed in one of our laboratories (C.J.R., see Materials and methods)

Full size image

Calcitropic hormones and estradiol levels

All subjects and controls had normal serum calcium concentration (Table 3), electrolytes and renal function. A trend toward lower urinary calcium excretion was observed in subjects (2.7±0.3 mmol/l; range: 1.3–6.3), although the difference when compared with the controls was not significant. Serum PTH, 25-OHD and 1,25(OH)2D concentrations did not differ between groups.

Table 3 Biochemical characteristics. Data are expressed as the mean±SEM
Full size table

Serum estradiol, measured during the follicular phase of the cycle (days 1–5), was approximately 35% lower overall in subjects than in controls (115.3±11 versus 177.3±22 pmol/l, P<0.05). Morever, six subjects had levels below 110 pmol/l, as compared to only one control. The difference in estradiol levels remained statistically significant by linear regression analysis after controlling for oral contraceptive use. BMD did not differ between those who were and were not taking oral contraceptives (both overall and only within subjects). Lower estradiol levels were twice as likely to be associated with lower bone mineral density than was oral contraceptive use (r=0.431; P=0.045). The between-groups difference in estradiol levels was independent of the difference between groups in BMI. Total serum estradiol levels correlated with BMD and BMAD at the femoral neck (r=0.50 and r=0.44 respectively, both P<0.05), but did not correlate with BMD or BMAD at the lumbar spine.

Because it is difficult to interpret serum estradiol levels in oral contraceptive users, estradiol levels in subjects and controls not using oral contraceptives were also compared. When only those women were considered (four subjects and seven controls), serum estradiol levels remained lower in the subjects as compared to controls, approaching significance (124.1±13 versus 194.9±24 pmol/l, P=0.06). Among the non-oral contraceptive users, total serum estradiol correlated with BMD at the lumbar spine (r=0.63, P<0.05). Oral contraceptive users in subjects and controls considered together had higher SHBG levels than non-oral contraceptive users (336±65 versus 176±23 nmol/l, P<0.05). We therefore calculated free, bioavailable estradiol levels, which were significantly lower overall in subjects than controls (0.6±0.1 versus 1.2±0.2 pmol/l, P<0.05). Free, bioavailable estradiol correlated with BMD and BMAD at the lumbar spine (r=0.54, P<0.01 and r=0.54, P<0.05, respectively) and femoral neck (r=0.60 and r=0.55, respectively, both P<0.01). When only the women not using oral contraceptives were considered, free, bioavailable estradiol levels were still lower in subjects than controls and approached significance (1.0±0.1 versus 1.4±0.1 pmol/l, P=0.06). Free, bioavailable estradiol levels in the non-oral contraceptive users correlated with BMD at the lumbar spine (r=0.63, P<0.05).

Biochemical markers of bone turnover

Serum BAP, a marker of bone formation present in preosteoblasts and osteoblasts, was above the upper end of the normal range in subjects while normal in controls, though the difference was not statistically significant. Serum osteocalcin, a marker of osteoblast differentiation synthesized by mature osteoblasts, was normal and did not differ between the groups. Mean urinary NTX excretion, although within the normal range, was significantly higher in subjects than controls. Linear regression analysis confirmed that urinary NTX was still higher in subjects than controls when oral contraceptive use was controlled for (r=0.466; P=0.033). Estradiol levels did not correlate with any of the markers of bone turnover, regardless of oral contraceptive use. However, BMI correlated inversely with urinary NTX (r=−0.52, P<0.05), a relationship that persisted after exclusion of oral contraceptive users (r=−0.90, P<0.01).

Discussion

The results of this study provide data and perspective on a group of subjects whose osteoporosis has been little studied and even less understood. This cohort, premenopausal women with osteoporosis and no well-defined etiology, has a counterpart in middle-aged men with idiopathic osteoporosis. The results indicate that women and men differ substantially in the findings that could help to explain this obscure syndrome. In contrast to our observations in men with IOP, we found no difference in serum IGF-I levels between patients and controls, nor did we detect any relationship between serum IGF-I and BMD at any site. The histological and biochemical data reported in men suggest osteoblast dysfunction that results in defective bone formation [3,4,10,24], perhaps related to low serum IGF-I concentrations. In contrast to studies of men [3], we found that IGF-I was normal in women with IOP and that there was no apparent relationship between IGF-I and BMD. While prospective histomorphometric data for premenopausal women with IOP are scant, the increase in the bone turnover markers, NTX and BAP, suggests that bone turnover is not suppressed in these women. These observations suggest that premenopausal women with unexplained osteoporosis may have a different pathogenesis than men with IOP.

Males with IOP have normal testosterone levels [4]. In contrast, the women, despite having regular menses, had significantly lower serum estradiol levels. Moreover, serum estradiol levels correlated with BMD at the femoral neck. An immediate explanation may rest with a confounder, namely the women who were on oral contraceptive agents. It is possible that the estrogen preparations in the oral contraceptives were not measured by the estradiol assay we used, yet suppressed gonadotrophin-releasing hormones, thus reducing ovarian estrogen production. However, when we controlled for oral contraceptive use, the differences in estrogen levels between subjects and controls remained statistically significant. Moreover, lower serum estradiol levels were more likely to be associated with lower BMD than was oral contraceptive use. Our observation that free, bioavailable estradiol levels, which should be less affected by oral contraceptive use, were significantly lower in subjects than controls also suggests that the lower serum estradiol levels observed in the subjects may truly reflect decreased estrogen synthesis. In addition, when oral contraceptive users were excluded, both total and free serum estradiol levels still tended to be lower in subjects than controls.

Our finding that premenopausal women with low BMD had lower estrogen levels and that BMD and estradiol levels were correlated with each other is consistent with reports by Sowers et al. indicating that in normally menstruating women, differences in sex hormone production may affect BMD [25,26]. In a nested case-control study of premenopausal women from the Michigan Bone Health Study, those with femoral neck BMD in the lowest 5th percentile were compared with women with femoral neck BMD within 1 SD of the normal mean [25]. All reported regular menses, and there was no clinical evidence of hypogonadism, nor were there any diseases to explain the lower BMD. However, cases had significantly lower follicular phase serum estradiol levels than controls [25]. Moreover, integrated urinary excretion of both estrogen and progesterone metabolites over two consecutive menstrual cycles was significantly lower in cases than controls [26], suggesting that marginally lower levels of these sex steroids in apparently normally cycling premenopausal women may substantially influence premenopausal bone mass.

We have found that weight was positively associated with BMD, independent of serum estradiol levels. This observation runs counter to the common knowledge that weight improves bone mass by fat being a source of aromatase activity and therefore estrogen [27]. In heavier women, more estrogen is formed in fat from circulating androgens [28]. However, weight per se can have a salutary effect on the skeleton, because it exerts positive gravitational and other mechanical forces on bone. This is consistent with a recent report that fat-free mass was a better predictor of variations in bone mass than fat mass in young women followed over 10 years [29]. In addition, Hui et al. found that bone loss in 130 healthy premenopausal women was associated with both weight loss and low estrone sulfate [30]. Most likely, both sex hormones and weight interact to impact low BMD in premenopausal women, as is the case in postmenopausal bone loss [31].

There is controversy about the significance of low BMD without fractures in younger women [32,33]. However, premenopausal women are increasingly likely to have BMD measured, whether for appropriate or inappropriate indications, and thus young women who fulfill WHO criteria for osteoporosis in postmenopausal women [34], but who do not have a history of fracture, are coming to medical attention. We included such young women in this study, recognizing that there are insufficient prospective data to know whether low BMD has the same predictive value for fracture in premenopausal as it does in postmenopausal women, and that low BMD is not necessarily synonymous with compromised bone strength. Moreover, low BMD could be related to inadequate peak bone mass, rather than excessive bone loss and such women may have normal microarchitecture, turnover and mineralization. Nevertheless, it is also possible that premenopausal women with low BMD, particularly those with increased bone turnover as we observed in this study, have microarchitectural changes or abnormalities of bone mineral or collagen that disrupt the structural integrity of bone and lead to decreased strength and increased fracture risk [35,36,37]. For example, otherwise healthy, premenopausal women with low-energy wrist fractures have significantly lower forearm [38], as well as spine and hip [39] BMD than appropriate controls. Defective bone quality in premenopausal women, in fact, might not even be detected by BMD. We recently observed that in nine young women with normal BMD and spontaneous fractures, the collagen cross-link ratio, a measure of matrix quality, was distinctly abnormal (Paschalis, Shane, Lyritis, Skarantavos, Boskey, unpublished data). Thus very low BMD or abnormal bone quality in premenopausal women could represent a pre-symptomatic phase of IOP.

This study has several limitations. The cross-sectional design and the small number of subjects may have limited power to detect between-groups differences in bone formation markers. Fracture ascertainment was based on history, which may introduce inaccuracies. The inclusion of women who had not sustained fractures may also be criticized. Estrogen status was assessed by a single measurement that may not reflect an integrated assessment over a typical cycle, and thus could obscure the true relationship between circulating estradiol and BMD. Finally, the use of oral contraceptives in approximately half the cases and controls may confound interpretation of estradiol and bone turnover markers. However, despite these limitations, we believe the results are of interest, add to the literature on this emerging problem and should direct future studies designed to elucidate the etiology and pathogenesis of IOP in women.