Bone-related circulating hormonal factors

Bone turnover markers

Importance of bone remodelling in osteoporosis

Bone remodelling is an important process that continuously renews the cortical and trabecular envelopes throughout life. It repairs micro-damage, maintains mineral homeostasis and ensures mechanical competence by modifying the micro-architecture. Bone remodelling is regulated by a variety of systemic and local factors as well as by nutritional factors (Fig. 1). Bone remodelling follows a sequence of events that includes activation, recruitment of osteoclasts to begin resorption, degradation and removal of bone, reversal, and formation of new bone by osteoblasts (Fig. 1). After this phase, a quiescent or resting period occurs. Remodelling begins with release of cytokines from osteoblast precursors, leading to recruitment of osteoclast precursors, which differentiate in multinucleated osteoclasts and subsequently attach to the bone surface. Resorption of the skeletal matrix releases growth factors such as IGF-I and transforming growth factor-β, which recruit lining cells and early osteoblast precursors that eventually form new bone (Fig. 1). Osteoblasts that do not undergo apoptosis remain on the bone surface as lining cells or become entrapped as osteocytes into the new bone matrix. These osteocytes respond to gravitational forces and can signal to lining cells to initiate remodelling. Antiresorptive and anabolic agents influence bone structure by affecting remodelling. Likewise, supply in Ca, vitamin D and protein influences bone remodelling. The presence of Ca and vitamin D inhibits bone resorption, and protein stimulates bone formation (Fig. 1). Ca, Pi and vitamin D are necessary for adequate mineralisation of the newly formed bone matrix (Fig. 1).

Fig. 1 Bone remodelling cycle. Bone turnover follows a sequence of events that includes activation, recruitment of osteoclasts (Ocl) to begin resorption, degradation and removal of bone, reversal, and formation of new bone by osteoblasts (Ob). After this phase a quiescent or resting period occurs. LC, lining cell; Oc, osteocyte.

In young adulthood, the amount of bone removed by the osteoclasts is equal to the amount of bone formed by the osteoblasts, so that bone mass is maintained. After the menopause and with ageing in both sexes, there is a progressive bone loss due to an imbalance between bone resorption and formation.

There is a substantial body of evidence that accelerated bone turnover is associated with a higher risk of fracture, independent of other parameters such as age or BMD (see below). BTM are not included in the fracture risk calculator (FRAX®), mainly for the reason that their relationships with other established risk factors included in the risk calculator are not clarified and because different markers and different methodologies for their assessment have been used across studies. This has led to the recommendation for the standardisation of BTM measurements in future studies and it is suggested that serum carboxy terminal telopeptide of collagen type I (s-CTX) is used as the reference bone resorption marker and serum procollagen type I N-terminal propeptide (sPINP) as the reference bone formation marker, measured by standardised assays⁽ Reference Vasikaran, Eastell and Bruyere ^{66 )}.

Bone markers reflecting bone formation and bone resorption

The BTM can be divided into two groups: markers of resorption and markers of formation (Fig. 2). The principal, specific markers of bone formation, measured in serum by immunoassays, are bone alkaline phosphatase (BAP), osteocalcin (OC) and PINP. Most assays are now adapted to automated platforms⁽ Reference Lee and Vasikaran ^{7 )} (Fig. 2). Markers of bone resorption are the pyridinium cross-links (pyridinoline and deoxypyridinoline) and their associated peptides (telopeptides), released during collagen breakdown⁽ Reference Kraenzlin, Seibel, Seibel, Robins and Bilezikian ^{6 )}. Markers of osteoclast number include tartrate-resistant acid phosphatase isoenzyme 5b (TRAP 5b), the osteoclast-specific isoform of total TRAP, and, potentially, cathepsin K (Fig. 2). These enzymes are expressed and released by osteoclasts⁽ Reference Henriksen, Tanko and Qvist ^{67 )}. TRAP 5b is the most commonly used marker of osteoclast number.

Fig. 2 Diagram of the relationship between bone resorption or formation and bone turnover markers. On the resorption side, tartrate-resistant acid phosphatase (TRAP) reflects osteoclast number and activity, whereas pyridium cross-links, carboxy terminal telopeptide (CTX) and N-terminal telopeptide (NTX) are markers of bone matrix resorption. On the formation side, collagen type I propeptides, such as procollagen type I N-terminal propeptide (PINP), bone alkaline phosphatase (BAP) and osteocalcin reflect mainly osteoblastic bone formation. Osteoid corresponds to the non-mineralised bone matrix.

TRAP 5b is predominantly expressed by osteoclasts, and immunoassays to detect this isoform have been developed⁽ Reference Halleen, Hentunen and Karp ^{68 –} Reference Kraenzlin, Lau and Liang ^{70 )}. In absolute terms, the response of TRAP 5b to antiresorptive therapy is not as great as the response of s-CTX, but within-subject variability is significantly lower. TRAP 5b has a lower diurnal variability than s-CTX and lower acute effect of feeding, possibly due to a longer circulating half-life, making it less susceptible to short-term influences⁽ Reference Hannon, Clowes and Eagleton ^{71 )}. Furthermore, TRAP 5b is not cleared by the kidney, and thus does not accumulate in renal failure⁽ Reference Halleen, Alatalo and Janckila ^{69 ,} Reference Hannon, Clowes and Eagleton ^{71 )}.

Predictive value of early changes in bone markers in response to pharmaceutical agents on later effects on bone mineral density/content and fragility fractures: bone turnover markers and fracture risk

BTM may be useful in the assessment of fracture risk. Findings of prospective studies have shown that BTM of resorption predict fracture risk in postmenopausal women and in men (Table 1)⁽ Reference Garnero, Hausherr and Chapuy ^{74 ,} Reference Garnero, Sornay-Rendu and Claustrat ^{83 –} Reference Garnero, Delmas, Seibel, Robins and Bilezikian ^{89 )}. In five prospective studies of postmenopausal women (Epidemiology of Osteoporosis (EPIDOS), Rotterdam, Os de Femmes de Lyon (OFELY), Hawaii Osteoporosis Study (HOS), and Malmö Osteoporosis Prospective Risk Assessment (OPRA)), there were significant associations of baseline levels of urinary or s-CTX, urinary free deoxypyridinoline, or serum TRAP 5b with fracture risk⁽ Reference Garnero, Hausherr and Chapuy ^{74 ,} Reference Garnero, Sornay-Rendu and Claustrat ^{83 –} Reference Meier, Nguyen and Center ^{87 )} (Table 1). An increase of these markers above the premenopausal range was associated, after adjustment for aBMD, with a twofold increase in risk for hip, vertebral and non-vertebral fractures, over a follow-up period of 1.8 to 5 years. The results for the relationship between bone formation markers and fracture risk are not consistent. In the French cohort study of elderly women (EPIDOS), no significant relationship between OC or BAP and fracture risk could be demonstrated during a 2-year follow-up. In contrast, in younger healthy postmenopausal women (OFELY and HOS cohorts), a significant association was found between increased BAP serum levels and risk of vertebral as well as non-vertebral fracture⁽ Reference Garnero, Hausherr and Chapuy ^{74 ,} Reference Garnero, Sornay-Rendu and Claustrat ^{83 ,} Reference Ross, Kress and Parson ^{84 )}.

Table 1 Relationship between increases in bone resorption rate and fracture risk

RR, relative risk; BMD, bone mineral density; EPIDOS, Epidemiology of Osteoporosis; CTX, carboxy terminal telopeptide of collagen type I; OFELY, Os de Femmes de Lyon; HOS, Hawaii Osteoporosis Study; DPD, free deoxypyridinoline; OPRA, Osteoporosis Prospective Risk Assessment; TRAP, tartrate-resistant acid phosphatase; DOES, Dubbo Osteoporosis Epidemiology Study; ICTP, C-terminal telopeptide generated by metalloproteinases.

* RR-BMD =  RR for fracture by 1 sd decrease in BMD.

† RR-marker =  RR for fracture by 1 sd increase in marker above the premenopausal normal range.

‡ For the highest quintile.

These results suggest that a combined approach, with both aBMD and indices of bone turnover, could improve fracture prediction in postmenopausal women. In fact, calculating the absolute risk of fracture by combining clinical risk factor (i.e. previous fracture), low aBMD and increased bone turnover (high CTX) resulted in a better hip fracture probability at 10 years than with low aBMD alone⁽ Reference Johnell, Odén and De Laet ^{90 )}. In elderly men high bone resorption was also associated with an increased risk of osteoporotic fracture⁽ Reference Meier, Nguyen and Center ^{87 )}.

Bone turnover markers and treatment monitoring

The main domain for the clinical use of BTM is the monitoring of osteoporosis therapy. The ultimate goal in treating patients with osteoporosis is to reduce their fracture risk. However, the short-term incidence of osteoporotic fractures is low, and the absence of fracture during treatment does not necessarily mean that the treatment is effective. Consequently, serial measurements of changes in aBMD as a surrogate marker of therapeutic efficacy are currently the standard approach to monitor osteoporosis therapy. Yet, changes in aBMD occur slowly and therapeutic effects are usually not detectable before 1–2 or more years of treatment⁽ Reference Delmas ^{80 ,} Reference Bonnick and Shulman ^{91 –} Reference Roux, Garnero and Thomas ^{94 )}. In contrast, BTM levels change much faster than aBMD in response to therapeutic interventions⁽ Reference Delmas ^{80 )}. Antiresorptive treatment generates a rapid decrease in bone resorption markers after only 2–4 weeks, reaching a plateau after 3 to 6 months of treatment⁽ Reference Delmas ^{80 ,} Reference Garnero, Shih and Gineyts ^{95 )} (Fig. 3(a)). The decrease in bone formation markers, reflecting the physiological coupling of bone formation to bone resorption, is delayed and reaches a plateau after 6–12 months. The magnitude of the decreases varies according to the type and dose of the drug and the marker used to assess the effect (Table 2).

Fig. 3 Schematic representation of changes in bone turnover markers (BTM) at 3–6 months and corresponding increase in bone mineral density (BMD) during treatment of osteoporosis. (a) Bone antiresorptive treatment (i.e. bisphosphonate): a more pronounced decrease in bone resorption marker at 3–6 months is associated with larger increases in BMD. (b) Bone formation-stimulating treatment (i.e. recombinant parathyroid hormone): a greater increase in bone formation marker is associated with a larger increase in BMD. * Mild response in changes in BTM and BMD. † Moderate response in changes in BTM and BMD. ‡ Pronounced response in changes in BTM and BMD. The diagrams show the relationship between early changes (3–6 months) in BTM and changes in BMD after 2 years of either bone resorption inhibition (a) or bone formation stimulation (b). Fig. 3(a) is based on quantitative data obtained by antiresorptive treatment with alendronate, documenting tertile changes at 6 months of the bone resorption marker urinary N-terminal telopeptide (u-NTX) and the changes in BMD as assessed by dual-energy X-ray absorptiometry (DXA) in total hip, trochanter and spine at 3 years in postmenopausal women⁽ Reference Greenspan, Resnick and Parker ^{98 )}. Fig. 3(b) is based on quantitative data obtained by anabolic treatment with parathyroid hormone documenting tertile changes at 3 months of the bone formation marker N-terminal procollagen I propeptide (PINP) and the change in areal and volumetric BMD at 12 months, as assessed in the spine by DXA and quantitative computed tomography in postmenopausal women⁽ Reference Greenspan, Resnick and Parker ^{98 ,} Reference Bauer, Garnero and Bilezikian ^{113 )}.

Table 2 Changes from baseline in serum tartrate-resistant acid phosphatase (s-TRAP), carboxy terminal telopeptide of collagen type I (s-CTX) and procollagen type I N-terminal propetide (s-PINP) in response to antiresorptive and bone formation-stimulating treatments*

p.o., Oral; i.v., intravenous; s.c., subcutaneous; PTH, parathyroid hormone.

* Data have been adapted from the references Naylor & Eastell⁽ Reference Naylor and Eastell ^{8 )}, Henriksen et al. ⁽ Reference Henriksen, Tanko and Qvist ^{67 )} and Bonjour et al. ⁽ Reference Bonjour, Benoit and Pourchaire ^{127 )}.

Evidence from interventional drug trials, where the placebo groups received Ca and vitamin D, as well as in nutritional interventions with fortified food, shows a reduction in bone resorption markers of 20–30 % (as outlined in more detail in the ‘Study designs to test the effects of food products on bone metabolism’ section). The decrease in BTM during antiresorptive treatment is inversely related to the subsequent increase in aBMD, predominantly at the lumbar spine⁽ Reference Delmas ^{80 ,} Reference Delmas, Hardy and Garnero ^{81 )}. Several studies of postmenopausal women treated with antiresorptive agents (hormone replacement therapy, raloxifen, bisphosphonates) have indicated that the degree of short-term reduction in BTM (after 3 to 6 months) correlates with the observed long-term increase in aBMD after 1 to 3 years of treatment⁽ Reference Garnero, Sornay-Rendu and Duboeuf ^{79 ,} Reference Delmas, Hardy and Garnero ^{81 ,} Reference Garnero, Shih and Gineyts ^{95 –} Reference Ravn, Hosking and Thompson ^{99 )}. Reduction in fracture risk is only partially explained by the documented changes in aBMD (4–28 %)⁽ Reference Bonnick and Shulman ^{91 ,} Reference Cummings, Karpf and Harris ^{100 –} Reference Watts, Geusens and Barton ^{104 )}. There is also some evidence that even a modest decrease in bone mass is nevertheless associated with some fracture risk reduction⁽ Reference Chapurlat, Palermo and Ramsay ^{105 )}. There is thus substantial evidence that bone mass or changes thereof does not explain satisfactorily either the skeletal fragility of osteoporosis or the effects of bone active agents⁽ Reference Heaney ^{106 )}. It is therefore possible that changes in other determinants of bone strength, including the rate of bone turnover, may be more predictive of anti-fracture efficacy than changes in aBMD. In fact, several studies confirmed that short-term reductions in bone turnover were associated with a reduction in vertebral and/or non-vertebral fracture risk in women treated with hormone replacement therapy, raloxifene, risedronate and alendronate⁽ Reference Bauer, Black and Garnero ^{88 ,} Reference Hochberg, Greenspan and Wasnich ^{102 ,} Reference Bjarnason, Sarkar and Duong ^{107 –} Reference Reginster, Sarkar and Zegels ^{109 )}.

In contrast to antiresorptive agents, PTH fragments administered intermittently in low doses increase bone remodelling, stimulating bone formation preferentially over bone resorption, and resulting in net gain of bone and reducing fracture risk⁽ Reference Cranney, Papaioannou and Zytaruk ^{110 –} Reference Kraenzlin and Meier ^{112 )}. Increases in BTM indicate a skeletal response to PTH. There is evidence that early measurement of bone formation markers (OC, BAP and particularly PINP) correlates positively with the subsequent BMD response to PTH (Fig. 3(b))⁽ Reference Bauer, Garnero and Bilezikian ^{113 –} Reference Lane, Sanchez and Genant ^{116 )}. These data suggest that serial BTM measurements could become useful in identifying skeletal responders to an anabolic therapy with PTH. Based on results from clinical studies and improved technical performance on automated platforms, several national and international guidelines recommend these markers in the follow-up of patients treated for postmenopausal osteoporosis⁽ Reference Kanis, McCloskey and Johansson ^{32 ,} Reference Bergmann, Body and Boonen ^{117 , 118 )}.

Usefulness of biomarkers of bone turnover in clinical nutrition research

In the process of generating scientific support for claims on foods, BTM were deemed as not sufficiently indicative of bone health to provide evidence for enhanced function or reduced osteoporosis risk⁽ Reference Prentice, Bonjour and Branca ^{29 )}. However, over the last 10 years there have been improvements in analytical performance and new knowledge of how biochemical markers can be of great utility in osteoporosis management. Because BTM measurement can assess response to anti-osteoporotic medications, the time has come to use BTM levels to reflect bone formation and resorption in response to foods and nutrients.

Postmenopausal women

In a 12-month intervention involving postmenopausal women (mean age 60–62 years), the consumption of Ca- and vitamin D-fortified milk and yogurt favourably influenced vitamin D status and bone metabolism, as expressed by changes in serum 25(OH)D, PTH, CTX and IGF-I, as compared with a control group maintaining their usual diet or a group receiving equivalent Ca as lactate–gluconate and bicarbonate salts⁽ Reference Manios, Moschonis and Trovas ^{119 )} (Fig. 4). A prospective cross-over trial in postmenopausal women (mean age 59 years) found that supplementation with non-fortified milk for 6 weeks induced a decrease in serum PTH and BTM, including CTX, and a non-significant increase in IGF-I⁽ Reference Bonjour, Brandolini-Bunlon and Boirie ^{120 )}. In a randomised study of postmenopausal women (mean age 57 years), consumption of a vitamin D- and Ca-fortified soft white cheese significantly reduced the sensitive biochemical marker of bone resorption TRAP 5b and increased serum IGF-I levels⁽ Reference Bonjour, Benoit and Rousseau ^{121 )}.

Fig. 4 Changes (%) in serum biochemical indexes at 5 months (a) and in bone mineral density (BMD) at 12 months (b) in response to fortified dairy products (▓) for 12 months as compared with controls (░) in postmenopausal women. Data are adapted from Manios et al. ⁽ Reference Manios, Moschonis and Trovas ^{119 )}. PTH, parathyroid hormone; CTX, carboxy terminal telopeptide of collagen type I; IGF-I, insulin-like growth factor-I. P= Probability level for treatment ×  time interaction effect as compared with the control group.

Elderly individuals

Favourable effects of milk, fortified or not with vitamin D or Ca, on bone turnover in elderly individuals have been documented in several intervention studies⁽ Reference Tenta, Moschonis and Koutsilieris ^{42 ,} Reference Bonjour, Brandolini-Bunlon and Boirie ^{120 ,} Reference Heaney, McCarron and Dawson-Hughes ^{122 –} Reference Manios, Moschonis and Panagiotakos ^{124 )}.

Interventions with dairy products such as yogurts⁽ Reference Heaney, Rafferty and Dowell ^{125 ,} Reference Bonjour, Benoit and Payen ^{126 )} or soft white cheese⁽ Reference Bonjour, Benoit and Payen ^{126 –} Reference Bonjour, Benoit and Pourchaire ^{128 )} have shown reductions in bone turnover. In institutionalised elderly individuals with vitamin D deficiency and secondary hyperparathyroidism, an 8-week consumption of vitamin D- and Ca-fortified yogurts saw increases in serum 25(OH)D, associated with a reduction in the BTM CTX and TRAP 5b, and a normalisation of previously high serum PTH⁽ Reference Bonjour, Benoit and Payen ^{126 )}. Elderly patients hospitalised for an osteoporotic fracture of the proximal femur often present with signs of undernutrition, particularly protein malnutrition⁽ Reference Bonjour, Schurch and Rizzoli ^{129 )}. In a randomised study, protein repletion by the daily consumption of a 20 g casein-enriched food v. an isoenergetic non-protein supplement led to a significant elevation in serum IGF-I by the end of the 6-month dietary intervention⁽ Reference Schürch, Rizzoli and Slosman ^{58 )}. The serum IGF-I response to 20 g/d protein repletion in patients with a recent hip fracture was observable by 7 d, with no further rise in serum IGF-I⁽ Reference Chevalley, Hoffmeyer and Bonjour ^{130 )}. Thus, in hip fracture patients, the long-term effects of various protein preparations on IGF-I can be predicted from changes observed as early as 1 week following protein supplementation⁽ Reference Chevalley, Hoffmeyer and Bonjour ^{130 )}.

Anorexia nervosa

In young women with anorexia nervosa, aBMD is reduced at several skeletal sites⁽ Reference Grinspoon, Thomas and Pitts ^{131 )}. Body weight, but not oestrogen use, is a significant predictor of aBMD in these patients. Together with low endogenous oestrogen levels and Ca deficiency, a low protein intake very probably also contributes to the bone deficit observed in anorexia nervosa. However, low Ca intake has not been consistently reported and 25(OH)D serum levels are usually normal⁽ Reference Warren ^{132 )}. The BTM of bone formation, serum OC and BAP, are significantly reduced in adolescents whereas in adults with anorexia nervosa, the markers of bone resorption are elevated while those of formation are suppressed⁽ Reference Soyka, Misra and Frenchman ^{133 ,} Reference Dominguez, Goodman and Sen Gupta ^{134 )}. The serum level of IGF-I is low in anorexia nervosa⁽ Reference Soyka, Grinspoon and Levitsky ^{135 ,} Reference Hotta, Fukuda and Sato ^{136 )}, an observation quite compatible with insufficient protein intake⁽ Reference Soyka, Grinspoon and Levitsky ^{135 )} although low protein intake has not been consistently recorded. Nutritional rehabilitation reversed the uncoupling of bone remodelling⁽ Reference Caillot-Augusseau, Lafage-Proust and Margaillan ^{137 )} and led to a significant increase in BTM of formation followed by increases in aBMD in as little as 3 months⁽ Reference Dominguez, Goodman and Sen Gupta ^{134 )}. This reversibility occurs before oestrogen levels rise and menses return, suggesting that nutritional factors are of paramount importance⁽ Reference Warren ^{132 ,} Reference Dominguez, Goodman and Sen Gupta ^{134 )}.

The serum level of IGF-I was reported to increase rapidly after the onset of intravenous hyperalimentation, followed by a progressive elevation in serum osteocalcin⁽ Reference Hotta, Fukuda and Sato ^{136 )}. IGF-I level changes were dependent on variations in the nutritional state⁽ Reference Soyka, Misra and Frenchman ^{133 )}. Increased GH levels accompanied by decreased IGF-I are seen in anorexia nervosa, suggesting an acquired resistance to GH that reversed with refeeding⁽ Reference Warren ^{132 )}. The administration of recombinant IGF-I alone has only a small bone-sparing effect⁽ Reference Grinspoon, Thomas and Miller ^{138 )}. This is not surprising, since the anabolic effects of IGI-I on bone formation are tempered by poor nutrition, particularly by insufficient protein intake⁽ Reference Bourrin, Ammann and Bonjour ^{139 )}.

Intensive exercise

In female athletes or ballet dancers, intensive exercise in the absence of appropriate nutrition can lead to hypothalamic dysfunction with delayed menarche and disruption of menstrual cyclicity, bone loss and increase risk of fracture⁽ Reference Warren and Perlroth ^{140 –} Reference Warren, Brooks-Gunn and Fox ^{142 )}. Nutritional restriction, more common when leanness confers advantages for performance, can play an important role in the disturbance of the female reproductive system when combined with intense physical activity. Insufficient energy intake with respect to energy expenditure impairs the secretion of gonadotropin-releasing hormone and thereby leads to a state of hypo-oestrogenism. This hormonal disturbance is reversed by nutritional rehabilitation leading to an increase in both luteinising hormone and follicle-stimulating hormone. These changes occur well before oestrogen rises and menses return, suggesting the dominant effect of nutritional factors⁽ Reference Warren ^{132 )}. However, the relative contribution of insufficient protein intake with low IGF-I remains to be assessed since it is often associated with reduced energy intake. The consequence on bone metabolism of acute experimental energy restriction has been investigated in young exercising female volunteers⁽ Reference Ihle and Loucks ^{143 )}. This study indicated that severe reduction in energy availability rapidly induces a negative uncoupling between bone formation and bone resorption⁽ Reference Ihle and Loucks ^{143 )}. The resulting bone loss may become irreversible with persistence of such a negative uncoupled remodelling, explaining the increased risk of fragility fractures in anorexia nervosa⁽ Reference Warren ^{132 ,} Reference Misra and Klibanski ^{144 )} as well as in certain athletes and ballet dancers⁽ Reference Warren and Perlroth ^{140 ,} Reference Warren, Brooks-Gunn and Fox ^{142 )}.

Induced weight loss in overweight/obese subjects

Reduction in energy balance induced by energy restriction is often associated with bone loss that varies according to sex, age, ovarian function and initial fat mass (for a recent review, see Shapses & Sukumar⁽ Reference Shapses and Sukumar ^{145 )}). The degree of bone loss also depends upon the kind of diet accompanying the energy restriction. Bone mineral loss during body-weight reduction may not be fully recovered with weight regained, at least in postmenopausal women⁽ Reference Villalon, Gozansky and Van Pelt ^{146 )}. Hence, it is important to understand the underlying pathophysiological mechanisms and thereby define the best strategies to prevent bone loss during energy restriction in overweight/obese individuals. Several studies have used bone health biomarkers, particularly with the objective to identify the key implicated nutrients.

Ca intake typically decreases with energy-restriction diets. In this condition, the reduced Ca intake is associated with elevation in both serum PTH and biochemical markers of bone remodelling⁽ Reference Jensen, Kollerup and Quaade ^{147 ,} Reference Ricci, Heymsfield and Pierson ^{148 )}. Consumption of Ca salt supplements can suppress the rise in both serum PTH and BTM⁽ Reference Jensen, Kollerup and Quaade ^{147 ,} Reference Ricci, Chowdhury and Heymsfield ^{149 –} Reference Riedt, Schlussel and von Thun ^{151 )}.

Dairy products may prevent bone loss during energy restriction not only by providing Ca but also by supplying proteins. Weight-reducing diets that are high in dairy products can suppress the rise in bone turnover and prevent bone loss⁽ Reference Bowen, Noakes and Clifton ^{152 ,} Reference Thorpe, Jacobson and Layman ^{153 )}. In postmenopausal women, a randomised 12-month controlled study showed that higher dietary protein intake (86 v. 60 g/d) during weight reduction increased serum IGF-I and attenuated loss of aBMD at several skeletal sites and volumetric BMD at the distal tibia⁽ Reference Sukumar, Ambia-Sobhan and Zurfluh ^{154 )}. The additional consumed proteins were from various food sources including dairy products, with a whey supplement providing 6 g protein/d⁽ Reference Sukumar, Ambia-Sobhan and Zurfluh ^{154 )}. Combining dairy foods with physical exercise⁽ Reference Ryan, Nicklas and Dennis ^{155 ,} Reference Silverman, Nicklas and Ryan ^{156 )} might be appropriate to prevent bone loss during weight loss in overweight/obese human adults⁽ Reference Josse and Phillips ^{157 )}. In a recent study, hypo-energetic diets higher in dairy Ca and protein combined with daily exercise favourably affected bone health biomarkers as follows: decrease in PTH; increase in bone formation markers (OC and PINP) with no change in bone resorption markers (CTX and N-terminal telopeptide (NTX)); increment in osteoprotegerin:receptor activator of NFκB ligand (RANKL) ratio⁽ Reference Josse and Phillips ^{157 )}.

Issues with randomised controlled trials in osteoporosis

An RCT is the clinical study design that best permits strong causal inference between a tested drug or nutrient and a given outcome⁽ Reference Blumberg, Heaney and Huncharek ^{18 ,} Reference Guyatt ^{20 )}. Previous reviews or position papers have emphasised the distinct differences in the evidence that can be obtained by using RCT in the context of drug testing, as compared with the development of nutrient requirement or dietary guidelines⁽ Reference Blumberg, Heaney and Huncharek ^{18 ,} Reference Biesalski, Aggett and Anton ^{23 ,} Reference Fenton and Fenton ^{24 ,} Reference Heaney, Marcus, Feldman and Nelson ^{158 ,} Reference Heaney, Marcus, Feldman and Nelson ^{159 )}. In the osteoporosis context, in order to demonstrate the efficacy of a new medication, regulatory agencies usually require RCT of 3 years’ duration, with fracture as the primary endpoint. In the nutrition field, all types of research are useful in setting dietary recommendations: observational studies, to document prevalence and generate hypotheses; preclinical experiments to identify physiological and biochemical pathways; qualitative studies to understand the human condition; and, finally, RCT, to establish a relationship of causality⁽ Reference Blumberg, Heaney and Huncharek ^{18 ,} Reference Fenton and Fenton ^{24 ,} Reference Heaney and Blumberg ^{25 ,} Reference Mann ^{160 )}.

Ethical consideration in randomised controlled trials and sample size

The demonstration that a drug is efficacious for reducing the burden of fragility fractures necessitates the design of RCT with very large sample sizes. It is especially the case when, for ethical reasons, the new therapy is tested in so-called ‘active comparator trials’, i.e. when the recently developed medication is compared with an approved drug with convincing antifracture efficacy in patients at high risk of fragility fractures⁽ Reference Bouxsein and Delmas ^{161 )}. Moreover, a placebo-controlled trial can neither be considered ethically acceptable when the new drug is tested in patients at low risk of fractures. This study design requires a much larger number of subjects, as compared with previous RCT conducted in patients at high risk of fragility fractures⁽ Reference Bouxsein and Delmas ^{161 )}. Thus, proper design and execution of long-duration large RCT aimed at demonstrating the antifracture efficacy of a drug impose requirements that are difficult to meet⁽ Reference Heaney, Marcus, Feldman and Nelson ^{158 )}. This difficulty also encompasses both the narrow response due to the placebo effect and the non-random loss of enrolled subjects that convert a randomised trial into a cohort study⁽ Reference Heaney, Marcus, Feldman and Nelson ^{158 )}. It can be expected to be magnified in long-duration RCT aimed at testing the antifracture efficacy of foods or nutrients in osteoporotic patients⁽ Reference Blumberg, Heaney and Huncharek ^{18 ,} Reference Heaney, Marcus, Feldman and Nelson ^{158 )}.

Surrogate endpoints for antifracture efficacy of new drugs

Therefore, the development of surrogate endpoints for antifracture efficacy of new drugs has been considered not only for foods and nutrients⁽ Reference Heaney, Marcus, Feldman and Nelson ^{158 ,} Reference Heaney, Marcus, Feldman and Nelson ^{159 )}, but also for drugs⁽ Reference Bouxsein and Delmas ^{161 )}. Taking into account not only the large sample size but also the very long duration of conducting RCT with fragility fracture as the primary outcome, it is worth considering the measurement of fracture-related data that can be obtainable within a few weeks or months after the onset of the intervention. In clinical investigation, it is recognised that the early bone remodelling response is a good predictor of fragility fracture⁽ Reference Heaney, Marcus, Feldman and Nelson ^{158 ,} Reference Heaney, Marcus, Feldman and Nelson ^{159 )}. Effective pharmaceutical agents, whether acting via anticatabolic or anabolic mechanisms, very promptly influence bone remodelling, as biochemically expressed by measuring BTM changes⁽ Reference Heaney, Marcus, Feldman and Nelson ^{158 )}. These pharmoco-kinetic and -dynamic characteristics, that combine an early response to intervention with a reliable predictability of fragility fractures, make BTM tools better suited to assess the effects of foods or nutrients in clinical research on osteoporosis than the delayed changes in DXA-measured aBMD or bone mineral content. Furthermore, the effect size is usually larger with BTM after a few months than the changes in BMD/bone mineral content that can be observed after 1, or more frequently, 2 years of intervention.

Short-duration randomised controlled trials in nutrition clinical research

In adulthood, the effects of foods on BTM can be assessed in appropriately designed RCT. Examples of such studies are described in the ‘Study designs to test the effects of food products on bone metabolism’ section. The short duration of these RCT limits the number of non-compliant participants. Furthermore, one may also expect that few subjects will be completely lost, i.e. without outcome measurement. This makes the ‘intention-to-treat’ data analysis more valuable for comparison with the ‘per protocol’ analysis that will only include treatment-adhering participants, and thereby provides an estimate of the effect size, some sort of ‘dose–response’, of the intervention.

A physiologically meaningful set of biochemical variables can be selected and assessed at several time points. For instance, when nutritional products are evaluated for their capacity to improve vitamin D status and bone metabolism, changes in serum 25(OH)D, PTH, bone formation and resorption markers should provide information on the dynamic interrelationships between these variables. Measuring the circulating level of IGF-I can assess the response to an increase in protein intake. These short-term RCT studies should provide not only qualitative information on the response but also quantitative data on both the magnitude and the pharmacokinetics of the effects.

The advantage of cross-over design in nutritional investigation

A powerful methodology is the cross-over design, also called ‘n of 1 RCT’⁽ Reference Montori and Guyatt ^{162 ,} Reference Dawson and Trapp ^{163 )}. Herein there are two intervention periods, separated by a ‘washout’ period, the duration of which depends upon how long the effect of the intervention lingers after its discontinuation. For some anti-osteoporotic agents such as bisphosphonate compounds, the effect can last months or years. For foods and nutrients acting on bone, the effects may be ‘washed out’ much more rapidly, i.e. within a few days or weeks, with the exception of large doses of vitamin D. Within a trial, each participant acts as his or her own control, and permits comparison between and within groups⁽ Reference Mills, Chan and Wu ^{164 )}. This design eliminates any genetic influence and reduces bias associated with imbalance in known and unknown confounding variables, as compared with randomised parallel-group controlled trials enrolling two unrelated groups of subjects⁽ Reference Mills, Chan and Wu ^{164 )}.

Comparative effectiveness research

An important recent trend in medical research is to apply ‘comparative effectiveness research’ (CER)⁽ Reference Iglehart ^{165 )}. CER is a marked departure from the efficacy standard that any new intervention should outperform placebo or randomised controls. There is particular emphasis placed on the leading question regarding clinical effectiveness of care. A CER study is aimed at determining the value of an intervention over another. It answers two relevant questions: (1) how effective is the intervention, if enrolled subjects follow it as prescribed?; (2) how likely is it that the enrolled participants follow it as prescribed?⁽ Reference Kaats, Preuss and Leckie ^{166 )}. In a classically designed study, the answers to these two questions would be given by comparing the results computed by ‘intention to treat’ with those analysed ‘per protocol’. CER is a recommended priority for nutrition research⁽ Reference Iglehart ^{165 )}. Although CER will pose challenges for medical and nutritional industries, it will present numerous opportunities to improve the information base supporting nutritional products, especially when they are integrated into other medical interventions⁽ Reference Kaats, Preuss and Leckie ^{166 )}. Recent CER nutrition studies have been reported on the bone effects of several vitamin D and Ca supplement regimens⁽ Reference Kaats, Preuss and Croft ^{167 ,} Reference Michalek, Preuss and Croft ^{168 )}.