Osteoporosis and related fracture represent a significant public health burden in many countries. In the past, excessive protein intake has been thought to be detrimental to bone health by inducing chronic metabolic acidosis, which could lead to hypercalciuria and accelerated mineral dissolution.1 However, short-term well-controlled feeding studies showed that a high-protein diet did not have adverse effects on calcium retention and bone metabolism.2–4 Furthermore, it has been suggested that adequate protein intake may benefit the skeleton of the elderly through providing amino acids and increasing calcium absorption5 and circulating insulin-like growth factor 1 (IGF-1).6, 7 Although epidemiologic studies with younger subjects have shown variable results on the effect of protein on bone mass and fracture risk,8–10 most studies with elderly subjects have shown that relatively high protein intakes were associated with reduced bone loss11–13 and reduced risk of hip fracture.14, 15 A recent systemic review showed that in cross-sectional studies with older subjects, protein intake positively correlated with bone mineral density (BMD) and could explain 1% to 2% of the variation in BMD.16 Two milk proteins intervention studies in frail hip fracture patients have shown positive effects in reducing bone loss.6, 17 To the best of our knowledge, there have been no well-designed randomized trials of sufficient duration or power to examine the effects of increased dietary protein intake on bone density in older community-dwelling women. Thus the primary aim of this study was to evaluate the effect of protein supplementation on BMD and strength, and the secondary aim was to examine the effect on calcium excretion and serum IGF-1 of older postmenopausal women. Since a randomized, placebo-controlled trial of elderly men and women showed that higher protein intake reduced bone loss only in individuals who received a calcium supplement,18 we also provided 600 mg of calcium to both the protein and the placebo groups in the test drink.
Subjects and Methods
Study subjects were recruited from April to September 2007 using a population-based approach in which a random selection of women (n = 6065) aged 70 to 80 years on the electoral roll in the metropolitan area of Perth, Western Australia, received a letter inviting them to join the study. Over 98% of women of this age are on the Western Australian electoral roll. Of the 829 women who responded to the letter, 256 attended clinic screening, and 219 women who met the inclusion criteria joined the study. The exclusion criteria were participation in another clinical trial during the last 12 weeks, previous osteoporotic fracture, currently or within last year taking medication for osteoporosis apart from calcium or vitamin D, taking steroid tablets in the past 3 months or have taken more than 7 g in total in lifetime, metabolic bone disease apart from osteoporosis, total-hip bone density more than 2 SD below the mean for age, lactose intolerance or dislike of milk products, high protein intake as assessed by a food frequency questionnaire (equivalent to protein intake of more than 1.5 g/kg of body weight per day), cognitive impairment (Mini Mental State score < 24), body mass index (BMI) greater than 35 kg/m2, malabsorption disorders, celiac disease, clinical hepatic or renal insufficiency, clinical diagnosis of diabetes, and participants who in the opinion of the investigator would not be likely to complete the study for any reason. All procedures followed were in accordance with institutional guidelines and were conducted at the Sir Charles Gairdner Hospital in Perth. The study was approved by the Sir Charles Gairdner Hospital Human Research Ethics Committee, and all participants provided written informed consent. The study was conducted in compliance with the ethical principles of the Declaration of Helsinki and the International Conference on Harmonization Good Clinical Practices Guidelines and registered with the Australian Clinical Trials Registry (Registration Number ACTRN012607000163404).
A 2-year randomized, double-blind, placebo-controlled, prospective parallel study was undertaken. Eligible participants were randomized to one of two treatment groups (protein and placebo groups). The study used a computer-generated randomization sequence with a block size of 10 to assign participants to either a protein drink or a placebo drink in a ratio of 1:1. The randomization code was generated by one of the investigators (CWB) who did not have direct contact with participants, and the code was kept at the School of Public Health, Curtin University. In addition to assigning participants to intervention, Curtin University staff labeled the drinks and organized the delivery of the study drinks to participants' homes. The study participants and researchers at the Sir Charles Gairdner Hospital responsible for recruitment and assessment of outcomes measures remained blinded as to group assignment.
The supplemental drinks were developed by an experienced food technologist (VS). The test drinks were provided in three flavors—coffee, chocolate, and strawberry—and the nutrient content of the test drinks is shown in Table 1. Test drinks were delivered to participants' homes every 3 months, and participants were instructed to take a daily test drink before breakfast for the duration of the study.
Table 1. Nutritional Composition of the Test Drink (per 250 mL)
Participants in the protein group received a 250-mL skim milk–based high-protein supplement drink reconstituted with cold water from a powder that provided 30 g of protein, 600 mg of calcium, and 3.2 kJ/mL of energy. The high-protein product used skim milk plus whey protein isolate (Alacen 894, Fonterra Brands, Ltd., Palmerston North, New Zealand) to increase protein content.
Participants in the placebo group received a 250-mL skim milk–based supplement drink reconstituted with cold water from a powder that provided identical calories and calcium (600 mg of calcium and 3.2 kJ/mL of energy) but only 2.1 g of protein The extra energy content in the placebo drink was supplied by carbohydrate. Alginate natural flavoring and natural emulsifying agents were used to provide a similar texture and flavor to the drinks. Adherence to the study drinks was established by counting empty tins returned at the clinic visits at 1 and 2 years.
Total hip BMD was the primary outcome measure of this study and was measured by dual-energy X-ray absorptiometry (DXA) using a Hologic Discovery A fan-beam densitometer (Hologic Corp., Waltham, MA, USA) at baseline and years 1 and 2. The coefficient of variation (CV) at this site was less than 2% in our laboratory.
Quantitative computed tomographic (QCT) scans of the hip were undertaken at baseline and 24 months using a Phillips Brilliance 64-slice spiral QCT scanner (Philips Medical Systems, Andover, MA, USA). Scans were acquired at 120 kVp and 100 mA with a 1-mm slice thickness and a pitch of 1. The patients were supine on the CT scanner table, lying on top of a QCT PRO (Mindways Software, Inc., Austin, TX, USA) calibration phantom and bolus bag so that the calibration phantom extended from the lumbar vertebrae to below the lesser trochanter (TR). The scans were analyzed using the QCT PRO analysis software (Mindways Software, Inc., Austin, TX, USA), and volumetric BMD (vBMD) and geometric engineering values were obtained from the analysis. These included femoral neck cross-sectional bone area (CSA) related to strength in compression, the polar cross-sectional moment of inertia (CSMI) related to strength in torsion, and the buckling ratio (BR) related to strength in buckling.
Serum and 24-hour urine samples were collected at baseline and at 1 and 2 years for assessment of the secondary outcomes of this study—serum IGF-1 and 24-hour urinary serum. Venous blood samples were collected following a 12-hour fast from the antecubital vein of the forearm. Blood was collected into BD Vacutainer SST tubes (BD, Belliver Industrial Estate, Plymouth, UK). Analyses of IGF-1 were performed on sera stored at −80°C, and serum IGF-1 concentration was determined using a solid-phase, enzyme-labeled chemiluminesent immunometric assay (IMMULITE 2000 IGF-1, Siemens Medical Solutions, Los Angeles, CA, USA).
The participants collected a 24-hour urine sample on the third day of the food recording period in a 5-L plastic collection bottle that contained 20 mL of 1 M HCl. They discarded the first urine specimen of the morning and collected all specimens for 24 hours. The urine sample was weighed, and a 5-mL sample was stored at −20°C until analyzed. Urinary calcium concentration was determined using an Architect c16000 Analyser with calcium reagent (Abbott Diagnostics, Abbott Laboratories, Abbott Park, IL, USA).
Dietary intake was assessed by a 3-day weighed food record (2 weekdays, 1 weekend day). Participants were asked to record everything they ate and drank for 3 consecutive days using either the electronic food scales provided or household measures. They watched a training video on how to complete their food record prior to undertaking the food record. When the food record was returned 1 week later, the participant was interviewed to clarify types and amounts of food or beverages recorded. The food record was analyzed using the AUSNUT99 database (Foodworks Professional Edition, Version 3.02, Xyris Software Pty Ltd, Highgate Hill, QLD, Australia) by nutritionists trained in dietary assessment. The net endogenous acid production (NEAP) was estimated with the following formula using energy-adjusted protein and potassium intakes19, 20:
Anthropometric measurements were performed with subjects in light clothes and without shoes. Physical activity level was assessed by the International Physical Activity Questionnaire (IPAQ) short form (www.ipaq.ki.se). Demographic information for participants, including health history, education, past occupation, and smoking history, was collected using a demographic questionnaire. The Mini Mental State Test was administered at baseline with the aim of excluding participants who demonstrated significant cognitive impairment.
Using a previously validated method,21 participants were asked to fill out an adverse-event diary in which each contact with a physician was recorded. At 6-month intervals, the diary was return to the study center at clinic visit or by mail. The adverse events were coded using the International Classification of Primary Care (ICPC2 Plus) system database of disease coding, a validated method of event recording developed for use in general practice.22
Sample size calculation
Power calculations were performed prior to commencement of the study. A sample size of 85 in each group was required to detect a difference of 3% on change in hip BMD, the primary outcome variable of the study, assuming an SD of 6% based on our previous study, at 90% power and 5% level of significance. A 3% increase was considered reasonable based on our previous epidemiologic studies.13 This number was increased to 110 per group (total of 220) to allow for a 30% predicted dropout rate that we reported in previous studies of a similar age group.
Descriptive statistics are reported as mean ± SD and differences as mean ± SEM for all variables unless otherwise stated. The normality of continuous variables was checked through the construction of histograms. Two variables that were not normally distributed (serum IGF-1 and 24-hour urinary calcium) were log-transformed. Baseline characteristics between the groups were compared by Student's t test or chi-square test as appropriate. The main intention-to-treat analysis included all subjects who entered the study and had at least one follow-up measurement to evaluate the effectiveness of protein supplementation while ignoring the lack of full compliance. Treatment, time, and interaction effects during the 24 months study period were examined using linear mixed-effects model analysis with time since baseline as the timeline, group effects as fixed, and subject effects as random and a first-order autoregressive covariance structure (AR1). If the linear mixed-model analysis indicated a significant treatment and time interaction, the treatment effects at years 1 and 2 were analyzed by analysis of covariance (ANCOVA) with baseline values and age as covariates. If the linear mixed-model analysis indicated a significant time effect, the time effects at 1 and 2 years were evaluated by one-factor repeated-measures ANOVA. In further analyses, baseline protein and carbohydrate intakes were added as covariates in addition to treatment effects on bone and biochemistry outcomes, and the significance of the interaction terms of baseline protein or carbohydrate intakes and treatment group was tested in the models. The normality and independence of the residuals and the homogeneity of variance of each model were checked by residual plots (normal probability plot and plot of residuals versus treatment, subject, and predicted values). Values of p < .05 (two-tailed) were regarded as statistically significant. All data were analyzed by PASW (Version 18; SPSS, Inc., Chicago, IL, USA).
Participant flow through the study is shown in Fig. 1. The 196 subjects who had at least one follow-up measurement were included in the analysis. There were no significant differences between the protein and placebo groups in any baseline characteristics (Table 2). At study entry, the mean age of participants was 74.3 ± 2.7 years, and the mean protein intake was 76 ± 17 g/d (1.1 ± 0.3 g/kg of body weight per day). Seventeen (8.7%) participants had protein intake below the Australian recommendation of the Estimated Average Requirement (EAR, 0.75 g/kg of body weight per day), and 55 (28.2%) had protein intake below the Recommended Dietary Intake (RDI, 0.94 g/kg of body weight per day) for women aged over 70 years.23 Only 5 subjects (2.6%) had protein intake below the World Health Organization (WHO) recommended population average requirement of 0.66 g/kg of body weight per day for adults.24 During the 2 years, 16 subjects (14.7%) in the protein group and 22 subjects (20.0%) in the placebo group withdrew from the study, and 16 subjects (14.7%) in the protein group and 17 subjects (15.5%) in the placebo group discontinued the test drink. There was no significant difference between the protein group and the control group in the number of subjects who discontinued the test drink or were lost to follow-up. The compliance rate with the test drink, as determined from empty test drink containers, was higher in the protein group (87.1%) than in the placebo group (80.8%, p = .025).
Table 2. Baseline Characteristics of Study Participants
Protein group (n = 101)
Placebo group (n = 95)
Note: There were no significant differences between the two groups in any variables.
74.2 ± 2.8
74.3 ± 2.6
159.8 ± 6.3
159.8 ± 5.7
66.8 ± 11.1
69.6 ± 11.3
Body mass index (kg/m2)
26.1 ± 3.8
27.2 ± 4.0
Mini Mental State score (unit)
28.3 ± 1.4
28.1 ± 1.5
Physical activity (MET-minutes/week)
449 ± 391
398 ± 376
Energy intake (kJ/d)
7056 ± 1552
7102 ± 1379
Protein intake (g/d)
76 ± 18
76 ± 16
Protein intake (g/kg of body weight per day)
1.1 ± 0.3
1.1 ± 0.3
Calcium intake (mg/d)
978 ± 376
1021 ± 439
Total-hip aBMD (mg/cm2)
838 ± 123
859 ± 125
Femoral neck aBMD (mg/cm2)
693 ± 93
710 ± 105
Insulin-like growth factor 1 (ng/mL)
106.3 ± 39.1
111.1 ± 38.6
(n = 94)
(n = 91)
Urinary calcium excretion (mmol/d)
2.9 ± 1.9
3.0 ± 1.6
(n = 99)
(n = 93)
Supplement effects on diet
The effects of the supplements on those completing all three diet records are shown in Table 3. With drink supplementation, average daily protein intake increased from 76 to 93 and 95 g/d in the protein group at years 1 and 2, respectively, which was significantly higher than that in the placebo group. Carbohydrate intake increased in the placebo group at 1 and 2 years compared with that of baseline and was higher than in the protein group at both time points both in absolute and precent energy terms. In the protein group, carbohydrate intake as a percent of energy fell at 1 and 2 years compared with baseline. Energy contribution from fat decreased in the protein group at years 1 and 2 and in the placebo group at year 2. Fiber and potassium intakes decreased in both groups at years 1 and 2. There was a significant increase in calcium intake with the supplementation in both the calcium and placebo groups at years 1 and 2. There was a significant reduction in estimated energy-adjusted NEAP at years 1 and 2 in the placebo group, and the placebo group had significantly lower NEAP at these two time points than the protein group.
Table 3. Change in Energy and Nutrient Intakes During the Study in Participants With Completed Dietary Intakes Data at Baseline and 1 and 2 Years
Note: Subject numbers for protein group n = 88, placebo group n = 85.
Significantly different from that of the placebo group at the same time point, p < .001.
Significantly different from that of baseline, p < .001.
Significantly different from that of the placebo group at the same time point, p < .05.
Significantly different from that of baseline, p < .05.
There were no significant differences between the protein and placebo groups in any bone variables at baseline (Tables 2 and 4). Total-hip and femoral neck areal BMD (aBMD) and vBMD fell significantly from baseline at 1 and 2 years in both groups (Fig. 2, Table 4). However, there were no detectable changes in QCT femoral neck cross-sectional area, bucking ratio, or polar CSMI (Table 4). There were no group-by-time interactions in the linear mixed-effects model analysis for any of the bone variables, indicating that there were no between-groups differences in the changes in bone density or strength over the study period. In further analyses, when baseline protein and energy intakes and their interaction terms with treatment group were added in the models, there was no significant baseline protein or energy intake and treatment interaction for all the bone variables examined, indicating that the treatment effect was not influenced by baseline protein or energy intakes. There was no significant correlation between the change in NEAP and change in total-body and hip BMD from baseline at years 1 and 2 (r = −0.133 to −0.006, p > .05).
Table 4. QCT Bone Density and Strength at Baseline and 2 Years
Baseline values (mean ± SD)
Difference (2 year–baseline, mean ± SE)
Protein (n = 67)
Placebo (n = 66)
Protein (n = 67)
Placebo (n = 66)
There were no significant group by time interaction in the linear mixed effects model analysis, indicating that there were no significant treatment effects.
There were significant treatment-to-time interactions in the linear mixed-effects model analysis for serum IGF-1 and 24-hour urinary calcium excretion. Compared with the placebo group, the protein group had significantly higher serum IGF-1 levels at years 1 (7.3% ± 2.5%, p = .004) and 2 (8.0% ± 3.3%, p = .016; Fig. 3). The serum IGF-1 level increased significantly from baseline in the protein group at 1 year and decreased significantly from baseline in the placebo group at 2 years. The 24-hour urinary calcium excretion increased significantly from baseline in both groups at 1 year but only in the protein group at 2 years (Fig. 3). Although the protein group had a greater increase in urinary calcium excretion than the placebo group at 2 years (13.4% ± 7.3%), this did not achieve statistical significance (p = .069). There was a significant positive correlation between the change in NEAP and change in urinary calcium excretion from baseline at year 2 (r = 0.292, p < .001) but not year 1 (r = −0.002, p = .977).
During the study period, there were no significant differences between the protein and placebo groups in the rate of incident cancer (protein group 5.0%, control group 5.3%), type 2 diabetes (protein group 3.0%, control group 1.1%), diarrhea (protein group 4.0%, control group 1.1%), esophageal reflux (protein group 2.0%, control group 5.3%), or fracture (protein group 3.0%, control group 3.2%). Two participants in the protein group reported constipation.
Our study showed that in older postmenopausal women with an average protein intake of 76 ± 17 g/d (1.1 ± 0.3 g/kg of body weight per day), there were no significant differences between the group that received the high-protein drink (which contained 30.1 g of protein per 250 mL) and the group that received the low-protein, high-carbohydrate drink (contained 2.1 g of protein per 250 mL) in changes in hip BMD and femoral neck bone strength over 2 years. The intervention had a biologic effect, as noted by the increased serum IGF-1 level in the protein group.
Effects on bone density
Protein represents 25% of bone by mass and 50% of bone by volume. During bone turnover, a significant proportion of amino acids in bone collagen cannot be reused. Therefore, it is expected that adequate protein intake is important for the maintenance of bone mass in the elderly, where high bone turnover rate exists. Besides providing amino acids as substrates for building matrix, protein intakes have been shown to be positively associated with the increased circulating levels of IGF-1, a recognized bone growth-promoting factor.6, 7 A number of cross-sectional and longitudinal studies with older subjects have shown that relatively high protein intakes were associated with reduced bone loss.11–13 However, in this study in older Western Australian women, 30 g of extra protein per day did not affect change in bone density or strength over 2 years. One possible explanation could be that the beneficial effects of protein on BMD observed in epidemiologic studies reflect long-term effects accumulated from early years. Another possible explanation for the lack of effect is the relatively high usual dietary protein intake of our study subjects of 1.1 g/kg of body weight per day, which is well above the EAR of 0.75 g/kg of body weight per day. This is consistent with the participants having a relatively adequate nutritional status, as evidenced by their BMI and nutritional intake. In a study in frail hip fracture patients that showed that 20 g of protein supplementation for 6 months could reduce bone loss at the proximal femur by 2.4% at 12 months, the average dietary protein intake was 45 to 51 g/d, much lower than the protein intake of 76 g/d in this study.6
Effects on calcium metabolism
The effect of dietary protein on calcium and bone metabolism has long been debated. In the past, it was generally considered that high protein intake could lead to chronic metabolic acidosis through the oxidation of the sulfur amino acids, which requires buffering in the bone and thus leads to increased urinary calcium excretion and bone loss.1 In this study, although the protein group had higher estimated NEAP at 1 and 2 years and there was a positive correlation between change in estimated NEAP and urinary calcium excretion from baseline to 2 years, there were no significant between-group differences in calcium excretion. However, this may have been due in part to the increase in calcium intake provided by the drinks that were given to both groups. Certainly at year 2 calcium excretion was higher in the protein group than in the placebo group, but this did not reach statistical significance. Moreover, this increase in urinary loss in the protein group did not result in detrimental effects on bone mass, suggesting that any increase in urinary calcium loss is ameliorated by the increased gut calcium absorption associated with increased protein intake5 rather than increased bone resorption. Indeed, previous short-term (1 week to 2 months) controlled feeding studies showed that a high-protein diet did not have adverse effects on calcium retention and bone metabolism.2–4 In conclusion, this study is the first long-term study demonstrating that increased protein intake does not have a beneficial or negative impact on bone mass over and above that of aging.
Effects on IGF-1
Our study observed a 7% to 8% increase in serum IGF-1 in the protein group in comparison with the placebo group after 1 to 2 years of intervention. IGF-1 is a bone growth-promoting factor,7 and it plays an important role in activating the osteoblast differentiation program and regulating 25-dihydroxyvitamin D3 1α-hydroxylase activity and the tubular reabsorption of phosphate in the kidney.25 Growth hormone secretion decreases with aging, and the circulating levels of IGF-1 in more than 30% of elderly people are lower than the young reference, which has implications in age-related osteoporosis.26 In this study, the increase in serum IGF-1 was not sufficient to result in a beneficial effect on bone mass, which could be due to the fact that the increase of IGF-1 by 7% to 8% observed in this study is rather moderate. In a study of hip fracture patients with lower baseline protein intake who received 20 g of protein supplementation daily and bone loss was reduced by 2.4% at 12 month, a 52% difference in serum IGF-1 after 6 months of supplementation was observed.6 Furthermore, an open-label randomized trial of 18-month milk drink intervention in 82 healthy schools girls showed an increase in serum IGF-1 in the milk group of 10%, similar to the effect size of this study, and a 2.6% greater gain in bone mineral content at the end of intervention.27 This suggests that a moderate increase in IGF-1 may be sufficient to produce beneficial effects on bone during growth but not in the aged, as in this study.
Strengths and limitations
The strengths of our study include the randomized, controlled design; study subjects who were representative of large numbers of individuals living in Western countries; and the high retention and compliance rates. A limitation of our study is that the study subjects had a relatively high dietary protein intake at baseline, which may explain the lack of effect of the protein intervention on the bone outcomes. Protein supplementation could be more effective in populations with low protein intake. Therefore, intervention study in such populations requires urgent attention. A further criticism may be that an increase in vegetable protein may have had a different effect.
In conclusion, in these healthy ambulant women with baseline protein intake well above the current Australian recommended EAR of 0.75 g/kg of body weight per day and the WHO recommended population average requirement of 0.66 g/kg of body weight per day, extra protein was not a critical beneficial or deleterious regulator of their bone mass or strength.
All the authors state that they have no conflicts of interest.
We are grateful to Fonterra Brands Limited for providing whey protein isolate (Alacen 894) free of charge and Anchor Foods (Fremantle, Western Australia) for providing processing assistance for the test drink powder. We would like to thank Tricia Knox and Linda Gregory of the PathWest Laboratory at the Royal Perth Hospital, Western Australia, for performing the biochemical analysis.
This study was supported by the Australian National Health Medical Research Council (Project Grant 458625) and the University of Western Australia Research Grants Scheme.
None of the funding agencies had any role in the conduct of the study; collection, management, analysis, or interpretation of the data; or preparation, review, or approval of the manuscript. KZ, XM, DAK, AD, VS, CWB, and RLP had full access to all the data in the study and take responsibility for the integrity of the data and the accuracy of the data analysis.
Authors' roles: Study design: KZ, DAK, AD, VS, CWB, and RLP. Study conduct: KZ, XM, DAK, AD, VS, and RLP. Data collection: KZ, XM, DAK, AD, VS, and RLP. Data analysis: KZ. Data interpretation: KZ and RLP. Drafting manuscript: KZ and RLP. Revising manuscript content: XM, DAK, AD, VS, and CWB. Approving final version of manuscript: KZ, XM, DAK, AD, VS, CWB, and RLP. KZ takes responsibility for the integrity of the data analysis.