Dietary Protein Intake and Risk of Osteoporotic Hip Fracture in Elderly Residents of Utah

Authors

  • Heidi J Wengreen,

    1. Department of Nutrition and Food Sciences, Utah State University, Logan, Utah, USA
    2. Center for Epidemiologic Studies, Utah State University, Logan, Utah, USA
    Search for more papers by this author
  • Ronald G Munger,

    Corresponding author
    1. Department of Nutrition and Food Sciences, Utah State University, Logan, Utah, USA
    2. Center for Epidemiologic Studies, Utah State University, Logan, Utah, USA
    • Department of Nutrition and Food Sciences, Utah State University, 4450 Old Main Hill, Logan, Utah 84322–4450, USA
    Search for more papers by this author
  • Nancy A West,

    1. Department of Nutrition and Food Sciences, Utah State University, Logan, Utah, USA
    2. Center for Epidemiologic Studies, Utah State University, Logan, Utah, USA
    Search for more papers by this author
  • D Richard Cutler,

    1. Center for Epidemiologic Studies, Utah State University, Logan, Utah, USA
    2. Department of Mathematics and Statistics, Utah State University, Logan, Utah, USA
    Search for more papers by this author
  • Christopher D Corcoran,

    1. Center for Epidemiologic Studies, Utah State University, Logan, Utah, USA
    2. Department of Mathematics and Statistics, Utah State University, Logan, Utah, USA
    Search for more papers by this author
  • Jianjun Zhang,

    1. Hipple Cancer Research Center, Department of Community Health, School of Medicine, Wright State University, Dayton, Ohio, USA
    Search for more papers by this author
  • Nancy E Sassano

    1. Center for Epidemiologic Studies, Utah State University, Logan, Utah, USA
    Search for more papers by this author

  • The authors have no conflict of interest.

Abstract

The role of protein intake in osteoporosis is unclear. In a case-control study in Utah (n = 2501), increasing level of protein intake was associated with a decreased risk of hip fracture in men and women 50–69 years of age but not in those 70–89 years of age. Protein intake may be important for optimal bone health.

Introduction: Protein is an important component of bone, but the role of dietary protein intake in osteoporosis and fracture risk remains controversial.

Material and Methods: The role of dietary protein intake in osteoporotic hip fracture was evaluated in a statewide case-control study in Utah. Patients, 50–89 years of age, with hip fracture (cases) were ascertained through surveillance of 18 Utah hospitals during 1997–2001. Age- and gender-matched controls were randomly selected. Participants were interviewed in their place of residence, and diet was assessed using a picture-sort food frequency questionnaire previously reported to give a useful measure of usual dietary intake in the elderly Utah population. The association between protein intake and risk of hip fracture was examined across quartiles of protein intake and stratified by age group for 1167 cases (831 women, 336 men) and 1334 controls (885 women, 449 men).

Results: In logistic regression analyses that controlled for gender, body mass index, smoking status, alcohol use, calcium, vitamin D, potassium, physical activity, and estrogen use in women, the odds ratios (OR) of hip fracture decreased across increasing quartiles of total protein intake for participants 50–69 years of age (OR: 1.0 [reference]; 0.51 [95% CI: 0.30–0.87]; 0.53 [0.31–0.89]; 0.35 [0.21–0.59]; p < 0.001). No similar associations were observed among participants 70–89 years of age. Results from analyses stratified by low and high calcium and potassium intake did not differ appreciably from the results presented above.

Conclusion: Higher total protein intake was associated with a reduced risk of hip fracture in men and women 50–69 years of age but not in men and women 70–89 years of age. The association between dietary protein intake and risk of hip fracture may be modified by age. Our study supports the hypothesis that adequate dietary protein is important for optimal bone health in the elderly 50–69 years of age.

INTRODUCTION

BONE FRACTURES CAUSED by osteoporosis have long plagued the elderly of industrialized countries and are now an emerging problem in developing countries.(1) Hip fractures are the most serious of all fractures related to osteoporosis.(2) Although the incidence of hip fracture is higher in women than in men, rates increase exponentially with age in both, and the age-specific rates in men lag those of women by only 5-7 years.(2, 3) The lifetime risk of suffering a hip fracture is 6.0% in white men and 17.5% in white women.(4) One year after hip fracture, 40% of patients cannot walk alone, 60% have difficulty with at least one essential activity of daily living, 27% have been admitted to a nursing home for the first time, and 20% have died.(5) Discovering new methods for the prevention of osteoporotic hip fractures will decrease this burden and may help many elders remain independent and have a higher quality of life in their later years.

Osteoporosis is characterized by low bone mass and microarchitectural deterioration of bone tissue resulting in increased bone fragility and susceptibility to fracture.(6) Although peak bone mass has a significant genetic component, modifiable risk factors, including nutrition, seem to play an important role in bone health throughout the lifecycle. Dietary interventions for the prevention and treatment of osteoporosis have long been focused on calcium and vitamin D, whereas other nutrients, including protein, have been less well studied.

Protein is an important component of bone,(7) and protein supplementation improves the medical outcome in elderly hip fracture patients,(8, 9) but the role of dietary protein intake in osteoporotic hip fractures remains controversial. Although several population-based observational studies have examined the relationship between dietary protein intake and risk of hip fracture or change in BMD in elderly adults, their results are conflicting.(10-16) Most epidemiological studies report that high dietary protein intake is positively associated with BMD.(12-14) However, only one previous study has reported a protective association between high dietary protein intake and risk of hip fracture.(11)

Although the inconsistency between dietary protein intake, BMD, and risk of fracture is not well understood, protein intake seems to play an important and complex role in bone growth, development, and maintenance. We examined the associations between dietary protein intake, including source of protein, and risk of osteoporotic hip fracture in a statewide case control study of older Utah residents. In addition, we evaluated possible modification of the protein intake-hip fracture association by age, gender, and other dietary components, including calcium and potassium intake.

MATERIALS AND METHODS

The Utah Study of Nutrition and Bone Health (USNBH) is a state-wide case-control study of risk factors for hip fracture in Utah residents 50-89 years of age. Cases were ascertained through surveillance of 18 Utah hospitals responsible for treating 98% of hip fracture cases in the state and were interviewed an average of 4.2 months after their hip fracture. Controls were randomly selected from the Utah Drivers License and Medicare databases and frequency-matched to cases by gender and 5-year age intervals. All study procedures were reviewed and approved by the institutional review board of each hospital and participating university. After obtaining written informed consent, an in-person interview was conducted at each participant's place of residence.

The percentage of subjects who refused to complete an interview was similar for cases and controls (23.2% and 24.0%, respectively); however, more cases than controls were deceased or too frail, ill, or demented to complete the interview (37.2% and 16.8%, respectively). An additional 2.8% of cases and 3.4% of controls could not be located. As a result, the overall participation rate was 36.9% for cases (n = 1338) and 55.8% for controls (n = 1360).

Data collection

Diet was assessed using a 137-item picture-sort food frequency questionnaire (FFQ). The Utah picture-sort FFQ was developed specifically for this population as an alternative to the more traditional paper-and-pencil format FFQ. Foods included in the Utah picture-sort FFQ were systematically selected from the 126-item Nurses' Health study FFQ, the National Cancer Institute (NCI) FFQ, and a list of commonly eaten foods as identified in studies of elderly in Utah. The picture-sort FFQ method, first developed by Kumanyika et al.,(17) engages respondents by having them sort cards with color photographs of foods into trays representing frequency of use over the past year (for controls) or the year before hip fracture (for cases). An interviewer records information about specific consumption patterns for foods within each frequency category. Spearman rank correlation coefficients between nutrient intakes estimated using the Utah picture-sort FFQ and the average of multiple 24-h recalls were comparable with those reported in the literature for traditional paper-and-pencil FFQs.(18) After adjustment for total energy intake, median Spearman rank correlation coefficients for all measured nutrients between intakes estimated using the picture-sort FFQ and the reference method were 0.50 for men 55-69 years of age (n = 51), 0.52 for men 70-85 years of age (n = 52), 0.55 for women 55-69 years of age (n = 54), and 0.46 for women 70-85 years of age (n = 51).(18) FFQ data were converted to usual daily intakes by transforming all consumption periods to consumption per day and multiplying by frequency of use and weight of the standard serving size according to the U.S. Food and Drug Administration's Dietary Guidelines for Americans.(19) Respondents with biologically implausible total energy intake (<2.51 MJ [600 kcal] or >20.92 MJ [5000 kcal]) were excluded from the analyses (n = 47). Basal metabolic rate (BMR) was estimated using the equations of Schofield(20) and was used to calculate the ratio of energy intake to basal metabolic rate (EI:BMR).

A detailed history of vitamin and mineral supplementation use was obtained during the interview. Daily intake from individual, combination, or multivitamin and mineral supplements was combined with dietary intake to give a measure of total daily intake of calcium, vitamin D, and potassium. Calcium users were defined as participants obtaining at least 200 mg/day of total calcium from supplemental sources. Vitamin D supplement users were defined as participants obtaining at least 100 IU/day of total vitamin D from supplemental sources.

Other characteristics assessed and included as covariates in the analyses were total energy intake, gender, weight, height, cognitive function (measured by the Mini-Mental State Examination [MMSE]), smoking status, alcohol use, physical activity level, and in women, estrogen use. Self-reported measures of weight and height were used to compute body mass index (BMI; kg/m2). The MMSE was scored and adjusted for sensory impairment, and respondents with an adjusted MMSE score of 17 or less, indicating severe cognitive impairment,(21) were excluded from the analyses (n = 33). Cases with hip fractures caused by high-impact trauma (such as motor vehicle accident or a fall from greater than chair height) were also excluded (n = 117). Smoking status was characterized as current smoker (smoked regularly in past year for controls or year before hip fracture for cases), former smoker, or never smoker. Similarly, alcohol use was characterized as current user (regularly drank one or more drinks per month over the past year for controls or year before hip fracture for cases), former user, or never user. Recent physical activity was classified by hours per week for a number of recreational (bicycling, swimming, exercise classes, weight lifting, jogging, tennis, dancing, skiing, calisthenics, or yoga) and household activities (vacuuming, scrubbing floors, weeding, cutting grass, snow shoveling) and then categorized into three levels (<7, 7-14, or >14 h/week). Women were characterized as current users, former users, or never users of estrogen. Estrogen use was defined as having ever used estrogen in the form of pills, a patch, or a cream other than for contraception.

Statistical analyses

Dietary intakes of specific nutrients were analyzed by using food composition data from the Food Processor dietary assessment program (Food Processor Nutrition Analyses & Fitness Software, Version 7.1, ESHA Research, Salem, OR, USA), USDA food composition tables (www.nal.usda.gov/fnic/foodcomp/Data/index.html), and product information from manufacturers. To quantify protein intake in relation to total energy intake, protein was expressed as a percentage of total energy intake (protein % E). Quartiles of total protein % E, animal protein % E, and vegetable protein % E for all participants were used to define levels of exposure. Willett and Stampfer's(22) residual method for energy adjustment was also applied and produced similar results to those presented here.

Logistic regression models were used to evaluate the relationships between the dietary protein variables and risk of hip fracture. Odds ratios were calculated using the lowest quartile of intake as the reference intake. All combinations of interactions between quartiles of protein intake, age, and gender were tested by comparing likelihood-ratio statistics for the logistic regression model with each of three two-way interactions to the model with only the main effects. Interactions between quartiles of protein intake and calcium intake, level of physical activity, estrogen replacement therapy status (women only), and years since menopause (women only) were also tested. Of the interactions examined, only the interactions between age and quartile of total protein intake and quartile of animal protein intake were significant (both p values < 0.0001). Thus, both men and women were included together in the analyses that were then stratified by age group (50-69 and 70-89 year olds). Because animal and vegetable protein intake were not highly correlated (r = −0.33), they were included together in logistic regression models testing the effect of animal and vegetable protein intake on risk of hip fracture. Multivariate logistic regression models were used to control for the possible confounding effects of BMI, smoking status, alcohol use, physical activity level, total calcium, vitamin D, and potassium intakes, and in women, estrogen. p trend indicates the p value for a test of monotonic trend of the association in which each participant is assigned the median nutrient level for the quartile of intake and nutrient intake is modeled as a continuous variable. All analyses were performed with SPSS software, version 10.0 (SPSS, Chicago, IL, USA) and SAS software, version 8 (SAS Institute, Cary, NC, USA).

RESULTS

Data on age, anthropometric, and lifestyle characteristics for 785 male participants (336 cases, 449 controls) and 1716 female participants (831 cases, 885 controls) are shown in Table 1. Participants with hip fracture were older, weighed less, were taller, and had a lower mean BMI than participants without hip fracture. Although men reported more physical activity than women, the mean reported physical activity for male cases and controls were not significantly different from each other. Female controls reported significantly more physical activity per week than female cases. In women, more cases than controls reported being current smokers and current regular users of alcohol; there were no significant differences in current smoking status or alcohol use between male cases and controls, although more men reported regular smoking or alcohol consumption than women. A high percent of all participants reported never consuming alcohol (72% and 80% of case and control women; 45% and 49% of case and control men), indicating low overall alcohol consumption in this population. Among women, more controls than cases reported currently using an estrogen supplement in the form of pills, a patch, or a cream.

Table Table 1.. Characteristics of 1167 Cases and 1334 Controls by Gender From the Utah Study of Nutrition and Bone Health
original image

Mean intakes of nutrients by gender and case-control status are shown in Table 2. The mean ratio of reported energy intake to BMR for case and control men and case and control women was 1.89, 1.70, 1.66, and 1.57, respectively. The EI:BMR ratios observed in either cases or controls of both genders were greater than those of most published studies(23) and the WHO recommended energy requirement for a sedentary lifestyle,(24) and provide no evidence of an under-reporting bias of usual dietary intake estimates. The means of percent of energy contributed by protein, carbohydrate, and fat were not significantly different between cases and controls within gender, although women consumed a larger percentage of total energy from carbohydrates than men (p < 0.001), and men consumed a larger percentage of total energy from fat than women (p < 0.001). Men and women consumed a similar percentage of total calories from protein (p = 0.465). Mean calcium consumption from food and supplements was not significantly different for cases and controls within gender, although women consumed more supplemental calcium than men (p < 0.001). Women reported higher mean potassium consumption compared with men (p < 0.001), and within gender, controls reported higher mean potassium consumption compared with cases, although this difference did not reach statistical significance among men. Mean reported intake from dietary and supplemental vitamin D were not significantly different between men and women or between cases and controls within gender.

Table Table 2.. Selected Daily Nutrient Intakes* for Participants by Gender and Case-Control Status, the Utah Study of Nutrition and Bone Health
original image

High calcium intake in the Utah population may be explained by high calcium supplement use (Table 1) and dairy food consumption (Table 3). Among women, 59% of cases and 61% of controls reported taking a calcium supplement and 62% and 64%, respectively, reported consuming at least one serving of milk per day. Among men, 35% of cases and 34% of controls reported taking a calcium supplement, and 67% and 65%, respectively, reported consuming at least one serving of milk per day (data not shown). Additionally, 68% and 71% of case and control women and 42% and 43% of case and control men, respectively, reported total calcium intakes above the Dietary Reference Intake (1200 mg/day) for this population.

Table Table 3.. Percentage Contribution (%) of Food Groups to Total Dietary Protein Intake and Servings Per Day (Serv/day)* by Gender and Case-Control Status: The Utah Study of Nutrition and Bone Health
original image

The contribution to total dietary protein intake by food group for cases and controls within gender is shown in Table 3. The red meat (beef, pork, lamb, venison), and dairy food groups were the top two protein-providing food groups for cases and controls of both genders, and together contributed more than 40% of total protein intake in the study population. In addition, 64% of total protein intake was derived from animal sources.

As previously noted, men and women consumed a similar percentage of total calories from protein with nonsignificant differences between cases and controls within gender (all ages combined; Table 2). However, among case and control women and case and control men 50-69 years of age, cases of both genders had significantly lower mean total protein % E compared with controls (female cases = 15.9%, female controls = 17.0%, p < 0.001; male cases = 15.2%, male controls = 16.3%, p = 0.003; Table 4). No difference in protein intake between cases and controls was observed among case and control women or case and control men 70-89 years of age (female cases = 15.6%, female controls = 15.5%, p = 0.710; male cases = 15.5%, male controls = 15.6%, p = 0.879).

Table Table 4.. Daily Protein Intakes as a Percent of Total Energy Intake (Protein %E) for Participants by Gender, Age Group (50-69 Year Olds; 70-89 Year Olds), and Case-Control Status: The Utah Study of Nutrition and Bone Health
original image

In logistic regression models that controlled for age, gender, and the two-way interaction between age and protein intake, the risk of hip fracture decreased with increasing quartile of total protein intake % E for participants 50-69 years of age (OR: 1.0 [reference]; 0.39 [95% CI 0.19-0.80]; 0.36 [95% CI 0.18-0.74]; 0.29 [95% CI 0.14-0.59], p trend = 0.001); no similar associations were observed among older participants (70-89 year olds).

Major risk factors for hip fracture including BMI, smoking status, alcohol use, physical activity, calcium, vitamin D, and potassium intake, and in women, estrogen use, were examined in multivariate analyses. ORs for risk of hip fracture reported by age group across increasing quartiles of percent protein intake adjusting for these additional potential confounders are shown in Table 5. Including these additional variables in the model attenuated the point estimates but did not change the direction of association between protein intake and risk of hip fracture observed in the analyses adjusted for only age and gender. In the more complex adjusted analyses, the risk of hip fracture also decreased across increasing quartiles of total protein intake for participants 50-69 years of age (OR: 1 [reference]; 0.51 [95% CI 0.30-0.87]; 0.53 [95% CI 0.31-0.89]; 0.35 [95% CI 0.21-0.59], p trend < 0.001); no similar association was observed among older participants 70-89 years of age (OR: 1 [reference]; 1.01 [95% CI 0.78-1.32]; 1.15 [95% CI 0.88-1.51]; 1.28 [95% CI 0.97-1.70], p trend = 0.06).

Table Table 5.. Multivariate Adjusted OR and 95% CI* for Risk of Hip Fracture in Utah Men and Women Combined by Quartile of Percent Energy Provided by Total Protein (Protein %E), Stratified by Age Group: The Utah Study of Nutrition and Bone Health
original image

Among participants 50-69 years of age, significant protective effects of protein consumption in the highest versus lowest quartile of intake were seen for total protein (OR: 0.35; 95% CI: 0.21, 0.59) and animal protein (OR: 0.43, 95% CI: 0.22, 0.82), with a marginal protective effect of vegetable protein (OR: 0.52, 95% CI: 0.27, 0.997). Among participants 70-89 years of age, the point estimate comparing risk of hip fracture in the highest versus lowest quartile of total protein was above 1.0 (OR: 1.28, 95% CI: 0.97, 1.70), and similar results were found for animal protein (OR: 1.54, 95% CI: 0.93, 2.55), but the 95% CIs were wide and included 1.0. The point estimate of the OR for high versus low vegetable protein intake among 70-89 year olds was 0.79 (95% CI: 0.48, 1.30), but this effect was also nonsignificant.

Results from analyses stratified by low versus high calcium intake or tertile of potassium intake did not differ appreciably from the results presented in Table 4. Total dietary protein intake was similarly inversely associated with risk of hip fracture among participants 50-69 years of age with calcium intake of either <800 or ≥800 mg/1000 kcal/day and in those with potassium intakes of <1500, 1500-1760, or >1760 mg/1000 kcal/day.

DISCUSSION

Dietary protein intake was significantly associated with a reduced risk of osteoporotic hip fracture in Utah men and women 50-69 years of age. For participants in this age group, the risk of hip fracture in the highest quartile of protein intake was 65% less than the risk among those in the lowest quartile of intake (OR = 0.35, 95% CI = 0.21-0.59); both animal and vegetable sources of protein seemed to contribute to this association. The protective association between protein intake and reduced risk of hip fracture in participants 50-69 years of age was not modified by potassium or calcium intakes. Among participants 70-89 years of age, protein intake was not significantly associated with a decreased or an increased risk of hip fracture. The association between dietary protein intake and risk of hip fracture thus seems to be modified by age. It is unclear whether the lack of a protective association in the oldest age group is a reflection of changes in functional aspects of protein nutrition and bone health or whether this is an artifact of methodological limitations of the case-control study design among the oldest old.

Osteoporotic hip fractures represent a complex phenotype with causal factors known to be related to BMD, microarchitecture, and geometry; propensity for falls related to balance, coordination, and sensation; and physical protection of the hip by muscle and adipose tissue. Sir Astley Cooper(25) was one of the first to describe the role of poor nutrition in hip fractures by noting in 1824 the atrophic skeletal state of hip fracture patients admitted to London hospitals. In children, malnutrition is a cause of poor bone development and maintenance,(26) and in adults, poor nutrition contributes to medical complications and mortality of hip fracture patients.(6, 8, 27, 28) A daily dietary supplement was found to reduce complications and deaths after hip fractures.(8) Malnutrition may also increase fracture risk by increasing the rate of falls caused by impaired muscle strength, coordination, and reaction time. Elderly persons who had suffered falls were found to have reduced muscle mass and a poor profile of serum proteins, including serum albumin, compared with others who had not fallen.(29) In the NHANES I survey, white women with reduced serum albumin had a higher risk of hip fracture.(30) Within the complex chain of events leading to hip fractures in the elderly, dietary protein intake is emerging as a specific nutritional factor with public health and clinical importance.

Protein accounts for about 30% of the mass of bone tissue making bone one of the most protein-dense tissues in the body.(31) Bone turnover is dependent on continuous dietary intake of protein(32); hence, it is surprising that the role of protein intake in bone health has largely been viewed in terms of how it may negatively influence calcium balance. Wachman and Bernstein(33) proposed that meat in the diet of modern, industrialized populations is an important source of acid ash, resulting in the acidification of urine and weakening bones from a life-long drain on the buffering capacity of the basic salts of bone; they recommended a diet high in alkaline ash foods including fruits, vegetables, vegetable protein, and milk. Several studies have found positive associations between intake of fruits and vegetables or potassium and BMD.(34-36) Other authors have echoed the viewpoint of Wachman and Bernstein,(33, 37-40) despite the fact that it is not consistent with studies of traditional hunter-gatherer populations and the fossil record indicating that ancestral human populations evolved over hundreds of thousands of years while consuming large amounts of animal food, providing on average, two-thirds of total energy intake and contributing to strong bones.(41, 42)

High-protein diets have been consistently shown to increase urinary calcium excretion in humans in studies conducted over the past 80 years, but the source of the excreted calcium remains unclear.(43) Most published reports have documented that protein intake does not affect intestinal calcium absorption, and bone is most often proposed as the source of calcium excreted in urine. Kerstetter et al.(43) showed, however, that high-protein diets induce hypercalciuria primarily because absorption of calcium is increased. Low-protein diets were shown to acutely reduce calcium absorption and led to abrupt increases in serum parathyroid hormone and calcitriol levels, resulting in hypocalciuria. In addition, Kerstetter et al.(43) pointed out that the previous calcium balance studies may have been inadequate to detect an effect of protein on calcium absorption because of the large interindividual variability in intestinal calcium absorption and urinary calcium reabsorption.

A direct, positive effect of dietary protein intake on bone structure and function apart from calcium balance is plausible. In ovariectomized rats, restriction of protein intake resulted in a marked decrease in the α1 type 1 collagen mRNA in bone tissues.(44) Lysine is involved in the cross-linking of both collagen and osteopontin, and abnormalities in the hydroxylation of lysine residues in collagen fibers have been described in osteoporotic bone.(45) Lysinuric protein intolerance, a defect in transport of cationic amino acids, results in decreased collagen synthesis, radiographic evidence of osteoporosis, and an elevated risk of fractures(46); this rare genetic disorder provides evidence that lysine and perhaps other specific amino acids may play important roles in normal bone health. Dietary supplements of lysine are also known to increase the intestinal absorption and renal conservation of calcium.(47) Adequate dietary protein improves muscle strength and movement coordination, thus decreasing the risk of falling; the increased muscularity may also provide increased protection around bones and joints.(9, 48) Prevalence of protein-energy malnutrition increases with advancing age and is common in older hip fracture patients.(6, 27) In clinical trials, oral protein supplementation reduced complications and mortality caused by hip fracture and increased serum levels of insulin-like growth factor (IGF) in elderly patients.(8, 9, 28)

Findings from the Utah study are in agreement with several population-based studies reporting positive effects of increased protein intake on bone health. Current protein intake in Japanese women was positively associated with mid-radial BMC.(49) In Taiwan, a survey of postmenopausal vegetarian women revealed that long-term adherents to a vegan vegetarian diet had a 2.5-fold higher risk of falling below a BMD threshold for fracture risk of the lumbar spine and a 4-fold higher risk of being classified as having osteopenia of the femoral neck.(50) In the Iowa Women's Health Study, Munger et al.(11) reported that women with higher intakes of total protein and animal protein had a lower risk of hip fracture than women with lower intakes. Hannan et al.(12) reported that men and women in the Framingham Study in the lowest quartile of protein intake had a greater amount of bone loss than women in the highest quartile. Kerstetter et al.(14) reported 3-4% higher total hip BMD in women in the highest quartile of protein intake compared with those in the lower two quartiles, regardless of level of calcium intake (>800 versus <800 mg/day). The association between dietary protein intake and bone health was examined in the Rancho Bernardo Study,(13) and a positive association was found between animal protein consumption and BMD in women; a negative association between vegetable protein intake and BMD was observed in both men and women. In a secondary analysis of data collected from a 3-year randomized, placebo-controlled trial of calcium and vitamin D supplementation, Dawson-Hughes et al.(51) reported that protein intake was positively associated with an increase in BMD in calcium-supplemented participants but not in the placebo group.

In contrast, other population-based studies have reported no association or an increase in risk of fractures in elders consuming high amounts of total or animal protein. In the Nurses Health Study, protein intake was positively associated with risk of forearm fractures but was not associated with risk of hip fracture.(10) Elderly white women in the Study of Osteoporotic Fractures who consumed diets with a high ratio of animal to vegetable protein content had an increased risk of hip fracture,(40) although the significance of the ratio of these sources of protein versus the total amount of protein (from animals, vegetables, or both) is uncertain.

The discrepant findings regarding risk of hip fracture for Utah participants 50-69 years of age compared with older participants (70-89 years of age) have several possible explanations. Other factors associated with bone loss, including vitamin D status, hormone levels, renal function, and calcium homeostasis(52) may modify the protective effect of protein intake on bone health late in life and the mechanisms of net bone loss related to bone formation and remodeling may vary before and after menopause.(53)

The inability to detect a protective effect of protein intake on hip fracture risk may also be because of methodologic limitations of case-control studies among the oldest old. Information was more difficult to obtain from the older cases compared with the older controls because of high rates of death and disabilities in the oldest cases. This selection bias may have attenuated the protein-hip fracture association in the oldest age groups. Whereas the interview refusal rates were similar for both case and controls 50-69 years and 80-89 years of age (23%), older cases were more likely than controls to be unable to complete the interview because of death, frailty, illness, or dementia (42% versus 25% of female cases and controls 80-89 years of age; 50% versus 20% of male cases and controls 80-89 years of age). Because low protein intake is known to be associated with morbidity and mortality after hip fracture,(6, 8, 9, 27, 28) the biased selection of the healthiest hip fracture cases in the oldest age group may have resulted in overestimation of protein intake among the oldest cases. This is an inherent difficulty in conducting case-control studies in the oldest old.

A beneficial effect of dietary protein on risk of hip fracture may exist in older participants but may have been difficult to detect because of limitations of the dietary assessment method. As with other semiquantitative FFQs, the picture-sort FFQ best ranks individuals according to usual nutrient intake. While this method is quick and relatively easy to administer and therefore popular in large population-based studies, it lacks the precision of other methods, such as multiple 24-h dietary recall interviews or food records designed to capture more detailed dietary data.(54) In a validation study that compared mean nutrient intakes reported from the Utah picture-sort FFQ to mean nutrient intakes reported from the average of three 24-h dietary recalls, women ≥70 years of age had lower correlations between methods for protein intake than did younger women or men (r = 0.41 for women 55-69 years of age; r = 0.01 for women 70-85 years of age; r = 0.30 for men 55-69 years of age; r = 0.50 for men 70-85 years of age).(18) Cognitive and physical limitations, such as poor memory and impaired vision and hearing, are more common with increasing age and may hinder the ability of the very old to accurately report nutrient intakes.

Recall bias is always a concern in case-control studies because information that is self-reported after the fact of a hip fracture may be biased and consequently distort results, although this is unlikely to have produced a spurious association between protein intake and hip fracture risk. Self-reports of height and weight were used in the Utah study, and whereas a study of 60- to 79-year-old persons in the United Kingdom found a high correlation (r = 0.98) between self-reported and clinically measured weight, obese persons were more likely to underestimate their weight.(55)

In conclusion, dietary protein intake was significantly associated with a reduced risk of osteoporotic hip fracture in Utah men and women 50-69 years of age, and intake of protein from both animal and vegetable sources seemed to contribute to this protective association. This association was consistent across low and high levels calcium and potassium intake. The relationship between protein intake and risk of hip fracture in the oldest participants, those 70-89 years of age, remains unclear. Investigators examining these associations in other populations should consider how age modifies the relationship between protein intake and risk of osteoporotic hip fracture.

Acknowledgements

We are grateful for the assistance of study participants and staff of Utah hospitals participating in the Utah Study of Nutrition and Bone Health including the following: (1) Alta View Hospital, American Fork Hospital, Orem Community Hospital, Utah Valley Regional Medical Center, Cottonwood Hospital, Davis Hospital and Medical Center, Dixie Regional Medical Center, Jordan Valley Hospital, Lakeview Hospital, LDS Hospital, Logan Regional Hospital, McKay-Dee Hospital Center, Mountain View Hospital, Ogden Regional Medical Center, Pioneer Valley Hospital, Salt Lake Regional Medical Center, St. Mark's Hospital, University of Utah Medical Center, Valley View Medical Center; (2) Health Insight, Inc., the Utah Department of Transportation, the U.S. Health Care Finance Administration; and (3) Camielle Alleman, Jason Anderson, Merrily Anderson, Terri Beaty, Sharon Bell, Newell Belnap, Steven Belnap, Sarah Booth, Todd Carpenter, Gene Charoonruk, Nedra Christensen, Julie Claspill, Dominique Cooney, Deborah Cutler, Jim Dawes, Ruth Gerritsen-McKane, Deborah Gustafson, PhD, Sandy Ezrine, Tom Jackson, Devin Johnson, Jason Jones, Colleen Mohr, Angela Poulton, Lani Rasley, Marty Slattery, PhD, Patty Smith, Marcia Willing, MD, Siew Sun Wong, Kijuana Wright, Jim Wyatt, and Frank Yanowitz, MD. This study was supported by U.S. National Institute of Arthritis and Musculoskeletal and Skin Diseases Grant R01 AR43391 (RGM) and funding from the Agricultural Experiment Station and the Office of the Vice President for Research of Utah State University.

Ancillary