Bone fractures are an important cause of morbidity and mortality among the elderly in the United States (Cummings et al. 1985, Kiel 1994). The annual occurrence of osteoporotic bone fractures is estimated as 1.3 million (NIH 1984). Although the large majority of these occur among women, fracture rates among men are not trivial. For example, men at age 50 have a 5% lifetime risk of sustaining a hip fracture (Gallagher et al. 1980).

A low calcium intake has been postulated to be an important predictor of fractures because bones are largely composed of calcium phosphate and a combination of calcium phosphate and calcium hydroxide called hydroxyapatite. There is evidence of an association between dietary calcium intake before adulthood and peak bone density (Cumming 1990, Dalen et al 1974, Horsman and Currey 1983, Weaver 1992), and conversely that calcium deficiency can lead to osteoporosis as a result of continued calcium loss through feces and urine (Nordon and Heaney 1990). It is less clear whether high calcium intake in adult years, such as an intake well above the recommended daily allowance of 800 mg/d, can reduce fracture risk.

The relationship between calcium intake and fracture risk has been investigated through ecologic studies (Abelow et al. 1992. Hegsted 1986, Matkovic et al. 1979), case-control studies (Cooper et al. 1988, Cumming and Klineberg 1994, Lau et al. 1988, Krieger et al. 1992), and prospective cohort investigations (Feskanich et al. 1997, Holbrook et al. 1988, Looker et al. 1993, Paganini-Hill et al. 1991, Wickham et al. 1989). However, epidemiologic research findings have not been consistent in spite of the known biological role of calcium in bone maintenance, raising doubts about whether a high calcium intake is a major contributor to fracture prevention in adult populations. The vast majority of the previous studies were conducted among women, whose fracture risk is more than three times that of men (Cummings et al.1989). Given the relative lack of data on men and the public health importance of osteoporotic bone fractures, we sought to investigate whether higher intakes of calcium can reduce the incidence of forearm and hip fracture among men in the Health Professionals Follow-up Study (HPFS).⁴ In addition, given the possibility of higher excretion of urinary calcium with higher protein diets (Hu et al. 1993), we examined the hypothesis that calcium from nondairy foods and supplements is more beneficial than calcium from dairy foods, which are good sources of protein.

SUBJECTS AND METHODS

Of the 51,529 men who returned the 1986 base-line questionnaire, we excluded 2770 with a history of fracture, 2136 who developed cancer during follow-up, 1899 with a history of cancer at the beginning of follow-up, 1612 with a reported daily energy intake outside the range of 800-4200 kcal or with an incomplete dietary questionnaire and 59 with missing covariate information at base line. After all exclusions, 43,063 participants remained in the base-line 1986 population.

To determine dairy calcium, we totaled the calcium intake from the following items on the food-frequency questionnaire: whole milk, skim or low fat milk, yogurt, ice cream, cottage cheese and other cheese. Calcium from nondairy sources was then defined as the difference between total calcium and dairy calcium. Total milk was defined as the sum of the reported servings (glasses, 240 mL) of whole milk and low fat or skim milk; the total score was categorized into five groups (1 glass per week, 2-6 glasses per week, 1 glass per day, >1-2.5 glasses per day and >2.5 glasses per day).

The validity of the food-frequency questionnaire was evaluated in a random sample of 127 men from the HPFS living in the Boston area. In that study, nutrient intakes as computed from the questionnaire were compared with nutrients from two 1-wk diet records spaced ~6 mo apart. Rimm et al. (1992) reported a correlation coefficient of 0.61 for total calcium, and Feskanich et al. (1993) reported correlation coefficients of 0.88 for skim or low fat milk and 0.67 for whole milk.

Age, body mass index (BMI, kg/m²), smoking status and physical activity from leisure-time activities were assessed on the 1986 base-line questionnaire.

Although no formal validation studies have been conducted among men, a validation study to confirm the accuracy of self-reported fractures by female nurses (Colditz et al. 1986) found exact agreement with medical records in all 30 cases. Valid self-reports are also likely in this population of male health professionals.

Both age-adjusted and multivariate logistic regression models were performed to assess the associations among calcium, milk and fractures. We used the odds ratio from the logistic model to estimate relative risk (RR). The lowest category of food or nutrient intake was used as the referent, and we conducted tests for trend across increasing categories of intake as a continuous variable with the median of each category used as the score.

RESULTS

The characteristics of the study population by quintiles of total calcium intake are presented in Table 1. Energy intake was directly related to calcium; therefore, the vitamin D, protein and phosphorus intakes in this table were adjusted for total energy to show their association with calcium independent of energy. Among the characteristics examined, only vitamin D appeared to be associated with calcium intake. Men in the highest quintile had over twice the mean vitamin D intake as those in the lowest quintile. The study population was moderately active [mean activity level of 24 metabolic equivalents (MET)/wk = ~4 h of jogging], and a low percentage were current smokers. The characteristics of the study population were also examined by fracture status. No differences were observed among forearm fracture cases, hip fracture cases and noncases, except for age, which was higher among the hip fracture cases.

Table 1. Characteristics of the study population by quintiles of total calcium intake at base line in 1986 [View Table]

We observed no clear association between total calcium intake and fracture incidence (Table 2). The RR for men with calcium intakes > 1227 mg/d compared with those consuming <512 mg/d were 0.98 [95% confidence interval (CI) = 0.59-161, P for trend = 0.78) for forearm fracture and 1.19 (95% CI = 0.42-3.35, P for trend = 0.58) for hip fracture after adjusting for age, BMI, alcohol intake, smoking, physical activity and total energy intake. Further adjustment for vitamin D, protein, caffeine and phosphorus intakes did not materially alter these associations. Tests for interaction between calcium and vitamin D were nonsignificant for both forearm (P = 0.91) and hip (P = 0.81) fractures.

Table 2. Relative risks (RR) with 95% confidence intervals (CI) for the associations between total calcium, dairy calcium, and nondairy calcium intakes and fractures of the forearm and hip in men [View Table]

We repeated these analyses excluding the 6707 men who were using calcium supplements in 1986 (16% of the study population) to reduce bias introduced by self-selection of calcium supplements by men with a family history of osteoporosis or a diagnosis of low bone density. Ten hip fracture cases and 29 forearm fracture cases occurred in calcium supplement users. The results of these analyses were essentially the same as those observed for the whole cohort.

We separated total calcium into its dairy and nondairy components but found no evidence that either one is associated with forearm fracture (Table 2). For hip fractures, we observed a decrease among men consuming >591 mg/d of dairy calcium compared with those with intakes of <134 mg/d (RR = 0.64, 95% CI = 0.24-1.69, P for trend = 0.12), but the results were not significant and are difficult to interpret with only eight hip fracture cases in the highest quintile of dairy calcium. As for total calcium, the dairy and nondairy calcium models were not greatly changed when calcium supplement users were excluded from analysis.

Milk consumption did not appear to be associated with either forearm or hip fracture. In multivariate analyses, men consuming >2.5 glasses per day (1 glass = 240 mL) experienced no apparent protective effect against either forearm (RR = 1.06, 95% CI = 0.69-1.62, P for trend = 0.82) or hip (RR = 0.97, 95% CI = 0.39-2.42, P for trend = 0.56) fracture compared with men consuming one or fewer glasses per week (Table 3).

Table 3. Relative risks (RR) with 95% confidence intervals (CI) for the association between frequency of milk consumption and fractures of the forearm and hip in men [View Table]

DISCUSSION

Internationally, the incidence of hip fracture is positively correlated with calcium intake (Abelow et al. 1992, Hegsted 1986). However, calcium and protein intake are strongly correlated in these population studies, and the hypercalciuria effect of protein (Hu et al. 1993) could account for these findings. In another population study within Yugoslavia (Matkovic et al. 1979), the risk of hip fractures in a dairy region was lower than that in another region with a low consumption of dairy foods. However, this difference could be attributed to other factors, such as differences in physical activity or consumption of dairy foods during childhood and adolescence.

Case-control and prospective studies that have examined the association between calcium intake and fractures in men have produced inconsistent results. Among those that support a protective association, a case-control study in Britain (Cooper et al. 1988) reported a significantly lower risk of hip fracture (RR = 0.16, 95% CI = 0.03-0.77) among men in the highest quintile (>1041 mg/d) of calcium intake compared with those in the lowest quintile (<500 mg/d). In another case-control study in Hong Kong (Lau et al. 1988) where calcium intakes are extremely low, 244 mg/d was associated with a 50% decreased risk of hip fracture compared with intakes of <75 mg/d in both men and women. In a prospective study that reported a protective effect from calcium (Holbrook et al. 1988), the RR for men in the highest tertile of intake (>440 mg/4.184 MJ) was 0.3 (P < 0.05) compared with those in the lowest tertile (283 mg/4.184MJ). However, the small number of fractures (n = 33) and the use of one base-line 24-h recall as a measure of diet for the entire 14-y follow-up limit the interpretation of these results.

Our findings are consistent with those reported by Wickam et al. (1989) in a nested case-control study. The RR was increased by 40% for men in the highest tertile of dietary intake (999 mg/d) compared with those in the lowest tertile (<694 mg/d). Paganini-Hill et al. (1991) also found no significant association between calcium intake and fractures in their prospective study of older men and women. The multivariate RR for participants consuming 876 mg of total calcium per day compared with those with daily intakes 405 mg was 1.11 for men. Findings from our study are also consistent with a recent analysis of prospective data from the Nurses' Health Study (Feskanich et al. 1997), in which the incidence of hip and forearm fractures was examined over 12 y among middle-aged and older women. Calcium intake among women not taking supplements showed no association with forearm fractures, whereas women consuming >900 mg/d had an increased risk of hip fracture (RR = 2.04, 95% CI = 1.12-3.71; P for trend = 0.07) compared with women consuming 450 mg/d. Other prospective observational studies conducted exclusively among women (Cummings et al. 1995, Michaelsson et al. 1995) did not show any significant reduction in risk of hip fracture with higher calcium intakes.

In a randomized placebo-controlled trial in France (Chappuy et al. 1992) among elderly women, the treatment group had a significant 43% reduction in hip fractures compared with the placebo group after 18 mo of 1.2 g of elemental calcium and 20 µg (800 IU) of cholecalciferol. The independent contributions of calcium and vitamin D towards this protective effect cannot be determined.

Results from clinical trials show that calcium supplementation can retard nonvertebral bone loss (Cumming 1990), particularly in older women and those consuming low calcium diets (Dawson-Hughes et al. 1990). Because women on average lose more of their peak bone mass over their lifetime than men (Law et al. 1991), calcium supplementation may be more beneficial, and therefore more easily detected in women than in men. Calcium supplementation during the course of an experimental trial can slow the rate of bone resorption and remodeling process and lead to short-term increases in bone mineral content. There is evidence, however, that these small differences may not continue to increase with time (Parfitt 1980, Slemenda et al. 1993), which could account for the apparent discordance between epidemiologic studies of calcium intake and fracture risk as opposed to supplementation trials of bone mineral density.

Improvement in bone mineral content has been found in some calcium intervention studies among adolescents (Andon et al. 1994, Chan et al. 1995, Lee et al. 1994, Lloyd et al. 1993) but not in others (Johnston et al. 1992). Some studies (Murphy et al. 1994, Nieves et al. 1995, Soroko et al. 1994) also suggest that the intake of calcium and calcium-rich dairy products such as milk is most beneficial for adult bone health when it occurs before adulthood. A possible explanation is that the body's utilization of calcium is optimal before adulthood (Abrams and Stuff 1994) when it is usually building skeletal mass at rapid rates (Kerstetter 1995).

We hypothesized that calcium from nondairy foods and supplements, in particular, might be protective against fractures and that dairy foods, which tend to be high in protein thereby leading to an increased urinary excretion of calcium (Hu et al. 1993), might negate the benefits of nondairy calcium and make the effect of total calcium on fractures appear null. Our results did not support this hypothesis. Instead, we observed a somewhat decreased risk of hip fracture with higher intakes of dairy calcium, although this association is likely the result of chance because it lacked significance and was not supported by a similar protective effect from milk, the major dairy food in the diet.

The overall lack of association between calcium intake and risk of fractures in this study may be due to misclassification. However, the food-frequency questionnaire has been compared with weighed food records and found to categorize subjects reasonably well according to individual long-term nutrient intakes (Rimm et al. 1992). Also, in this same cohort, calcium intake strongly predicted lower risk of kidney stones (Curhan et al. 1993). Bias due to confounders is not likely to seriously distort associations in this study because our regression models were adjusted for many of the other risk factors for osteoporotic fractures, including age, physical activity, cigarette smoking, BMI and intakes of vitamin D, alcohol, caffeine, phosphorus and protein. Biologically, intestinal absorption of calcium is influenced by the availability of vitamin D (Murray et al. 1993, Norman 1990). However, in our study, vitamin D intake was not associated with the risk of forearm and hip fracture; although we had limited power to detect interactions, it did not appear to interact with calcium.

Hip fractures in men begin to escalate after the age of 70 y. The median age at hip fracture in our cohort was 67 y, and 25% of the cases were men <60 y of age. It is possible that long-term calcium intake is a protective factor only among those who fracture when elderly. Dawson-Hughes et al. (1990) found calcium to have diferential effects between younger and older posmenopausal women.

Other important factors to consider are the sample size, duration of follow-up and variability in exposure. Although the data analyzed included 331,234 person-years of follow-up, the number of hip fracture cases was relatively small and the confidence intervals were wide, thus limiting the interpretation of these results. The range of calcium intake was considerable; mean total calcium intake in the uppermost quintile was 3.2 times higher than that for the lowest quintile.

In summary, our data do not support a strong protective effect of calcium intake, from dairy or nondairy sources, against forearm and hip fractures among men. However, further follow-up will be needed before a modest benefit can be excluded.