Risk Factors for Hip Fracture in a Japanese Cohort



Risk factors for hip fracture were determined from a Japanese cohort. A cohort of 4573 people (mean age 58.5 ± 12.2) who participated in the Adult Health Study in 1978–1980 were subsequently followed by biennial examinations up to 1992. Fifty-five incident hip fractures not due to traffic accidents were identified by medical records during the follow-up period. Poisson regression analysis showed that baseline low body mass index (BMI), regular alcohol intake, prevalent vertebral fracture, and having five or more children significantly increased the risk of hip fracture, and low milk intake and later age at menarche were marginally associated with increased fracture risk, after multivariable adjustment. Regular alcohol intake doubled the risk of hip fracture (relative risk 1.91, 95% confidence interval 1.07–3.42). Those individuals who had a vertebral fracture had 2.6 times higher risk than those who did not. The risk was 2.5 times higher among women who had five or more children than women with one or two. Body height, health status, marital status, intake of fish, coffee, tea, Japanese tea, smoking, exposure to atomic bomb radiation, and age at menopause were not associated with hip fracture. Relative risk for hip fracture decreased with decreasing number of preventable risk factors (low BMI, low milk intake, and regular alcohol intake). We conclude that many factors, such as BMI, milk intake, alcohol intake, prevalent vertebral fracture, age at menarche, and number of children, are related to the risk of hip fracture, and prevention programs need to focus on reducing preventable risk factors.


Osteoporosis is highly prevalent among the elderly and is characterized by low bone mass and increased skeletal fragility, resulting in increased risk of bone fracture. Among the fractures associated with osteoporosis, hip fracture causes profound physical impairment and a reduction in quality of life in the elderly; about 70% of hip fracture patients require assistance in walking even after their discharge,(1) and about 10% of the bed-ridden elderly in Japan are the result of a hip fracture.(2) A few case-control studies have been reported on risk factors of hip fracture in Japan.(3,4) However, the findings from case-control studies might be affected by recall bias and selection bias, especially in very elderly or demented individuals. There has been no report on risk factors associated with hip fracture from a cohort study in Japan. We have examined the risk factors associated with hip fracture in a Japanese cohort, known as the Adult Health Study (AHS), which is a long-term longitudinal study that began in 1958.


The AHS was established in 1950 to observe the late health effects of radiation exposure among atomic bomb survivors in Hiroshima and Nagasaki. The original AHS cohort consisted of about 10,000 atomic bomb survivors and about 10,000 controls selected from residents in Hiroshima and Nagasaki using the 1950 national census supplementary schedules and the Atomic Bomb Survivors Survey. Since 1 July 1958, the AHS subjects have been followed through biennial medical examinations. The participation rate has been around 71 to 86% throughout the examination period. Attrition due to death ranges from 2 to 5%, and that due to migration out of the cities ranges between 9 and 13% at each examination cycle. Details concerning recruitment and examination of participants have been reported previously.(5,6)

At the baseline, a total of 4869 AHS subjects aged 32 years or over underwent the physical examination in Hiroshima in the 1978–1980 examination cycle and responded to the mail questionnaire survey conducted in 1979–1981. Of these, 4573 (1586 men and 2987 women) were used for the analyses; 285 who lacked measurements of height and weight and 11 who were diagnosed as having hip fracture in the 1978–1980 examination were excluded.

They underwent general physical examination, blood tests, chest X-ray examination, and measurements of height and weight at the Radiation Effects Research Foundation (RERF) outpatient clinic. Those cohort members who could not come to the clinic were visited by a physician at their home or in the hospital. Height and weight were measured with the participant in light clothing and without shoes. Body mass index (BMI) was calculated as weight (kg)/height (m)2. Vertebral fracture was diagnosed by lateral and posterior–anterior chest and spinal X-ray examination. Diagnoses of disease were coded using the International Classification of Diseases (ICD).

In the questionnaire survey conducted in 1979–1981, both men and women were questioned about self-rated health, marital status, current tobacco use, current alcohol consumption, and dietary habits, such as food preference (Japanese food, Western food, or Chinese food) and the frequency of intake of 18 selected foods. For women, the survey included questions about reproductive histories (age at menarche, age at menopause, menstrual cycle, number of children, and lactation period, etc.).

The diagnosis of hip fracture was made by a physician based on the history of hip fracture taken at the biennial AHS examinations. Medical charts and radiographs were reviewed for case validation. The diagnoses on death certificates were also checked for diagnoses of hip fracture. Pathologic fractures or fractures due to traffic accidents or falling from a high place were excluded.

Follow-up of all subjects began in the 1978–1980 examination cycle. The accumulation of each subject's person-years of risk ended at the earliest date of initial diagnosis of hip fracture, the date of death, or the date of the last examination prior to 30 June 1992.

Statistical methods

The risk of hip fracture was examined by Poisson regression analysis, assuming the number, dij, of incident cases of hip fracture to be an independent Poisson variable with the expected value of E(dij) = PYijλij, where PYij and λij are the person-years at risk and the hip fracture risk, or hazard, in the ijth stratum produced by making a cross-classification of all categories of the variables; sex (si), age at baseline (ai), BMI (bi), alcohol consumption (alcoi), milk intake (milki), prevalent vertebral fracture (VFi), and follow-up period (pj), along with age at menarche (mi) and number of children (ci) for women. The form of λ in the ij-th stratum was assumed to be a log-linear function, namely:

equation image

adding β7 × mi + β8 × ci for women, where ageij, is the age that the ij-th stratum members attained at the jth follow-up period, defined by ageij = ai + pj. And the terms, bi, ai, and pj are mean values of BMI, age at baseline, and follow-up period for each stratum, respectively. The parameters were estimated using the maximum likelihood method. The significance of the parameter estimates was based on the likelihood ratio test.

Table Table 1. Characteristics of the Study Subjects in 1978–1980 Examination Cycle
original image

There were numerous unknown responses to each item on the questionnaire, as is usual with a mail survey. For example, 310 (6.8%) were missing for alcohol intake, 759 (16.6%) for milk intake, and 387 (13.0%) for age at menarche in women. In the Poisson regression analysis, these missing values were grouped as one category for each item without excluding them from the study sample to avoid a loss of statistical power.


Characteristics of the 4573 subjects are shown in Table 1. In the 1978–1980 examination, the mean ages of the subjects and 1 SD were 58.2 ± 13.2 years for men and 58.6 ± 11.6 years for women. Approximately 62% of men and 13% of women smoked currently, and 76% of men and 34% of women drank alcohol regularly. Regular use of alcohol was defined as consuming almost everyday. Total daily alcohol consumption was estimated using data on the alcohol content of commonly consumed beverages. Among those who drank alcohol regularly, daily ethanol consumption, expressed in terms of 100% alcohol, was 35 g for men and 9 g for women on average. The age at menarche was 15.1 years on average. Approximately 69.4% of women were postmenopausal and the average age at menopause was 47.7 years. During the follow-up period of 1978–1992, 55 (6 men and 49 women) had hip fracture, after excluding one fracture due to a traffic accident and one due to falling from a high place.

The age- and sex-specific incidence of hip fracture is shown in Fig. 1. The risk of hip fracture was higher in women than in men, and it increased 1.78-fold with each 5-year increase in age for women.

Figure FIG. 1.

Age- and sex-specific incidence of hip fracture in a Japanese cohort. Follow-up period was from 1978 through 1992.

A multivariate analysis, including men and women, showed that age, gender, BMI, milk intake, alcohol intake, and prevalent vertebral fracture were independently related to risk of hip fracture (Table 2). The risk of hip fracture increased 1.11-fold with a decrease in BMI of 1 kg/m2. Intake of milk almost every day (≥5 times/week) marginally decreased the risk of fracture (relative risk [RR], 0.54; 95% confidence interval [CI], 0.25–1.07). The test for trend revealed a suggestive linear relationship between frequency of milk intake and the fracture risk. Regular alcohol intake doubled the risk of hip fracture (RR, 1.91; 95% CI, 1.07–3.42). Those participants who had been previously diagnosed as having a vertebral fracture had a 2.6 times higher risk than those without vertebral fracture (RR, 2.64; 95% CI, 1.37–4.98).

Self-rated health, marital status, food preference, intake of tofu (food containing soy bean protein and calcium), fish, coffee, tea, Japanese tea, smoking, and exposure to atomic-bomb radiation were not related to the hip fracture risk. No relationship was found between the risk and baseline disease: specifically, cerebrovascular diseases, paralysis, epilepsy, Parkinson's syndrome, cataract, hypotension, hypertension, anemia, hyperthyroidism, or hypothyroidism. There was a trend toward an increased risk of hip fracture among those with senile dementia. However, since this condition was uncommon in the 1978–1980 examination cycle and the statistical power was limited, we cannot draw a conclusion. The associations between the risk of hip fracture and BMI, intake of milk, alcohol intake, and prevalent vertebral fracture were not modified by gender.

In analyses limited to the women, relative risks of age, BMI, alcohol intake, and prevalent vertebral fracture did not change (Table 3). Women who had five or more children had a 2.5 times higher risk than those women with one or two children (RR, 2.53; 95% CI, 1.07–6.38). Relative risk for nulliparous women seemed to be higher, but not significantly so. A test for trend showed a significant linear quadratic relationship (p = 0.03) between number of children and the fracture risk. Women of age 17 or more at menarche had marginally higher risk as compared with those of age 14 or less (RR, 2.17; 95% CI, 0.89–5.79). Length of breast-feeding, age at menopause, years from menarche and menopause, menstrual cycle, or dysmenorrhea were not associated with hip fracture risk.

Among the six risk factors shown in Table 2, three, namely low BMI, alcohol intake, and low milk intake, can be modified through intervention, and the others (prevalent spinal fracture, giving birth to 5 or more children, and later menarche) cannot, although the latter three are modifiable if intervention is begun earlier in life. After excluding those who answered “unknown” to any of the questions on the six risk factors, a total of 1818 women were used to determine to what extent a decrease in the number of preventable risk factors contributed to a decrease in the relative risks for hip fracture. When compared with women who had two or three preventable risk factors and any of the nonpreventable risk factors, the relative risks decreased with decreasing preventable risk factors whether non-preventable risk factors did or did not exist (Fig. 2). It is suggested that the risk of hip fracture could be halved by modifying the preventable risk factors even for women with several nonpreventable risk factors. No hip fracture developed among the 190 women who had none of the six risk factors.


This Japanese cohort study showed that lower BMI, prevalent vertebral fracture, alcohol intake, giving birth to five or more children, and later age at menarche increased the risk of hip fracture, whereas intake of milk might prevent hip fracture. Reducing preventable risk factors (low BMI, regular alcohol intake, and low milk intake) decreased the risk of hip fracture.

There have been several reports on the incidence of hip fracture in Japan.(7–10) In all studies, the incidence of hip fracture was estimated as the number of hospital or clinic admissions based on the age- and sex-specific population for each region in a specific year. To the authors' knowledge, the present study is the first one to obtain the incidence of hip fracture from a cohort study in Japan. The incidence in other reports in Japan are similar in magnitude to the rates from this study.

Most studies show that weight and BMI are associated with bone mass and occurrence of hip fracture,(11–19) which agrees with our results. In American and European people, hip fracture risk increases with body height.(11–13) However, we did not find a significant association between risk of hip fracture and height. The mechanism by which height influences hip fracture may be mechanical, that is, a stronger impact would occur because of the greater distance from the hip to ground level. Mean height was 162 cm in men and 150 cm in women of the AHS cohort we studied. Since they are smaller in height than Europeans and Americans, body height may not be associated with the occurrence of hip fracture. Height might also influence risk through its association with hip axis length, which predicts hip fractures.(20,21)

Table Table 2. Multivariable Model for Risk Factors for Hip Fracture (Men and Women)
original image
Table Table 3. Multivariable Model for Risk Factors for Hip Fracture (Women)
original image

An association between alcohol intake and hip fracture has not been reported consistently.(11,14,16) Grisso et al.(14) reported that the risk of hip fracture was high among black women who drank alcohol seven or more times per week. Our study revealed a similar finding. Regular alcohol intake may increase the risk of hip fracture by increasing the risk of falls.

The effect of calcium intake on bone mass has been observed in the intervention studies.(22–24) However, no consistent conclusion has been obtained as to whether calcium intake prevents hip fracture. On the one hand, several studies observed no relationship between calcium intake and the risk of hip fracture.(11,15,25) On the other hand, the risk of hip fracture was reported to be low among those with a high calcium intake in a longitudinal follow-up study(26) and in a case-control study.(27) In the latter two studies, the daily calcium intake of the subjects was 391 and 171 mg on average, respectively. We found that the intake of milk almost every day lowered the risk of hip fracture to one half. Daily calcium intake of the AHS subjects was about 550 mg on average in 1982 (unpublished data), which is almost the same as the average calcium intake in Japanese found in the National Nutrition Survey.(28) A threshold effect for calcium intake is suggested; bone mass increases with increasing calcium intake up to the threshold, beyond which further calcium intake causes small or no additional increment in bone mass. In a population with relatively low calcium intake, calcium intake may increase the bone mass and consequently lower the risk of hip fracture.

We found that those individuals who had a vertebral fracture showed a high risk of hip fracture. Prevalent spinal fracture is associated with low bone mass and fragility of bone. Recent reports show that those individuals who have had a fracture at any site have a higher risk of a subsequent fracture of the spine or hip than those individuals without fracture, even after adjustment for bone mass.(11,29,30) These results suggest that a history of fracture may reflect not only bone mass but also bone strength and increased risk of fall.

Figure FIG. 2.

Relative risk of hip fracture according to the number of preventable and nonpreventable risk factors in women. Relative risk was adjusted to age. The preventable risk factors are as follows: BMI of less than 20, drinking milk once or less per week, and drinking alcohol regularly. The nonpreventable risk factors include any of the followings: having vertebral fracture, age at menarche of 17 or over, and giving birth to five or more.

A large-scale case-control study(16) showed that late menarche and a shorter duration from menarche to menopause is a risk factor for hip fracture. We found a marginal increase in the hip fracture risk among those AHS participants whose age at menarche was 17 or over, but no association was found between fracture risk and time from menarche to menopause or age at menopause. In our study, multiparity with five or more children was related to the risk of hip fracture, but no relationship was found between the risk and the period of breast feeding. Statistical power for the analyses of the lactation period was limited, since 65% of the values were missing. The findings on the association between fracture risk and child bearing or breast feeding have been inconsistent.(15,31) A possible explanation for the controversial results may depend on the intake of calcium during pregnancy and lactation. According to the National Nutrition Survey, average calcium intake was 270 mg/day in 1950 and 338 mg/day in 1955(28) in Japan. Although calcium absorption increases during pregnancy and lactation, Japanese women may not have taken enough calcium to compensate for a negative balance of calcium in their early adulthood. Nulliparity was not significantly related to the risk of hip fracture in our study. However, a test for trend showed a significant linear quadratic relationship between number of children and the risk of hip fracture.

The risk of fracture among smokers has been reported to be higher than among nonsmokers,(17) however many studies have reported no substantial effect of smoking on hip fracture.(11,12,14,16,18) We did not find a significant influence of smoking on the occurrence of hip fracture.

Caffeine increases the excretion of calcium in urine. Intake of greater than two or three cups of coffee per day increased the risk of fracture.(19) Since the highest quartile was defined as five or more times per week in our questionnaire, we cannot evaluate those who drank a great amount of coffee.

Stroke, paralysis, Parkinson's disease, and senile dementia are reported to be risk factors of hip fracture,(11,18) but no relationship with these diseases was observed in our study except for senile dementia. Since the average age of the subjects in our study was about 58 years at the baseline examination and the prevalence of these diseases was very low, the analytical power was insufficient for assessment of these disorders as risk factors.

A few studies have shown an association between hip fracture and the number of risk factors.(11,18) From the public health perspective, we focused on the risk factors that can be modified through intervention. Our study suggested that reducing preventable risk factors lowered the fracture risk even in the presence of nonpreventable risk factors.

In our study, none of the risk factors were modified by gender. However, the number of men who suffered hip fractures during the follow-up period was too small to draw a conclusion on effect modification of risk factors by gender.

There are several limitations to the present study. This is a cohort study including atomic bomb survivors and their controls, who were selected from residents in Hiroshima. Although no effect of atomic bomb radiation was found on the hip fracture risk, they are not representative of the Japanese people of the same generation. Incident cases of hip fracture were identified by a physician at the biennial examinations. One third of hip fracture cases were ascertained by hip X-ray or medical charts of hospitals, but the remaining cases were diagnosed based on self-reports. We think that hip fracture is a quite distinct event, and the recall for hip fracture is relatively accurate, which is supported by a paper that self reports are relatively accurate for hip fracture and confirmed fracture reports mentioning “hip” were almost always fractures of the proximal femur.(32) Even though an AHS subject is unable to come for the examination after sustaining hip fracture, information on hip fracture can be obtained in a home visit health examination. However, we cannot rule out a bias by nonparticipation in the AHS examination. Since the incidence rates of hip fracture among the AHS population are essentially identical to other reports in Japan, the number of missing cases of hip fracture must be small, if such cases exist. Although uncertainty remains regarding the accuracy of self reports and the bias of nonparticipation cannot be ruled out, this study provides a unique opportunity to look at risk factors of hip fracture, since there has been no cohort study followed for more than 10 years in Japan. Most risk factors we detected in this study are factors affecting bone mass. Since the average age of the cohort was 58 years old, it was difficult to detect risk factors related to falls, which occur more frequently as one becomes older.

In summary, this cohort study revealed associations among BMI, intake of milk, alcohol consumption, history of bone fracture, age at menarche, and number of children and hip fracture risk in men and women. Reducing preventable risk factors such as BMI, calcium intake and alcohol intake is useful to prevent hip fracture in Japanese.


We thank the Department of Epidemiology, the Radiation Effects Research Foundation (RERF), for providing the mail survey data (RP 14–78). This study was supported by RERF and partly by Grants-in-Aid from the Research Society for Metabolic Bone Diseases, Japan. This publication is based on research performed at RERF, Hiroshima, Japan. RERF is a private, nonprofit foundation funded equally by the Japanese Ministry of Health and Welfare and the U.S. Department of Energy through the National Academy of Sciences.