A 5-year follow-up study investigated calcaneal bone mineral density (BMD) and changes in BMD in relation to fracture occurrence. The subjects comprised two cohorts born in 1914 and 1910 living in the city of Jyväskylä in central Finland. One hundred and three men (82%) and 188 women (73%), aged 75, and 57 men (74%) and 136 women (65%), aged 80, of the eligible population participated in the baseline bone measurements. The follow-up bone measurements were obtained for 59 men (68%) and 119 women (66%), aged 80 years, and for 21 men (53%) and 61 women (48%), aged 85 years. During the follow-up period, 8 men and 36 women from the younger and 11 men and 24 women from the older cohort sustained at least one fracture. When the baseline levels of BMD were related to fracture occurrence, the results clearly showed that with increased BMD values the probability of fracture decreased. Where men and women had similar BMD values, they also had a similar fracture probability. Except for one woman in the older cohort, none of those who had initial BMD values more than 1 standard deviation above the mean for their age developed a fracture during the follow-up period. The mean annual decrease in BMD was greater in the women (2.5–2.7%) than in the men (0.8–1.0%). The BMD change tended to associate with fracture occurrence only in the 75-year-old women (p = 0.075). The results suggest that calcaneus BMD can be used as a predictor of fracture occurrence in 75− to 80-year-old men and women. However, associating fractures with the change in BMD was difficult due to the limited number of survivors and initial differences in BMD values.
It has been well documented that appendicular bone mineral density (BMD) can be used as a predictor of fractures in different age groups.(1–6) The predictive value of the extent of bone loss in the elderly, especially after the age of 70, is not, however, known with certainty.(7–9) Nevitt et al.(10) showed in their study that a decrease of 1 standard deviation (SD) in the bone density of the distal radius was associated with an increased risk of nonspinal, wrist, and humerus fractures and that of the femoral neck with nonspine, humerus, and hip fractures among women aged 80 and over. A recent follow-up study(11) of women aged 75+ in France showed that femoral neck BMD as well as fall-related factors such as neuromuscular and visual impairments are independent predictors of the risk of hip fracture in elderly mobile women. However, by the age of 80+ about two-thirds of white women have a bone density at least 2 SD below the mean for young normal women.(12) The low level of BMD in very old people may limit the ability of BMD to predict fractures at these ages. In addition, much less is known regarding age-related bone loss in men,(13–16) despite the fact that the incidence of fracture is increasing in later life in men as well.(17–19)
The present paper reports a 5-year follow-up study with the aim of investigating the predictive value of baseline calcaneal BMD on fracture occurrence and the association of the change in BMD with fractures at any bone site in 75− and 80-year-old men and women.
MATERIALS AND METHODS
The study was part of the Evergreen project, a large gerontological research program. The target population was drawn from the national population register. It consisted of all men and women who were 75 years old in 1989 (born in 1914) and 80 years old in 1990 (born in 1910) and were resident in the city of Jyväskylä (population 70,000, located in the province of Central Finland). The age and gender distribution of the population aged 65 or over in Jyväskylä is similar to that of the general urban population of Finland.
The baseline laboratory examinations commenced on October 23, 1989, and were completed by the end of December, 1989 for the cohort born in 1914 (75-year-olds); and the corresponding examinations for the cohort born in 1910 (80-year-olds) began in early February, 1990 and were completed by the end of March, 1990. The follow-up measurements for the younger cohort started on October 12, 1994 and finished on December 12, 1994, and for the older cohort the follow-up began on January 23, 1995 and ended at the end of February, 1995.
The target population for the 75-year-olds originally comprised 388 individuals (127 men and 261 women) of whom 103 men (82%) and 188 women (73%) participated in the baseline bone measurements.(20) The corresponding figures for the 80-year-olds were 291 persons (78 men and 213 women). Bone measurements were obtained for 57 men (74%) and 136 women (65%).(6)
The final sample size for the follow-up was 268 individuals (87 men and 181 women) for the younger cohort and 167 individuals (40 men and 127 women) for the older cohort. Bone measurements were obtained for 59 men (68%) and 119 women (66%) aged 80 years, and for 21 men (53%) and 61 women (48%) aged 85 years. The rate of participation and nonparticipation in the study is given in Table 1. Informed consent was obtained in advance from all the participants.
Table Table 1. Rate of Participation of Men and Women Born in 1914 and 1910 in the Baseline and Follow-up Study
Of the men who participated in baseline or follow-up interviews, 5% were institutionalized and 6% were unable to get about outdoors. Of the women who participated in the baseline interview, about the same proportion was institutionalized (6%) and unambulatory (5%) while at follow-up the corresponding numbers for women were 9 and 11%, respectively. Only a few of the subjects who participated in BMD measurements at baseline or follow-up lived in institutions (1–3%) or were unable to get about outdoors (0–2%).
Assessment of fractures
The occurrence of fractures was followed up after the baseline measurements for 5 years in both age groups. Information about accidental falls and fractures was first obtained either by a postal questionnaire or a telephone call as well as a home interview during three different time periods: baseline to May 1991, May 1991 to March 1993, and March 1993 to February 1995. If the subjects were living in a long-term care institution, the institution's personnel were asked for the information. Only the falls that had caused a visit to a doctor were included in this study which made it possible to check the consequences from the medical records at the local health center in the city of Jyväskylä and the Central Hospital of Central Finland. The information about fractures was checked against radiographic records at the same institutions. To ensure validity, the medical records of those persons who reported no fractures, or who had died, or who could not be contacted, or failed to respond were also checked. In addition, when the subjects participated in the follow-up measurements, they were asked about any falls during the preceding year.
At the follow-up, the subjects were asked about medication affecting bone metabolism. Nine women had taken estrogen for at least 1 year during menopause (1–20 years, mean 6 years); everyone had discontinued the therapy at least 5 years before the baseline study. Two men had been taking estrogen for 8 and 20 years and still continued the medication because of prostate cancer. Furthermore, six women continued to take cortisol (1–12 years, mean 6 years). Thyroxin was taken by 10 women (2–44 years, mean 20 years) and 1 man (14 years) because of hypothyreosis. Five of the women who had been taking estrogen replacement therapy had fallen at least once during the follow-up period but none had sustained a fracture. Both men taking estrogen had fallen without fractures. Of those taking thyroxin, two women had fallen and both had sustained a fracture. Of the six women who were taking cortisol, two had fallen without fractures.
Bone mineral measurements
The volume density BMD (g/cm3) was measured in the right calcaneus by a single photon absorption method as described earlier.(20) The coefficient of variation between repeated measurements for different days was 1.5%.
Standard methods were used for measuring body height and total body mass. Body fat content and lean body mass were determined using bioelectrical impedance(21) with the manufacturer's equation (RJL Systems, Inc., Detroit, MI, U.S.A.). The physical characteristics of the subjects are summarized in Table 2.
Table Table 2. Physical Characteristics and BMD Values of Men and Women in 75-Year-Old (Born 1914) and in 80-Year-Old (Born 1910) Groups at Baseline and at the End of the 5-Year Follow-up (Mean ± SD)
Student's t-test (two-tailed) for paired and for independent samples was applied to examine the significance of the differences between the means of the two measurements and two groups. The multivariate analysis of variance (MANOVA) procedure followed by simple contrasts was performed to determine the differences in BMD between the groups with different fracture status: (1) nonfracture (NF); (2) prospective fracture (PF = fracture occurred after the baseline bone measurements); (3) retrospective fracture (RF = fracture occurred before the baseline bone measurements); and (4) retrospective fracture plus prospective fracture (RPF). The analysis of covariance (ANCOVA) was applied in order to control the effect of the differences in initial BMD on the association between the changes in BMD and the fracture occurrences. The relative risks (RR) for fractures and falling with 95% confidence intervals (CI) were calculated as risk ratios of low BMD (at least 1 SD below group mean) to higher BMD for both genders and ages. Cox proportional regression and logistic regression were used to estimate the relationship of fractures with BMD, age, body weight, and height in men and women. Hazard ratios estimated from the Cox model were converted to the probability of remaining fracture-free during the follow-up time. Furthermore, survival analysis was used to examine the relationship between 5-year mortality and BMD level, falling, and fractures. The statistical analyses were done using a SPSS-X software package (SPSS Software Inc., Chicago, IL, U.S.A.).
The baseline BMD values for all those who participated in the baseline study and whose fractures were followed were 0.149 ± 0.030 g/cm3 in 75-year-old men, 0.128 ± 0.035 g/cm3 in 75-year-old women, 0.148 ± 0.032 g/cm3 in 80-year-old men, and 0.117 ± 0.035 g/cm3 in 80-year-old women.
The medical records (including X-rays) showed that 11 men (9%) and 43 women (18%), aged 75, and 13 men (18%) and 35 women (19%), aged 80, of those interviewed had sustained at least one fracture during the 5-year period. Of the fractured, 8 men (73%) and 36 women (84%) from the younger cohort and 11 men (85%) and 24 women (69%) from the older cohort had participated in the baseline bone measurements. Of these, 2 men and 25 women in the younger cohort, and 4 men and 9 women in the older cohort were able to participate in the follow-up bone measurements. The most common fracture sites for the men were femoral neck and wrist/hand, and for the women radius/ulna and femoral neck (Table 3).
Table Table 3. Number of Fractures During 5-Year Follow-up at Different Bone Sites in 75-Year-Old (Born 1914) and 80-Year-Old (Born 1910) Men and Women
When the baseline BMD levels were related to fracture occurrence, the results showed that with increased BMD values the probability of developing a fracture decreased. Where men and women had similar BMD values, they also had a similar fracture probability (Fig. 1). The fracture group tended to have a lower baseline BMD than the nonfracture group (0.130 ± 0.022 vs. 0.151 ± 0.031 g/cm3, p = 0.055 in 75-year-old men; 0.106 ± 0.026 vs. 0.134 ± 0.035 g/cm3, p < 0.001 in 75-year-old women; 0.127 ± 0.023 vs. 0.153 ± 0.032 g/cm3, p = 0.014 in 80-year-old men; 0.108 ± 0.026 vs. 0.119 ± 0.036 g/cm3, p = 0.162 in 80-year-old women). The results also showed that in either gender or age group, those who had baseline BMD values more than 1 SD above the mean for their age did not sustain a fracture during the follow-up period, except for one woman in the older cohort. However, the RR for fractures in those who had a BMD value 1 SD or more below the group mean compared with those with higher BMD values was significant in 75-year-old women (RR 3.48, CI 2.02–5.99).
The results of logistic regression for fractures showed that low BMD (p < 0.001) and higher age (p = 0.022) increased significantly the probability of fractures when the entered variables were BMD, age, gender, body weight, and height. When BMD alone was in the model, 64.4% of the subjects were allocated correctly in the different fracture groups (fractures vs. no fractures). When the models were constructed separately for men and women and the entered variables were BMD, age, body weight, and height, BMD remained significant in both men and women (p < 0.001) and age in women (p = 0.004). Again, BMD alone explained 61% of fracture occurrence in women and 70% in men. Cox proportional regression showed similar results to the logistic regression. When BMD, age, body height and weight were included, the survival function for fractures was similar for men and women (Fig. 2).
To determine whether low BMD was the cause of an individual sustaining a fracture and putting that individual at greater risk of having a second and subsequent fracture, the women were assigned to four groups according to their fracture history (see statistical analyses). The results showed that BMD values were significantly lower in the RF and PF groups compared with the NF group in both cohorts of women and the BMD values were lowest in the RPF group in the younger cohort (Fig. 3). No significant differences in BMD were found between the RF and PF groups at either age. Due to the small number of fracture cases, the same analysis was not performed among the men.
A fall was the primary cause for fractures in all the groups studied. Twenty-three percent of the men and 36% of the women born in 1914, and 30% of the men and 34% of the women born in 1910 reported at least one fall during the follow-up period. In the group of fallers, the proportion of those who had sustained fractures was 41% (men) and 48% (women) for the younger cohort, and 55% (men) and 49% (women) for the older cohort. Those who had low BMD tended to fall more often than those with higher BMD values. However, the association was statistically significant only in 75-year-old women (RR for falling 1.51, CI 1.02–2.22 among those with a BMD value 1 SD or more below the group mean vs. those with higher BMD values). Self-reported falling during the preceding year was not significantly related to baseline BMD in any of the groups. In addition, the survival analyses showed that the baseline BMD level, falling, and fractures were not associated with mortality during the follow-up period. Among the fallers during the follow-up, however, body height was lost significantly more in the younger cohort women and in the older cohort men and women compared with their nonfaller counterparts (1.3 vs. 0.8%, p = 0.013, 0.8 vs. 0.4%, p = 0.017 and 1.4 vs. 0.6%, respectively).
The changes in BMD in those men and women participating in the 5-year follow-up are shown in Table 2 and Figs. 4 and 5. There was a significant loss in BMD in both gender and age groups. The mean annual loss was 0.8% for the men and 2.5% for the women between 75 and 80 years of age, and 1.0% for the men and 2.7% for the women between 80 and 85 years of age. The women had a higher decline in BMD compared with the men in both age groups.
The decrease in BMD during the follow-up period tended to associate with fracture occurrence in the younger women (p = 0.075, Fig. 6). When the baseline BMD was controlled by ANCOVA, the results showed that the differences in the change in BMD between the fracture and nonfracture groups disappeared (p = 0.210). In the older women, the change in BMD was not associated with fracture occurrence (Fig. 7). The small number of men participating in both bone measurements was insufficient to allow reliable calculations of the association between the changes in BMD and fracture occurrence. There were no significant differences between fallers and nonfallers in the BMD change during the follow-up.
There was a significant decrease during the 5-year follow-up in body height and lean body mass in both genders and age groups (Table 2). Body weight also declined among the women in both age groups, whereas body fat content did not show significant changes in either the gender or age groups. There were no differences between the fractured and nonfractured regarding percentage change in body weight during the follow-up period. However, the younger cohort women who had sustained a fracture had lost significantly more body height than their nonfracture counterparts (1.6 vs. 0.8%, p = 0.002).
In this population-based follow-up study, we found that when the absolute levels of the baseline BMD were related to subsequent fracture occurrence, the probability of fractures clearly decreased with increased BMD values in the calcaneus. Where men and women had similar BMD values, they also had a similar probability for fracture occurrence. If the initial BMD was greater than the level of about 1 SD above the mean for 75-year-old females, no fracture was sustained during the follow-up period. However, the relative risk of fractures was significant for women with BMD 1 SD below the group mean. This indicates that even among the very old, the absolute levels of calcaneus BMD remain useful in the prediction of all kinds of fractures.
In previous studies, the predictive power of bone measurements at various bone sites for subsequent fractures in both men and women has been commonly found for subjects below 80 years of age, but not above where most people have very low BMD values.(8,10,22) Our results showed that those men and women aged 75 and in the age groups combined who had sustained fractures had significantly lower baseline BMD values compared with those without fractures. In the 80-year-old group, this also held true for the men but not for the women, whose BMD values in most cases represented high fracture risk. Although it seems clear that old women with very high BMD have low fracture risk, it may be difficult to differentiate the rest of these women, the majority, with respect to fracture risk.
Previous studies have shown that body composition has a profound impact on bone mass and fracture risk.(23,24) In the present study, however, BMD and age remained the only significant predictors when body weight and height were included in a logistic regression model for fractures. The importance of BMD was further supported by the multivariate Cox model which showed no differences between men and women in remaining fracture-free. Lean body mass and body fat were not taken into account in the models because of the eventual problems with multicollinearity. An indirect estimation of body composition by bioimpedance could also evoke some problems, although the reproducibility of the technique is generally good.
The question has always remained open as to whether patients who have suffered one fracture are at an increased risk of a subsequent fracture.(25) Despite methodological concerns, prolonged immobilization following fractures has been suggested as a cause of bone loss and functional limitation.(26,27) After bone loss, BMD is maintained at lower levels at least for a certain period of time, thus increasing the risk of bone fracture. When we compared BMD values between the RF and PF women, there were no differences between those two groups for either age cohort. Consequently, our results suggest that a priori lower bone mineral values mainly contribute to the incidence of new fractures.
About half of those reporting having fallen during the follow-up had sustained a fracture. There was also a tendency of those with low BMD to fall more frequently. The actual falling may not, however, be different because in this case only falls resulting in a visit to a doctor were recorded and those with lower BMD may be more susceptible to injuries when falling. In fact, self-reported falling during the preceding year, which also included less severe falls, was not more common among those with low BMD. With low BMD, however, the prevention of falls becomes a critical issue for fracture prevention at very advanced age. Furthermore, a recent study shows that the fall-related factors such as neuromuscular and visual impairments are independent predictors of the risk of hip fracture beyond the BMD level.(11)
We found that there was a significant loss in the calcaneus BMD after the age of 75 and 80 in both men and women. The annual change for women was similar to a U.S. study conducted by Ensrud et al.(28) and about three times greater compared with the men in both age groups. There were no significant differences between 75− and 80-year-old men in the cross-sectional comparison. The relatively small group of men surviving in the oldest age group had rather higher BMD and body weight than the other men.
The decrease in BMD during the follow-up period did not, however, show a clear association with fracture occurrence. The change in body height was associated with fractures and falling but other body composition variables were not. Gärdsell et al.(8) also found in their longitudinal study that bone loss over the years did not differ between fracture and nonfracture women; the initial bone mass was the better predictor. The reason for the lack of association in our study is partly because during the 5-year follow-up of those in the population who had sustained a fracture, 36% of the men and 12% of the women among the 75-year-olds, and 38% of the men and 26% of the women among the 80-year-olds died. Only 25% of the men and 69% of the women in the younger cohort and 36% of the men and 38% of the women in the older cohort were able to participate in the follow-up bone measurements out of those who had sustained fractures and had participated in the baseline bone measurements. These results indicate that the effects of the small number of the subjects participating in the bone measurements should be taken into consideration when judging the findings. Furthermore, although the changes in BMD and some of the anthropometric variables were statistically significant, they may be small with respect to clinical consequences. The importance of the change depends probably more on the level of the initial values.
In conclusion, the results suggest that calcaneus BMD can be used as a predictor of fracture occurrence in 75− to 80-year-old men and women. In the case of both genders, where BMD values were similar, the likelihood of fracture was also similar even though the total number of fractures was higher in women. However, it was difficult to associate fractures with the change in BMD because of the limited number of survivors and initial differences in BMD values.
The study was financially supported by grants from the Academy of Finland, the Ministry of Social Affairs and Health, the Ministry of Education, and the city of Jyväskylä. The authors thank the staff at the Health Centre of the city of Jyväskylä and at the Central Hospital of Central Finland, and the late Esko Partio, M.D., Sirkku Suvikorpi, M.Sc., and Sanna Sihvonen, P.T., for assisting in the collection of the fracture data.