- Email: [email protected]
Association between a family history of fractures and bone mineral density in early postmenopausal women
Bone Vol. 24, No. 5 May 1999:507–512
Association Between a Family History of Fractures and Bone Mineral Density in Early Postmenopausal Women M. J. GRAINGE,1 C. A. C. COUPLAND,1 S. J. CLIFFE,1 C. E. D. CHILVERS,1 and D. J. HOSKING2 1
Division of Public Health Medicine and Epidemiology, School of Community Health Sciences, University of Nottingham Medical School, Nottingham, UK 2 Division of Mineral Metabolism, Nottingham City Hospital, Nottingham, UK
Key Words: Bone density; Fractures; Heredity; Osteoporosis; Postmenopausal; Risk factors.
The aim of this analysis was to measure the strength of the association between a family history of fractures and bone mineral density (BMD), and to determine what definition of family fracture history best predicts BMD. Five hundred and eighty postmenopausal women aged 45–59 at recruitment completed a risk factor questionnaire. Women were asked to recall details of fractures sustained by any female relative. BMD measurements taken at five sites were used. The data were analysed using linear regression, adjusting for age. Two hundred and ninety-seven (52.8%) women reported a family history of fractures, and they had a significantly lower BMD at two of the sites measured (p < 0.05). The associations with BMD were most significant when only counting fractures that occurred in the subject’s mother or a sister as a result of low trauma, with no restrictions made on age at the time of fracture and site of fracture (p < 0.01 at three sites; 0.01 < p < 0.05 at two sites). Women with a family history according to this definition had a 4.6% reduction in BMD at the femoral neck. When T scores were used to categorize women as either osteopenic/osteoporotic (T < 21) or normal at the femoral neck, the sensitivity of using this definition was 39% and the specificity was 74%. The small group of women that reported a low-trauma hip fracture in a mother or sister (n 5 23) had a mean femoral neck BMD which was 8.9% lower than that of the remainder of the sample, although this difference was less statistically significant than when low trauma fractures at any site were counted. Of these 23 women, 70% were osteopenic or osteoporotic, compared with 57% of those reporting a low-trauma fracture at any site and 47% of the sample as a whole. The sensitivity of this definition, however, was low (6%). From these analyses it can be concluded that the definition of family fracture history that best predicts BMD in postmenopausal women is a fracture at any age in a mother or sister resulting from low trauma, although the sensitivity and specificity of using a family history of fractures by itself to screen for low BMD were poor. (Bone 24:507–512; 1999) © 1999 by Elsevier Science Inc. All rights reserved.
Introduction Studies of monozygotic and dizygotic twins1,5,13,16 have shown strong correlations in bone mass between pairs of twins and have also provided evidence that this association is due at least partly to genetic factors. Other studies8,9,19 have reported that strong and significant correlations in bone mass exist among firstdegree relatives. Three further studies6,14,15 have found that relatives of patients with osteoporosis have a significantly lower bone mass when compared with control groups. Although all these studies have produced findings that suggest that osteoporosis in a relative is a risk factor for low bone mass, in practice the only information that would be available to clinicians would be from a simple question asking patients to report any family history of osteoporosis. The question of whether such a measure correlates with bone density was addressed by Soroko et al.,17 who found that a positive reported family history of osteoporosis was associated with significantly lower bone mineral density (BMD) at the lumbar spine for women and total hip for men. In this study, we investigate whether a simple question on fracture history in female relatives would be able to predict BMD at various bone sites. In particular, our aim was to identify what definition of family fracture history had the strongest association with BMD, thereby suggesting a definition that could be used both in the clinic to predict low bone density, and by researchers who want to incorporate a self-reported measure of family history into a study of bone density. Materials and Methods The study participants comprised two groups of women. The first group included postmenopausal women participating in the Early Postmenopausal Interventional Cohort (EPIC) study. This is a multicenter clinical trial carried out in four study centers (Copenhagen, Denmark; Hawaii, United States; Nottingham, UK; and Portland, Oregon, United States). The trial was established to test the efficacy of the bisphosphonate alendronate in preventing postmenopausal bone loss. The subjects used for the analysis reported in the present study came from the Nottingham center. Recruitment in Nottingham took place between September 1992 and June 1993, using a two-stage sampling procedure. First, 60 general practices were randomly selected from all those in the Nottingham Health district. The next stage was to sample at
Address for correspondence and reprints: Mr. M. J. Grainge, Division of Public Health Medicine and Epidemiology, School of Community Health Sciences, University of Nottingham Medical School, Queens Medical Centre, Nottingham, NG7 2UH, UK. E-mail: [email protected] © 1999 by Elsevier Science Inc. All rights reserved.
507
8756-3282/99/$20.00 PII S8756-3282(99)00016-2
508
M. J. Grainge et al. Family history of fracture and BMD
Bone Vol. 24, No. 5 May 1999:507–512
Table 1. Distribution of BMD values at five bone sites
AP spine Femoral neck Greater trochanter Total radius Whole body
n
Mean
SD
Minimum
Maximum
562 562 562 562 559
0.961 0.755 0.676 0.521 1.043
0.141 0.115 0.102 0.050 0.091
0.584 0.473 0.383 0.356 0.751
1.380 1.320 1.100 0.696 1.361
random 150 women aged 45–59 from each participating practice. These women were invited to attend information sessions, followed by a screening clinic. The main inclusion criteria were that subjects had to be postmenopausal, in good general health, and must not have taken hormone replacement therapy (HRT) within the last 3 months. Also, no more than 10% of subjects were permitted to have a spinal BMD below 0.8 g/cm2. Full details of the recruitment procedure have been published elsewhere.4 The second group of women included in this study comprised a random sample of women who had been ineligible or unwilling to take part in the EPIC trial, and had been recontacted. Inclusion/ exclusion criteria for the epidemiological study were that the subjects still had to be postmenopausal, but women previously excluded owing to medical history, low bone density, or the use of HRT were now included. It was intended in this way to construct a final sample of postmenopausal women more typical of those in the general Nottingham population. All women in the study provided written informed consent, and the study was approved by the Nottingham City Hospital Ethics Committee. All subjects completed an interviewer-administered risk factor questionnaire, which contained the question, “Have any women in your family had a broken, cracked, or fractured bone?” For each fracture reported, the bone fractured, the age at which the fracture occurred, and the circumstance from which the fracture arose were recorded. Subjects were also asked the question, “Have any of the women in your family suffered from osteoporosis (brittle bones)?” They were prompted to recall female relatives who had classic signs of osteoporosis (i.e., a humped back and loss of height). The age at which it began was recorded for each relative reported. The questionnaire also contained sections on menstrual history, HRT use, and other potential risk factors for osteoporosis. Bone mineral density was measured using dual-energy X-ray absorptiometry (Hologic 2000; Waltham, MA). Measurements taken at five sites were used: antero-posterior spine (AP spine), femoral neck, greater trochanter, radius/ulna (total radius), and whole body. For subjects in the EPIC trial, their baseline scan prior to commencing therapy was used in the analyses reported here. For the other subjects, a single scan was carried out around the same time that the questionnaire was administered. The coefficients of variation calculated for women in the EPIC trial are those which are used to assess the reproducibility of the local scanner. These subjects had their BMD measured twice within the space of 14 days, and the pairwise coefficients of variation were 1.48% for the AP spine, 1.77% for the femoral neck, 1.36% for the trochanter, 0.98% for the total radius, and 0.90% for the whole body. Height and weight were also measured at the time of the BMD scan. Linear regression was used to analyze the data using the Statistical Analysis System (SAS Institute, Cary, NC), with BMD measurements at the five different sites as the dependent variables. All analyses were adjusted for age, as older subjects will tend to have older relatives who will have had more opportunity for fracture. Different definitions of family fracture history were used as explanatory variables. Restrictions were made according to the age at which the fracture occurred (under
50, between 50 and 70, and over 70), whether it was sustained by a first-degree female relative (mother or sister) or other female relative, and whether the fracture was considered to be the result of only low trauma (such as a fall from standing). For all definitions, family fracture history was treated as a two-level factor, indicating the presence or absence of a family history. In further analyses, the effect of only counting hip fractures was investigated, as was the effect of controlling for other BMD risk factors: years since menopause, height, weight, duration of HRT use, and a two-level factor indicating whether the subject was part of the main EPIC trial. Finally a residual analysis was carried out to test model assumptions and to identify any unduly influential observations. A p value , 0.05 was used to denote statistical significance. The BMD measurements at the femoral neck were converted into T scores representing the number of standard deviations above or below the mean BMD for healthy young women, using reference ranges derived from phases 1 and 2 of the National Health and Nutrition Examination Survey (NHANES) III study.11 These were then categorized to classify subjects as either osteoporotic (T # 22.5), osteopenic (22.5 , T , 21), or normal (T $ 21) at the femoral neck in accordance with the World Health Organization (WHO) definition.20 As only a small number of women (n 5 13) were osteoporotic (T # 22.5), these women were combined with the osteopenic group, to provide the low BMD (T , 21) group for the present analysis. Odds ratios associated with low BMD were then produced for family fracture history, adjusting for age by the use of logistic regression. The same classification was also used to calculate the sensitivity and specificity of using family fracture history as a diagnostic tool for predicting low bone density. Results Of the 7564 women originally approached, 428 (5.6%) were randomized into the EPIC trial.4 Of these, 349 (81.5%) agreed to participate in the epidemiology study. Six hundred and two women who had not been included in EPIC were recontacted, of whom 231 (38.4%) were eligible and agreed to take part. Of the remainder, 156 (25.9%) were found to be ineligible and 215 (35.7%) either refused or did not respond. Of the 580 participants recruited from the two groups, 562 women (336 from the EPIC study) gave a definite positive or negative answer to the family fracture history question, and were used in this analysis. The participants had a median age of 53 years (percentiles: 25th 5 50, 75th 5 57) and were an average (median) of 6 years past the menopause (percentiles: 25th 5 3 years, 75th 5 11 years). Two hundred and twenty-four women (40.4%) had used HRT, and the median duration of use was 20 months (percentiles: 25th 5 6 months, 75th 5 49 months). The distribution of BMD values at each of the five bone sites is presented in Table 1. A total of 263 (46.8%) women had low BMD (T , 21) the femoral neck. Two hundred and ninety-seven (52.8%) women reported 524 fractures in any female relative, 172 (30.6%) women reported
Bone Vol. 24, No. 5 May 1999:507–512
M. J. Grainge et al. Family history of fracture and BMD
509
Table 2. Results of age-adjusted regression analyses to examine effects of family fracture history on BMD at five sites for fractures sustained at any age by: (a) any female relative, (b) mother, (c) sister, (d) mother or sister, and (e) other female relative Fractures sustained by: (a) Any female relative n (%) Mean BMD AP spine Femoral neck Greater troch. Total radius Whole body (b) Mother n (%) Mean BMD AP spine Femoral neck Greater troch. Total radius Whole body (c) Sister n (%) Mean BMD AP spine Femoral neck Greater troch. Total radius Whole body (d) Mother or Sister n (%) Mean BMD AP spine Femoral neck Greater troch. Total radius Whole body (e) Other female relative n (%) Mean BMD AP spine Femoral neck Greater troch. Total radius Whole body
FFH 1ve
FFH 2ve
297 (52.8)
265 (47.2)
0.954 0.744 0.670 0.517 1.035
0.969 0.767 0.683 0.524 1.052
172 (30.6)
390 (69.4)
0.947 0.734 0.663 0.516 1.029
0.967 0.764 0.682 0.523 1.049
62 (11.0)
500 (89.0)
0.945 0.744 0.664 0.515 1.027
0.963 0.756 0.678 0.521 1.045
211 (37.5)
351 (62.5)
0.947 0.737 0.663 0.517 1.030
0.970 0.766 0.684 0.523 1.050
141 (25.1)
421 (74.9)
0.962 0.748 0.678 0.514 1.035
0.961 0.757 0.675 0.523 1.045
Differencea (SE)
t-statistic
p
20.0147 (0.0117) 20.0227 (0.0095) 20.0128 (0.0085) 20.0070 (0.0039) 20.0173 (0.0075)
21.26 22.39 21.50 21.78 22.31
0.208 0.017 0.134 0.075 0.021
20.0197 (0.0126) 20.0296 (0.0102) 20.0192 (0.0092) 20.0068 (0.0043) 20.0192 (0.0081)
21.56 22.89 22.08 21.59 22.38
0.119 0.004 0.038 0.113 0.018
20.0179 (0.0187) 20.0124 (0.0151) 20.0138 (0.0136) 20.0063 (0.0063) 20.0174 (0.0081)
20.96 20.82 21.01 21.00 21.46
0.337 0.412 0.311 0.318 0.145
20.0228 (0.0120) 20.0294 (0.0097) 20.0211 (0.0088) 20.0064 (0.0041) 20.0205 (0.0077)
21.90 23.02 22.40 21.57 22.66
0.058 0.003 0.017 0.118 0.008
0.0008 (0.0135) 20.0092 (0.0110) 0.0026 (0.0099) 20.0090 (0.0046) 20.0105 (0.0087)
0.06 20.84 0.26 21.97 21.21
0.951 0.400 0.792 0.049 0.226
a (FFH 1ve) 2 (FFH 2ve). KEY: FFH 1ve, positive family fracture history; FFH 2ve, negative family fracture history; SE, standard error; t-statistic, difference/standard error.
233 fractures in their mother, and 62 (11.0%) women reported 86 fractures in a sister. Two hundred and eleven (37.5%) women reported 319 fractures sustained by their mother or a sister, and 182 (32.4%) women reported 277 fractures in their mother or a sister resulting from low trauma. Sixty-five women reported 79 fractures in their mother or a sister sustained before the age of 50, of which 54 (68.4%) occurred as a result of low trauma. One hundred and three women reported 131 fractures in their mother or a sister between the ages of 50 and 70, of which 122 (93.1%) were a result of low trauma; and 78 women reported 96 fractures after the age of 70 of which 92 (95.8%) were a result of low trauma. This indicates that fractures under age 50 were proportionately less likely to have occurred following low trauma than those after age 50. Subjects reporting a fracture in any female relative had a lower BMD at all sites, significantly so at the femoral neck and the whole body (p , 0.05; 3.0% reduction in BMD at the femoral neck) (Table 2). When only maternal fractures were counted, the differences in BMD between those with and without a family history increased at all sites except the total radius (p , 0.01 for differences in BMD at the femoral neck; 0.01 , p , 0.05 at the
greater trochanter and whole body; 3.9% reduction in BMD at the femoral neck). The differences in BMD were smaller when only fractures in a sister were counted, and were not statistically significant, partly because of the small number of fractures reported in sisters. However, when these fractures were combined with fractures in a mother, the associations were marginally more significant than when only maternal fractures were counted at four of the sites, and were very similar at the total radius (p , 0.01 at the femoral neck and whole body; 0.01 , p , 0.05 at the greater trochanter; 3.9% reduction in BMD at the femoral neck). A history of fractures in other female relatives was only significantly associated with BMD at the total radius (p 5 0.049). When fractures in a mother or sister were restricted to those which occurred as a result of low trauma, the differences in BMD between those with and without a family history increased at all five sites and were statistically significant at all sites (p , 0.01 at three sites; 0.01 , p , 0.05 at two sites; 4.6% reduction in BMD at the femoral neck) (Table 3). Particularly large differences in BMD at all sites were observed when only low-trauma fractures of the hip in a mother or sister were counted (8.9%
510
M. J. Grainge et al. Family history of fracture and BMD
Bone Vol. 24, No. 5 May 1999:507–512
Table 3. Results of age-adjusted regression analyses to examine effects of family fracture history on BMD at five sites for fractures occurring in a mother or sister that were: (a) sustained with low trauma, (b) sustained with low trauma at the hip, (c) sustained with low trauma after age 50, (d) sustained with low trauma before age 70 Low-trauma fractures in a mother or sister at: (a) Any site n (%) Mean BMD AP spine Femoral neck Greater troch. Total radius Whole body (b) The hip n (%) Mean BMD AP spine Femoral neck Greater troch. Total radius Whole body (c) Any site after age 50 n (%) Mean BMD AP spine Femoral neck Greater troch. Total radius Whole body (d) Any site before age 70 n (%) Mean BMD AP spine Femoral neck Greater troch. Total radius Whole body
FFH 1ve
FFH 2ve
182 (32.4)
380 (67.6)
0.942 0.732 0.658 0.514 1.025
0.970 0.766 0.685 0.524 1.051
23 (4.1)
539 (95.9)
0.888 0.695 0.630 0.509 0.995
0.964 0.757 0.678 0.521 1.045
151 (27.2)
405 (72.8)
0.939 0.728 0.658 0.512 1.024
0.968 0.765 0.683 0.523 1.049
125 (22.5)
431 (77.5)
0.937 0.728 0.653 0.515 1.019
0.967 0.762 0.683 0.522 1.049
Differencea (SE)
t-statistic
p
20.0283 (0.0124) 20.0342 (0.0100) 20.0263 (0.0090) 20.0093 (0.0042) 20.0259 (0.0079)
22.28 23.41 22.91 22.21 23.27
0.023 ,0.001 0.004 0.027 0.001
20.0766 (0.0293) 20.0629 (0.0238) 20.0482 (0.0214) 20.0122 (0.0100) 20.0493 (0.0188)
22.61 22.64 22.25 21.23 22.63
0.009 0.008 0.025 0.220 0.009
20.0284 (0.0132) 20.0369 (0.0106) 20.0253 (0.0096) 20.0115 (0.0044) 20.0248 (0.0084)
22.15 23.47 22.63 22.58 22.95
0.032 ,0.001 0.009 0.010 0.003
20.0298 (0.0141) 20.0344 (0.0114) 20.0304 (0.0103) 20.0073 (0.0048) 20.0304 (0.0077)
22.11 23.02 22.96 21.53 23.39
0.035 0.003 0.003 0.127 ,0.001
a (FFH 1ve) 2 (FFH 2ve). KEY: FFH 1ve, positive family fracture history; FFH 2ve, negative family fracture history; SE, standard error; t-statistic, difference/standard error.
reduction in BMD at the femoral neck). Owing to the small number of women (n 5 23) with such a family history, however, the standard errors of these differences were high, so the results were less statistically significant than when low-trauma fractures at any site were counted. When low-trauma fractures occurring in a mother or sister were stratified by age at occurrence of fracture (under 50 years, between 50 and 70 years, and over 70 years), evidence of a consistently lower BMD in women with a positive family history was apparent in all three age groups across all five sites, although the results were not always statistically significant owing to reduced numbers (data not shown). When fractures occurring before the age of 50 years were excluded, the association between family history and BMD was less significant than when no age restriction was made at the AP spine, the greater trochanter, and the whole body, with the significance levels very similar at the femoral neck (Table 3). When a similar exclusion was made for fractures occurring after the age of 70 years, the associations were less significant than they were for fractures at any age at the AP spine, the femoral neck, and the total radius. The significance levels for a history of fractures at any site in a mother or sister following low trauma occurring at any age remained very similar after adjustment was made for other BMD risk factors. The reduction in BMD at the femoral neck associated with a positive family history in this case was 4.2%, slightly lower than the 4.6% before adjustment.
The 119 (21.3%) women reporting a positive family history of osteoporosis (as distinct from fracture) did not have a significantly lower BMD than those without such a family history, even when restrictions were made according to the closeness of the relative and age at onset. When women were classified as osteopenic/osteoporotic (T , 21) or normal (T $ 21) based on the femoral neck BMD, the age-adjusted odds ratio (OR) for low BMD (T , 21) associated with a low-trauma fracture at any site in a mother or sister was 1.88 [95% confidence interval (CI) 5 1.30 –2.70]. When fractures were limited to those occurring at the hip, the OR increased, although this was accompanied by a wider CI (OR 5 2.54; 95% CI 5 1.02– 6.34). A history of low-trauma hip fracture in a mother or sister was a better predictor of low BMD at the femoral neck than a history of low-trauma fracture at any site: 70% of those with a family history of hip fracture had low BMD compared to 57% of those with a fracture at any site (Table 4) and 47% of the sample as a whole. The sensitivity for detecting low BMD was low for both definitions, but particularly so when the hip fracture restriction was made. Discussion We have found that a history of fracture in a female relative was associated with a reduction in BMD in postmenopausal women aged 45–59. The association between a family history of frac-
Bone Vol. 24, No. 5 May 1999:507–512
M. J. Grainge et al. Family history of fracture and BMD
Table 4. Sensitivity and specificity for diagnosing low BMD (T , 21) at femoral neck for fractures in a mother or sister sustained by low trauma at (a) any site and (b) the hip Femoral neck BMD Low-trauma fracture (a) mother or sister at any site Positive history: n (%) Negative history: n (%) Sensitivity Specificity (b) mother or sister at the hip Positive history: n (%) Negative history: n (%) Sensitivity Specificity
Low (T , 21)
Normal (T . 21)
103 (56.6a) 160 (42.1) 39.2%
79 (43.4) 220 (57.9b) 73.6%
16 (69.6a) 247 (45.8) 6.1%
7 (30.4) 292 (54.2b) 97.7%
a
Positive predictive value. Negative predictive value.
b
tures and BMD was strongest when fractures were restricted to those occurring in the subject’s mother or any of her sisters, and were sustained with low trauma. The small group of women whose mother or sister suffered a hip fracture as a result of minimal trauma had a particularly low BMD. Because the original participants in the EPIC study were a select group of women who were willing to participate in and were eligible for a clinical trial, and only around 50% of the recontacted sample that were eligible agreed to participate in the present study, the question as to whether this sample is entirely representative of all postmenopausal women aged 45–59 in Nottingham might be raised. Evidence that this sample does reflect the general population is given by the fact that the mean BMD at the femoral neck of the women in our sample who were aged 50 –59 is comparable with the mean BMD at this site for women aged 50 –59 who participated in phase 1 of the NHANES III study (0.743 g/cm2 vs. 0.732 g/cm2).12 Our sample also had a mean BMD at the AP spine which was similar to that of a reference sample of British women aged 50 –59 (0.949 g/cm2 vs. 0.929 g/cm2).7 We were unable to compare the proportion of women who reported a positive family history of fracture with that of any other population of the same age, owing to the limited number of studies that have looked at self-reported family history. However, the proportion of women reporting a lowtrauma fracture in a mother or sister was similar in those who participated in the main EPIC trial and those who did not. This comparison gives some indication that family history is unlikely to have been heavily influenced by sampling strategy or response patterns, although it is certainly possible that women who were aware of a family history of osteoporosis would have been more willing to participate in a study of osteoporosis. This would not bias the study results unless these women were aware of their BMD status; the study investigators were aware of neither family history nor BMD status at recruitment, and only 12 women who were excluded from the EPIC trial owing to low BMD knew their BMD at that time. One problem inherent in this type of study is the possibility of a misclassification of responses between those with a positive and negative family history. This would underestimate the true effect of this measure upon BMD. In particular, the finding that differences in BMD increased in significance when only counting fractures obtained by a mother or sister could be partly due to respondents being more likely to recall fractures in close relatives, as well as to a genetic effect. Thirty-seven women may have been aware of a personal history of osteoporosis, as they had reported a low-trauma fracture of the wrist (n 5 35) or
511
vertebra (n 5 2) after the age of 20. When these women along with the 12 women who had knowledge of a low BMD were removed, the size of the associations between family history and BMD remained similar. This suggests that recall was unlikely to be differential according to knowledge of a personal history of osteoporosis. A large number of significance tests were carried out, because outcome measurements at five different skeletal sites were used. There were strong correlations between the five BMD measurements and also the different definitions of family fracture history. Multiple comparison procedures such as the Bonferroni method would not, therefore, be valid here. Instead, interpretation of results took account of findings at all five sites, whereas the different measures of family fracture history which were used were all derived from preplanned hypotheses. The present findings are consistent with those of Evans et al.,6 who found that 35 relatives of patients who had suffered an osteoporotic fracture had a significantly lower BMD at the spine than did a control group. In addition, Seeman et al.14 found evidence that premenopausal daughters of osteoporotic women had a significantly lower bone mineral content at the spine than controls of the same age. Both of these studies used the presence of a vertebral fracture to define osteoporosis. A more recent study reported by Seeman et al.15 found that daughters of women with hip fractures had a lower BMD than controls at the femoral neck and femoral shaft. The present study, in particular, supports the findings of Soroko et al.,17 that self-reported family history was significantly associated with BMD. Their results were from a study of 600 men and 877 women aged 60 – 89, who were part of the Rancho Bernardo, California, cohort. A subject was classified as having a family history of osteoporosis if his or her mother or father, or a sister (although not brother) had either been diagnosed as having osteoporosis, developed kyphosis, for suffered a low-trauma fracture after the age of 50. Our finding that associations with BMD were strongest for fractures occurring in a mother or sister as compared with more distant relatives is what would be expected owing to the closer similarity of their genetic pool. Recent research has identified a number of genes which may predispose individuals to osteoporosis, perhaps the most widely studied of these has been the vitamin D receptor gene.3 The present study found no evidence that restrictions should be made on the age at which low-trauma fractures occur when determining whether a woman has a family history of fracture. Two studies2,18 which examined family history in conjunction with a number of other risk factors for low bone density only counted fractures sustained after the ages of 50 and 45, respectively. Low-trauma fractures that occur in women at a young age could still be a consequence of low peak bone mass, and they would therefore be at greater risk of osteoporosis once postmenopausal bone loss has started to occur. Seeman et al.14 used the presence of a nontraumatic vertebral compression fracture to define osteoporosis. Most fractures reported in our study were at the hip or wrist, with hardly any at the spine. Fractures of the vertebrae do not receive medical attention in most cases,20 so they would be difficult to ascertain in self-report studies. Our subjects, however, were asked if they knew of any relatives who they thought had suffered from osteoporosis. It was this question that attempted to identify evidence of osteoporosis at the spine, by enquiring about features such as a humped back or loss of height. However, unlike the fracture history question, this question was found to be very poor at predicting BMD. When only low-trauma hip fractures in a mother or sister were counted, a measure was produced that had a very strong association with BMD at all sites. Two other studies2,17 com-
512
M. J. Grainge et al. Family history of fracture and BMD
pared the effects of family history on bone mass either when considering hip fractures alone or when including all fractures. Both of these studies found the magnitude of the effect to be similar for the two definitions of family fracture history. With such a small number of women reporting a low-trauma hip fracture in our study, estimates were imprecise and the particularly large effect for this measure could have been due to chance. However, there were differences between our study and the other two, since both Bauer et al.2 and Soroko et al.17 used older women (60 years and older), whereas the latter17 study also included paternal fractures. If a question on family fracture history were to be used in an osteoporosis clinic, then important additional information comes from Soroko et al.17 They found that a paternal history of osteoporosis was associated with a reduction in BMD in women at both the hip and the spine. Whether fracture history, specifically, in fathers predicts BMD, and, if so, how it should be defined are matters for further investigation. From the present findings, it is recommended that as a measure of family fracture history, low-trauma fractures in a mother or sister without restrictions on age and site of fracture should be used by researchers in future risk factor studies. By using this definition which has the most statistically significant association with BMD, the proportion of variation in BMD explained by family history will be maximized. The sensitivity and specificity values suggest that family fracture history is unlikely to be a good diagnostic tool by itself when used in a clinic setting. It is possible, however, that it could be useful in combination with other risk factors—in particular, an early menopause and low body weight—for this purpose. Recent guidelines for the diagnosis and treatment of osteoporosis issued by the European Foundation for Osteoporosis10 and the Royal College of Physicians (unpublished) both suggested that risk factors have a role in identifying which individuals should receive a BMD scan, and that a family history of fracture is one of the risk factors which should be used. We found that the prevalence of low BMD in this early postmenopausal age group was noticeably higher among women reporting a low-trauma hip fracture in their mother or sister, and it may be that this small group of women in particular should be recommended for a BMD scan.
Acknowledgments: The authors thank all the subjects who participated in the study. They thank Anne Williams, who carried out much of the interviewing; Pat San, who gave continuing help throughout the study; and other members of the Nottingham EPIC study group, who have helped out with the epidemiology study. The EPIC and EPIC add-on study were both funded by a grant from Merck & Co., Inc.
Bone Vol. 24, No. 5 May 1999:507–512 3. Cooper, G. S. and Umbach, D. M. Are vitamin D receptor polymorphisms associated with bone mineral density? A meta-analysis. J Bone Miner Res 11:1841–1849; 1996. 4. Coupland, C. A. C., Cliffe, S. J., Lyons, A. R., Tolley, K., Hosking, D. J., and Chilvers, C. E. D. Costs of recruiting women for an osteoporosis prevention trial. J Epidemiol Biostat 2:179 –183; 1997. 5. Dequeker, J., Nijs, J., Verstraeten, A., Geusens, P., and Gevers, G. Genetic determinants of bone mineral content at the spine and radius: a twin study. Bone 8:207–209; 1987. 6. Evans, R. A., Marel, G. M., Lancaster, E. K., Kos, S., Evans, M., and Wong, S. Y. P. Bone mass is low in relatives of osteoporotic patients. Ann Intern Med 109:870 – 873; 1988. 7. Hall, M. L., Heavans, J., Cullum, I. D., and Ell, P. J. The range of bone density in normal British women. Br J Radiol 63:266 –269; 1990. 8. Jouanny, P., Guillemin, F., Kuntz, C., Jeandel, C., and Pourel, J. Environmental and genetic factors affecting bone mass. Arthritis Rheum 38:61– 67; 1995. 9. Kahn, S. A., Pace, J. E., Cox, M. L., Gau, D. W., Cox, S. A. L., and Hodkinson, H. M. Osteoporosis and genetic influence: a three-generation study. Postgrad Med J 70:798 – 800; 1994. 10. Kanis, J. A., Delmas, P., Burckhardt, P., Cooper, C., and Torgerson, D. Guidelines for diagnosis and management of osteoporosis. Osteoporos Int 7:390 – 406; 1997. 11. Looker, A. C., Orwoll, E. S., Johnston, C. C., Jr., Lindsay, R. L., Wahner, H. W., Dunn, W. L., Calvo, M. S., Harris, T. B., and Heyse, S. P. Prevalence of low femoral bone density in older U.S. adults from NHANES III. J Bone Miner Res 12:1761–1768; 1997. 12. Looker, A. C., Wahner, H. W., Dunn, W. L., Calvo, M. S., Harris, T. B., Heyse, S. P., Johnston, C. C., Jr., and Lindsay, R. L. Proximal femur bone mineral levels of US adults. Osteoporos Int 5:389 – 409; 1995. 13. Pocock, N. A., Eisman, J. A., Hopper, J. L., Yeates, M. G., Sambrook, P. N., and Eberl, S. Genetic determinants of bone mass: a twin study. J Clin Invest 80:706 –710; 1987. 14. Seeman, E., Hopper, J. L., Bach, L. A., Cooper, M. E., Parkinson, E., McKay, J., and Jerums, G. Reduced bone mass in daughters of women with osteoporosis. N Engl J Med 320:554 –558; 1989. 15. Seeman, E., Tsalamandris, C., Formica, C., Hopper, J. L., and McKay, J. Reduced femoral neck bone density in the daughters of women with hip fractures: the role of low peak bone density in the pathogenesis of osteoporosis. J Bone Miner Res 9:739 –743; 1994. 16. Slemenda, C. W., Christian, J. C., Williams, C. J., Norton, J. A., and Johnston, C. C., Jr. Genetic determinants of bone mass in adult women: A reevaluation of the twin model and the potential importance of gene interaction on heritability estimates. J Bone Miner Res 6:561–567; 1991. 17. Soroko, S. B., Barrett-Connor, E., Edelstein, S. L., and Kritz-Silverstein, D. Family history of osteoporosis and bone mineral density at the axial skeleton: The Rancho Bernardo Study. J Bone Miner Res 9:761–769; 1994. 18. Spector, T. D., Edwards, A. C., Thompson, P. W. Use of a risk factor and dietary calcium questionnaire in predicting bone density and subsequent bone loss at the menopause. Ann Rheum Dis. 51:1252–1253; 1992. 19. Tylavsky, F. A., Bortz, A. D., Hancock, R. L., and Anderson, J. J. B. Familial resemblance of radial bone mass between premenopausal mothers and their college-age daughters. Calcif Tissue Int 45:265–272; 1989. 20. WHO Study Group. Assessment of fracture risk and its applications to screening for postmenopausal osteoporosis. Geneva: World Health Organization; 1994.
References 1. Arden, N. K., Baker, J., Hogg, C., Baan, K., and Spector, T. D. The heritability of bone mineral density, ultrasound of the calcaneus and hip axis length: A study of postmenopausal twins. J Bone Miner Res 11:530 –534; 1996. 2. Bauer, D. C., Browner, W. S., Cauley, J. A., Orwoll, E. S., Scott, J. C., Black, D. M., Tao, J. L., and Cummings, S. R. Factors associated with appendicular bone mass in older women. Ann Intern Med. 118:657– 665; 1993.
Date Received: June 25, 1998 Date Revised: December 16, 1998 Date Accepted: December 16, 1998
Association Between a Family History of Fractures and Bone Mineral Density in Early Postmenopausal Women M. J. GRAINGE,1 C. A. C. COUPLAND,1 S. J. CLIFFE,1 C. E. D. CHILVERS,1 and D. J. HOSKING2 1
Division of Public Health Medicine and Epidemiology, School of Community Health Sciences, University of Nottingham Medical School, Nottingham, UK 2 Division of Mineral Metabolism, Nottingham City Hospital, Nottingham, UK
Key Words: Bone density; Fractures; Heredity; Osteoporosis; Postmenopausal; Risk factors.
The aim of this analysis was to measure the strength of the association between a family history of fractures and bone mineral density (BMD), and to determine what definition of family fracture history best predicts BMD. Five hundred and eighty postmenopausal women aged 45–59 at recruitment completed a risk factor questionnaire. Women were asked to recall details of fractures sustained by any female relative. BMD measurements taken at five sites were used. The data were analysed using linear regression, adjusting for age. Two hundred and ninety-seven (52.8%) women reported a family history of fractures, and they had a significantly lower BMD at two of the sites measured (p < 0.05). The associations with BMD were most significant when only counting fractures that occurred in the subject’s mother or a sister as a result of low trauma, with no restrictions made on age at the time of fracture and site of fracture (p < 0.01 at three sites; 0.01 < p < 0.05 at two sites). Women with a family history according to this definition had a 4.6% reduction in BMD at the femoral neck. When T scores were used to categorize women as either osteopenic/osteoporotic (T < 21) or normal at the femoral neck, the sensitivity of using this definition was 39% and the specificity was 74%. The small group of women that reported a low-trauma hip fracture in a mother or sister (n 5 23) had a mean femoral neck BMD which was 8.9% lower than that of the remainder of the sample, although this difference was less statistically significant than when low trauma fractures at any site were counted. Of these 23 women, 70% were osteopenic or osteoporotic, compared with 57% of those reporting a low-trauma fracture at any site and 47% of the sample as a whole. The sensitivity of this definition, however, was low (6%). From these analyses it can be concluded that the definition of family fracture history that best predicts BMD in postmenopausal women is a fracture at any age in a mother or sister resulting from low trauma, although the sensitivity and specificity of using a family history of fractures by itself to screen for low BMD were poor. (Bone 24:507–512; 1999) © 1999 by Elsevier Science Inc. All rights reserved.
Introduction Studies of monozygotic and dizygotic twins1,5,13,16 have shown strong correlations in bone mass between pairs of twins and have also provided evidence that this association is due at least partly to genetic factors. Other studies8,9,19 have reported that strong and significant correlations in bone mass exist among firstdegree relatives. Three further studies6,14,15 have found that relatives of patients with osteoporosis have a significantly lower bone mass when compared with control groups. Although all these studies have produced findings that suggest that osteoporosis in a relative is a risk factor for low bone mass, in practice the only information that would be available to clinicians would be from a simple question asking patients to report any family history of osteoporosis. The question of whether such a measure correlates with bone density was addressed by Soroko et al.,17 who found that a positive reported family history of osteoporosis was associated with significantly lower bone mineral density (BMD) at the lumbar spine for women and total hip for men. In this study, we investigate whether a simple question on fracture history in female relatives would be able to predict BMD at various bone sites. In particular, our aim was to identify what definition of family fracture history had the strongest association with BMD, thereby suggesting a definition that could be used both in the clinic to predict low bone density, and by researchers who want to incorporate a self-reported measure of family history into a study of bone density. Materials and Methods The study participants comprised two groups of women. The first group included postmenopausal women participating in the Early Postmenopausal Interventional Cohort (EPIC) study. This is a multicenter clinical trial carried out in four study centers (Copenhagen, Denmark; Hawaii, United States; Nottingham, UK; and Portland, Oregon, United States). The trial was established to test the efficacy of the bisphosphonate alendronate in preventing postmenopausal bone loss. The subjects used for the analysis reported in the present study came from the Nottingham center. Recruitment in Nottingham took place between September 1992 and June 1993, using a two-stage sampling procedure. First, 60 general practices were randomly selected from all those in the Nottingham Health district. The next stage was to sample at
Address for correspondence and reprints: Mr. M. J. Grainge, Division of Public Health Medicine and Epidemiology, School of Community Health Sciences, University of Nottingham Medical School, Queens Medical Centre, Nottingham, NG7 2UH, UK. E-mail: [email protected] © 1999 by Elsevier Science Inc. All rights reserved.
507
8756-3282/99/$20.00 PII S8756-3282(99)00016-2
508
M. J. Grainge et al. Family history of fracture and BMD
Bone Vol. 24, No. 5 May 1999:507–512
Table 1. Distribution of BMD values at five bone sites
AP spine Femoral neck Greater trochanter Total radius Whole body
n
Mean
SD
Minimum
Maximum
562 562 562 562 559
0.961 0.755 0.676 0.521 1.043
0.141 0.115 0.102 0.050 0.091
0.584 0.473 0.383 0.356 0.751
1.380 1.320 1.100 0.696 1.361
random 150 women aged 45–59 from each participating practice. These women were invited to attend information sessions, followed by a screening clinic. The main inclusion criteria were that subjects had to be postmenopausal, in good general health, and must not have taken hormone replacement therapy (HRT) within the last 3 months. Also, no more than 10% of subjects were permitted to have a spinal BMD below 0.8 g/cm2. Full details of the recruitment procedure have been published elsewhere.4 The second group of women included in this study comprised a random sample of women who had been ineligible or unwilling to take part in the EPIC trial, and had been recontacted. Inclusion/ exclusion criteria for the epidemiological study were that the subjects still had to be postmenopausal, but women previously excluded owing to medical history, low bone density, or the use of HRT were now included. It was intended in this way to construct a final sample of postmenopausal women more typical of those in the general Nottingham population. All women in the study provided written informed consent, and the study was approved by the Nottingham City Hospital Ethics Committee. All subjects completed an interviewer-administered risk factor questionnaire, which contained the question, “Have any women in your family had a broken, cracked, or fractured bone?” For each fracture reported, the bone fractured, the age at which the fracture occurred, and the circumstance from which the fracture arose were recorded. Subjects were also asked the question, “Have any of the women in your family suffered from osteoporosis (brittle bones)?” They were prompted to recall female relatives who had classic signs of osteoporosis (i.e., a humped back and loss of height). The age at which it began was recorded for each relative reported. The questionnaire also contained sections on menstrual history, HRT use, and other potential risk factors for osteoporosis. Bone mineral density was measured using dual-energy X-ray absorptiometry (Hologic 2000; Waltham, MA). Measurements taken at five sites were used: antero-posterior spine (AP spine), femoral neck, greater trochanter, radius/ulna (total radius), and whole body. For subjects in the EPIC trial, their baseline scan prior to commencing therapy was used in the analyses reported here. For the other subjects, a single scan was carried out around the same time that the questionnaire was administered. The coefficients of variation calculated for women in the EPIC trial are those which are used to assess the reproducibility of the local scanner. These subjects had their BMD measured twice within the space of 14 days, and the pairwise coefficients of variation were 1.48% for the AP spine, 1.77% for the femoral neck, 1.36% for the trochanter, 0.98% for the total radius, and 0.90% for the whole body. Height and weight were also measured at the time of the BMD scan. Linear regression was used to analyze the data using the Statistical Analysis System (SAS Institute, Cary, NC), with BMD measurements at the five different sites as the dependent variables. All analyses were adjusted for age, as older subjects will tend to have older relatives who will have had more opportunity for fracture. Different definitions of family fracture history were used as explanatory variables. Restrictions were made according to the age at which the fracture occurred (under
50, between 50 and 70, and over 70), whether it was sustained by a first-degree female relative (mother or sister) or other female relative, and whether the fracture was considered to be the result of only low trauma (such as a fall from standing). For all definitions, family fracture history was treated as a two-level factor, indicating the presence or absence of a family history. In further analyses, the effect of only counting hip fractures was investigated, as was the effect of controlling for other BMD risk factors: years since menopause, height, weight, duration of HRT use, and a two-level factor indicating whether the subject was part of the main EPIC trial. Finally a residual analysis was carried out to test model assumptions and to identify any unduly influential observations. A p value , 0.05 was used to denote statistical significance. The BMD measurements at the femoral neck were converted into T scores representing the number of standard deviations above or below the mean BMD for healthy young women, using reference ranges derived from phases 1 and 2 of the National Health and Nutrition Examination Survey (NHANES) III study.11 These were then categorized to classify subjects as either osteoporotic (T # 22.5), osteopenic (22.5 , T , 21), or normal (T $ 21) at the femoral neck in accordance with the World Health Organization (WHO) definition.20 As only a small number of women (n 5 13) were osteoporotic (T # 22.5), these women were combined with the osteopenic group, to provide the low BMD (T , 21) group for the present analysis. Odds ratios associated with low BMD were then produced for family fracture history, adjusting for age by the use of logistic regression. The same classification was also used to calculate the sensitivity and specificity of using family fracture history as a diagnostic tool for predicting low bone density. Results Of the 7564 women originally approached, 428 (5.6%) were randomized into the EPIC trial.4 Of these, 349 (81.5%) agreed to participate in the epidemiology study. Six hundred and two women who had not been included in EPIC were recontacted, of whom 231 (38.4%) were eligible and agreed to take part. Of the remainder, 156 (25.9%) were found to be ineligible and 215 (35.7%) either refused or did not respond. Of the 580 participants recruited from the two groups, 562 women (336 from the EPIC study) gave a definite positive or negative answer to the family fracture history question, and were used in this analysis. The participants had a median age of 53 years (percentiles: 25th 5 50, 75th 5 57) and were an average (median) of 6 years past the menopause (percentiles: 25th 5 3 years, 75th 5 11 years). Two hundred and twenty-four women (40.4%) had used HRT, and the median duration of use was 20 months (percentiles: 25th 5 6 months, 75th 5 49 months). The distribution of BMD values at each of the five bone sites is presented in Table 1. A total of 263 (46.8%) women had low BMD (T , 21) the femoral neck. Two hundred and ninety-seven (52.8%) women reported 524 fractures in any female relative, 172 (30.6%) women reported
Bone Vol. 24, No. 5 May 1999:507–512
M. J. Grainge et al. Family history of fracture and BMD
509
Table 2. Results of age-adjusted regression analyses to examine effects of family fracture history on BMD at five sites for fractures sustained at any age by: (a) any female relative, (b) mother, (c) sister, (d) mother or sister, and (e) other female relative Fractures sustained by: (a) Any female relative n (%) Mean BMD AP spine Femoral neck Greater troch. Total radius Whole body (b) Mother n (%) Mean BMD AP spine Femoral neck Greater troch. Total radius Whole body (c) Sister n (%) Mean BMD AP spine Femoral neck Greater troch. Total radius Whole body (d) Mother or Sister n (%) Mean BMD AP spine Femoral neck Greater troch. Total radius Whole body (e) Other female relative n (%) Mean BMD AP spine Femoral neck Greater troch. Total radius Whole body
FFH 1ve
FFH 2ve
297 (52.8)
265 (47.2)
0.954 0.744 0.670 0.517 1.035
0.969 0.767 0.683 0.524 1.052
172 (30.6)
390 (69.4)
0.947 0.734 0.663 0.516 1.029
0.967 0.764 0.682 0.523 1.049
62 (11.0)
500 (89.0)
0.945 0.744 0.664 0.515 1.027
0.963 0.756 0.678 0.521 1.045
211 (37.5)
351 (62.5)
0.947 0.737 0.663 0.517 1.030
0.970 0.766 0.684 0.523 1.050
141 (25.1)
421 (74.9)
0.962 0.748 0.678 0.514 1.035
0.961 0.757 0.675 0.523 1.045
Differencea (SE)
t-statistic
p
20.0147 (0.0117) 20.0227 (0.0095) 20.0128 (0.0085) 20.0070 (0.0039) 20.0173 (0.0075)
21.26 22.39 21.50 21.78 22.31
0.208 0.017 0.134 0.075 0.021
20.0197 (0.0126) 20.0296 (0.0102) 20.0192 (0.0092) 20.0068 (0.0043) 20.0192 (0.0081)
21.56 22.89 22.08 21.59 22.38
0.119 0.004 0.038 0.113 0.018
20.0179 (0.0187) 20.0124 (0.0151) 20.0138 (0.0136) 20.0063 (0.0063) 20.0174 (0.0081)
20.96 20.82 21.01 21.00 21.46
0.337 0.412 0.311 0.318 0.145
20.0228 (0.0120) 20.0294 (0.0097) 20.0211 (0.0088) 20.0064 (0.0041) 20.0205 (0.0077)
21.90 23.02 22.40 21.57 22.66
0.058 0.003 0.017 0.118 0.008
0.0008 (0.0135) 20.0092 (0.0110) 0.0026 (0.0099) 20.0090 (0.0046) 20.0105 (0.0087)
0.06 20.84 0.26 21.97 21.21
0.951 0.400 0.792 0.049 0.226
a (FFH 1ve) 2 (FFH 2ve). KEY: FFH 1ve, positive family fracture history; FFH 2ve, negative family fracture history; SE, standard error; t-statistic, difference/standard error.
233 fractures in their mother, and 62 (11.0%) women reported 86 fractures in a sister. Two hundred and eleven (37.5%) women reported 319 fractures sustained by their mother or a sister, and 182 (32.4%) women reported 277 fractures in their mother or a sister resulting from low trauma. Sixty-five women reported 79 fractures in their mother or a sister sustained before the age of 50, of which 54 (68.4%) occurred as a result of low trauma. One hundred and three women reported 131 fractures in their mother or a sister between the ages of 50 and 70, of which 122 (93.1%) were a result of low trauma; and 78 women reported 96 fractures after the age of 70 of which 92 (95.8%) were a result of low trauma. This indicates that fractures under age 50 were proportionately less likely to have occurred following low trauma than those after age 50. Subjects reporting a fracture in any female relative had a lower BMD at all sites, significantly so at the femoral neck and the whole body (p , 0.05; 3.0% reduction in BMD at the femoral neck) (Table 2). When only maternal fractures were counted, the differences in BMD between those with and without a family history increased at all sites except the total radius (p , 0.01 for differences in BMD at the femoral neck; 0.01 , p , 0.05 at the
greater trochanter and whole body; 3.9% reduction in BMD at the femoral neck). The differences in BMD were smaller when only fractures in a sister were counted, and were not statistically significant, partly because of the small number of fractures reported in sisters. However, when these fractures were combined with fractures in a mother, the associations were marginally more significant than when only maternal fractures were counted at four of the sites, and were very similar at the total radius (p , 0.01 at the femoral neck and whole body; 0.01 , p , 0.05 at the greater trochanter; 3.9% reduction in BMD at the femoral neck). A history of fractures in other female relatives was only significantly associated with BMD at the total radius (p 5 0.049). When fractures in a mother or sister were restricted to those which occurred as a result of low trauma, the differences in BMD between those with and without a family history increased at all five sites and were statistically significant at all sites (p , 0.01 at three sites; 0.01 , p , 0.05 at two sites; 4.6% reduction in BMD at the femoral neck) (Table 3). Particularly large differences in BMD at all sites were observed when only low-trauma fractures of the hip in a mother or sister were counted (8.9%
510
M. J. Grainge et al. Family history of fracture and BMD
Bone Vol. 24, No. 5 May 1999:507–512
Table 3. Results of age-adjusted regression analyses to examine effects of family fracture history on BMD at five sites for fractures occurring in a mother or sister that were: (a) sustained with low trauma, (b) sustained with low trauma at the hip, (c) sustained with low trauma after age 50, (d) sustained with low trauma before age 70 Low-trauma fractures in a mother or sister at: (a) Any site n (%) Mean BMD AP spine Femoral neck Greater troch. Total radius Whole body (b) The hip n (%) Mean BMD AP spine Femoral neck Greater troch. Total radius Whole body (c) Any site after age 50 n (%) Mean BMD AP spine Femoral neck Greater troch. Total radius Whole body (d) Any site before age 70 n (%) Mean BMD AP spine Femoral neck Greater troch. Total radius Whole body
FFH 1ve
FFH 2ve
182 (32.4)
380 (67.6)
0.942 0.732 0.658 0.514 1.025
0.970 0.766 0.685 0.524 1.051
23 (4.1)
539 (95.9)
0.888 0.695 0.630 0.509 0.995
0.964 0.757 0.678 0.521 1.045
151 (27.2)
405 (72.8)
0.939 0.728 0.658 0.512 1.024
0.968 0.765 0.683 0.523 1.049
125 (22.5)
431 (77.5)
0.937 0.728 0.653 0.515 1.019
0.967 0.762 0.683 0.522 1.049
Differencea (SE)
t-statistic
p
20.0283 (0.0124) 20.0342 (0.0100) 20.0263 (0.0090) 20.0093 (0.0042) 20.0259 (0.0079)
22.28 23.41 22.91 22.21 23.27
0.023 ,0.001 0.004 0.027 0.001
20.0766 (0.0293) 20.0629 (0.0238) 20.0482 (0.0214) 20.0122 (0.0100) 20.0493 (0.0188)
22.61 22.64 22.25 21.23 22.63
0.009 0.008 0.025 0.220 0.009
20.0284 (0.0132) 20.0369 (0.0106) 20.0253 (0.0096) 20.0115 (0.0044) 20.0248 (0.0084)
22.15 23.47 22.63 22.58 22.95
0.032 ,0.001 0.009 0.010 0.003
20.0298 (0.0141) 20.0344 (0.0114) 20.0304 (0.0103) 20.0073 (0.0048) 20.0304 (0.0077)
22.11 23.02 22.96 21.53 23.39
0.035 0.003 0.003 0.127 ,0.001
a (FFH 1ve) 2 (FFH 2ve). KEY: FFH 1ve, positive family fracture history; FFH 2ve, negative family fracture history; SE, standard error; t-statistic, difference/standard error.
reduction in BMD at the femoral neck). Owing to the small number of women (n 5 23) with such a family history, however, the standard errors of these differences were high, so the results were less statistically significant than when low-trauma fractures at any site were counted. When low-trauma fractures occurring in a mother or sister were stratified by age at occurrence of fracture (under 50 years, between 50 and 70 years, and over 70 years), evidence of a consistently lower BMD in women with a positive family history was apparent in all three age groups across all five sites, although the results were not always statistically significant owing to reduced numbers (data not shown). When fractures occurring before the age of 50 years were excluded, the association between family history and BMD was less significant than when no age restriction was made at the AP spine, the greater trochanter, and the whole body, with the significance levels very similar at the femoral neck (Table 3). When a similar exclusion was made for fractures occurring after the age of 70 years, the associations were less significant than they were for fractures at any age at the AP spine, the femoral neck, and the total radius. The significance levels for a history of fractures at any site in a mother or sister following low trauma occurring at any age remained very similar after adjustment was made for other BMD risk factors. The reduction in BMD at the femoral neck associated with a positive family history in this case was 4.2%, slightly lower than the 4.6% before adjustment.
The 119 (21.3%) women reporting a positive family history of osteoporosis (as distinct from fracture) did not have a significantly lower BMD than those without such a family history, even when restrictions were made according to the closeness of the relative and age at onset. When women were classified as osteopenic/osteoporotic (T , 21) or normal (T $ 21) based on the femoral neck BMD, the age-adjusted odds ratio (OR) for low BMD (T , 21) associated with a low-trauma fracture at any site in a mother or sister was 1.88 [95% confidence interval (CI) 5 1.30 –2.70]. When fractures were limited to those occurring at the hip, the OR increased, although this was accompanied by a wider CI (OR 5 2.54; 95% CI 5 1.02– 6.34). A history of low-trauma hip fracture in a mother or sister was a better predictor of low BMD at the femoral neck than a history of low-trauma fracture at any site: 70% of those with a family history of hip fracture had low BMD compared to 57% of those with a fracture at any site (Table 4) and 47% of the sample as a whole. The sensitivity for detecting low BMD was low for both definitions, but particularly so when the hip fracture restriction was made. Discussion We have found that a history of fracture in a female relative was associated with a reduction in BMD in postmenopausal women aged 45–59. The association between a family history of frac-
Bone Vol. 24, No. 5 May 1999:507–512
M. J. Grainge et al. Family history of fracture and BMD
Table 4. Sensitivity and specificity for diagnosing low BMD (T , 21) at femoral neck for fractures in a mother or sister sustained by low trauma at (a) any site and (b) the hip Femoral neck BMD Low-trauma fracture (a) mother or sister at any site Positive history: n (%) Negative history: n (%) Sensitivity Specificity (b) mother or sister at the hip Positive history: n (%) Negative history: n (%) Sensitivity Specificity
Low (T , 21)
Normal (T . 21)
103 (56.6a) 160 (42.1) 39.2%
79 (43.4) 220 (57.9b) 73.6%
16 (69.6a) 247 (45.8) 6.1%
7 (30.4) 292 (54.2b) 97.7%
a
Positive predictive value. Negative predictive value.
b
tures and BMD was strongest when fractures were restricted to those occurring in the subject’s mother or any of her sisters, and were sustained with low trauma. The small group of women whose mother or sister suffered a hip fracture as a result of minimal trauma had a particularly low BMD. Because the original participants in the EPIC study were a select group of women who were willing to participate in and were eligible for a clinical trial, and only around 50% of the recontacted sample that were eligible agreed to participate in the present study, the question as to whether this sample is entirely representative of all postmenopausal women aged 45–59 in Nottingham might be raised. Evidence that this sample does reflect the general population is given by the fact that the mean BMD at the femoral neck of the women in our sample who were aged 50 –59 is comparable with the mean BMD at this site for women aged 50 –59 who participated in phase 1 of the NHANES III study (0.743 g/cm2 vs. 0.732 g/cm2).12 Our sample also had a mean BMD at the AP spine which was similar to that of a reference sample of British women aged 50 –59 (0.949 g/cm2 vs. 0.929 g/cm2).7 We were unable to compare the proportion of women who reported a positive family history of fracture with that of any other population of the same age, owing to the limited number of studies that have looked at self-reported family history. However, the proportion of women reporting a lowtrauma fracture in a mother or sister was similar in those who participated in the main EPIC trial and those who did not. This comparison gives some indication that family history is unlikely to have been heavily influenced by sampling strategy or response patterns, although it is certainly possible that women who were aware of a family history of osteoporosis would have been more willing to participate in a study of osteoporosis. This would not bias the study results unless these women were aware of their BMD status; the study investigators were aware of neither family history nor BMD status at recruitment, and only 12 women who were excluded from the EPIC trial owing to low BMD knew their BMD at that time. One problem inherent in this type of study is the possibility of a misclassification of responses between those with a positive and negative family history. This would underestimate the true effect of this measure upon BMD. In particular, the finding that differences in BMD increased in significance when only counting fractures obtained by a mother or sister could be partly due to respondents being more likely to recall fractures in close relatives, as well as to a genetic effect. Thirty-seven women may have been aware of a personal history of osteoporosis, as they had reported a low-trauma fracture of the wrist (n 5 35) or
511
vertebra (n 5 2) after the age of 20. When these women along with the 12 women who had knowledge of a low BMD were removed, the size of the associations between family history and BMD remained similar. This suggests that recall was unlikely to be differential according to knowledge of a personal history of osteoporosis. A large number of significance tests were carried out, because outcome measurements at five different skeletal sites were used. There were strong correlations between the five BMD measurements and also the different definitions of family fracture history. Multiple comparison procedures such as the Bonferroni method would not, therefore, be valid here. Instead, interpretation of results took account of findings at all five sites, whereas the different measures of family fracture history which were used were all derived from preplanned hypotheses. The present findings are consistent with those of Evans et al.,6 who found that 35 relatives of patients who had suffered an osteoporotic fracture had a significantly lower BMD at the spine than did a control group. In addition, Seeman et al.14 found evidence that premenopausal daughters of osteoporotic women had a significantly lower bone mineral content at the spine than controls of the same age. Both of these studies used the presence of a vertebral fracture to define osteoporosis. A more recent study reported by Seeman et al.15 found that daughters of women with hip fractures had a lower BMD than controls at the femoral neck and femoral shaft. The present study, in particular, supports the findings of Soroko et al.,17 that self-reported family history was significantly associated with BMD. Their results were from a study of 600 men and 877 women aged 60 – 89, who were part of the Rancho Bernardo, California, cohort. A subject was classified as having a family history of osteoporosis if his or her mother or father, or a sister (although not brother) had either been diagnosed as having osteoporosis, developed kyphosis, for suffered a low-trauma fracture after the age of 50. Our finding that associations with BMD were strongest for fractures occurring in a mother or sister as compared with more distant relatives is what would be expected owing to the closer similarity of their genetic pool. Recent research has identified a number of genes which may predispose individuals to osteoporosis, perhaps the most widely studied of these has been the vitamin D receptor gene.3 The present study found no evidence that restrictions should be made on the age at which low-trauma fractures occur when determining whether a woman has a family history of fracture. Two studies2,18 which examined family history in conjunction with a number of other risk factors for low bone density only counted fractures sustained after the ages of 50 and 45, respectively. Low-trauma fractures that occur in women at a young age could still be a consequence of low peak bone mass, and they would therefore be at greater risk of osteoporosis once postmenopausal bone loss has started to occur. Seeman et al.14 used the presence of a nontraumatic vertebral compression fracture to define osteoporosis. Most fractures reported in our study were at the hip or wrist, with hardly any at the spine. Fractures of the vertebrae do not receive medical attention in most cases,20 so they would be difficult to ascertain in self-report studies. Our subjects, however, were asked if they knew of any relatives who they thought had suffered from osteoporosis. It was this question that attempted to identify evidence of osteoporosis at the spine, by enquiring about features such as a humped back or loss of height. However, unlike the fracture history question, this question was found to be very poor at predicting BMD. When only low-trauma hip fractures in a mother or sister were counted, a measure was produced that had a very strong association with BMD at all sites. Two other studies2,17 com-
512
M. J. Grainge et al. Family history of fracture and BMD
pared the effects of family history on bone mass either when considering hip fractures alone or when including all fractures. Both of these studies found the magnitude of the effect to be similar for the two definitions of family fracture history. With such a small number of women reporting a low-trauma hip fracture in our study, estimates were imprecise and the particularly large effect for this measure could have been due to chance. However, there were differences between our study and the other two, since both Bauer et al.2 and Soroko et al.17 used older women (60 years and older), whereas the latter17 study also included paternal fractures. If a question on family fracture history were to be used in an osteoporosis clinic, then important additional information comes from Soroko et al.17 They found that a paternal history of osteoporosis was associated with a reduction in BMD in women at both the hip and the spine. Whether fracture history, specifically, in fathers predicts BMD, and, if so, how it should be defined are matters for further investigation. From the present findings, it is recommended that as a measure of family fracture history, low-trauma fractures in a mother or sister without restrictions on age and site of fracture should be used by researchers in future risk factor studies. By using this definition which has the most statistically significant association with BMD, the proportion of variation in BMD explained by family history will be maximized. The sensitivity and specificity values suggest that family fracture history is unlikely to be a good diagnostic tool by itself when used in a clinic setting. It is possible, however, that it could be useful in combination with other risk factors—in particular, an early menopause and low body weight—for this purpose. Recent guidelines for the diagnosis and treatment of osteoporosis issued by the European Foundation for Osteoporosis10 and the Royal College of Physicians (unpublished) both suggested that risk factors have a role in identifying which individuals should receive a BMD scan, and that a family history of fracture is one of the risk factors which should be used. We found that the prevalence of low BMD in this early postmenopausal age group was noticeably higher among women reporting a low-trauma hip fracture in their mother or sister, and it may be that this small group of women in particular should be recommended for a BMD scan.
Acknowledgments: The authors thank all the subjects who participated in the study. They thank Anne Williams, who carried out much of the interviewing; Pat San, who gave continuing help throughout the study; and other members of the Nottingham EPIC study group, who have helped out with the epidemiology study. The EPIC and EPIC add-on study were both funded by a grant from Merck & Co., Inc.
Bone Vol. 24, No. 5 May 1999:507–512 3. Cooper, G. S. and Umbach, D. M. Are vitamin D receptor polymorphisms associated with bone mineral density? A meta-analysis. J Bone Miner Res 11:1841–1849; 1996. 4. Coupland, C. A. C., Cliffe, S. J., Lyons, A. R., Tolley, K., Hosking, D. J., and Chilvers, C. E. D. Costs of recruiting women for an osteoporosis prevention trial. J Epidemiol Biostat 2:179 –183; 1997. 5. Dequeker, J., Nijs, J., Verstraeten, A., Geusens, P., and Gevers, G. Genetic determinants of bone mineral content at the spine and radius: a twin study. Bone 8:207–209; 1987. 6. Evans, R. A., Marel, G. M., Lancaster, E. K., Kos, S., Evans, M., and Wong, S. Y. P. Bone mass is low in relatives of osteoporotic patients. Ann Intern Med 109:870 – 873; 1988. 7. Hall, M. L., Heavans, J., Cullum, I. D., and Ell, P. J. The range of bone density in normal British women. Br J Radiol 63:266 –269; 1990. 8. Jouanny, P., Guillemin, F., Kuntz, C., Jeandel, C., and Pourel, J. Environmental and genetic factors affecting bone mass. Arthritis Rheum 38:61– 67; 1995. 9. Kahn, S. A., Pace, J. E., Cox, M. L., Gau, D. W., Cox, S. A. L., and Hodkinson, H. M. Osteoporosis and genetic influence: a three-generation study. Postgrad Med J 70:798 – 800; 1994. 10. Kanis, J. A., Delmas, P., Burckhardt, P., Cooper, C., and Torgerson, D. Guidelines for diagnosis and management of osteoporosis. Osteoporos Int 7:390 – 406; 1997. 11. Looker, A. C., Orwoll, E. S., Johnston, C. C., Jr., Lindsay, R. L., Wahner, H. W., Dunn, W. L., Calvo, M. S., Harris, T. B., and Heyse, S. P. Prevalence of low femoral bone density in older U.S. adults from NHANES III. J Bone Miner Res 12:1761–1768; 1997. 12. Looker, A. C., Wahner, H. W., Dunn, W. L., Calvo, M. S., Harris, T. B., Heyse, S. P., Johnston, C. C., Jr., and Lindsay, R. L. Proximal femur bone mineral levels of US adults. Osteoporos Int 5:389 – 409; 1995. 13. Pocock, N. A., Eisman, J. A., Hopper, J. L., Yeates, M. G., Sambrook, P. N., and Eberl, S. Genetic determinants of bone mass: a twin study. J Clin Invest 80:706 –710; 1987. 14. Seeman, E., Hopper, J. L., Bach, L. A., Cooper, M. E., Parkinson, E., McKay, J., and Jerums, G. Reduced bone mass in daughters of women with osteoporosis. N Engl J Med 320:554 –558; 1989. 15. Seeman, E., Tsalamandris, C., Formica, C., Hopper, J. L., and McKay, J. Reduced femoral neck bone density in the daughters of women with hip fractures: the role of low peak bone density in the pathogenesis of osteoporosis. J Bone Miner Res 9:739 –743; 1994. 16. Slemenda, C. W., Christian, J. C., Williams, C. J., Norton, J. A., and Johnston, C. C., Jr. Genetic determinants of bone mass in adult women: A reevaluation of the twin model and the potential importance of gene interaction on heritability estimates. J Bone Miner Res 6:561–567; 1991. 17. Soroko, S. B., Barrett-Connor, E., Edelstein, S. L., and Kritz-Silverstein, D. Family history of osteoporosis and bone mineral density at the axial skeleton: The Rancho Bernardo Study. J Bone Miner Res 9:761–769; 1994. 18. Spector, T. D., Edwards, A. C., Thompson, P. W. Use of a risk factor and dietary calcium questionnaire in predicting bone density and subsequent bone loss at the menopause. Ann Rheum Dis. 51:1252–1253; 1992. 19. Tylavsky, F. A., Bortz, A. D., Hancock, R. L., and Anderson, J. J. B. Familial resemblance of radial bone mass between premenopausal mothers and their college-age daughters. Calcif Tissue Int 45:265–272; 1989. 20. WHO Study Group. Assessment of fracture risk and its applications to screening for postmenopausal osteoporosis. Geneva: World Health Organization; 1994.
References 1. Arden, N. K., Baker, J., Hogg, C., Baan, K., and Spector, T. D. The heritability of bone mineral density, ultrasound of the calcaneus and hip axis length: A study of postmenopausal twins. J Bone Miner Res 11:530 –534; 1996. 2. Bauer, D. C., Browner, W. S., Cauley, J. A., Orwoll, E. S., Scott, J. C., Black, D. M., Tao, J. L., and Cummings, S. R. Factors associated with appendicular bone mass in older women. Ann Intern Med. 118:657– 665; 1993.
Date Received: June 25, 1998 Date Revised: December 16, 1998 Date Accepted: December 16, 1998