Abstract

Weight loss in the elderly increases bone loss and the risk of fractures, especially at the hip and spine. The influence of weight change on non-weight-bearing parts of the skeleton is less well known. The purpose of this study was to investigate an association between weight change during the peri- and postmenopausal years and forearm bone mineral density (BMD). Among 8,856 women aged 45–60 years attending the first Health Study of Nord-Trøndelag, Norway (HUNT I, 1984–1986), a random sample of 2,795 women was invited to forearm densitometry (single x-ray absorptiometry technology) at HUNT II (1995–1997), after a mean period of 11.3 years. A total of 2,005 women (mean age: 65.1 years) were eligible. The mean weight had increased 3.4 kg; the gain was greater in the youngest women. A total of 382 women (19.1%) had lost and 1,331 women (66.3%) had gained weight. Weight change explained little of the BMD variance, 0.7% and 0.4% for weight loss and weight gain, respectively. Weight loss was an independent and statistically significant negative predictor of BMD, adjusted for body weight, age, age at menopause, smoking, and ovarian hormone treatment, particularly among women with a baseline body mass index greater than 25 kg/m2. No independent association between weight gain and forearm BMD was found.

Bone mineral density (BMD) is positively associated with body weight, and low body weight is a risk factor of fractures. Weight loss has been shown to increase axillary bone loss in the elderly (1). Weight loss also increases the risk of fractures, especially at the hip (24). The bone effects of intentional weight loss do not seem to differ from unintentional weight loss (5). Among important risk factors for forearm fractures are low BMD, a history of falling, and previous fractures (6, 7). Very few lifestyle factors have been linked to increased risk of wrist fractures; one exception is physical activity that is positively associated with Colles' fracture, possibly because of the greater risk of falling (8, 9).

The influence of weight change on the risk of forearm fractures in women is not clear. Self-reported weight gain between the ages of 25 and 50 years has been found to decrease the risk of forearm fractures in elderly women (5). In retrospective data, weight gain may even increase the forearm fracture risk (10). Other studies have found no association between weight change and wrist fractures (11).

As a non-weight-bearing part of the skeleton, the distal radius is less influenced by changing mechanical strain due to weight gain or loss. Body weight is, however, a strong and positive predictor of forearm BMD in postmenopausal women (12, 13). Whether there is an effect of weight change on BMD, independent of body weight at baseline, or of weight at follow-up is not known.

The distal forearm consists of both cortical and cancellous bone. At the ultradistal radius, cancellous bone makes up to 80 percent of the total (14). Cancellous bone has a higher metabolic activity than does cortical bone and is more influenced by factors such as menopause-related, reduced-estrogen availability. More cancellous than cortical bone mass is lost with aging (15, 16). Menopause is frequently associated with weight gain, although the physiologic causes are unclear. Weight gain is related partially to the decline in sex hormones, partially to normal aging, and partially to a more sedentary lifestyle (1719). Menopause may also be associated with changes in body composition and fat distribution (20).

The average body weight in general has increased over the last decades, especially in the United States, but also worldwide, with increasing weight-related morbidity (21, 22). Data from Norway also show the same pattern (23). On the other hand, weight loss is also associated with health hazards, especially in an elderly population (24).

The purpose of this study was to investigate whether there is an association between weight change and forearm BMD during the peri- and postmenopausal years in a population-based cohort of Norwegian women followed for more than 11 years. A more specific purpose was to explore whether the association between weight gain or weight loss and forearm BMD was independent of body weight, or if the BMD response to weight changes differed according to body size at baseline.

MATERIALS AND METHODS

The Health Study of Nord-Trøndelag, Norway (HUNT), is a multipurpose health survey focusing on the total population in the rural county of Nord-Trøndelag in mid-Norway (25). The first data collection (HUNT I) took place between 1984 and 1986; the second observation (HUNT II) was during 1995–1997. On both occasions, all county citizens over the age of 19 years were invited to participate, and the total potential numbers included 85,100 and 92,936, respectively. The overall participation rate was 88.1 percent in 1984–1986 and 71.2 percent in 1995–1997, with somewhat better rates for women than for men (26, 27).

With the exception of body weight and height that were measured at both baseline and follow-up, all other variables used for this analysis are from the follow-up study (HUNT II). At both evaluations, data collection was carried out by practically identical protocols. A comprehensive health questionnaire was sent by mail along with an invitation for a general health examination (27). The questionnaire form included questions about general health and lifestyle, and the participants were requested to bring it to the physical examination. A second, more detailed questionnaire concerning illnesses, symptoms, medical treatment, socioeconomic factors, lifestyle and, for women, reproductive factors was distributed after the examination, to be completed at home and returned by mail. The physical examinations at both HUNT I and HUNT II involved measurements of blood pressure, height, weight, and blood glucose. Height (in centimeters) and weight (in kilograms) were measured without shoes and with participants wearing only light clothing. Forearm densitometry was included as part of one of several substudies, but only in HUNT II.

Among all participants at HUNT I (baseline), a total of 8,856 women were aged 45–60 years, and their age-cohort participation rate was 95.3 percent. Forearm bone density was included at the physical examination in HUNT II for preplanned subsets of the female population: 35 percent of those aged 50–59 years, 5 percent of the age group 60–64 years, and 100 percent of the women aged 65 years or older. This involved 2,795 women from the baseline cohort, of whom 2,188 (78.3 percent) women attended and had forearm bone density measurements.

All women in this study were menopausal at their examination in HUNT II, and menopause was defined as no menstrual bleeding within the last 12 months prior to the examination. The lifestyle and reproductive variables used for this study are based on the questionnaires used in HUNT II, because no reproductive data were collected at baseline. All women who reported treatment for hyper- or hypothyroidism, which totaled 174, were excluded from the analyses. Additional exclusions included the eight women for whom weight and height data from one or both of the examinations were missing. Finally, three women had no BMD data registered. These exclusions left 2,005 women whose mean age was 65.1 (range: 55–71) years as eligible for the analyses. No data were available with which to determine whether or not weight change between baseline and follow-up was intentional.

Bone mineral density

BMD (in grams/centimeter squared, transformed to milligrams/centimeter squared for these analyses) was measured at the distal forearm and ultradistal radius of the nondominant arm by three different single energy x-ray bone densitometers (DTX 100; Osteometer Meditech A/S, Copenhagen, Denmark). Daily calibration of the densitometers was performed with equipment-specific phantoms. A cross-calibration study among the three devices was carried out and did not reveal significant differences. All sites for measurement were manually controlled and corrected. The densitometry procedure and quality assessments have been described in detail previously (13, 28).

Analysis

Weight change was calculated as the difference in kilograms between body weight at follow-up (HUNT II) and at baseline (HUNT I). Negative values therefore indicated weight loss, and positive values meant weight gain. The choice to study change in body weight instead of change in body weight for height (body mass index) or relative weight change (percentage) was based on the fact that BMD is more strongly related to absolute weight than to other anthropometric measures. This was confirmed in preliminary analyses and is in accordance with other studies (12). A change in weight of 1-kg gain or loss was defined as stable body weight. For most purposes, weight change was stratified in weight loss, stable weight, and weight gain. For some of the bivariate analyses, however, weight change was categorized into five groups: weight loss; stable weight; and weight gain of 1.1–4.0, 4.1–7.0, and >7.0 kg. These cutpoints were chosen because they provided weight change groups with adequate numbers for stratified analyses. Body mass index was calculated as body weight (kg)/height (m)2, measured at baseline and in follow-up. Women's age at follow-up was used in these analyses. The interval between baseline and follow-up was quantified in years to one decimal point. Other covariates from follow-up included smoking, history of cancer or diabetes, ovarian hormone treatment (estradiol with or without progesterone/progestin), and age at menopause. A total of 534 women (26.6 percent) did not answer the ovarian hormone treatment question. These women did not differ significantly in BMD or anthropometric data from those who reported no use of ovarian hormone treatment, but they were older than the women reporting ever ovarian hormone treatment use. It was assumed that nonresponse was because the question was irrelevant for these women; thus, for the multivariate analyses, the nonresponders were categorized as never users. The other ovarian hormone treatment categories were current and prior use. Smoking was calculated as pack-years, that is, (daily cigarettes × years of smoking)/20. The history of smoking was categorized as current, prior, and never smokers as reported at follow-up.

Bivariate associations for continuous data were analyzed by independent and paired sample t tests. Odds ratios for dichotomous variables were calculated by logistic regression. In order to determine the separate associations of weight gain and weight loss with BMD, we used two multiple linear regression models, one for weight loss and the other for weight gain. The models included final weight, age, ovarian hormone treatment, pack-years of smoking, and age at menopause; all are variables with a linear association with BMD. The interval between baseline and follow-up was not a statistically significant predictor and was omitted from the models. Similar analyses were also performed, adjusting for baseline instead of final body weight and for medical conditions that could induce weight change, such as diabetes and cancer.

The analyses documented an interaction between body weight and weight change. In order to assess whether the effect of weight loss or gain differed according to body mass index at baseline, we carried out multiple linear regression in stratified models of body mass index less than or equal to 25.0 kg/m2 and body mass index greater than 25.0 kg/m2. This cutoff represented the median baseline body mass index. All models were tested for collinearity and heteroscedasticity. The final multiple regression analyses were performed without residual outliers (±3 standard deviations), at the maximum 21 cases (1 percent). The results are presented as regression coefficients (change in BMD (milligrams/centimeter squared) per unit change of the explanatory variable). All statistical tests were two sided, and the analyses were carried out with SPSS, version 13.0, software (29).

Ethics

The HUNT study was approved by the Regional Committee for Medical Research Ethics and by the Norwegian Data Inspectorate.

RESULTS

Selected characteristics of the prospective cohort of the 2,005 women who had forearm BMD measurements at follow-up are shown in table 1. The mean body weight in this population of midlife women had increased by 3.4 kg over the 11.3 years of follow-up, and 66.3 percent of the women had gained weight. A total of 382 women (19.1 percent) had lost more than 1 kg of body weight. These women were older at the BMD examination than were the women with stable or increased weight; the mean age was 67.1 years (95 percent confidence interval: 66.6, 67.5) for those who had lost weight and 65.5 years (95 percent confidence interval: 65.2, 65.7) for those with stable and increased weight (pooled together). However, the women who had lost weight had the highest mean body weight at baseline. Their mean weight loss was 5.2 kg that corresponded to an average relative loss of 7.3 percent of body weight. The mean forearm BMD was statistically significantly lower in women with a history of weight loss or stable weight than the BMD for women with weight gain (table 1). No difference in age at menopause between the weight-change categories was found with the differences in time since menopause being explained by the differences in age between these groups. Women with weight gain reported statistically significantly more ovarian hormone therapy use than did women with stable weight (p = 0.003). However, compared with women with weight loss, women with weight gain were not statistically significantly more likely to use ovarian hormone treatment (p = 0.07). Current smoking was more frequent in the groups of women with either weight loss or stable weight than in the women who had weight gain (table 1). In a logistic regression model adjusting for age, the odds ratio of weight loss among current smokers was 1.95 (95 percent confidence interval: 1.5, 2.5) compared with never smokers. A diagnosis of diabetes was more frequent in the women who had lost weight. No association was present between weight change and a history of cancer (table 1).

TABLE 1.

Characteristics of 2,005 women by category of body weight change between baseline at HUNT* I (1984–1986) and follow-up at HUNT II (1995–1997), an average period of 11.3 years


 

Total (n = 2,005)
 

Category of weight change, from baseline to follow-up (11.3 years)
 
  
p value
 

 
 Weight loss§ (n = 382)
 
Stable weight§ (n = 292)
 
Weight gain§ (n = 1,331)
 
 
Continuous data, mean (SD*     
Distal BMD* (mg/cm2385.7 (68.9) 373.9 (68.7) 373.7 (67.0) 391.7 (68.6) 0.001 
Ultradistal BMD (mg/cm2287.9 (62.5) 277.4 (63.5) 277.0 (61.8) 293.4 (61.8) 0.001 
Age at baseline (years) 54.5 (4.9) 55.6 (4.2) 55.4 (4.6) 53.9 (5.1) 0.001 
Age at follow-up (years) 65.8 (5.1) 67.1 (4.5) 66.5 (4.7) 65.2 (5.3) 0.001 
Duration of baseline follow-up (years) 11.3 (1.3) 11.5 (1.3) 11.2 (1.3) 11.3 (1.2) 0.003 
Weight at baseline (kg) 68.0 (11.1) 70.4 (13.0) 66.7 (10.9) 67.6 (10.4) 0.001 
Weight at follow-up (kg) 71.4 (12.2) 65.2 (12.0) 66.8 (11.0) 74.2 (11.5) 0.001 
Weight change (kg) 3.4 (6.2) −5.2 (4.1) 0.1 (0.7) 6.6 (4.4) 0.001 
BMI* at baseline (kg/m225.7 (4.0) 26.6 (4.8) 25.3 (4.0) 25.5 (3.7) 0.001 
BMI at follow-up (kg/m227.4 (4.4) 25.2 (4.5) 25.8 (4.1) 28.3 (4.1) 0.001 
BMI change (kg/m21.7 (2.3) −1.4 (1.6) 0.5 (0.6) 2.9 (1.7) 0.001 
Age at menopause (years) 49.3 (4.1) 49.1 (3.8) 49.4 (3.8) 49.3 (4.2) NS* 
Years since menopause (no.) 16.5 (6.5) 17.9 (5.9) 17.2 (5.6) 15.9 (6.8) 0.001 
Smoking (pack-years) 9.8 (10.9) 11.3 (11.5) 9.7 (9.8) 9.4 (11.0) 0.04 
Categorical data      
Current smokers (%) (n = 453) 22.7 33.2 27.7 18.5 0.001 
OHT* (current or prior)# (%) (n = 292) 19.9 16.7 12.8 22.2 NS 
Diabetes mellitus (%) (n = 92) 4.6 10.5 3.4 3.2 0.001 
History of cancer (%) (n = 146)
 
7.3
 
7.6
 
6.8
 
7.3
 
NS
 

 

Total (n = 2,005)
 

Category of weight change, from baseline to follow-up (11.3 years)
 
  
p value
 

 
 Weight loss§ (n = 382)
 
Stable weight§ (n = 292)
 
Weight gain§ (n = 1,331)
 
 
Continuous data, mean (SD*     
Distal BMD* (mg/cm2385.7 (68.9) 373.9 (68.7) 373.7 (67.0) 391.7 (68.6) 0.001 
Ultradistal BMD (mg/cm2287.9 (62.5) 277.4 (63.5) 277.0 (61.8) 293.4 (61.8) 0.001 
Age at baseline (years) 54.5 (4.9) 55.6 (4.2) 55.4 (4.6) 53.9 (5.1) 0.001 
Age at follow-up (years) 65.8 (5.1) 67.1 (4.5) 66.5 (4.7) 65.2 (5.3) 0.001 
Duration of baseline follow-up (years) 11.3 (1.3) 11.5 (1.3) 11.2 (1.3) 11.3 (1.2) 0.003 
Weight at baseline (kg) 68.0 (11.1) 70.4 (13.0) 66.7 (10.9) 67.6 (10.4) 0.001 
Weight at follow-up (kg) 71.4 (12.2) 65.2 (12.0) 66.8 (11.0) 74.2 (11.5) 0.001 
Weight change (kg) 3.4 (6.2) −5.2 (4.1) 0.1 (0.7) 6.6 (4.4) 0.001 
BMI* at baseline (kg/m225.7 (4.0) 26.6 (4.8) 25.3 (4.0) 25.5 (3.7) 0.001 
BMI at follow-up (kg/m227.4 (4.4) 25.2 (4.5) 25.8 (4.1) 28.3 (4.1) 0.001 
BMI change (kg/m21.7 (2.3) −1.4 (1.6) 0.5 (0.6) 2.9 (1.7) 0.001 
Age at menopause (years) 49.3 (4.1) 49.1 (3.8) 49.4 (3.8) 49.3 (4.2) NS* 
Years since menopause (no.) 16.5 (6.5) 17.9 (5.9) 17.2 (5.6) 15.9 (6.8) 0.001 
Smoking (pack-years) 9.8 (10.9) 11.3 (11.5) 9.7 (9.8) 9.4 (11.0) 0.04 
Categorical data      
Current smokers (%) (n = 453) 22.7 33.2 27.7 18.5 0.001 
OHT* (current or prior)# (%) (n = 292) 19.9 16.7 12.8 22.2 NS 
Diabetes mellitus (%) (n = 92) 4.6 10.5 3.4 3.2 0.001 
History of cancer (%) (n = 146)
 
7.3
 
7.6
 
6.8
 
7.3
 
NS
 
*

HUNT, the Health Study of Nord-Trøndelag, Norway; SD, standard deviation; BMD, bone mineral density; BMI, body mass index; NS, not statistically significant (p > 0.05); OHT, ovarian hormone treatment.

Unless noted, all data were measured at HUNT II (follow-up). Bone mineral density was measured at the forearm.

Two samples of the t test or chi-square test of differences between the categories of weight loss and weight gain.

§

Categories of weight change: weight loss, >1-kg loss; stable weight, ±1 kg; weight gain, >1-kg increase.

Smoking in pack-years among 946 current or previous smokers at follow-up.

#

History of ovarian hormone treatment among 1,471 women who answered the question.

Body weight at baseline and follow-up in two strata of baseline body mass index (≤25.0 and >25.0 kg/m2) and by categories of weight change is presented in table 2. The women with baseline body mass index greater than 25.0 kg/m2 had lost 2.1 kg more weight (95 percent confidence interval: 1.4, 2.8) than did the women with body mass index less than or equal to 25.0 kg/m2 at baseline. This explained the difference in total weight change between the two strata of baseline body mass index. Figure 1 shows the mean BMD at the distal forearm according to five categories of weight change, stratified by baseline body mass index (≤25.0 and >25.0 kg/m2). Women with baseline body mass index greater than 25.0 kg/m2 had on average higher BMD than did the women with baseline body mass index less than or equal to 25.0 kg/m2. Weight gain, however, seemed to be positively associated with BMD for both body mass index strata before controlling for other covariates. The most noticeable BMD difference between the two body mass index strata was found in women who had lost weight. Women in the lowest body mass index stratum had almost 10 percent lower BMD than did women with baseline body mass index greater than 25.0 kg/m2. It should, however, be noted that the difference in body weight at the time of densitometry was 17.2 kg between these two groups (table 2). The women with weight loss in the lower baseline body mass index group had statistically significantly lower BMD than did women with stable weight or weight gain within the same body mass index stratum. The same pattern was found for BMD measured at the ultradistal radius.

TABLE 2.

Body weight at baseline (HUNT* I, 1984–1986) and follow-up (HUNT II, 1995–1997) and the magnitude of weight change with 95% confidence intervals in the categories of weight loss, stable weight, and weight gain between baseline and follow-up, an average period of 11.3 years, and by two strata of body mass index at baseline


Weight change
 

Baseline BMI* ≤25 kg/m2
 
    
Baseline BMI >25 kg/m2
 
    
 No.
 
Body weight at baseline and follow-up and weight change (kg)
 
   No.
 
Body weight at baseline and follow-up and weight change (kg)
 
   

 
 Baseline
 
Follow-up
 
Change
 
95% confidence interval
 
 Baseline
 
Follow-up
 
Change
 
95% confidence interval
 
Weight loss 162 59.3 55.2 −4.0 −4.4, −3.6 220 78.6 72.5 −6.1 −6.8, −5.5 
Stable weight 154 59.1 59.2 0.1 0.0, 0.2 138 75.1 75.2 0.1 0.0, 0.2 
Weight gain 672 60.5 67.1 6.5 6.2, 6.8 659 74.8 81.6 6.7 6.4, 7.1 
Total
 
988
 
60.1
 
63.9
 
3.8
 
3.4, 4.1
 
1,017
 
75.7
 
78.7
 
3.0
 
2.6, 3.5
 

Weight change
 

Baseline BMI* ≤25 kg/m2
 
    
Baseline BMI >25 kg/m2
 
    
 No.
 
Body weight at baseline and follow-up and weight change (kg)
 
   No.
 
Body weight at baseline and follow-up and weight change (kg)
 
   

 
 Baseline
 
Follow-up
 
Change
 
95% confidence interval
 
 Baseline
 
Follow-up
 
Change
 
95% confidence interval
 
Weight loss 162 59.3 55.2 −4.0 −4.4, −3.6 220 78.6 72.5 −6.1 −6.8, −5.5 
Stable weight 154 59.1 59.2 0.1 0.0, 0.2 138 75.1 75.2 0.1 0.0, 0.2 
Weight gain 672 60.5 67.1 6.5 6.2, 6.8 659 74.8 81.6 6.7 6.4, 7.1 
Total
 
988
 
60.1
 
63.9
 
3.8
 
3.4, 4.1
 
1,017
 
75.7
 
78.7
 
3.0
 
2.6, 3.5
 
*

HUNT, the Health Study of Nord-Trøndelag, Norway; BMI, body mass index.

Categories of weight change: weight loss, >1-kg loss; stable weight, ±1 kg; weight gain, >1-kg increase.

FIGURE 1.

The curves show the mean distal-forearm bone mineral density by categories of weight change during the 11.3 years between baseline at HUNT I (1984–1986) and follow-up at HUNT II (1995–1997). (Vertical bars denote 95% confidence intervals). These data are stratified according to body mass index (BMI) at baseline. The solid curve represents the mean bone mineral density in 1,013 women with a baseline body mass index greater than 25 kg/m2, and the dashed line shows the mean bone mineral density in 992 women with a baseline body mass index less than or equal to 25 kg/m2. The light, dotted bars show the mean weight at baseline, and the dark, diagonally striped bars show the mean weight at follow-up in the categories of weight change, with 95% confidence intervals (vertical bars). HUNT, the Health Study of Nord-Trøndelag, Norway.

FIGURE 1.

The curves show the mean distal-forearm bone mineral density by categories of weight change during the 11.3 years between baseline at HUNT I (1984–1986) and follow-up at HUNT II (1995–1997). (Vertical bars denote 95% confidence intervals). These data are stratified according to body mass index (BMI) at baseline. The solid curve represents the mean bone mineral density in 1,013 women with a baseline body mass index greater than 25 kg/m2, and the dashed line shows the mean bone mineral density in 992 women with a baseline body mass index less than or equal to 25 kg/m2. The light, dotted bars show the mean weight at baseline, and the dark, diagonally striped bars show the mean weight at follow-up in the categories of weight change, with 95% confidence intervals (vertical bars). HUNT, the Health Study of Nord-Trøndelag, Norway.

In the multiple linear regression models shown in table 3 and table 4, there was an inverse association between BMD and both weight loss and weight gain, when adjustment was made for body weight at follow-up along with other covariates. Weight change explained less than 1 percent of the total BMD variance with contributions of 0.7 and 0.4 percent for weight loss and gain, respectively. When performing the same analysis but adjusting for baseline instead of follow-up body weight, we found that weight gain was positively associated with BMD, but the association was not statistically significant. For weight loss, the association remained inverse and statistically significant. This indicates that weight loss has a negative effect on forearm BMD, whereas the effect of weight gain is explained by the body weight at baseline.

TABLE 3.

Multiple linear regression of bone mineral density measured at the distal forearm and ultradistal radius in a model of 382 women with weight loss greater than 1 kg during the 11.3 years between baseline (HUNT* I, 1984–1986) and follow-up (HUNT II, 1995–1997)


Predictors
 

Distal forearm, R2 = 26.0%
 
 
Ultradistal radius, R2 = 22.3%
 
 
 BMD* (mg/cm2)
 
p value
 
BMD (mg/cm2)
 
p value
 
Constant 455.1 <0.001 368.9 <0.001 
Weight loss (kg) −1.6 <0.033 −1.5 <0.042 
Body weight at BMD (kg) 1.8 <0.001 1.6 <0.001 
Age at BMD (years) −4.7 <0.001 −4.4 <0.001 
Age at menopause (years) 2.3 <0.003 2.2 <0.003 
Smoking (pack-years) −0.6 <0.067  NS* 
OHT* (never/prior/current)
 
13.2
 
<0.039
 

 
NS
 

Predictors
 

Distal forearm, R2 = 26.0%
 
 
Ultradistal radius, R2 = 22.3%
 
 
 BMD* (mg/cm2)
 
p value
 
BMD (mg/cm2)
 
p value
 
Constant 455.1 <0.001 368.9 <0.001 
Weight loss (kg) −1.6 <0.033 −1.5 <0.042 
Body weight at BMD (kg) 1.8 <0.001 1.6 <0.001 
Age at BMD (years) −4.7 <0.001 −4.4 <0.001 
Age at menopause (years) 2.3 <0.003 2.2 <0.003 
Smoking (pack-years) −0.6 <0.067  NS* 
OHT* (never/prior/current)
 
13.2
 
<0.039
 

 
NS
 
*

HUNT, the Health Study of Nord-Trøndelag, Norway; BMD, bone mineral density; NS, not statistically significant; OHT, ovarian hormone therapy.

The complete model is presented.

TABLE 4.

Multiple linear regression of bone mineral density measured at the distal forearm and ultradistal radius in a model of 1,331 women with weight gain greater than 1 kg during 11.3 years between baseline (HUNT* I, 1984–1986) and follow-up (HUNT II, 1995–1997)


Predictors
 

Distal forearm, R2 = 21.0%
 
 
Ultradistal radius, R2 = 20.0%
 
 
 BMD* (mg/cm2)
 
p value
 
BMD (mg/cm2)
 
p value
 
Constant 542.6 <0.001 411.6 <0.001 
Weight gain (kg) −1.4 <0.001 −1.0 <0.010 
Body weight at BMD (kg) 1.5 <0.001 1.3 <0.001 
Age at BMD (years) −4.9 <0.001 −4.2 <0.001 
Age at menopause (years) 1.3 <0.001 1.2 <0.001 
Smoking (pack-years) −0.6 <0.001 −0.6 <0.001 
OHT* (never/prior/current)
 
5.9
 
<0.030
 
7.3
 
<0.030
 

Predictors
 

Distal forearm, R2 = 21.0%
 
 
Ultradistal radius, R2 = 20.0%
 
 
 BMD* (mg/cm2)
 
p value
 
BMD (mg/cm2)
 
p value
 
Constant 542.6 <0.001 411.6 <0.001 
Weight gain (kg) −1.4 <0.001 −1.0 <0.010 
Body weight at BMD (kg) 1.5 <0.001 1.3 <0.001 
Age at BMD (years) −4.9 <0.001 −4.2 <0.001 
Age at menopause (years) 1.3 <0.001 1.2 <0.001 
Smoking (pack-years) −0.6 <0.001 −0.6 <0.001 
OHT* (never/prior/current)
 
5.9
 
<0.030
 
7.3
 
<0.030
 
*

HUNT, the Health Study of Nord-Trøndelag, Norway; BMD, bone mineral density; OHT, ovarian hormone therapy.

The complete model is presented.

Because of the interaction between baseline weight and weight change (p = 0.04), multiple linear regression was performed by use of models stratified for baseline body mass index (≤25.0 kg/m2 vs. >25.0 kg/m2). An independent and inverse association between weight loss and BMD was found in women with baseline body mass index greater than 25.0 (regression coefficient: −1.84, p = 0.04). This association was not present for women with baseline body mass index less than or equal to 25.0 kg/m2. However, in the latter group of women, there was an inverse and statistically significant association between weight gain and BMD, but only when adjustment was made for body weight at follow-up. This observation is thus not independent of baseline weight. These results indicate that the negative effect of weight loss on forearm BMD can be demonstrated only in women who were overweight (body mass index >25 kg/m2) at baseline.

A history of cancer was not a confounder in the multiple linear regression models, whereas a diagnosis of diabetes was a statistically significant positive predictor of BMD in women who had lost weight (data not shown). When eliminating the 92 women reporting diabetes from the multivariate analyses, we found that the inverse association between weight loss and BMD became stronger.

DISCUSSION

This population-based study shows that women's weight loss through the peri- and postmenopausal years adversely affects forearm BMD. This was particularly the case for women with a relatively high body mass index at baseline. The apparent inverse association between weight gain and BMD was explained by a lower weight at baseline. In women having lost weight, however, the inverse association between weight change and BMD was independent of body weight whether at baseline or at follow-up. Predicted BMD at a given body weight will thus be lower in women with a history of substantial weight loss than in women with a smaller loss, despite a higher mean weight in the former group. This indicates that weight loss is negative for BMD, even in non-weight-bearing bones.

The strengths of this study are its population-based design and high participation rate. The women invited to densitometry, although a subset of the age-stratified population, were selected at random. They did not differ concerning anthropometric measures from other women of similar age who participated in HUNT. A study of nonresponders in HUNT II did not reveal any important selection biases except for a slightly higher prevalence of current smokers in those who did not attend (26, 27). In this study, smokers had lower body weight than did nonsmokers, and smoking was more prevalent among women with weight loss. However, smoking was included as a covariate in the multivariate analyses, and thus, we do not think that selection represented a major bias to the final result. This study involved only middle-aged and elderly women, and therefore the results cannot be generalized to the entire adult population.

We have no information about the causes for weight loss, whether it was deliberate or due to medical conditions or treatment. The women with weight loss greater than 1 kg between baseline and follow-up had the highest mean weight at baseline, and women with a baseline body mass index greater than 25.0 kg/m2 had lost more than did those with a body mass index less than or equal to 25.0 kg/m2. This may indicate that weight loss was intentional for the majority of women losing weight, although the effect of regression toward the mean should not be ignored. A recent study rejected the hypothesis that intentional weight loss is less detrimental than unintentional weight loss, at least in elderly women (5). The possibility remains, however, that during deliberate weight loss the women will increase physical activity and be more conscientious about nutrient intake than during unintentional weight loss and that these behavior changes could minimize the negative effect of weight loss on the skeleton. This was not explored in our study.

Some medical conditions could explain weight change, and we excluded women reporting treatment for thyroid diseases from all analyses. The women with diabetes in this age span mainly have diabetes type 2 (30), and these women had slightly higher BMD as has been shown in other cross-sectional studies (31). However, more rapid bone loss in diabetic women has recently been reported (32). Weight loss was more frequent in women with diabetes than in those without it. Elimination of the women reporting diabetes from the analyses increased the inverse effect of weight loss on BMD. Hence, the possibility remains that subclinical illness could explain both weight loss and low BMD, as has been discussed by Ensrud et al. (5). These authors also observed that hip fracture risk was unchanged over time when following a cohort of elderly women with a history of both intentional and unintentional weight loss (5). One would have expected the association between weight loss and fractures to diminish with time if subclinical pathology had provided the explanation for the weight loss.

The interval between the two weight measurements in this study was rather large. With only these two points in time, we are not able to assess the influence of rapid weight loss compared with slower loss or the effects of weight cycling that seems to be associated with reduced BMD at the radius in obese premenopausal women (33).

Weight loss has been shown to induce a higher rate of bone loss in premenopausal women (34, 35) and in the elderly (1, 36). In this study, BMD was not measured at baseline, and the influence of weight change on bone loss could not be assessed. Body weight is a strong positive predictor of BMD and of peak bone mass (37). Further, women with low body mass index experience more rapid bone loss (36, 38). The results of this study indirectly suggest that weight loss increases bone loss. The rather unexpected result that the independent negative effect of weight loss on forearm BMD could be demonstrated only among women with relatively high body size at baseline may be explained by somewhat higher relative weight loss than in women of smaller size. It is also possible that the influence of weight loss on bone is more discernible in women with an increased BMD related to their higher body weight. These women, however, do still benefit from a relatively high body weight that explains that their forearm BMD values were significantly higher than for the more slender women with a history of weight loss.

When considering these results, one should take into account that the impact of weight loss on BMD was rather small. For a predicted difference of 1 standard deviation in forearm BMD in women with baseline body mass index greater than 25 kg/m2, the difference in weight loss must be 33 kg, given the same body weight at follow-up. This would, however, increase the forearm fracture risk by 50 percent (7). Bauer et al. (12) found that each 10-kg weight loss was associated with 3.9 percent lower bone mass (12). This is comparable with our results that showed that a 10-kg loss was associated with 4.1 percent lower BMD.

Mechanisms are unclear for the adverse effect of weight loss on the skeleton. Reduced mechanical strain due to reduced body weight is one hypothesis. This should, however, not count as much for the forearm as for the weight-bearing parts of the skeleton. Reduced conversion of circulating androgens to estrogen due to less adipose tissue is another explanation. Serum leptin could represent a plausible mediator between body fat and bone (39). There are, however, conflicting results about whether this mechanism relates to the mature skeleton and the distal radius (40, 41). Alternatively, weight loss could induce changes in insulin-like growth factors and sex-hormone binding globulin that would alter the bioavailability of estrogen (42, 43).

About two thirds of the women in this cohort had gained weight, with the younger women gaining more than the older women. Although we found a positive crude association between weight gain and forearm BMD, there was no effect of weight gain independent of body weight. Trovas et al. (44) found a positive correlation between weight gain and lumbar spine BMD; however, these authors did not adjust for body weight or body mass index as either final weight or at baseline. When assessing the effect of weight change on BMD, we found it important to adjust for body weight. The positive crude association between weight gain and BMD found in this study was explained mainly by age, as the younger women, with more weight gain, had higher BMD than did older women. The negative association of weight gain with BMD that appeared when adjustment was made for current weight may indicate that BMD is associated with weight history throughout adult life. It is possible that the women with a history of midlife weight gain follow a different profile of bone change than do women whose weight remains stable in midlife. Their baseline BMD might have been lower, but bone loss was reduced because of weight gain, as has been found in other studies (1, 36). This was seen particularly in women whose baseline body mass index was less than or equal to 25 kg/m2.

In summary, weight gain during the peri- and postmenopausal years was common in this cohort of 2,005 middle-aged women followed over a period of more than 11 years. Weight loss had a small, but inverse and independent effect on forearm bone mineral density. This was found particularly in women with a body mass index greater than 25 kg/m2 at baseline. This study did not document an independent effect of weight gain on forearm BMD; the observed inverse association was explained by their lower body weight at baseline.

The study was supported by grants from the Norwegian Women's Public Health Association, the Norwegian Research Council, the Norwegian Osteoporosis Foundation, and the Association of Health and Rehabilitation.

The authors are grateful to W. Dana Flanders and Jerilynn C. Prior who assisted with revisions of the manuscript.

The Health Study of Nord-Trøndelag is a collaboration among the HUNT Research Centre, the Norwegian University of Science and Technology, the Norwegian Institute of Public Health, and the Nord-Trøndelag County Council.

Conflict of interest: none declared.

References

1.
Hannan MT, Felson DT, Dawson-Hughes B, et al. Risk factors for longitudinal bone loss in elderly men and women: the Framingham Osteoporosis Study.
J Bone Miner Res
 
2000
;
15
:
710
–20.
2.
Fahramand BY, Michaëlsson K, Baron JA, et al. Body size and hip fracture risk.
Epidemiology
 
2000
;
11
:
214
–19.
3.
Meyer HE, Tverdal A, Falch JA. Risk factors for hip fracture in middle-aged Norwegian women and men.
Am J Epidemiol
 
1993
;
137
:
1203
–11.
4.
Ensrud KE, Cauley J, Lipschutz R, et al. Weight change and fractures in older women. Study of Osteoporotic Fractures Research Group.
Arch Intern Med
 
1997
;
157
:
857
–63.
5.
Ensrud KE, Ewing SK, Stone KL, et al. Intentional and unintentional weight loss increase bone loss and hip fracture risk in older women.
J Am Geriatr Soc
 
2003
;
51
:
1740
–7.
6.
Vogt MT, Cauley JA, Tomaino MM, et al. Distal radius fractures in older women: a 10 year follow-up study of descriptive characteristics and risk factors. The Study of Osteoporotic Fractures.
J Am Geriatr Soc
 
2002
;
50
:
97
–103.
7.
Nguyen TV, Center JR, Sambrook PN, et al. Risk factors for proximal humerus, forearm, and wrist fractures in elderly men and women.
Am J Epidemiol
 
2001
;
153
:
587
–95.
8.
Silman AJ. Risk factors for Colles' fracture in men and women: results from the European Prospective Osteoporosis Study.
Osteoporos Int
 
2003
;
14
:
213
–18.
9.
Keegan THM, Kelsey JL, King AC, et al. Characteristics of fallers who fracture at the foot, distal forearm, proximal humerus, and shaft of the tibia/fibula compared with fallers who do not fracture.
Am J Epidemiol
 
2004
;
159
:
192
–203.
10.
Gunnes M, Lehmann AH, Mellström D, et al. The relationship between anthropometric measurements and fractures in women.
Bone
 
1996
;
19
:
407
–13.
11.
Margolis KL, Ensrud KE, Schreiner PJ, et al. Body size and the risk for clinical fractures in older women.
Ann Intern Med
 
2000
;
133
:
123
–7.
12.
Bauer DC, Browner WS, Cauley JA, et al. Factors associated with appendicular bone mass in older women.
Ann Intern Med
 
1993
;
118
:
657
–65.
13.
Forsmo S, Langhammer A, Forsen L, et al. Forearm bone mineral density in an unselected population—the HUNT study, Norway.
Osteoporos Int
 
2005
;
16
:
562
–7.
14.
Augat P, Fuerst T, Genant HK. Quantitative bone mineral assessment at the forearm: a review.
Osteoporos Int
 
1998
;
8
:
299
–310.
15.
Prior J, Vigna YM, Barr SI, et al. Ovulatory premenopausal women lose cancellous spinal bone: a five year prospective study.
Bone
 
1996
;
18
:
261
–7.
16.
Mundy GR. Bone remodelling and mechanisms of bone loss in osteoporosis. In: Meunier PJ, ed. Osteoporosis, diagnosis and management. London, United Kingdom: Martin Dunitz,
1998
:17–35.
17.
Macdonald HM, New SA, Campbell MK, et al. Longitudinal changes in perimenopausal and early postmenopausal women: effects of dietary energy intake, energy expenditure, dietary calcium intake and hormone replacement therapy.
Int J Obes Relat Metab Disord
 
2003
;
27
:
669
–76.
18.
Dubnov G, Brzezinski A, Berry EM. Weight control and the management of obesity after menopause: the role of physical activity.
Maturitas
 
2003
;
44
:
89
–101.
19.
Sternfeldt B, Wang H, Quesenberry CP Jr, et al. Physical activity and changes in weight and waist circumference in midlife women: findings from the Study of Women's Health across the Nation.
Am J Epidemiol
 
2004
;
160
:
912
–22.
20.
Jensen LB, Vestergaard P, Hermann AP, et al. Hormone replacement therapy dissociates fat mass and bone mass, and tends to reduce weight gain in early postmenopausal women: a randomized controlled 5-year clinical trial of the Danish Osteoporosis Prevention Study.
J Bone Miner Res
 
2003
;
18
:
333
–42.
21.
Mokdad AH, Serdula MK, Dietz WH, et al. The spread of the obesity epidemic in the United States, 1991–98.
JAMA
 
1999
;
282
:
1519
–22.
22.
Kopelman PG. Obesity as a medical problem.
Nature
 
2000
;
404
:
635
–43.
23.
Drøyvold WB, Holmen J, Krüger Ø, et al. Leisure time physical activity and change in body mass index: an 11-year follow-up study of 9357 normal weight healthy women 20–49 years old.
J Womens Health
 
2004
;
13
:
55
–62.
24.
Wedick NM, Barrett-Connor E, Knoke JD, et al. The relationship between weight loss and all-cause mortality in older men and women with and without diabetes mellitus: the Rancho Bernardo Study.
J Am Geriatr Soc
 
2002
;
50
:
1810
–15.
25.
The Nord-Trøndelag Health Study. Verdal, Norway: Norwegian University of Science and Technology,
2005
. (http://www.hunt.ntnu.no/index.php?side=english).
26.
Langhammer A, Johnsen R, Holmen J, et al. Cigarette smoking gives more respiratory symptoms among women than among men. The Nord-Trøndelag Health Study (HUNT).
J Epidemiol Community Health
 
2000
;
54
:
917
–22.
27.
Holmen J, Midthjell K, Krüger Ø, et al. The Nord-Trøndelag Health Study 1995–97 (HUNT 2): objectives, contents, methods and participation.
Norsk Epidemiologi
 
2003
;
13
:
19
–32. (http://www.medisin.ntnu.no/ism/nofe/norepid/2003(1)%2005-Holmen.pdf).
28.
Forsmo S, Schei B, Langhammer A, et al. How do reproductive and lifestyle factors influence bone density in distal and ultradistal radius of early postmenopausal women? The Nord-Trøndelag Health Survey, Norway.
Osteoporosis Int
 
2001
;
12
:
222
–9.
29.
SPSS for Windows, version 13.0. Chicago, IL: SPSS, Inc,
2005
. (http://www.spss.com/spss/).
30.
Midthjell K, Krüger Ø, Holmen J, et al. Rapid changes in the prevalence of obesity and known diabetes in an adult Norwegian population. The Nord-Trøndelag Health Surveys 1984–86 and 1995–97.
Diabetes Care
 
1999
;
22
:
1813
–20.
31.
Strotmeyer ES, Cauley JA, Schwartz AV, et al. Diabetes is associated independently of body composition with bone mineral density and bone volume in older white and black men and women. The Health, Aging and Body Composition Study.
J Bone Miner Res
 
2004
;
19
:
1084
–91.
32.
Schwartz AV, Sellmeyer DE, Strotmeyer ES, et al. Diabetes and bone loss at the hip in older black and white adults.
J Bone Miner Res
 
2005
;
20
:
596
–603.
33.
Fogelholm M, Sievänen H, Heinonen A, et al. Association between weight cycling history and bone mineral density in premenopausal women.
Osteoporos Int
 
1997
;
7
:
354
–8.
34.
Salamone LM, Cauley JA, Black DM, et al. Effect of a lifestyle intervention on bone mineral density in premenopausal women: a randomized trial.
Am J Clin Nutr
 
1999
;
70
:
97
–103.
35.
Hui SL, Perkins AJ, Zhou L, et al. Bone loss at the femoral neck in premenopausal white women: effects of weight change and sex-hormone levels.
J Clin Endocrinol Metab
 
2002
;
87
:
1539
–43.
36.
Nguyen TV, Sambrook PN, Eisman JA. Bone loss, physical activity and weight change in elderly women: the Dubbo Osteoporosis Epidemiology Study.
J Bone Miner Res
 
1998
;
13
:
1458
–67.
37.
Hawker GA, Forsmo S, Cadarette SM, et al. Correlates of forearm bone mineral density in young Norwegian women: the Nord-Trøndelag Health Study.
Am J Epidemiol
 
2002
;
156
:
418
–27.
38.
Ravn P, Cizza G, Bjarnason NH, et al. Low body mass index is an important risk factor for low bone mass and increased bone loss in early postmenopausal women.
J Bone Miner Res
 
1999
;
14
:
1622
–7.
39.
Westerterp KR. Weight loss and bone mineral content. (Editorial). Obes Res
2002
;10:559.
40.
Rauch F, Blum WF, Klein K, et al. Does leptin have an effect on bone in adult women?
Calcif Tissue Int
 
1998
;
63
:
456
–8.
41.
Pasco JA, Henry MJ, Kotowicz MA, et al. Serum leptin levels are associated with bone mass in nonobese women.
J Clin Endocrinol Metab
 
2001
;
86
:
1884
–7.
42.
Keen RW. Effects of lifestyle interventions on bone health.
Lancet
 
1999
;
354
:
1923
–4.
43.
Pfeilschifter J, Scheidt-Nave C, Leidig-Bruckner G, et al. Relationship between circulating insulin-like growth factor components and sex hormones in a population-based sample of 50- to 80-year-old men and women.
J Clin Endocrinol Metab
 
1996
;
81
:
2534
–40.
44.
Trovas G, Lyritis GP, Galanos A, et al. Influence of weight gain on spine mineral density in postmenopausal women.
Calcif Tissue Int
 
1999
;
64
:
380
–3.