Bone mineral density (BMD) remains the most important index for diagnosing, managing, and studying osteoporosis/osteopenia [1-3] despite the emerging concept of “bone quality”. For example, the World Health Organization continues to diagnose osteoporosis according to BMD. The associations between anthropometric measurements and BMD have been extensively investigated both cross-sectionally and longitudinally although nutritional status has also been recognized as clinically relevant. However, which anthropometric index associated with BMD according to sex and age has not been identified.

Analyses revealed that low body weight and body mass index (BMI) were associated with lower BMD in older men.[4,5] Cross-sectional investigations [6-9] suggested positive correlations between height and BMD, with greater height associated with higher BMD. However, unlike cross-sectional findings, a longitudinal study showed that low body weight and/or low BMI were not significant predictors of future reductions in BMD.[10] Likewise, baseline height was not shown to affect subsequent BMD changes [10,11] although in these studies weight was not adjusted by height or other known confounding factors such as uric acid,[12] serum lipids,[13] or anemia.[14] To examine discrepancies between baseline anthropometric measurements and changes in BMD, as well as integrating data from sporadic reports of correlations among confounding factors, simultaneous analyses of cross-sectional and longitudinal data collected by a centralized source are needed.

By utilizing one of the world’s largest cohorts with data on BMD measurements, we examined the associations of BMD with baseline height and weight and their interactions and changes both cross-sectionally and longitudinally in Japanese adult men and women widely ranging in age in consideration of possible confounding factors. We intended to identify anthropometric measurements that can be reliably used in association with BMD.

This is the first report of an examination of relationships between anthropometric measurements and forearm BMD both cross-sectionally and longitudinally in a large cohort comprised of men and women of various ages in consideration of many covariates including laboratory test results. Results could lead to an anthropometric index to be used in clinical practice to prevent bone loss and fractures in adults of all ages regardless of sex. Furthermore, a population might be identified whose BMD should be monitored regularly.

Cross-sectional studies clarified that height alone correlated little with BMD, with weight correlating better with BMD than BMI especially in elder men and women with or without adjustment by confounding factors. On the other hand, the longitudinal study did not show that anthropometric values were predictors of future BMD in either sex. Initially, this study attempted to address which anthropometric index is better correlated with BMD and which confounding factors would be important in osteoporosis research or clinical practice. In studies examining correlations between bone density and various factors, weight and height are sometimes used simultaneously for the adjustment of BMD. However, results might not necessarily be statistically appropriate considering the potential correlation between height and weight.

This cohort was representative of the Japanese population for various reasons. The mean height and weight of participants were very close to those of the general population in Japan as reported in The National Health and Nutrition Survey in Japan, 2014.[15] Persons with diseases were not excluded, although participants could be considered “healthy” because all laboratory findings other than low-density lipoprotein cholesterol were within reference ranges. The mean BMD_BL values in every 5-year strata were almost identical to those in a previous smaller-scale report from Fukuoka Prefecture, a region far from our study site with a different climate especially during daylight hours, which included 1,786 healthy women and 625 healthy men aged 20 to 79 years (Supplementary Fig. 1).[16]

In cross-sectional analyses, we found that baseline height was not strongly correlated with BMD in any sex or age group, and that since weight was better correlated than BMI with BMD, it should be used to adjust for BMD. In groups of younger men and women, height and BMD were not correlated. In a cross-sectional study of a smaller number of young and old Taiwanese men,[18] it was reported that height was not a predictor of lumbar spine and proximal femur BMD, similar to our results. The correlation coefficient with BMD was only slightly higher with BMI than weight, and the advantage of BMI weakened when adjusted by other non-anthropometric covariates. On the other hand, there was a very weak positive correlation (correlation coefficient approx, 0.2) between unadjusted BMD and height in elder men/women, so BMD was better correlated with body weight than BMI. We assumed that height was a confounding factor in these age groups. In a cross-sectional study of Turkish postmenopausal women that investigated relationships between anthropometric indices or age and T-scores for femur or vertebra, the correlation coefficients were higher in the order of weight, BMI, and height.[6] Although we did not control for medications or medical histories unlike their study, results were similar in our cohort. In a previous study, BMI was used instead of weight for multivariate models for osteoporosis and low BMD, with the explanation that BMI can be used regardless of height although validation was lacking.[19] Another group explained that weight was a better predictor of BMD than BMI, but their study only involved elder women.[20] Our results could be generalized in that the concept was applicable regardless of age or sex.

The strength of correlations between height and weight has varied according to groups or cohorts. For instance, a study of elderly Chinese men and women attempted to solve the problem of collinearity between height and weight by performing principal component analyses; the conclusion was that BMI in elderly people was affected more by weight than by height, suggesting that height could be excluded from covariates for adjustment of central BMD values.[8] Our results for the elder groups supported their conclusion. On the other hand, this finding could not be applied to groups of younger men and women because correlations between height and weight were weaker than in the elderly. However, we still consider that BMI is not a more significant factor in BMD than weight.

Our findings indicated that low weight might be a better indicator of osteopenia than low BMI. As shown in Figure 2, associations between the prevalence of osteopenia and weight or BMI categories were largely dependent upon sex and age. In the elder women group, weight or BMI and the prevalence of BMD had an almost linear correlation with an approximate expression of y=−0.0164x+1.362 (R²=0.9407) and y=−0.027x+1.063 (R²=0.9609), respectively. In contrast, among younger women, the regression lines started to rise from around 60 kg and 22 kg/m² with an approximate expression of y=−0.0164x+0.98 (R²=0.9274) and y=−0.0269x+0.782 (R²=0.7669), respectively. Increased prevalence of osteopenia was not shown in younger men regardless of weight or BMI. However, there was a progressive increase in osteopenia in elderly men whose weight and BMI were approximately 50 kg and ≤19.5 kg/m², respectively.

Our results showed that body weight could better detect osteopenia in the non-weight-bearing forearm than low BMI, which might reflect undernutrition. There is little evidence in the literature that body weight loading affects distant bones via humoral factors and biochemical mechanism(s) explaining these phenomena could not be clarified from this epidemiological study. However, our comprehensive research across all ages in both sexes should contribute to a better understanding of the pathophysiology of osteopenia/osteoporosis by incorporating factors related to activity and nutrition into consideration. Furthermore, in short individuals the BMI is calculated to be high, likely placing individuals in obese or “normal range” categories. Translating these results into clinical practice could lead to increased detection of bone loss in younger women weighing <60 kg and elder men weighing <50 kg, who had previously been overlooked.

Several variables that are reportedly [12-14] and significantly associated with BMD were elucidated as independent variables in multivariate analyses such as uric acid and HDL-C in all 4 groups and hemoglobin in the elder women group. It is well known that these blood factors are influenced by nutritional intake and exercise. We did not consider evaluating the relationship between these factors and BMD in detail in this study, and additional research is necessary to prove these associations.

In the longitudinal analysis, we examined the association between anthropometric indices at baseline and subsequent changes in BMD. In all 4 categories, neither weight nor BMI, regardless of adjustments by confounders, were associated with subsequent changes in BMD. These results derived from data on a large number of participants varying in age were not fundamentally different from those of preceding studies in which age categories of participants were limited.[10] The results represented that weight and BMI at one point in time were not predictive of subsequent BMD, meaning that overweight might not contribute to bone maintenance. This might suggest that factors other than anthropometrics have a strong influence on BMD change.

The limitations of this research are as follows. First, although lumbar DXA should be the gold standard for osteoporosis diagnosis, we used the forearm for BMD measurements in the present study. However, multiple reports showed that DXA of the forearm, which is not weight-bearing, detected osteoporosis diagnosed by central DXA. [21-23] It should be noted that bone density measured by DXA showed a significant correlation of r=0.619 between the radius and vertebrae in healthy Japanese women.[24] In addition to the advantages of cost, portability, and ease of use, we thought that measuring forearm BMD as a screening method was worthwhile. Second, lifestyle habits, such as diet and activity which may differ between generations and sex, can affect the formation of the physique during the growth period or the progression of bone loss. In particular, Japanese people experienced a period of turmoil before and after World War II. Decade-long periods of observations are needed to determine if the present results were robust. Third, the medical checkups were arbitrarily applied, and there could be self-selection biases. Participants diagnosed with osteoporosis after their ‘baseline’ might have stopped follow-up. Despite these possibilities, this database is one of the largest global registries with BMD values, anthropometric measurements, and results of laboratory examinations.

In conclusion, our cross-sectional analyses showed that among the anthropometric indices, weight but not height or BMI was most closely correlated with non-bearing forearm BMD regardless of sex or age. Therefore, in screening for osteopenia, we might not need to take body composition into consideration but should only consider weight. From the results of our longitudinal analysis, not all anthropometric indices can be used as predictors of future reductions in BMD. The results could be helpful in understanding how anthropometrics affect BMD at loaded and unloaded sites and also lead to an anthropometric index to be used in clinical practice to prevent bone loss and fractures in both men and women of any age including young adults.