Background
Osteoporosis is a common skeletal disease characterized by the loss of bone mass and the risk of brittle fracture. It has become a major global public health problem, affecting about 200 million people around the world [
1]. Osteoporosis could be clinically diagnosed by measuring the bone mineral density (BMD), which is a highly heritable trait [
2]. Some modifiable environmental factors such as body weight [
3], physical activity [
4], sleep behavior [
5], and inflammatory disease [
6] could have an impact on bone mass gain and the development of osteoporosis. Other non-modifiable factors included sex, age, and genetic factors. The genome-wide association studies (GWASs) have discovered hundreds of genetic loci that are associated with BMD, osteoporosis, and fracture in the past decade [
7]. However, there is yet a considerable proportion of the variance in bone mass that cannot be explained by known genetic and environmental determinants, e.g., the role of the growth during the prenatal period on adult osteoporosis was often underestimated.
The Developmental Origins of Health and Disease (DOHaD) hypothesis was initially proposed in the 1990s [
8], which proposed that the growth during the prenatal period might play a critical role in an individual’s short- and long-term health [
9]. Although the relationship between birth weight and risk of osteoporosis in later life has been investigated in observational studies, the results were controversial. Some studies detected a positive association between birth weight and bone mass [
10‐
13], but some other studies found an inverse association [
14‐
16]; these conflicting results might be in part due to the limited sample size and other unmeasured confounding factors. With the development of genomic medicine, the Mendelian randomization (MR) approach somehow provides a possibility to reduce the confounding bias or reverse causation by leveraging genetic variants to instrument a potential exposure and therefore leads to more robust results than the conventional observational approach [
17]. Recently, by conducting genome-wide association analyses of own birth weight versus own genetic factors and offspring birth weight versus mother genetic factors, Warrington et al. have successfully partitioned the genetic effects on birth weight into direct fetal genetic component and indirect maternal genetic component [
18]. Consequently, by using the maternal and fetal genetic effects separately, Moen et al. suggested that the maternal intrauterine environment, as proxied by maternal genetic variants that influence offspring birthweight, was unlikely to be a major determinant of adverse cardiometabolic outcomes, but offspring birth weight determined by own genetic factors was found to be associated with offspring cardiometabolic risk factors [
19].
Therefore, the aim of this study is to investigate the effect of birth weight on adult BMD by integrating the observational data and both fetal and maternal genetic evidence. We first performed multivariable regression analyses to assess the association of birth weight with BMD-related parameters measured by dual-energy X-ray absorptiometry (DXA) and by quantitative ultrasound (QUS) in the UK Biobank. We also applied the MR approach to test whether the observational relationship between birth weight and adult BMD could be reflected by either indirect maternal or direct fetal genetic effects or both.
Discussion
In this study, the observational analyses suggested that higher birth weight was associated with increased adult spine bone area but decreased spine BMCadjArea, spine BMD, and heel BMD. By leveraging fetal-only effect SNPs as the genetic instrumental variables, the birth weight determined by own genetic factors was found to be positively associated with offspring bone area but negatively associated with BMD. Furthermore, by using the parent-offspring pairs, our results suggested that the maternal intrauterine environment, as proxied by maternal genetic variants that influence offspring birthweight, was also a determinant of adult BMD. In other words, the observational relationship between birth weight and BMD could be reflected by both the direct fetal and indirect maternal genetic effects.
Lower birth weight often occurs along with inferior intrauterine environments, for example, malnutrition or stress during pregnancy. The Developmental Origin of Health and Disease (DOHaD) hypothesis proposes that exposure to a hostile uterine environment would change the fetus’ structure, function, and metabolism in ways that lead to significantly adverse consequences on an individual’s short- and long-term health, followed by the reduced fetal growth rate, low body weight at birth, and the high risk of future related diseases [
51,
52]. The hypothesis was applied well in the studies of cardiovascular diseases [
9,
53]. Other epidemiologic observations also showed that lower birth weight was associated with an increased risk of stroke, type 2 diabetes mellitus (T2DM), obesity, and hypertension [
54‐
57].
A previous study has tried to investigate the relationship between birth weight and osteoporosis risk, but as they listed as a limitation, they failed to analyze the DXA data, which is the gold standard for the diagnosis of osteoporosis [
58]. In our study design, the observation that higher birth weight was associated with lower adult BMD was consistently shown in two independent study samples: the DXA scan data (20,367 and 21,364 participants) and QUS measurement data (177,675 samples). With DXA scan data, we could also look at the association between birth weight and BMC/bone area. It is better to pay close attention to BMC/bone area because Dennison et al. demonstrated that birth weight had greater contributions to bone size and mineral content than to bone mineral density [
10], and birth weight was previously detected to be associated with the forearm and tibial bone size in the old population aged 65–73 years [
59]. In our study, we found that higher birth weight was associated with higher adult LS BMC and higher LS area. Interestingly, the effect of birth weight on bone size was much larger than the effect on BMC after standardization. At the site of the lateral spine (LaS), the association between birth weight and LaS BMC was not observed, but birth weight was positively associated with the LaS area. These results suggested that higher birth weight would lead to lower adult BMD; this might, to some extent, be because of the larger effect of birth weight on bone size than on BMC. Coincidentally, Steer et al. reported similar results in teenagers that birth weight was positively related to bone size but inversely related to cortical BMD, but they suggested that the relationship was mediated by the effects on late onset of puberty and bone resorption [
14]. In addition, we included an alternative parameter BMCadjArea (BMC adjusted for bone area) in our analysis, and as expected, the results for BMCadjArea were consistent with the result for BMD, because BMCadjArea is conceptually similar to BMD.
Mendelian randomization (MR) is an approach to test the causality of an observed relationship between an exposure and an outcome [
17]. Recently, some studies have tried to leverage the MR method to detect the relationship between birth weight and non-communicable diseases (i.e., atrial fibrillation, type 2 diabetes, and cardiovascular diseases) to address the DOHaD hypothesis [
60‐
62]. However, these MR analyses only took the offspring genetic variants to instrument the birth weight. An optimal way is to also take the maternal genotypes that effect the offspring’s birth weight to proxy the intrauterine environment to test aspects of DOHaD [
63]. Benefiting from the study by Warrington et al., the genetic effects on birth weight were partitioned into direct fetal genetic component and indirect maternal genetic component [
18]. They identified a total of 205 autosomal SNPs associated with birth weight. Of which, 63 autosomal SNPs that have a fetal-only effect on own birth weight represent the effect of own genotype on own birth weight after adjusting the maternal effect on birth weight, 31 autosomal SNPs that had a maternal-only effect on offspring birth weight represent the intrauterine environment, and 41 autosomal SNPs have both fetal and maternal effects [
18]. By leveraging these SNPs with different effects, we found that higher fetal-only effect SNP-determined birth weight was associated with lower LS/FN BMD and heel BMD. We also found that higher maternal-only effect SNP-determined offspring birth weight was associated with lower adult BMD after adjusting for offspring genetic effects but did not find any association evidence in father-offspring pairs, which is suggestive of DOHaD mechanisms (related to lower birthweight) through the effect on bone size. Here, we calculated the unweighted GRS in mothers because we were ignorant of the effect size of each allele to avoid the inflation in type 1 error [
19]. Previously, Moen et al. suggested that the maternal intrauterine environment, as proxied by maternal genetic variants that influence offspring birth weight, was unlikely to be a major determinant of adverse cardiometabolic outcomes, but offspring birth weight determined by own genetic factors was found to be associated with offspring cardiometabolic risk factors [
19]. However, in our study, we suggested that both indirect maternal and direct fetal genetic effects reflected the observational relationship between higher birth weight and lower adult bone mass.
Acknowledged as a limitation of this study, we excluded the individuals with extreme birth weight (very low birth weight < 1.5 kg, low birth weight > 1.5 kg and < 2.5 kg, and birth weight > 4.5 kg) in our study to exclude likely pre-term births and fetal macrosomia. Hovi et al. found that adults born with very low birth weight (< 1.5 kg) had lower LS BMD compared with the participants born at normal birth weight at ages 18.5 to 27.1 [
64]. When we checked back in the UK Biobank dataset, we found that the participants born with very low birth weight had lower lumbar spine BMD than the participants born with normal birth weight (1.166 vs 1.179 g/cm
2). Therefore, the results should be interpreted with caution when facing the extreme value of birth weight.
Publisher’s Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.