Background

Bone is an active tissue that maintains itself by continuous formation and reabsorption [1]. Osteoporosis is a condition in which the density of the bone decreases due to the increased activity of the osteoclasts [2]. A great variance is observed in the prevalence of osteoporosis in different ethnic groups [3]. Age and gender are the two major contributing factors in the occurrence of osteoporosis. Worldwide, one out of three women over the age of 50 experiences osteoporotic fractures in comparison to one in five men of the same age group [4]. Genetic and environmental factors play a crucial role in the etiology of osteoporosis [5, 6]. Calcium intake and exercise are the main risk factors for osteoporosis [5]. It is very well established that along with the environmental factors, individual genetics plays a key role in the development of osteoporosis, e.g., (i) low bone density is found in the female offspring of the osteoporotic women [7], (ii) male offspring of idiopathic osteoporotic men have low bone mineral density [8], and (iii) studies of female twins have shown heritability of bone mineral density (BMD) to be 57 to 92% [9, 10].

Amongst all the genes studied in osteoporosis, the vitamin D receptor (VDR) gene polymorphism is the most important in the etiology of the disease [11, 12]. VDR gene polymorphisms have been reported to be associated with the development of several bone diseases, multiple sclerosis, vitamin D-dependent rickets type II, and other complex diseases [13]. However, the mechanism by which the VDR gene influences bone mass has not been fully elucidated.

In human, VDR gene is found on the chromosome 12 (12q12-q14) with 11 exons and spans ~ 75 kb genomic DNA. The most studied VDR gene polymorphisms are BsmI, ApaI, FokI, and TaqI. Although several studies between osteoporosis and VDR gene polymorphisms have been published, the results are contradictory [14, 15]. This may be due to the differences in the designing of the studies, less number of samples, differences in ethnicities, or various other environmental factors. So, the aim of the present study was to find an association between VDR gene polymorphisms and osteoporosis risk.

Methods

Different databases (PubMed, Google Scholar, SpringerLink, and Science direct) were searched up to December 31, 2018, with the keywords “vitamin D receptor gene,” “BsmI,” “ApaI,” “FokI,” “TaqI,” and “VDR,” along with “osteoporosis.” The retrieved studies were conducted between 1995 and 2018, and we examined all the retrieved papers thoroughly to determine their suitability for inclusion in the current meta-analysis.

Inclusion and exclusion criteria

Studies found suitable to be included in the present study should have (a) a case-control study and (b) reported the sample size and distribution of genotypes. Similarly, a study should be excluded if (a) the study was conducted on the animal model, (b) the study that has replication of data, (c) only cases were reported, and (d) book chapters or review articles.

Data extraction

From the selected articles, we extracted different information like (a) last name of the first author, (b) year of publication of the study, (c) country where the study was conducted, and (d) number of genotypes in different groups. We also checked whether the genotype distributions of control population of all the included studies were in agreement with Hardy–Weinberg equilibrium (HWE) by using the goodness of fit chi-squared test. All the data from the different papers were retrieved by the two authors (UY and PK) and if any discrimination was found, it was resolved by the consultation with the corresponding author.

Statistical analysis

Meta-analysis was done according to the method given in Rai et al. [16]. Briefly, statistical analysis of different vitamin D receptor gene polymorphisms and risk of osteoporosis were estimated by pooling the odds ratio (OR) with its corresponding 95% confidence intervals (CI). Heterogeneity was tested using Q statistics (a p value of less than 0.05 was considered significant). The I2 statistics was also used to assess the discrepancy between studies. If the heterogeneity was higher (p value of Q test < 0.05 or I2 > 50%) than the random effect model [17] that was applied, fixed effect model [18] was used. The heterogeneity may arise due to the differences in ethnicities or variation in study design or outcome. The funnel plot of precision by log odds ratio and standard error by log odds ratio was assessed for the possible publication bias, and if the funnel plot was found asymmetric, it denoted a publication bias [19]. The linear regression method of Egger was used to measure the asymmetry in the funnel plot [20], and a statistically significant publishing bias was considered to be a p value of < 0.05. The meta-analysis was conducted by Open Meta-Analyst program [21].

Results

PRISMA (Preferred Reporting Items for Systematic Reviews and Meta-Analyses) guideline was followed in the present meta-analysis. Flow chart of article selection was shown in Fig. 1 with specific reasons. Eighty-one studies were found to be eligible for inclusion in the present meta-analysis after applying the inclusion and exclusion criteria. Out of 81 included studies, BsmI, ApaI, FokI, and TaqI polymorphisms were investigated in 65, 31, 18, and 26 studies respectively.

Fig. 1
figure 1

Flow diagram of study search and selection process

Full size image

Eligible studies

For BsmI, a total of 65 studies with 6880 cases and 8049 controls were included in the meta-analysis [22,23,24,25,26,27,28,29,30,31,32,33,34,35,36,37,38,39,40,41,42,43,44,45,46,47,48,49,50,51,52,53,54,55,56,57,58,59,60,61,62,63,64,65,66,67,68,69,70,71,72,73,74,75,76,77,78,79,80,81,82,83,84,85,86].

For ApaI, a total of 31 studies with 3763 cases and 3934 controls were found eligible for the meta-analysis [24, 28, 30, 38, 44, 45, 48, 51, 56, 63, 64, 66, 69, 71, 73, 75, 77, 79, 81, 83,84,85, 87,88,89,90,91,92,93,94,95].

For FokI, meta-analysis which has a total of 18 studies with 1895 cases and 1722 controls were included in the meta-analysis [38, 45, 50, 56, 61, 67, 70, 71, 73, 75, 79, 81, 84, 96,97,98,99,100].

For TaqI, a total of 26 studies including 2458 cases and 2895 controls were found eligible for inclusion in the meta-analysis [24, 28, 30, 38, 45, 48, 51, 56, 63, 64, 69, 71, 73, 75, 77, 79, 81, 83,84,85,86, 92, 93, 95, 101, 102].

Meta-analysis

BsmI meta-analysis

In allele contrast model, high heterogeneity was observed with insignificant association (ORbvs.B = 0.89, 95% CI = 0.78–1.01, p = 0.09, I2 = 82.02%, Pheterogeneity = < 0.001). No significant association was found in any other genetic models—for dominant model (bb + Bb vs. BB) OR = 0.81, 95% CI = 0.68–0.97, p = 0.02; for homozygote model (bb vs. BB) OR = 0.77, 95% CI = 0.60–0.99, p = 0.04; for co-dominant model (Bb vs. BB) OR = 0.85, 95% CI = 0.73–0.98, p = 0.03; and for recessive model (BB + Bb vs. bb) OR = 0.88, 95% CI = 0.74–1.06, p = 0.20. Heterogeneity was high in all the genetic models except in the co-dominant model (Table 1; Fig. 2).

Table 1 Summary estimates for the odds ratio (OR) of BsmI in various allele/genotype contrasts, the significance level (p value) of heterogeneity test (Q test), and the I2 metric
Full size table
Fig. 2
figure 2

Random effect forest plot of allele contrast model (b vs. B) of VDR BsmI polymorphism. Results of individual and summary OR estimates, and 95% CI of each study were shown. Horizontal lines represented 95% CI, and dotted vertical lines represent the value of the summary OR

Full size image

Ethnicity was used for the sub-group analysis. Out of 65 studies, 37 belong to Caucasians, 22 were Asian, and 6 were of other origins. High heterogeneity was observed in all genetic models in all sub-groups. No significant association was found in any sub-group analyses in any genetic models (Table 1; Fig. 2).

ApaI meta-analysis

Insignificant association with high heterogeneity was found in the allele contrast model (ORavs.A = 1.01, 95% CI = 0.87–1.17, p = 0.86, I2 = 74.82%, Pheterogeneity = < 0.001). No significant association was found in any other genetic models—for dominant model (aa+Aa vs. AA) OR = 0.95, 95% CI = 0.78–1.14, p = 0.60; for homozygote model (aa vs. AA) OR = 0.97, 95% CI = 0.72–1.30, p = 0.84; for co-dominant model (Aa vs. AA) OR = 0.92, 95% CI = 0.81–1.04, p = 0.21; and for recessive model (AA+Aa vs. aa) OR = 1.02, 95% CI = 0.81–1.28, p = 0.83 (Table 2; Fig. 3).

Table 2 Summary estimates for the odds ratio (OR) of ApaI in various allele/genotype contrasts, the significance level (p value) of heterogeneity test (Q test), and the I2 metric
Full size table
Fig. 3
figure 3

Random effect forest plot of allele contrast model (a vs. A) of VDR ApaI polymorphism

Full size image

The ethnicity-based sub-group analyses were conducted. Out of 31 studies, 15 were Caucasians, 12 were Asians, and 4 were of other origin. High heterogeneity was observed in Caucasian studies while low heterogeneity was found in Asian and other studies. Insignificant association was found in all sub-group analyses and in all the genetic models except for the recessive model of the other studies (AA+Aa vs. aa) OR = 1.49, 95% CI = 1.00–2.23, p = 0.04 (Table 2; Fig. 3).

FokI meta-analysis

In the dominant model of FokI polymorphism, significant association was found (ORff + Ffvs.FF = 1.19, 95% CI = 1.04–1.36, p = 0.01, I2 = 39.36%). No significant association was observed in any other genetic models—allele contrast model ORfvs.F = 1.13, 95% CI 0.95–1.34, p = 0.15, I2 = 61.8%, Pheterogeneity = < 0.001; homozygote model (ff vs. FF) OR = 1.38, 95% CI = 0.92–2.05, p = 0.11; co-dominant model (Ff vs. FF) OR = 1.12, 95% CI = 0.97–1.30, p = 0.11; and recessive model (FF + Ff vs. ff) OR = 1.34, 95% CI = 0.94–1.91, p = 0.10) (Table 3; Fig. 4).

Table 3 Summary estimates for the odds ratio (OR) of FokI in various allele/genotype contrasts, the significance level (p value) of heterogeneity test (Q test), and the I2 metric
Full size table
Fig. 4
figure 4

Fixed effect forest plot of dominant model (ff + Ff vs. FF) of VDR FokI polymorphism

Full size image

Studies were further analyzed by sub-group analysis on the basis of ethnicity. Out of 18, ten studies belong to Caucasians, five were Asians, and three were of other ethnicity. High heterogeneity was found in Asian and other studies; while in the Caucasian studies, low heterogeneity was observed. No significant association was found in any sub-group in any genetic model (Table 3; Fig. 4).

TaqI meta-analysis

High heterogeneity with insignificant association was found in the allele contrast model of TaqI polymorphism (ORtvs.T = 1.10, 95% CI = 0.91–1.32, p = 0.30, I2 = 77.26%, Pheterogeneity = < 0.001). Insignificant association was found in the other four genetic models—dominant model (tt + Tt vs. TT) OR = 1.09, 95% CI = 0.84–1.41, p = 0.48; for homozygote model (tt vs. TT) OR = 1.20, 95% CI = 0.85–1.69, p = 0.29; for co-dominant model (Tt vs. TT) OR = 1.04, 95% CI = 0.82–1.33, p = 0.70; and for recessive model (TT + Tt vs. tt) OR = 1.16, 95% CI = 0.91–1.48, p = 0.20 (Table 4; Fig. 5).

Table 4 Summary estimates for the odds ratio (OR) of TaqI in various allele/genotype contrasts, the significance level (p value) of heterogeneity test (Q test), and the I2 metric
Full size table
Fig. 5
figure 5

Fixed effect forest plot of recessive model (TT + Tt vs. tt) of VDR TaqI polymorphism

Full size image

The studies were further analyzed on the basis of ethnicity for sub-group analysis. Out of 26 studies, 17 belong to Caucasians, six were Asians, and three were of other ethnicity. High heterogeneity was observed in all groups, i.e., Asian, Caucasian, and other studies. Insignificant results were found in all the sub-groups of all the genetic models except for the recessive model of the Caucasian population (TT + Tt vs. tt) OR = 1.35, 95% CI = 1.11–1.63, p = 0.002 (Table 4; Fig. 5).

Sensitivity analysis

To conduct sensitivity analysis, all the studies deviated from the Hardy–Weinberg equilibrium (p < 0.05) were omitted. In BsmI, 21 studies [27, 30, 34, 38, 39, 44, 48,49,50,51,52, 58, 60, 62, 64, 66, 68, 70, 71, 76, 80] were deviated from the HWE. Meta-analysis, after removal of these 21 studies, showed no significant association with osteoporosis risk in the main analysis (ORbvs.B = 0.99, 95% CI = 0.85–1.15, p = 0.92, I2 = 77.48%) or in any sub-groups (Asian subgroup ORbvs.B = 0.99, 95% CI = 0.66–1.50, p = 0.99, I2 = 83.65%; Caucasian subgroup ORbvs.B = 0.96, 95% CI = 0.83–1.11, p = 0.65, I2 = 69.61%; and other studies subgroup ORbvs.B = 1.24, 95% CI = 0.64–2.43, p = 0.51, I2 = 86.53%). When these 21 studies were removed, heterogeneity was decreased in both the overall and in the sub-group meta-analyses except in the Asian studies.

In total of 18 FokI studies, control population in five studies [56, 70, 79, 99, 100] was not in HWE. When these studies were removed from the analysis, insignificant association was found in the main analysis (ORfvs.F = 1.12, 95% CI = 0.99–1.26, p = 0.05, I2 = 46.48%), and no association was found in any sub-group. Removal of these studies decreases the heterogeneity both in the overall and in sub-group meta-analyses.

The control samples of nine ApaI studies [28, 30, 44, 48, 51, 56, 71, 83, 94] were not in HWE. Result of meta-analysis after removal of these nine studies showed no association between ApaI polymorphism and osteoporosis risk in the main/overall analysis (ORavs.A = 1.07, 95% CI = 0.90–1.27, p = 0.39, I2 = 73.94%) and Caucasian population (ORavs.A = 0.85, 95% CI = 0.63–1.16, p = 0.32, I2 = 78.62%) but the Asian population (ORavs.A = 1.42, 95% CI = 1.03–1.96, p = 0.03, I2 = 77.61%) and subgroup other studies (recessive model ORAA + Aavs.aa = 1.49, 95% CI = 1.00–2.23, p = 0.04, I2 = 52.4%) showed statistically significant association with osteoporosis. Heterogeneity was also decreased both in the overall and sub-group meta-analyses.

Out of 26 TaqI studies, control samples of the four studies [28, 56, 77, 101] were deviated from the HWE. Results of meta-analysis of 22 studies (after elimination of 4 studies deviated from HWE) did not show any association between TaqI polymorphism and osteoporosis risk either in total studies (ORtvsT = 1.05, 95% CI = 0.85–1.29, p = 0.63, I2 = 78.86%) or in any sub-group. Moreover, after removal of these 4 studies, there was an increase in the heterogeneities in overall and sub-group meta-analyses except the Asian population.

Publication bias

In all the genetic models in the overall and in sub-group meta-analyses for all polymorphisms, the funnel plots were symmetrical (Fig. 6; Tables 1–4) except recessive model of the other studies in FokI and co-dominant model of the Asian studies in ApaI polymorphisms. Similarly, no publication bias was found in any genetic model in overall meta-analyses of all the four polymorphisms by the Egger’s test except recessive model of the other studies in FokI and co-dominant model of the Asian studies in ApaI polymorphism (Tables 1–4).

Fig. 6
figure 6

Funnel plots for FokI. a Precision by log odds ratio. b Standard error by log odds ratio for BsmI. c Precision by log odds ratio. d Standard error by log odds ratio for ApaI. e Precision by log odds ratio. f Standard error by log odds ratio for TaqI. g Precision by log odds ratio. h Standard error by log odds ratio

Full size image

Discussion

The vitamin D receptors are the members of the nuclear hormone receptor (NR1I) family and expressed in different organs like the intestine, thyroid, and kidney in humans [103]. It is primarily responsible for the endocrine action of vitamin D that regulates calcium homeostasis and reduces the risk of osteoporosis. VDR is translocated from the cytoplasm to the nucleus when activated by binding of its ligand 1α,25-dihydroxyvitamin D3 (1,25(OH)2D3) [104]. Several studies have documented that the onset of osteoporosis is caused by VDR gene polymorphisms [81]. VDR gene polymorphisms are also associated with other diseases like breast cancer [105], diabetes [106], myocardial infarction [107], and metabolic syndrome and inflammation [108].

Meta-analysis is a well-established statistical tool used for combining the data of small sample-sized individual studies. Meta-analysis increases the power of the study and decreases type I and II errors. During the past two decades, a number of meta-analyses were published which assessed the polymorphism of small effect genes as risk factor for different diseases and disorders, e.g., Down syndrome [16], neural tube defects [109], Glucose 6-phosphate dehydrogenase deficiency [110], depression [111], schizophrenia [112], Alzheimer [113], breast cancer [114], colorectal cancer [115], esophageal cancer [116], and prostate cancer [117].

During literature search, we identified seven meta-analyses [15, 118,119,120,121,122,123] investigating the relationship between VDR gene polymorphisms and osteoporosis. BsmI, ApaI, FokI, and TaqI polymorphisms were included in seven, four, two, and two meta-analyses respectively. BsmI polymorphism studies were included in all seven meta-analyses. In six meta-analyses, no significant association was found between osteoporosis susceptibility and BsmI polymorphism [15, 118,119,120,121,122]. Zhang et al [123] conducted a meta-analysis of the risk of osteoporosis in postmenopausal women with 36 studies including 7192 subjects and found a marginally significant association (ORbvs.B = 1.2; CI = 1.00–1.46; p = 0.052). In all the meta-analyses, a low between study heterogeneity was found in all the studies except the study conducted by Yu et al [120]. ApaI polymorphism was included in four meta-analyses [118, 120, 122, 123]. Zintzaras et al [118], Yu et al [120], Wang et al [122], and Zhang et al [123] included seven, six, three, and eighteen studies, respectively, in their meta-analyses, and all four studies reported no association between ApaI polymorphism and osteoporosis risk. Zintzaras et al [118] and Zhang et al [123] conducted meta-analyses of three and 18 studies of FokI polymorphism, and no significant association was found between FokI polymorphism and osteoporosis. Both groups [118, 123] also conducted meta-analyses of TaqI polymorphism studies and again reported no association between TaqI polymorphism and osteoporosis susceptibility.

In the present meta-analysis, four common VDR gene polymorphisms (BsmI, ApaI, FokI, and TaqI) were included. A total of 65 (14929 samples), 31 (7697 samples), 18 (3617 samples), and 26 (5353 samples) studies for BsmI, ApaI, FokI, and TaqI polymorphisms, respectively, were included. We found a significant association in the dominant model of FokI polymorphism (ff + Ff vs. FF OR = 1.19, 95% CI = 1.04–1.36, p = 0.01) with low heterogeneity (I2 = 39.36). No association was found in sub-group analysis on the basis of ethnicity in any genetic model except in the Caucasian population in the recessive model of TaqI polymorphism (TT + Tt vs. tt OR = 1.35, 95% CI = 1.11–1.63, p = 0.002) with moderate heterogeneity (I2 = 50.07). The frequency of different VDR gene polymorphisms varies in different ethnic/regional populations. Due to this, the effect of these polymorphisms might vary from population to population.

The present meta-analysis has few demerits like (i) used crude odds ratio, (ii) only genetic polymorphisms considered, and other factors such as environmental factors or food habits that are not included which might have important roles in the etiology of osteoporosis. With these limitations, the present study has some strength like (i) this is the largest meta-analysis conducted both in number of included studies and number of sample size (81 studies; 19268 samples) and (ii) included all common VDR polymorphisms (BsmI, ApaI, FokI, and TaqI).

Conclusion

In conclusion, we found that the dominant model of FokI polymorphism is associated with osteoporosis, and also the recessive model of TaqI polymorphism is a risk factor for the osteoporosis in the Caucasian population. The other polymorphisms (BsmI and ApaI) have no role in the osteoporosis in total or in the stratified populations. In addition, it has been suggested that different gene-gene and gene-environment interactions should also be considered in future case-control studies, which could clarify the genetics of osteoporosis.