Blood lipid levels in patients with osteopenia and osteoporosis:a systematic review and meta-analysis

Considering the controversial relationship between blood lipid levels and osteopenia and osteoporosis (OP), we performed this meta-analysis. Using specific keywords and related words, we searched PubMed, Embase, and Cochrane Library databases. The Newcastle–Ottawa Scale form was used to evaluate the quality of the literature. According to the inclusion and exclusion criteria, we systematically screened the literature to extract relevant information and data. ReVman 5.3 and Stata 13.0 software were used for statistical analysis. Results were expressed as the mean difference (MD) and 95% confidence interval (95% CI). The heterogeneity test was conducted according to I2 and Q tests. Egger’s test was used to quantitatively evaluate publication bias. This analysis involved 12 studies (12,395 subjects). The quality of the literature was acceptable. Among subjects who were not taking lipid-lowering drugs, total cholesterol (TC) (MD = 0.11 mmol/L, 95%CI: − 0.03, 0.25; I2 = 21%; P = 0.36), triglycerides (TG) (MD =  − 0.01 mmol/L, 95%CI: − 0.09, 0.07; I2 = 6%; P = 0.34), and low-density lipoprotein cholesterol (LDL-C) (MD = 0.10 mmol/L, 95%CI: 0.00, 0.19; I2 = 0%; P = 0.74) in the osteopenia were not significantly increased/decreased. There were no significant differences in LDL-C (MD = 0.02 mmol/L, 95%CI: − 0.09, 0.13; I2 = 0%; P = 0.74) in postmenopausal women in osteopenia. TG (MD =  − 0.04 mmol/L, 95%CI: − 0.14,0.07; I2 = 49%; P = 0.07) was unchanged in the osteoporosis (OP) group in subjects without taking lipid-lowering drugs. HDL-C was elevated in OP group (MD = 0.05 mmol/L, 95%CI: 0.03, 0.07; I2 = 31%; P = 0.15) but not in osteopenia group (MD = 0.01 mmol/L, 95%CI: − 0.01, 0.02; I2 = 38%; P = 0.14) in all subjects. HDL-C was elevated in patients with OP.


Introduction
Osteoporosis (OP) is a disease characterized by reduced bone mineral density (BMD) and an increased risk of osteoporotic fractures. It is one of the most common metabolic diseases in the elderly population, which poses a serious health concern worldwide [1] because osteoporotic fractures are the leading cause of disability and death [2]. The onset of this disease is associated with various factors, including aging, gender, insufficient calcium intake, vitamin D deficiency, low body mass index (BMI), decreased physical activity [3], and hyperthyroidism [4]. Among them, nutrition is one of the most important factors. When malnutrition occurs, the raw materials for bone formation are limited, which can lead to the development of osteopenia and even OP. BMI is an indicator used to assess nutritional status. Previous studies have shown that subjects with a high BMI have a low risk of OP [5]. It has been suggested that blood lipid levels also reflect the nutrition level to some extent, but in patients with osteopenia or OP, the results of various studies on blood lipid levels have been inconclusive. Gu et al. [6] found that compared with the healthy subjects, total cholesterol (TC) and low-density lipoprotein cholesterol (LDL-C) were significantly increased in adult patients with osteopenia and OP. Ersoy's [7] study indicated that TC and LDL-C were decreased in postmenopausal women with OP. However, Li et al. [8] found that TC, triglyceride (TG), and LDL-C levels were not significantly different in postmenopausal women with OP. Therefore, the blood lipid levels in patients with osteopenia, or OP remain unknown. In addition, whether an increase in blood lipid levels can protect patients from the development of OP is not well understood now. The aim of this meta-analysis was to extract blood lipid indicators in patients with osteopenia, or OP from case-control studies and to determine their levels.

Search strategy
Two researchers used subject words and free words, including "bone mineral density", "osteoporosis", "fractures", "bone", and "lipids", to search PubMed, Embase, and Cochrane databases (Supplementary material 1). We also manually checked the references listed in the retrieved references to prevent missing studies. The search was limited to English language and deadline was February 2020. For articles with only titles and abstracts, we requested articles by email. The inclusion and exclusion criteria were as follows.

Inclusion criteria
i) The article involved at least one of osteopenia and osteoporosis, and the diagnostic criteria were clear. The World Health Organization (WHO) classification system was applied for defining osteoporosis as a T-score ≤ 2.5 and osteopenia as −2.5 < T-score < −1.
ii) The article involved at least one of the following blood lipids: TC, TG, high-density lipoprotein cholesterol (HDL-C), or LDL-C. iii) The study was a case-control study. iv) The literature provided original data or relevant data that could be obtained based on data conversion.

Exclusion criteria
i) The subjects had acute cardiovascular/cerebrovascular diseases or malignant tumors. ii) Subjects with diseases which may affect bone metabolism or calcium absorption, such as serious malabsorption syndrome, enteritidis, hyperparathyroidism, chronic infectious arthritis, osteomalacia and diabetes. iii) Subjects receiving medications such as glucocorticoid, estrogen, androgen, calcitonin, and antidiabetic agents. iv) The diagnosis of osteopenia and osteoporosis was not clear. v) The data were incomplete or could not be obtained by conversion. vi) The study was a review, case report, commentary, or animal/cell-based.

Literature screening
Two researchers independently screened the literature according to the inclusion and exclusion criteria and then compared their findings with each other. First, duplicate studies, animal/cell experiments, case reports, and commentaries were removed. Second, the studies were screened based on their titles and abstracts. Last, the full text was read to decide whether the study should be included. If there were objections, the studies were discussed with a third party to decide whether to include them.

Information extraction
We read the full text of each literature and extracted the relevant information, which included the name of the first author, publication year, country the studied was performed in, the average age of subjects, the sex of patients and lipidlowering drugs. In addition, the occurrence of osteopenia or OP, and blood lipid levels of each group were extracted.

Quality assessment
The quality of the literature included was evaluated according to the Newcastle-Ottawa Scale (NOS) form. A score of 1-3 points was considered poor quality, 4-6 points indicated moderate quality, and a score of 7-9 points was considered high quality. The specific scoring criteria were as follows: Selection (adequate case definition, case representativeness, selection of controls, definition of controls), Comparability (comparability of cases and controls based on the design or analysis), and Exposure (ascertainment of exposure, same method of ascertainment for cases and controls, nonresponse rate) (Supplementary material 2).

Statistical analysis
The meta-analysis was performed using ReVman 5.3 and Stata 13.0 software. All blood lipid indicators were considered continuous variables, and the effect amount was shown as the mean difference (MD) and 95% confidence interval (95%CI). An inverse variance model was used as the statistical model. Heterogeneity analysis was performed by Q and I 2 tests. If P > 0.1 in the Q test or I 2 < 50% in the I 2 test, it was considered homogeneous. If P < 0.1 or I 2 > 50%, it was regarded as heterogeneous, and the source of heterogeneity was identified. If the heterogeneity was large, a random-effects model was used; whereas, if the heterogeneity was small, a fixed-effects model was used. Subgroup analysis was also used to identify the source of heterogeneity. It was based on subject characteristics [(postmenopausal women, unclassified women: include menopausal and non-menopausal women, or in other words, women who are not sure whether they are menopausal or not, men or adults: include women and men.)] and lipid-lowering drugs (taking or not taking) to do subgroup meta-analysis. If the heterogeneity was significantly reduced, the characteristics were the source of heterogeneity. Of note, if there was only one study in one subgroup, the result was not listed. Egger's test was conducted using Stata software to quantitatively evaluate publication bias. If P > 0.05, publication bias was considered non-existent.

Literature search results and basic characteristics of the included literature
We retrieved 6680, 11,899, and 2476 articles from Pub-Med, Embase, and Cochrane libraries, respectively. Therefore, a total of 21,055 articles were imported into Endnote to manage the references. A stepwise screening was performed according to a previous method (Fig. S1). A total of 12 studies were finally included that involved 12,395 subjects. Among them, eight studies included data from patients with osteopenia, and 12 studies included data from patients with osteoporosis. The basic characteristics of included studies are shown in Table 1.

Quality assessment
According to the NOS form, we systematically evaluated the quality of the included articles. Six articles scored six points, five articles scored seven points, and one article scored eight points. Overall, the quality of the included literature was acceptable.  [17] 2009 Turkey 60.0 ± 6.0 F Postmenopausal 107 6 √ Verit [18] 2007 Turkey 54.0 ± 3.6 F Postmenopausal 97 6 √

HDL-C
A total of six studies (9360 subjects) measured HDL-C levels in patients with osteopenia. The total effect amount was MD = 0.01 mmol/L (95%CI: − 0.01, 0.02; I 2 = 38%; P = 0.14) in the fixed-effects model (Fig. 3). There was no difference in HDL-C levels between the osteopenia group and the control group. We found that P = 0.734 in Egger's test, indicating that there was no publication bias (Fig. S2c).

Lipid levels in patients with osteoporosis
The results showed neither characteristics of the subjects or lipid-lowering drugs was not the source of heterogeneity. There was no publication bias because P = 0.731 in Egger's test (Fig. S4a).

HDL-C
Ten studies involving 7424 subjects measured HDL-C levels in patients with osteoporosis. The MD was 0.05 mmol/L (95%CI: 0.03, 0.07; I 2 = 31%; P = 0.15) in the fixed effect model, which indicated that HDL-C levels were higher in the osteoporosis group than in the control group (Fig. 6). P = 0.627 in Egger's test, which indicated that there was some publication bias (Fig. S4c).

Discussion
We used the Egger's test to quantitatively evaluate publication bias. All P values were > 0.05 in the results, indicating that there was no publication bias for each indicator. The quality of the included literature was evaluated by the NOS score, which were all above median level, indicating that the quality of the included literature was acceptable, and the results of the meta-analysis were relatively credible. This meta-analysis involved 12 studies and a total of 12,395 subjects. The studies included two events (osteopenia and OP) and four indicators (TC, TG, HDL-C, and LDL-C). The larger number of subjects strived for a stable result. Some results in this meta-analysis were heterogeneous, so we performed subgroup analyses based on two factors. The one factor was the characteristics of the subjects, which may be due to the actions of hormone. Postmenopausal women are more prone to OP because of lack of estrogen protection. The other factor was whether or not take lipid-lowering drugs, because it affects blood lipid levels. Even if subgroup analyses were conducted, some sources of heterogeneity were still not found, and we considered the possible reason was mixed effects of multiple confounding factors. A meta-analysis conducted by Chen [19] in 2018 found that HDL-C and TC levels were higher in patients with postmenopausal OP, which differed from the results of our study. Chen's study [33] included menopausal women, however, the source of heterogeneity was not found, though a large heterogeneity of HDL-C and TC was present. The present meta-analysis is comprised of adults, including menopausal women. Accordingly, the heterogeneity of HDL-C was observed to be small, while that of TC was large. Moreover, the source of heterogeneity of TC could not be determined through the subgroup analysis.
Trimpou et al. [20] observed the necrosis of the femoral head under an electron microscope and found that the number and size of fat cells were significantly increased, suggesting that hypercholesterolemia is an independent risk factor for osteoporotic fractures. A recent meta-analysis, including 33 studies (16 cohort studies, 7 case-control studies, and ten randomized controlled trials) showed that statins reduced the risk of total and hip fractures. The use of statins has been associated with an increase in total hip BMD and lumbar spine and was found to increase the expression of bone formation markers, such as osteocalcin [21].
The pathophysiological relationship between blood lipid levels and BMD remains unclear. Most studies report that eating a high-fat diet reduces bone strength, changes the microstructure of the cancellous bone compartment, and changes the bone marrow environment, and low-level inflammation may play a role in these processes [22]. TC and its metabolites have been reported to affect the functional activity of osteoblasts in vitro and in vivo [23]. Elevated serum lipids may cause bone blood vessels to accumulate in the subendothelial matrix and may inhibit the differentiation and mineralization of bone cells.
This study has some limitations. First, the heterogeneity of some results obtained in this study was large, and the source of heterogeneity according to subject characteristics or lipid-lowering drugs was not identified. Considering the number of included studies, the sources of heterogeneity may involve many aspects. Second, most of the participants included in this meta-analysis were postmenopausal women, but there are limited studies on men. However, the male population is also a group that needs our attention. We look forward to more individual studies  on male OP to perform a future meta-analysis. Third, we included case-control studies with weaker levels of evidence. Long-term cohort studies are needed to determine the effects of elevated blood lipids on osteopenia, OP, and fractures.
In conclusion, the relationship between lipids and osteopenia/OP group was different among various populations. There was no difference in TC, TG and LDL-C between the osteopenia/OP and control groups in subjects who were not taking lipid-lowering drugs. In postmenopausal women, there was no difference between osteopenia and control groups. HDL-C was elevated in all subjects in the OP group but not in the osteopenia group. This suggests that HDL-C may be used as an indicator to predict osteopenia, osteoporosis, or even fractures. In future studies, blood lipid indicators can be grouped, and the incidence of osteopenia and OP can be determined after longer follow-up periods, which will increase the significance of the results. In addition to serving as a protective lipid marker for cardiovascular and cerebrovascular diseases, HDL-C has been shown to increase in osteoporosis, indicating that blood lipid levels should be controlled and maintained in an appropriate range to reduce the risk of osteopenia or OP while protecting against cardiovascular/cerebrovascular diseases. That is to say, in regard to cardiovascular and cerebrovascular diseases, the higher the HDL-C, the better. However, with a rise in HDL-C, the risk of OP may consequently increase as well. Therefore, a suitable range for HDL should be considered in view of protecting heart and brain vessels without increasing the risk of OP.