Background

Osteoarthritis knee (KOA) is a chronic arthritic disease characterized by degenerative lesions and osteophytes in the knee cartilage [1, 2]. Clinical manifestations include pain, restricted motion, joint deformity, and bone friction sounds [3, 4]. With the population aging, the incidence of KOA is anticipated to rise from 13.8 to 15.7% by 2032, placing a huge burden on families and society [5, 6].

Several risk factors have been established to be caused with KOA, including age, gender, previous knee injury, occupational performance, and overweight or obesity. Overweight or obesity has been found to have a temporal causal association with the development and progression of KOA in early cohort studies [7,8,9]. Overweight or obesity is measured in various ways, such as body mass index (BMI), body fat percentage, waist circumference, hip circumference, waist–hip ratio (WHR), and basal metabolic rate (BMR). Although previous studies have found that all these indicators are risk factors for KOA, their causal relationship is not yet clear [10,11,12].

Mendelian randomization (MR) is a causal inference approach that uses genetic variation as an instrumental variable (IV); it is based on the principle of using the random division and combination of gametes during sexual reproduction to simulate the random assignment process to the subject of the study [13, 14]. Katan was the first to formulate a MR method for exploring the direct increase in cancer risk cause with low serum cholesterol levels [15]. In recent years, it has been widely used in the study of causal associations in a variety of diseases [16,17,18]. MR uses IVs, usually single nucleotide polymorphisms (SNPs), which are reliably caused with exposure and do not vary with caused lifestyle or socio-economic factors, and have the potential to confound traditional observational associations [19, 20]. Therefore, our study used MR to explore the causal relationship between weight, BMI, body fat percentage, waist circumference, hip circumference, WHR, and BMR in KOA. Data from multiple datasets for the same indicator were combined using meta-analysis. Through exploring the causal association between body composition measurements and KOA, it can help to make relevant interventions in the clinic to effectively prevent the development of KOA, and to make the patients with KOA have better regression.

Methods

Study design

This study used MR to explore the causal relationship between weight, BMI, body fat percentage, body fat percentage, waist circumference, hip circumference, WHR, and BMR in KOA. Three assumptions need to be met in order to minimise bias due to unobserved confounding, measurement error, and reverse causality. They are (1) relevance, where the IV is strongly correlated with the exposure factor; (2) independence, whereby the IV is not correlated with the confounding factor; and (3) exclusion restriction, there is no causal association between the instrument variable and outcome independent of the exposure [21]. An overview of the study design is shown in Fig. 1. This study is reported following the STROBE-MR guidelines.

Fig. 1
figure 1

Overview of the design of this Mendelian randomization (MR) study on body composition measurements and osteoarthritis knee

Full size image

Data sources

All analysed data are available in the IEU OpenGWAS project for this study (https://gwas.mrcieu.ac.uk/). Exposure factors were body composition measurements, including weight, BMI, body fat percentage, waist circumference, hip circumference, WHR, and BMR. The principle of selection was that the same exposure factor was selected, systematic and comprehensive search for datasets on body composition measurements, with a screening process that: (1) has a clear data source (e.g., GIANT, MRC-IEU, Neale Lab, Within family GWAS, etc.); and (2) uses the most recent year of data for the same data source. SNPs were from individuals of European origin, including both males and females. The KOA outcome factor was derived from 29 999 696 SNPs obtained from 403 124 European populations, which were sequenced by Tachmazidou et al. and published in the UK Biobank consortium [22]. Detailed information is shown in Table 1.

Table 1 Overview of the data sources of the instrumental variables used in the MR study
Full size table

Genetic instrument selection

To avoid strong linkage disequilibrium between SNPs, then genome-wide significant SNPs with independent and highly correlated exposure factors, as well as outcome variables were selected as IVs. The genome-wide information from the Thousand Genomes Project was used as a reference to screen for IVs without linkage effects [23]: (1) the parameters of weight, BMI, body fat percentage, waist circumference, hip circumference, WHR, and BMR datasets with genome-wide significance were set to P < 5 × 10− 8; (2) the linkage disequilibrium parameter (r2) was set to 0.001; and (3) the genetic distance was set to 10 MB, to screen for IVs without linkage effects. Then, IVs that were apparently caused with KOA were excluded from the screened IVs (P < 0.05). At the same time, the data were pre-processed so as to ensure consistency in effects equivalence and effect sizes. Finally, the strength of causal association of the genetic instruments for each putative risk factor was quantified by the F statistic (F = β2/se2) for all SNPs, to assess the power of the SNPs, If the F-statistic is much greater than 10, the likelihood of weak IV bias is small [30, 35]. Our study further demonstrated a causal relationship between waist circumference, hip circumference, and KOA in different datasets. However, the results of WHR in both datasets suggested that there was no causal relationship with KOA. In the previous studies, the findings on WHR were inconsistent. Holliday et al. found that WHR was not caused with KOA [35], but Lohmander et al. found an RR (risk ratio) of 2.2 for WHR [36]. On the other hand, Gandhi et al. found the RR of being obese [10], as determined by WHR if classified as obese by the BMI criteria, was 1.04 for men and 1.23 for women, suggesting that the causal relationship between WHR and KOA might be influenced by gender factors and requires further study.

In the two datasets for BMR analysis, the Neale Lab consortium’s dataset was found to have a statistically significant pleiotropic analysis, and the MRC-IEU consortium’s dataset suggested a causal relationship between BMR and KOA, with a combined OR of 1.36 (1.27–1.46). Therefore, BMR might have a causal effect in KOA. However, BMR was influenced by a number of factors, such as body surface area, growth stage, gender, nutrition, and functional status, thus was expected further analysis.

There were some limitations should be mentioned in this study. The sources of datasets for different physical measures were inconsistent, with four datasets present for some indicators and two datasets for others, which might have impacted the results. The discrepancy in findings due to gender differences evident in previous studies of the WHR, and failing to analyse gender separately, might also be insufficient in other measures. The datasets were all from European populations, so the findings may be applicable only in European populations and be of limited use for other populations. MR assumed a linear relationship between exposure factors and outcome factors, and did not apply if there was no linear relationship between the two.

Conclusion

In summary, our study used MR to explore the causal relationships between weight, BMI, body fat percentage, waist circumference, hip circumference, WHR, and BMR in KOA. Additionally, we used meta-analysis to combine the results of different datasets and to enhance the strength of their causal associations. We found that weight and BMR might have a causal effect on KOA, but WHR did not. BMI, body fat percentage, waist circumference, and hip circumference had a causal relationship with KOA. Additionally, body fat percentage might be a better indicator of KOA than BMI.