Mendelian randomization study of body composition and depression in people of East Asian descent highlights possible setting-specific causation. BMC Medicine

The demographics of 3398 individuals with depressive symptoms in CKB and 96,979 controls with valid genetic data are summarized in Table 1. In general, those affected by depressive symptoms were younger, more likely to be female, and live in rural areas of China.

Within CKB, BMI GRS was strongly associated with BMI, 1.5% (F = 637.87) and 3.1% (F= 1815.94) difference in males and females respectively. Similarly, WHR was strongly associated with GRS explaining 0.4% (WHR).F= 191.74) difference in males and 0.9% (F= 516.41) difference in females. BMI GRS within each region was strongly correlated with BMI (Additional file 2 : Table S5), but WHR GRS was not as strongly correlated across regions. For example, in men from Qingdao, a unit higher BMI GRS was strongly associated with higher BMI. [beta: 0.57, 95% confidence intervals (95% CI): 0.45, 0.70]But WHR GRS was not associated with higher WHR [beta: 0.25, 95% CI: 0.03, 0.47].

Observationally, higher adiposity was associated with lower odds of depressive symptoms in CKB

Observationally, higher BMI was associated with lower odds of depressive symptoms (Table 2). For example, one-standard deviation (1-SD) (3.50 kg/m²) was associated with higher BMI 0.88 [95% CI: 0.85, 0.91] All individuals had lower odds of depressive symptoms (Table 2, Figure 1). The observed associations were similar in men and women and when stratified by urban and rural areas (Table 2, Fig. 1). Adjusting the observational analyzes for SES and smoking status slightly reduced the effect estimates towards the null, but high adiposity remained inversely associated with depressive symptoms (Additional file 2 : Table S6).

Using a binary obesity measure as a predictor in these models did not change our findings. Observationally, overweight and obese individuals were at 0.87 [95% CI: 0.80, 0.94] and 0.84 [95% CI: 0.74, 0.95] Relatively low odds of depressive symptoms compared to normal BMI (Additional file 2 : Table S7).

Similar results were seen when WHR was considered, with 0.91 being associated with a 1-SD higher WHR (increase of 0.07). [95% CI: 0.88, 0.94] All individuals had lower odds of depressive symptoms (Table 2, Additional file 3 : Figure S1).

MR analyzes provide evidence that higher BMI leads to less depression in people of East Asian descent

Two-sample MR in people of East Asian ancestry using GWAS summary statistics

Two-sample MR provided evidence that higher genetically instrumented BMI was associated with lower odds of depression. Results were consistent for the comprehensive search outcome and the clinical depression outcome, but not with the symptom-based measure of depression where the effect estimate was zero (Figure 2, Additional file 2 : Table S8). For example, using the clinical depression definition, a unit higher genetically instrumented BMI was associated with an odds ratio for depression of 0.96. [95% CI: 0.93, 0.98].

Genetically instrumented WHR was associated with reduced odds of both comprehensive and clinical depression outcomes, but confidence intervals exceeded zero (Figure 2 , Additional file 2 : Table S8).

Results using more pleiotropy resistant methods were directionally consistent, but all confidence intervals exceeded zero (Additional file 2 : Table S9). There was no strong evidence of horizontal pleiotropy with MR-Egger (Additional file 2 : Table S9).

A specimen MR in China Kadoorie Biobank

A one-sample MR in 100,377 individuals of East Asian ancestry provided evidence for a causal role of high adiposity in depressive symptoms (Table 2). Across all individuals, a 1-SD higher BMI was genetically associated with 0.81. [95% CI: 0.66, 0.99] Lower odds of depressive symptoms (Table 2, Figure 2).

One-sample MR methods also provided circumstantial evidence that higher WHR was associated with lower odds of depressive symptoms. For example, a genetically determined 1-SD higher WHR was associated with 0.74. [95% CI: 0.55, 1.00] All individuals had lower odds of depressive symptoms (Table 2, Additional file 3 : Figure S1).

Differential effects were observed when stratifying by international residence

Using summary data from five meta-analyses of depression, we were able to compare the association of high BMI with depression among people of East Asian ancestry living in East Asia (mainly China) with those of similar ancestry but living in any country. Be it Britain or America. Here, as in the main analyses, a genetic-instrument unit was associated with a higher BMI of 0.97. [95% CI: 0.94, 0.99] Lower likelihood of depression in people living in East Asia. However, this was reduced to zero in people of East Asian ancestry living in the UK and USA (Figure 2 , Additional file 2 : Table S8).

When controlling for WHR the results were similar, with higher genetic WHR being associated with lower odds of depression in people of East Asian ancestry living in East Asia, but not in those living in the UK or USA (Figure 2, Additional file 2 : Table S8).

Differential effects were observed when stratifying by gender and place of residence within China

We were able to test whether gender and place of residence moderated the association between high BMI and depressive symptoms within China using data from the CKB.

As observed, associations were similar in men and women and when stratified by urban and rural areas, higher BMI and higher WHR were consistently associated with lower odds of depressive symptoms (Figure 1, Table 2 and Additional file 2 : Table S6). Categorizing individuals with normal or obese BMI provided consistent results, with depressive symptoms (OR: 0.74, 95% CI: 0.58, 0.95 and OR: 0.86, 95% CI: 0.75, 1.00 with obese men and women, respectively. ; Additional file 2 : Table S7).

One-sample MR methods demonstrated that the inverse association between BMI and depressive symptoms was stronger in men than in women (P_sex= 0.024; Table 2, Figure 1). Here, a 1-SD higher BMI was genetically associated with 0.53 [95% CI: 0.35, 0.81] Lower odds of depressive symptoms in men. In contrast, among women, there was no conclusive evidence of an inverse association between BMI and depressive symptoms (OR: 0.92, 95% CI: 0.73, 1.17; Table 2 ).

Comparing all individuals from urban and rural areas, genetic analyzes suggested that higher BMI was associated with lower odds of depressive symptoms only in rural areas (Table 2, Fig. 1), although the overall estimate for urban areas was in a positive direction. The confidence interval exceeded zero (OR: 1.10 95% CI: 0.76, 1.59; P_area= 0.049).

In urban areas, stronger sex-specific effects were observed, with higher genetic BMI associated with lower odds of depressive symptoms only in men (OR: 0.43 [95%CI: 0.19, 0.95]; Table 2, Figure 1). In contrast, one-sample MR estimates from urban women were in the opposite direction (P_sex= 0.009), associated with higher odds of depressive symptoms with higher BMI, although confidence intervals exceeded zero (OR: 1.42, 95% CI: 0.94, 2.16).

Furthermore, there was evidence that higher BMI was associated with lower odds of depressive symptoms among all individuals in rural areas. A genetically determined 1-SD higher BMI was associated with 0.70 [95% CI: 0.55, 0.90] Lower odds of depressive symptoms (Table 2, Figure 1). Estimates were consistent in both men (OR: 0.57, 95% CI: 0.35, 0.94) and women (OR: 0.75, 95% CI: 0.56, 1.00) with higher BMI associated with lower odds of depressive symptoms.

Analyzes were performed within each of the 10 regions in the CKB, showing higher variation between estimates in rural areas, particularly for female-only analyses, although P_{heterogeneity}> 0.05 in all cases (I²= 49.3%; Additional file 2 : Table S10). At the regional level, Suzhou (P= 0.03) and Henan women (P= 0.004, Additional file 3 : Figure S2), with the outcome of major depression, we observed an inverse relationship with BMI in Henan women.P= 0.03) and men from Zhejiang (P= 0.03, Additional file 3 : Figure S3).

Findings for WHR were less conclusive with directionally consistent results observed in men and women but wide confidence intervals that exceeded zero, limiting the interpretation of our findings (Table 2 , Additional file 3 : Fig. S1). Similarly, there was no conclusive evidence for an association between WHR and depressive symptoms in individuals from urban or rural areas (Table 2 , Additional file 3 : Fig. S1).

Secondary analysis

Two-sample MR analysis

We performed several two-sample MR sensitivity analyzes and observed the following results:

1. 1.

   80 BMI genetic variants identified in a Japanese population using repeated BMI two-sample MR analysis. Here, our findings were directionally consistent, with higher BMI associated with lower odds of depression, although confidence intervals exceeded zero (Additional file 2 : Table S11).

2. 2.

   Except for SNPs where the minor allele frequency was close to 0.50 (> 0.45 and < 0.55) compared using all SNPs (Additional file 2 : Table S2; Additional file 3 : Figs S4-S13) provided consistent results.

3. 3.

   Allele uptake between people of European and East Asian ancestry was consistent except for SNPs with different characteristics, although confidence intervals were wide (Additional file 2 : Table S3; Additional file 3 : Figs S4-S13).

4. 4.

   removing MC4RA SNP that appeared outlier and is in a significant BMI-related gene in people of European ancestry also did not change our results (Additional file 2 : Table S4).

A sample MR analysis

We performed several secondary one-sample MR analyzes on CKB and observed the following results:

1. 1.

   We tested for evidence of non-linearity between BMI and depressive symptoms in CKB, using the CF approach. This provided some evidence for a non-linear relationship in urban-dwelling individuals, particularly in women, suggesting that both high and low BMI are associated with higher odds of depressive symptoms (Figure 3a, Additional file 2 : Table S12). There was no evidence of a non-linear relationship in individuals from rural areas (Figure 3b, Additional file 2 : Table S12).

2. 2.

   Analyzes were repeated with a more stringent measure of depression: major depression defined using the CIDI-SF. Although wide confidence intervals were observed, both observational and genetic one-sample results were directional, consistent with the main analyzes (Additional file 2 : Table S13).

3. 3.

   Finally, we repeated the genetic one-sample analysis excluding Qingdao because the prevalence of depressive symptoms in this region was <1% in both men (0.55%) and women (0.81%). The results were consistent when Qingdao was excluded (Additional file 2 : Table S14).