Abstract
Background
Observational studies have demonstrated a strong bidirectional association between frailty and depression, but it remains unclear whether this association reflects causality. This study aimed to examine the bidirectional causal relationship between frailty and depression.
Methods
Using genome‐wide association study summary data, two‐sample Mendelian randomization was performed to test for the potential bidirectional causality between frailty, as defined by both the frailty index and the frailty phenotype, and depression. Several frailty‐related traits were additionally investigated, including weaker hand grip strength, slower walking pace and physical inactivity. Findings were replicated using an independent depression data source and verified using multiple sensitivity analyses.
Results
Genetically predicted higher frailty index (odds ratio [OR], 1.86; P < 0.001), higher frailty phenotype score (OR, 2.79; P < 0.001), lower grip strength (OR, 1.23; P = 0.003), slower walking pace (OR, 1.55; P = 0.027) and physical inactivity (OR, 1.44; P = 0.003) all were associated with a higher risk of depression. As for the reverse direction, genetic liability to depression showed consistent associations with a higher frailty index (beta, 0.167; P < 0.001) and a higher frailty phenotype score (beta, 0.067; P = 0.001), but not with other frailty‐related traits that were investigated. The results were stable across sensitivity analyses and across depression datasets.
Conclusions
Our findings add novel evidence supporting the bidirectional causal association between frailty and depression. Improving balance and muscle strength and increasing physical activity may be beneficial in both depression and frailty.
Keywords: aging, depression, frailty, Mendelian randomization
Introduction
The condition of frailty is gaining increasing attention as the aging population rapidly expands. Frailty is generally recognized as a state characterized by loss of resistance to stressors and declined functional capacity across multiple physiological systems. 1 It has been linked to a range of adverse outcomes including poor quality of life and increased morbidity and mortality. 1 Growing evidence from epidemiological studies suggests a strong bidirectional association between frailty and depression, with each condition associated with a higher prevalence and incidence of the other. 2 , 3 , 4 Genome‐wide association studies (GWASs) of frailty further indicated a role of mental health and highlighted pathways related to brain function in aging. 5 , 6 , 7 However, the direct causal association between frailty and depression has been scarcely explored at a population level to our knowledge. There is still uncertainty as to whether a bidirectional causal association exists, or alternatively, whether the co‐existence is due to confounding or common risk factors such as aging. Establishing the causality between frailty and depression is crucial, as this would help to understand the disease aetiology, inform the development of effective interventions and ultimately reduce the increasing disease burden.
Mendelian randomization (MR) is a powerful approach that can help to infer causal relationships by leveraging genetic variants as instruments. 8 MR takes advantage of the random allocation of genetic variants at conception and therefore is less subject to confounding and reverse causation than typical observational studies. 8 This method has been performed in studies looking into the causal associations between frailty and multiple age‐related diseases. 9 , 10 , 11 , 12 Here, we conducted a bidirectional MR study to elucidate the potential causal association between frailty, as defined by both the frailty index and the frailty phenotype, and the risk of depression and vice versa. Several frailty‐related traits were additionally examined, including weak hand grip strength, slow walking pace and physical inactivity.
Methods
Study design
Figure 1 A shows an overview of the study design. This study used non‐overlapping GWAS summary data within a standard two‐sample MR framework. We first explored the bidirectional causal relationships between frailty‐related traits and depression using depression data from the Psychiatric Genomics Consortium (PGC). Significant associations were then replicated using depression data from the FinnGen study. The study adhered to the STROBE‐MR (Strengthening the Reporting of Observational Studies in Epidemiology using Mendelian Randomization) guidelines in terms of reporting and analytical processes. 13 All data used in this study are publicly available, and ethical approval and informed consent have been obtained by the original studies.
Figure 1.
Schematic overview of study design. (A) A two‐sample bidirectional Mendelian randomization (MR) framework. (B) Three assumptions in the MR setting. PGC, Psychiatric Genomics Consortium; PRESSO, Pleiotropy RESidual Sum and Outlier.
Data sources
Frailty
Summary data for the frailty index were retrieved from the most recent GWAS meta‐analysis of the UK Biobank and TwinGene, including 175 226 individuals of European ancestry. 6 The frailty index is a commonly used definition for frailty that is based on the accumulation of age‐related deficits. 14 According to this phenotype, frailty is measured on a continuous score between 0 and 1, calculated by summing up the number of health deficits and dividing it by the total number of deficits measured (at least 30), covering signs, symptoms, disabilities and diseases. 14
Summary data for the frailty phenotype were obtained from a large‐scale GWAS with 386 565 participants of European ancestry enrolled in the UK Biobank. 7 The frailty phenotype is another dominant frailty assessment instrument, consisting of five criteria (weight loss, exhaustion, low physical activity, slow walking speed and weak grip strength). 15 In the original GWAS, the frailty phenotype was analysed as an ordinal score ranging from 0 to 5 to reduce information loss and improve statistical power, rather than as a usually modelled binary variable with ≥3 criteria defined as frailty.
It is worth noting that there is a lack of consensus regarding the definition of frailty, with ongoing debates about the conceptual framework to be measured. Despite this, the frailty index and the frailty phenotype are two most known operational definitions of frailty that have been validated as good predictors of many adverse outcomes in large‐scale cohorts. Although the frailty index and the frailty phenotype have distinct conceptual underpinnings, they share commonalities in terms of determinants and identification of frailty.
Frailty‐related traits
Weak grip strength, slow walking pace and physical inactivity, which are major components of the frailty phenotype, were investigated as secondary phenotypes. Summary data for grip strength were obtained from a recently conducted GWAS in 334 925 participants of European ancestry from the UK Biobank. 16 Grip strength was measured using the Jamar J00105 dynamometer device for both hands in turn. Due to the high correlation between grip strength and body size, relative grip strength was calculated as an average of both hand measurements divided by weight. Summary data for walking pace were obtained from a GWAS in the UK Biobank, comprising 450 967 participants of European ancestry. 17 Walking pace was assessed using the self‐reported electronic questionnaire asking, ‘How would you describe your usual walking pace?’ Participants whose answers were ‘slow’, ‘steady/average’ or ‘brisk’ were classified for analyses. Summary data for physical activity were acquired from the largest GWAS meta‐analysis of self‐reported moderate‐to‐vigorous physical activity during leisure time (hereafter referred to as physical activity), which includes 661 399 European ancestry participants from up to 50 studies (nearly 0.5 million from the UK Biobank). 18 Owing to the different questionnaires used to capture physical activity across studies and binomial nature of the distribution of physical activity, physical activity was analysed as a dichotomous phenotype.
Depression
Given that the GWAS summary data for frailty and related traits were mainly derived from the UK Biobank, we used GWAS summary data for major depressive disorder (MDD) released by the PGC with the UK Biobank removed to maintain the two‐sample design. 19 We extracted genetic instruments for MDD using the data with 135 458 cases and 344 901 controls of European ancestry, when MDD was treated as the exposure. Owing to the restricted access to the 23andMe data, GWAS summary data without 23andMe were used, including 59 851 cases and 113 154 controls, when MDD was taken as the outcome. The studies included in the PGC used a range of methods for assessing MDD. All cases met standard criteria for a lifetime diagnosis of MDD primarily through structured assessments by trained interviewers, clinician‐administered checklists or medical record review. 19 For replication, we used the independent GWAS summary data for depression (F5_DEPRESSIO) from the Release 8 of the FinnGen study, including 38 225 patients and 299 886 control subjects. Depression was defined using the International Classification of Diseases.
All genetic associations have been adjusted for age, sex and principal components in the above GWASs.
Instrument selection and data harmonization
To select genetic instruments, genome‐wide significant (P < 5 × 10−8) single‐nucleotide polymorphisms (SNPs) strongly associated with each studied phenotype were used. SNPs were then clumped for independence by linkage disequilibrium (r 2 < 0.001 within 10 000‐kb clumping distance) using the European reference panel of the 1000 Genome Project. Where SNPs for the exposure were not available in the outcome datasets, we replaced them with SNPs in high linkage disequilibrium (r 2 > 0.8). We further removed ambiguous SNPs with intermediate effect allele frequencies (>0.42). SNP effects were harmonized between the exposure and the outcome data by alleles. Detailed information on SNPs used as instruments is provided in Table S1 .
Statistical analysis
For the primary analysis, we calculated the Wald ratio for each tested SNP and combined them using the multiplicative random‐effect inverse‐variance weighted (IVW) approach to obtain the MR estimates. 20 The IVW method provides the most precise and robust estimates when three pivotal assumptions regarding instruments are satisfied. These assumptions are as follows: Genetic variants (1) are closely associated with the exposure (relevance assumption), (2) are not associated with confounders (independence assumption) and (3) influence the outcome only through the exposure (exclusion restriction assumption, known as the ‘no pleiotropy’) (Figure 1 B ). We tested the three MR assumptions using the following approaches.
To validate relevance assumption, the proportion of variance (R 2) in each exposure explained by SNPs and F‐statistics were calculated to quantify the strength of instruments. 21 An F‐statistic > 10 suggested a sufficiently strong instrument. For independence assumption, we considered smoking and obesity as major confounders of the association between frailty and depression. We therefore carried out a sensitivity analysis with additional exclusion of SNPs showing genome‐wide significant associations with smoking‐ (i.e., smoking status and lifetime smoking) and obesity‐related phenotypes (i.e., body mass index, waist‐to‐hip ratio and body fat percentage) by manually searching the GWAS catalogue. 22 The exclusion restriction assumption was tested using pleiotropy‐robust methods, including weighted median, 23 MR‐Egger regression 24 and MR‐PRESSO (Pleiotropy RESidual Sum and Outlier). 25 These tests vary in their assumptions such that a consistent effect across multiple approaches should be more robust against bias from horizonal pleiotropy.
MR‐Egger intercept test and Cochran's Q statistic were adopted to check the evidence of horizontal pleiotropy and heterogeneity, respectively. Leave‐one‐out analysis was implemented to examine whether overall estimates were driven by a single SNP. As some bidirectional associations were observed, we further used the Steiger filtering, which removed SNPs explaining more variance of the outcome than the exposure to ensure the correct direction of inferred causal associations. 26
Estimates are reported as beta where the outcome was continuous/ordinal (e.g., frailty index) and as odds ratio (OR) where the outcome was binary/dichotomous (e.g., depression). All analyses were conducted using the ‘TwoSampleMR’ and ‘MRPRESSO’ packages in R software (Version 4.1.1). Statistical power for MR analyses was estimated using the mRnd webtool. 27 To address multiple testing, a conservative Bonferroni‐corrected threshold (P < 0.005) by the IVW method was adopted to indicate statistical significance, because we assessed the associations between five frailty‐related traits and depression in both directions.
Results
The number of SNPs used as genetic instruments ranged from 15 (the frailty index) to 87 (grip strength), explaining 0.34–1.79% of the phenotypic variance. F‐statistics for all studied phenotypes are >30, suggesting the good strength of genetic instruments (Table S2 ). Most tested associations are well powered, having >80% power to detect a minimum beta/OR ranging from 0.033/1.03 to 0.315/1.37 (Table S3 ).
Using the PGC data, genetically predicted higher frailty index (OR, 1.86; P < 0.001), higher frailty phenotype score (OR, 2.79; P < 0.001), lower grip strength (OR, 1.23; P = 0.003) and physical inactivity (OR, 1.44; P = 0.003) were significantly associated with a higher risk of MDD (Table 1 ). We also found that genetically predicted slower walking pace (OR, 1.55; P = 0.027) was suggestively associated with MDD risk, although the association did not survive multiple comparisons. In the reverse direction, genetic liability to MDD showed significant associations with a higher frailty index (beta, 0.167; P < 0.001) and a higher frailty phenotype score (beta, 0.067; P = 0.001), but not with other frailty‐related traits (Table 1 ). These results remained significant using pleiotropy‐robust methods including weighted median and MR‐PRESSO, with the exception for MR‐Egger. This is consistent with the lower precision of MR‐Egger than other methods. Although heterogeneity was detected by Cochran's Q statistic for some associations, the MR‐Egger intercept showed limited influence of horizontal pleiotropy in any of the analyses.
Table 1.
Mendelian randomization results for the bidirectional associations between five frailty‐related phenotypes and major depressive disorder using the depression data from the Psychiatric Genomics Consortium
| Exposure | Outcome | Method | Beta (95% CI) a | OR (95% CI) b | P value | P value for Cochran's Q/Egger intercept |
|---|---|---|---|---|---|---|
| Frailty index | MDD | IVW | 0.621 (0.312, 0.929) | 1.86 (1.37, 2.53) | <0.001 c | 0.137 |
| Weighted median | 0.554 (0.190, 0.917) | 1.74 (1.21, 2.50) | 0.003 | |||
| MR‐Egger | 0.598 (−2.179, 3.374) | 1.82 (0.11, 29.19) | 0.681 | 0.708 | ||
| MR‐PRESSO | 0.621 (0.312, 0.929) | 1.86 (1.37, 2.53) | 0.002 | |||
| Frailty phenotype | MDD | IVW | 1.027 (0.594, 1.461) | 2.79 (1.81, 4.31) | <0.001 c | <0.001 |
| Weighted median | 0.846 (0.404, 1.288) | 2.33 (1.50, 3.62) | <0.001 | |||
| MR‐Egger | 0.946 (−0.851, 2.743) | 2.58 (0.43, 15.54) | 0.310 | 0.928 | ||
| MR‐PRESSO | 1.100 (0.722, 1.477) | 3.00 (2.06, 4.38) | <0.001 | |||
| Weak grip strength | MDD | IVW | 0.207 (0.069, 0.344) | 1.23 (1.07, 1.41) | 0.003 c | <0.001 |
| Weighted median | 0.235 (0.068, 0.401) | 1.26 (1.07, 1.49) | 0.006 | |||
| MR‐Egger | 0.489 (−0.117, 1.095) | 1.63 (0.89, 2.99) | 0.118 | 0.351 | ||
| MR‐PRESSO | 0.182 (0.052, 0.312) | 1.20 (1.05, 1.37) | 0.008 | |||
| Slow walking pace | MDD | IVW | 0.438 (0.051, 0.826) | 1.55 (1.05, 2.28) | 0.027 | <0.001 |
| Weighted median | 0.675 (0.223, 1.126) | 1.96 (1.25, 3.08) | 0.003 | |||
| MR‐Egger | 0.552 (−1.155, 2.260) | 1.74 (0.31, 9.56) | 0.529 | 0.894 | ||
| MR‐PRESSO | 0.438 (0.051, 0.826) | 1.55 (1.05, 2.28) | 0.031 | |||
| Physical inactivity | MDD | IVW | 0.362 (0.122, 0.603) | 1.44 (1.13, 1.83) | 0.003 c | 0.197 |
| Weighted median | 0.500 (0.196, 0.805) | 1.65 (1.22, 2.24) | 0.001 | |||
| MR‐Egger | 0.609 (−0.725, 1.942) | 1.85 (0.48, 6.97) | 0.390 | 0.720 | ||
| MR‐PRESSO | 0.362 (0.122, 0.603) | 1.44 (1.13, 1.83) | 0.012 | |||
| MDD | Frailty index | IVW | 0.167 (0.110, 0.223) | 1.18 (1.12, 1.25) | <0.001 c | <0.001 |
| Weighted median | 0.185 (0.124, 0.246) | 1.20 (1.13, 1.28) | <0.001 | |||
| MR‐Egger | 0.041 (−0.235, 0.317) | 1.04 (0.79, 1.37) | 0.772 | 0.370 | ||
| MR‐PRESSO | 0.167 (0.110, 0.223) | 1.18 (1.12, 1.25) | <0.001 | |||
| MDD | Frailty phenotype | IVW | 0.067 (0.026, 0.107) | 1.07 (1.03, 1.11) | 0.001 c | <0.001 |
| Weighted median | 0.081 (0.042, 0.121) | 1.08 (1.04, 1.13) | <0.001 | |||
| MR‐Egger | 0.076 (−0.123, 0.275) | 1.08 (0.88, 1.32) | 0.460 | 0.926 | ||
| MR‐PRESSO | 0.086 (0.049, 0.123) | 1.09 (1.05, 1.13) | <0.001 | |||
| MDD | Grip strength | IVW | −0.009 (−0.045, 0.028) | 0.99 (0.96, 1.03) | 0.643 | <0.001 |
| Weighted median | −0.010 (−0.040 0.019) | 0.99 (0.96, 1.02) | 0.486 | |||
| MR‐Egger | −0.113 (−0.287, 0.062) | 0.89 (0.75, 1.06) | 0.216 | 0.242 | ||
| MR‐PRESSO | −0.018 (−0.045, 0.009) | 0.98 (0.96, 1.01) | 0.211 | |||
| MDD | Walking pace | IVW | −0.007 (−0.035, 0.021) | 0.99 (0.97, 1.02) | 0.604 | <0.001 |
| Weighted median | −0.016 (−0.041, 0.009) | 0.98 (0.96, 1.01) | 0.206 | |||
| MR‐Egger | −0.001 (−0.139, 0.135) | 1.00 (0.87, 1.14) | 0.979 | 0.936 | ||
| MR‐PRESSO | −0.024 (−0.046, 0.001) | 0.98 (0.95, 1.00) | 0.047 | |||
| MDD | Physical activity | IVW | −0.001 (−0.081, 0.080) | 1.00 (0.92, 1.08) | 0.982 | <0.001 |
| Weighted median | 0.001 (−0.072, 0.075) | 1.00 (0.93, 1.08) | 0.977 | |||
| MR‐Egger | 0.053 (−0.357, 0.462) | 1.05 (0.70, 1.59) | 0.803 | 0.796 | ||
| MR‐PRESSO | −0.027 (−0.102, 0.047) | 0.97 (0.90, 1.05) | 0.482 |
Abbreviations: CI, confidence interval; IVW, inverse‐variance weighted; MDD, major depressive disorder; MR, Mendelian randomization; OR, odds ratio; PRESSO, Pleiotropy RESidual Sum and Outlier.
Betas are presented for the analyses of continuous/ordinal outcomes (i.e., frailty index, frailty phenotype, grip strength and walking pace).
ORs are presented for the analyses of binary/dichotomous outcomes (i.e., MDD and physical activity).
Significant associations with P < 0.005 after multiple testing.
Replication analysis broadly supported these findings, albeit with relatively wide confidence intervals, which are probably due to the smaller number of cases with depression in the FinnGen study than the PGC (Table 2 ). Sensitivity analysis excluding SNPs associated with smoking‐ and/or obesity‐related phenotypes did not change the pattern of the primary findings (Table S4 ). There were not any SNPs removed by the Steiger filtering, indicating the correct orientation of the inferred relationships. Leave‐one‐out analysis showed that no single SNPs have substantial impact on the overall results (data not shown).
Table 2.
Mendelian randomization results for the replication analyses using the depression data from the FinnGen study
| Exposure | Outcome | Method | Beta (95% CI) a | OR (95% CI) b | P value | P value for Cochran's Q/Egger intercept |
|---|---|---|---|---|---|---|
| Frailty index | Depression | IVW | 0.605 (0.287, 0.923) | 1.83 (1.33, 2.52) | <0.001 | 0.002 |
| Weighted median | 0.683 (0.374, 0.993) | 1.98 (1.45, 2.07) | <0.001 | |||
| MR‐Egger | 1.070 (−0.400, 2.540) | 2.92 (0.67, 12.68) | 0.179 | 0.537 | ||
| MR‐PRESSO | 0.605 (0.287, 0.923) | 1.83 (1.33, 2.52) | 0.003 | |||
| Frailty phenotype | Depression | IVW | 0.945 (0.434, 1.455) | 2.57 (1.54, 4.28) | <0.001 | <0.001 |
| Weighted median | 0.553 (0.057, 1.049) | 1.74 (1.06, 2.85) | 0.029 | |||
| MR‐Egger | 1.492 (−0.814, 3.798) | 4.44 (0.44, 44.60) | 0.217 | 0.637 | ||
| MR‐PRESSO | 0.733 (0.312, 1.154) | 2.08 (1.37, 3.17) | 0.002 | |||
| Weak grip strength | Depression | IVW | 0.119 (−0.023, 0.260) | 1.13 (0.98, 1.30) | 0.100 | <0.001 |
| Weighted median | 0.097 (−0.055, 0.249) | 1.10 (0.95, 1.28) | 0.209 | |||
| MR‐Egger | 0.473 (−0.175, 1.121) | 1.60 (0.84, 3.07) | 0.157 | 0.276 | ||
| MR‐PRESSO | 0.120 (−0.014, 0.255) | 1.13 (0.99, 1.29) | 0.084 | |||
| Slow walking pace | Depression | IVW | 0.619 (0.268, 0.971) | 1.86 (1.31, 2.64) | 0.001 | <0.001 |
| Weighted median | 0.338 (−0.067, 0.753) | 1.40 (0.93, 2.12) | 0.110 | |||
| MR‐Egger | −0.421 (−2.241, 1.399) | 0.66 (0.11, 4.05) | 0.652 | 0.259 | ||
| MR‐PRESSO | 0.619 (0.268, 0.971) | 1.86 (1.31, 2.64) | 0.001 | |||
| Physical inactivity | Depression | IVW | 0.284 (−0.006, 0.574) | 1.33 (0.99, 1.78) | 0.055 | 0.002 |
| Weighted median | 0.234 (−0.042, 0.511) | 1.26 (0.96, 1.67) | 0.097 | |||
| MR‐Egger | −0.004 (−1.699, 1.691) | 1.00 (0.18, 5.42) | 0.996 | 0.740 | ||
| MR‐PRESSO | 0.189 (−0.054, 0.433) | 1.21 (0.95, 1.54) | 0.153 | |||
| Depression | Frailty index | IVW | 0.117 (0.031, 0.203) | 1.12 (1.03, 1.23) | 0.008 | 0.008 |
| Weighted median | 0.057 (−0.030, 0.144) | 1.06 (0.97, 1.15) | 0.201 | |||
| MR‐Egger | 0.371 (−0.900, 1.632) | 1.45 (0.41, 5.16) | 0.592 | 0.711 | ||
| MR‐PRESSO | 0.090 (0.003, 0.177) | 1.09 (1.00, 1.19) | 0.097 | |||
| Depression | Frailty phenotype | IVW | 0.084 (0.050, 0.118) | 1.09 (1.05, 1.13) | <0.001 | 0.140 |
| Weighted median | 0.074 (0.035, 0.112) | 1.08 (1.04, 1.12) | <0.001 | |||
| MR‐Egger | 0.165 (−0.151, 0.481) | 1.18 (0.86, 1.62) | 0.345 | 0.631 | ||
| MR‐PRESSO | 0.084 (0.050, 0.118) | 1.09 (1.05, 1.13) | 0.002 |
Abbreviations: CI, confidence interval; IVW, inverse‐variance weighted; MR, Mendelian randomization; OR, odds ratio; PRESSO, Pleiotropy RESidual Sum and Outlier.
Betas are presented for the analyses of continuous outcomes (i.e., frailty index).
ORs are presented for the analyses of binary outcomes (i.e., depression).
Discussion
Using large‐scale GWAS summary data within an MR framework, we observed robust associations of genetically predicted higher frailty index, higher frailty phenotype score, lower grip strength, slower walking pace and physical inactivity with an increased risk of depression. In contrast, genetic liability to depression was robustly associated with a higher frailty index and a higher frailty phenotype score, but not with other frailty‐related traits being examined. To our knowledge, this study is the first one to apply the MR technology to evaluate the bidirectional causal association between frailty and depression.
Previous systematic reviews and meta‐analyses have shown a strong bidirectional relationship between depression and frailty among older adults, but the evidence is mostly drawn from cross‐sectional and case–control studies. 2 , 3 , 4 Most existing studies have focused on investigating only one direction (either from depression to frailty or from frailty to depression), with very few examining the bidirectional association simultaneously, especially in a population‐based prospective design. Cao et al. found that pre‐frail and frail people, as defined by the frailty index and phenotype, are at a higher risk of developing depressive symptoms compared with those who are robust, based on a cohort study of 5303 older Chinese adults. 28 However, no effect of depressive symptoms on subsequent incidence of frailty was observed. Another 5‐year follow‐up study with 167 729 participants living in Netherlands demonstrated that depression and anxiety are reciprocally associated with frailty, as measured by the frailty index, in both younger and older adults. 29 Our MR results supported a robust bidirectional causal association between frailty and depression, comparably using the frailty index and the frailty phenotype and in two independent datasets. Notably, the data used in this study are solely from participants of European ancestry. The inconsistent findings with the Chinese study may be partially explained by ethnic/racial differences.
Our study also revealed that lower grip strength, slower walking pace and physical inactivity have detrimental causal effects on depression, which confirms previous MR findings. 30 , 31 Bidirectional analyses suggested that these associations are more likely to be unidirectional, rather than bidirectional, as documented in some observational studies. 32 , 33 , 34 According to the frailty phenotype, depression seems to increase physical frailty symptoms, but not the individual component of the frailty phenotype. This could be due to depression being more closely related to impairments of whole‐body physical fitness and function, as opposed to limited functional activity captured by a single measure. Another possible explanation is the overlap in some diagnostic criteria, such as fatigue and unintentional weight loss, which may drive the association between depression and the frailty phenotype. It is important to note that standard MR analysis assumes a linear exposure–outcome association, which may have limited our ability to detect potential non‐linear associations with these secondary phenotypes. For example, a recent cohort study in 13 208 Chinese participants identified an L‐shaped dose–response association between depressive symptoms and subsequent decreased grip strength. 32
Several mechanisms have been proposed to support the bidirectional association between frailty and depression. Chronic inflammation, oxidative stress and mitochondrial dysfunction may act as the overlapping biological pathways in both directions. 4 Additionally, an unhealthy lifestyle, including cigarette smoking, excessive drinking, unbalanced diet and low physical activity, may contribute to this bidirectional association. 35 Notably, such a pattern of the bidirectional association is commonly observed in frailty and comorbidities. Previous MR studies also provided support for the bidirectional causal associations of frailty with various cardiometabolic diseases and hearing loss. 11 , 12
In terms of implications, our principal findings identified a bidirectional causal relationship between frailty and depression. This means that the beneficial effect of means to prevent or treat one of both conditions may protect against the other. As the present study found that lower grip strength, slower walking pace and physical inactivity causally contribute to depression and all these factors are potentially modifiable, improving balance and muscle strength and increasing physical activity may represent effective interventions for both depression symptoms and frailty.
The key strength of this study is the MR design, which diminishes unmeasured confounding and reverse causation and thus improves causal inference for the bidirectional association between frailty and depression. Another strength is the application of a discovery‐replication design to enhance the confidence. This study also has several limitations. First, we cannot interpret the estimates from MR studies in the same way as non‐genetic/classic observational studies, especially for binary exposures (e.g., depression). 36 Nevertheless, MR remains a robust method to test the null causal hypothesis. Therefore, the main purpose of this study was to determine whether a causal relationship exists rather than to estimate the magnitude of the effect. Second, SNPs associated with frailty‐related traits were mainly identified from the UK Biobank, which is a non‐probability sample and subject to ‘healthy volunteer’ selection bias (in which participants tend to be more health conscious than the average). 37 It is unclear whether these SNPs are representative in other populations. Further MR studies using genetic instruments identified from a more representative sample to validate our findings would be desirable. Third, although multiple sensitivity analyses were performed to detect and correct for horizontal pleiotropy (a major concern for MR), such bias cannot be fully ruled out, because horizontal pleiotropies may widely spread across the genome. Fourth, it should be noted that depression is one of the components included in the frailty index. Due to the use of summary data, we were unable to conduct a sensitivity analysis that excluded depression‐related variables. However, we believe this would not have a substantial impact on our analysis. The GWAS of the frailty index consisted of only a small percentage (approximately 2%) of participants who reported depression and/or anxiety. Moreover, items that may be directly related to depression accounted for only a portion of the definition of the frailty index, with 7 out of 49 items in the UK Biobank and 3 out of 44 items in the TwinGene being related to depression. Finally, as previously mentioned, all participants in our study were of European ancestry, thus limiting generalizability of our findings to other ethnicities.
In conclusion, this study supports a bidirectional causal association between depression and frailty. Our data also suggest causal effects of weaker grip strength, slower walking pace and physical inactivity on depression. These findings may open important perspectives for the development of intervention strategies to lower the burden of high co‐existence of depression and frailty.
Conflict of interest statement
All authors declare that they have no conflict of interest.
Supporting information
Table S1. SNPs used as genetic instruments for each studied phenotype.
Table S2. Statistics for genetic instruments.
Table S3. Power calculation results.
Table S4. Results for the sensitivity analysis excluding SNPs associated with smoking‐ and/or obesity‐related phenotypesa.
Acknowledgements
This study was supported by the Key Discipline of Zhejiang Province in Public Health and Preventative Medicine (First Class, Category A), Hangzhou Medical College. We would like to thank the Psychiatric Genomics Consortium and other relevant consortiums and investigators for making their GWAS summary statistics publicly available.
Zhu J., Zhou D., Nie Y., Wang J., Yang Y., Chen D., et al (2023) Assessment of the bidirectional causal association between frailty and depression: A Mendelian randomization study, Journal of Cachexia, Sarcopenia and Muscle, 14, 2327–2334, 10.1002/jcsm.13319
Contributor Information
Min Yu, Email: myu@cdc.zj.cn.
Yingjun Li, Email: 2016034036@hmc.edu.cn.
Data availability statement
All data used in this study are publicly available.
References
- 1. Hoogendijk EO, Afilalo J, Ensrud KE, Kowal P, Onder G, Fried LP. Frailty: implications for clinical practice and public health. Lancet 2019;394:1365–1375. [DOI] [PubMed] [Google Scholar]
- 2. Buigues C, Padilla‐Sanchez C, Garrido JF, Navarro‐Martinez R, Ruiz‐Ros V, Cauli O. The relationship between depression and frailty syndrome: a systematic review. Aging Ment Health 2015;19:762–772. [DOI] [PubMed] [Google Scholar]
- 3. Vaughan L, Corbin AL, Goveas JS. Depression and frailty in later life: a systematic review. Clin Interv Aging 2015;10:1947–1958. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 4. Soysal P, Veronese N, Thompson T, Kahl KG, Fernandes BS, Prina AM, et al. Relationship between depression and frailty in older adults: a systematic review and meta‐analysis. Ageing Res Rev 2017;36:78–87. [DOI] [PubMed] [Google Scholar]
- 5. Mekli K, Stevens A, Marshall AD, Arpawong TE, Phillips DF, Tampubolon G, et al. Frailty Index associates with GRIN2B in two representative samples from the United States and the United Kingdom. PLoS ONE 2018;13:e0207824. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 6. Atkins JL, Jylhava J, Pedersen NL, Magnusson PK, Lu Y, Wang Y, et al. A genome‐wide association study of the frailty index highlights brain pathways in ageing. Aging Cell 2021;20:e13459. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 7. Ye Y, Noche RB, Szejko N, Both CP, Acosta JN, Leasure AC, et al. A genome‐wide association study of frailty identifies significant genetic correlation with neuropsychiatric, cardiovascular, and inflammation pathways. Geroscience 2023; Epub ahead of print. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 8. Smith GD, Ebrahim S. ‘Mendelian randomization’: can genetic epidemiology contribute to understanding environmental determinants of disease? Int J Epidemiol 2003;32:1–22. [DOI] [PubMed] [Google Scholar]
- 9. Liu W, Zhang L, Fang H, Gao Y, Liu K, Li S, et al. Genetically predicted frailty index and risk of stroke and Alzheimer's disease. Eur J Neurol 2022;29:1913–1921. [DOI] [PubMed] [Google Scholar]
- 10. Xiao G, Wang H, Hu J, Liu L, Zhang T, Zhou M, et al. Estimating the causal effect of frailty index on vestibular disorders: a two‐sample Mendelian randomization. Front Neurosci 2022;16:990682. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 11. Liu Y, Qian P, Guo S, Liu S, Wang D, Yang L. Frailty and hearing loss: from association to causation. Front Aging Neurosci 2022;14:953815. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 12. Zhu J, Zhou D, Wang J, Yang Y, Chen D, He F, et al. Frailty and cardiometabolic diseases: a bidirectional Mendelian randomisation study. Age Ageing 2022;51. [DOI] [PubMed] [Google Scholar]
- 13. Skrivankova VW, Richmond RC, Woolf BAR, Davies NM, Swanson SA, VanderWeele TJ, et al. Strengthening the reporting of observational studies in epidemiology using mendelian randomisation (STROBE‐MR): explanation and elaboration. BMJ 2021;375:n2233. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 14. Mitnitski AB, Mogilner AJ, Rockwood K. Accumulation of deficits as a proxy measure of aging. ScientificWorldJournal 2001;1:323–336. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 15. Fried LP, Tangen CM, Walston J, Newman AB, Hirsch C, Gottdiener J, et al. Frailty in older adults: evidence for a phenotype. J Gerontol A Biol Sci Med Sci 2001;56:M146–M156. [DOI] [PubMed] [Google Scholar]
- 16. Tikkanen E, Gustafsson S, Amar D, Shcherbina A, Waggott D, Ashley EA, et al. Biological insights into muscular strength: genetic findings in the UK Biobank. Sci Rep 2018;8:6451. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 17. Timmins IR, Zaccardi F, Nelson CP, Franks PW, Yates T, Dudbridge F. Genome‐wide association study of self‐reported walking pace suggests beneficial effects of brisk walking on health and survival. Commun Biol 2020;3:634. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 18. Wang Z, Emmerich A, Pillon NJ, Moore T, Hemerich D, Cornelis MC, et al. Genome‐wide association analyses of physical activity and sedentary behavior provide insights into underlying mechanisms and roles in disease prevention. Nat Genet 2022;54:1332–1344. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 19. Wray NR, Ripke S, Mattheisen M, Trzaskowski M, Byrne EM, Abdellaoui A, et al. Genome‐wide association analyses identify 44 risk variants and refine the genetic architecture of major depression. Nat Genet 2018;50:668–681. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 20. Burgess S, Butterworth A, Thompson SG. Mendelian randomization analysis with multiple genetic variants using summarized data. Genet Epidemiol 2013;37:658–665. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 21. Burgess S, Thompson SG, Collaboration CCG . Avoiding bias from weak instruments in Mendelian randomization studies. Int J Epidemiol 2011;40:755–764. [DOI] [PubMed] [Google Scholar]
- 22. Buniello A, MacArthur JAL, Cerezo M, Harris LW, Hayhurst J, Malangone C, et al. The NHGRI‐EBI GWAS Catalog of published genome‐wide association studies, targeted arrays and summary statistics 2019. Nucleic Acids Res 2019;47:D1005–D1012. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 23. Bowden J, Davey Smith G, Haycock PC, Burgess S. Consistent estimation in Mendelian randomization with some invalid instruments using a weighted median estimator. Genet Epidemiol 2016;40:304–314. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 24. Bowden J, Davey Smith G, Burgess S. Mendelian randomization with invalid instruments: effect estimation and bias detection through Egger regression. Int J Epidemiol 2015;44:512–525. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 25. Verbanck M, Chen CY, Neale B, Do R. Detection of widespread horizontal pleiotropy in causal relationships inferred from Mendelian randomization between complex traits and diseases. Nat Genet 2018;50:693–698. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 26. Hemani G, Tilling K, Davey Smith G. Orienting the causal relationship between imprecisely measured traits using GWAS summary data. PLoS Genet 2017;13:e1007081. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 27. Brion MJ, Shakhbazov K, Visscher PM. Calculating statistical power in Mendelian randomization studies. Int J Epidemiol 2013;42:1497–1501. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 28. Cao L, Zhou Y, Liu H, Shi M, Wei Y, Xia Y. Bidirectional longitudinal study of frailty and depressive symptoms among older Chinese adults. Front Aging Neurosci 2022;14:791971. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 29. Borges MK, Jeuring HW, Marijnissen RM, van Munster BC, Aprahamian I, van den Brink RHS, et al. Frailty and affective disorders throughout adult life: a 5‐year follow‐up of the Lifelines Cohort Study. J Am Geriatr Soc 2022;70:3424–3435. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 30. Choi KW, Chen CY, Stein MB, Klimentidis YC, Wang MJ, Koenen KC, et al. Assessment of bidirectional relationships between physical activity and depression among adults: a 2‐sample Mendelian randomization study. JAMA Psychiatry 2019;76:399–408. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 31. Li N, Zhou R, Zhang B. Handgrip strength and the risk of major depressive disorder: a two‐sample Mendelian randomisation study. Gen Psychiatr 2022;35:e100807. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 32. Lian Y, Wang GP, Chen GQ, Jia CX. Bidirectional associations between handgrip strength and depressive symptoms: a longitudinal cohort study. J Am Med Dir Assoc 2021;22:1744–1750.e1. [DOI] [PubMed] [Google Scholar]
- 33. Sanders JB, Bremmer MA, Deeg DJ, Beekman AT. Do depressive symptoms and gait speed impairment predict each other's incidence? A 16‐year prospective study in the community. J Am Geriatr Soc 2012;60:1673–1680. [DOI] [PubMed] [Google Scholar]
- 34. Zhang D, Pettee Gabriel K, Sidney S, Sternfeld B, Jacobs D Jr, Whitaker KM. Longitudinal bidirectional associations of physical activity and depressive symptoms: the CARDIA study. Prev Med Rep 2021;23:101489. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 35. Veronese N, Solmi M, Maggi S, Noale M, Sergi G, Manzato E, et al. Frailty and incident depression in community‐dwelling older people: results from the ELSA study. Int J Geriatr Psychiatry 2017;32:e141–e149. [DOI] [PubMed] [Google Scholar]
- 36. Burgess S, Labrecque JA. Mendelian randomization with a binary exposure variable: interpretation and presentation of causal estimates. Eur J Epidemiol 2018;33:947–952. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 37. Schoeler T, Speed D, Porcu E, Pirastu N, Pingault JB, Kutalik Z. Participation bias in the UK Biobank distorts genetic associations and downstream analyses. Nat Hum Behav 2023; Epub ahead of print. [DOI] [PMC free article] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.
Supplementary Materials
Table S1. SNPs used as genetic instruments for each studied phenotype.
Table S2. Statistics for genetic instruments.
Table S3. Power calculation results.
Table S4. Results for the sensitivity analysis excluding SNPs associated with smoking‐ and/or obesity‐related phenotypesa.
Data Availability Statement
All data used in this study are publicly available.