Effect of weight on depression using multiple genetic instruments

This is a self-archived version of an original article. This version may differ from the original in pagination and typographic details. Author(s): Title: Year: Version: Copyright: Rights: Rights url: Please cite the original version: CC BY 4.0 https://creativecommons.org/licenses/by/4.0/ Effect of weight on depression using multiple genetic instruments © 2024 the Authors Published version Viinikainen, Jutta; Böckerman, Petri; Willage, Barton; Elovainio, Marko; Kari, Jaana T.; Lehtimäki, Terho; Pehkonen, Jaakko; Pitkänen, Niina; Raitakari, Olli Viinikainen, J., Böckerman, P., Willage, B., Elovainio, M., Kari, J. T., Lehtimäki, T., Pehkonen, J., Pitkänen, N., & Raitakari, O. (2024). Effect of weight on depression using multiple genetic instruments. PLoS ONE, 19(2), Article e0297594. https://doi.org/10.1371/journal.pone.0297594 2024 RESEARCH ARTICLE Effect of weight on depression using multiple genetic instruments Jutta ViinikainenID 1 *, Petri Bo ¨ckerman 1,2,3 , Barton Willage 4 , Marko Elovainio 5,6 , Jaana T. KariID 1 , Terho Lehtima ¨ki 7,8,9 , Jaakko Pehkonen 1 , Niina Pitka ¨nen 10 , Olli Raitakari 10,11,12 1Jyva ¨skyla ¨University School of Business and Economics, University of Jyva ¨skyla ¨, Jyva ¨skyla ¨, Finland, 2Labour Institute for Economic Research LABORE, Helsinki, Finland, 3IZA Institute of Labor Economics, Bonn, Germany, 4Department of Economics, University of Colorado—Denver, Denver, Colorado, United States of America, 5Department of Psychology and Logopedics, University of Helsinki, Helsinki, Finland, 6Finnish Institute for Health and Welfare, Helsinki, Finland, 7Department of Clinical Chemistry, Fimlab Laboratories, Tampere, Finland, 8Faculty of Medicine and Health Technology, Tampere University, Tampere, Finland, 9Finnish Cardiovascular Research Center Tampere, Tampere University, Tampere, Finland, 10 Research Centre of Applied and Preventive Cardiovascular Medicine, University of Turku, Turku, Finland, 11 Centre for Population Health Research, University of Turku and Turku University Hospital, Turku, Finland, 12 Department of Clinical Physiology and Nuclear Medicine, Turku University Hospital, Turku, Finland *[email protected] Abstract A striking global health development over the past few decades has been the increasing prevalence of overweight and obesity. At the same time, depression has become increasingly common in almost all high-income countries. We investigated whether body weight, measured by body mass index (BMI), has a causal effect on depression symptoms in Finland. Using data drawn from the Cardiovascular Risk in Young Finns Study (N = 1,523, mean age 41.9, SD 5), we used linear regression to establish the relationship between BMI and depression symptoms measured by 21-item Beck’s Depression Inventory. To identify causal relationships, we used the Mendelian randomization (MR) method with weighted sums of genetic markers (single nucleotide polymorphisms, SNPs) as instruments for BMI. We employ instruments (polygenic risk scores, PGSs) with varying number of SNPs that are associated with BMI to evaluate the sensitivity of our results to instrument strength. Based on linear regressions, higher BMI was associated with a higher prevalence of depression symptoms among females (b = 0.238, p = 0.000) and males (b = 0.117, p = 0.019). However, the MR results imply that the positive link applies only to females (b = 0.302, p = 0.007) but not to males (b = -0.070, p = 0.520). Poor instrument strength may explain why many previous studies that have utilized genetic instruments have been unable to identify a statistically significant link between BMI and depression-related traits. Although the number of genetic markers in the instrument had only a minor effect on the point estimates, the standard errors were much smaller when more powerful instruments were employed. PLOS ONE PLOS ONE | https://doi.org/10.1371/journal.pone.0297594 February 23, 2024 1 / 11 a1111111111 a1111111111 a1111111111 a1111111111 a1111111111 OPEN ACCESS Citation: Viinikainen J, Bo¨ckerman P, Willage B, Elovainio M, Kari JT, Lehtima¨ki T, et al. (2024) Effect of weight on depression using multiple genetic instruments. PLoS ONE 19(2): e0297594. https://doi.org/10.1371/journal.pone.0297594 Editor: Billy Morara Tsima, University of Botswana School of Medicine, BOTSWANA Received: March 16, 2023 Accepted: January 9, 2024 Published: February 23, 2024 Copyright: ©2024 Viinikainen et al. This is an open access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited. Data Availability Statement: The dataset supporting the conclusions of this article was obtained from the Cardiovascular Risk in Young Finns Study (YFS) and Statistics Finland register data after submission and approval of our study plan by the YFS coordinators. The YFS dataset comprises health-related participant data, and their use is restricted under the regulations on professional secrecy (Act on the Openness of Government Activities, 612/1999) and on sensitive personal data (Personal Data Act, 523/1999, implementing the EU data protection directive 95/ 46/EC). Due to these legal restrictions, the data Introduction In recent decades, there has been a notable increase in the prevalence of overweight and obesity, while at the same time, depressive disorders have emerged as a leading cause of disease burden worldwide [1–3]. The question of whether increased body weight and increased depression are causally linked has become increasingly pressing. Previous research, primarily from the United States (US), has established a correlation between measures of body weight, such as body mass index (BMI), and depression [4,5], but identifying the direction of causality remains challenging. To draw policy-relevant conclusions, prior studies examining the impact of weight on mental health have used the BMI of a biological relative as an instrument for one’s own BMI [6,7]. These studies found evidence that weight causally affects mental health. Recently, researchers have begun to investigate the causal effects of weight on depression symptoms using genetic information as the instrument. Most of these studies are based on US data and have not found a significant effect of BMI on mental health [8–10]. However, a high BMI has been linked to weaker mental health among the US elderly [11], and evidence from Finland and the United Kingdom suggests that excessive body weight may increase the risk of depression [12,13]. In the European Union (EU), 1 in 6 people experiences a mental health problem. Among these countries, Finland has the highest estimated incidence of mental disorders, with nearly 1 in 5 people affected [14]. Additionally, the proportion of the population classified as overweight (BMI �25) is higher in Finland (59%) compared to the EU average (53%) [15], and the prevalence of obesity has been increasing [16]. Using data from Finland, this study examines how BMI is linked to depression symptoms among prime working-age individuals. To investigate this research question, we use instruments based on genetic markers, single nucleotide polymorphisms (SNPs), to identify causal effects. Recently, genome-wide association studies (GWASs) have identified an increasing number of SNPs related to BMI. Aggregating multiple weight-predictive SNPs into a single number yields a polygenic risk score, (PGS) which indicates an individual’s genetic susceptibility to high BMI. A higher number of SNPs in a PGS enhances the strength of the instrument, which mitigates the potential weak instrument problem associated with the statistical approach we employ. However, it also amplifies the risk that the SNPs affect the outcome through pathways other than BMI, thereby violating the key identifying assumptions. Consequently, the secondary aim of this study is to compare estimates using different instruments to shed light on the sensitivity of the results to which instruments are used. Materials and methods Study sample The Cardiovascular Risk in Young Finns Study (YFS) is an ongoing study that began in 1980 with a total of 3,596 participants between 3 and 18 years of age. The participants were randomly chosen from five Finnish university regions [17]. The main goal of the study was to determine the contribution of childhood lifestyle, biological, and psychological measures to the risk of cardiovascular diseases in adulthood. Risk factors associated with cardiovascular disease, such as information on BMI, have also been repeatedly collected from the survey participants. Moreover, information on psychosocial traits has been collected over the study. We perform secondary analysis using YFS data from 2011, i.e., 31 years after the study was launched when the participants were, on average, 41.9 years of age (SD = 5 years). To obtain information on participants’ parental background, the YFS was linked to the Longitudinal Population Census (LPC) of Statistics Finland from the year 1980 using unique personal PLOS ONE Weight and depression PLOS ONE | https://doi.org/10.1371/journal.pone.0297594 February 23, 2024 2 / 11 from this study cannot be stored in public repositories or otherwise made publicly available. Data access may be permitted on a case-by-case basis upon request only. Data sharing beyond the group is done in collaboration with the YFS group and requires a data-sharing agreement with YFS representatives and appropriate contracts with Statistics Finland. The linked YFS-FLEED-LPC data can only be used in the Statistics Finland remote access system (FIONA). Investigators can submit an expression of interest to the chairman of the publication committee (Prof. Mika Ka¨ho¨nen, Tampere University, Kalevantie 4, 33100 Tampere, Finland, Email: [email protected], Phone: +358503186297). Funding: The Young Finns Study has been financially supported by the Academy of Finland: grant numbers 356405, 322098, 286284, 134309 (Eye), 126925, 121584, 124282, 129378 (Salve), 117787 (Gendi), and 41071 (Skidi); the Social Insurance Institution of Finland; Competitive State Research Financing of the Expert Responsibility area of Kuopio; Tampere and Turku University Hospitals (grant number X51001); Juho Vainio Foundation; Paavo Nurmi Foundation; Finnish Foundation for Cardiovascular Research; Finnish Cultural Foundation; Sigrid Juselius Foundation; Tampere Tuberculosis Foundation; Emil Aaltonen Foundation; Yrjo¨Jahnsson Foundation; Signe and Ane Gyllenberg Foundation; Jenny and Antti Wihuri Foundation; Diabetes Research Foundation of Finnish Diabetes Association; EU Horizon 2020 (grant 755320 for TAXINOMISIS and grant 848146 for To Aition); European Research Council (grant number 742927 for MULTIEPIGEN project); Tampere University Hospital Supporting Foundation; Society of Finnish Clinical Chemistry; Cancer Foundation Finland; BETTER4U (Preventing obesity through Biologically and bEhaviorally Tailored inTERventions for you; project number: 101080117); Jane and Aatos Erkko Foundation. The use of linked data was supported by Palkansaajasa¨a¨tio¨and OP Group Research Foundation. ME was supported by the Academy of Finland (339390).The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript. Competing interests: The authors have declared that no competing interests exist. identifiers. The current analysis is based on a sample size of 1,523. The reduced size compared to the original sample is mainly due to missing information for certain variables. Fig 1 depicts a flowchart of the study sample showing the data sources and the exclusion criteria of individuals at each stage. The YFS was approved initially by the ethical committee of the University of Turku (3/ 1978) and subsequently by the ethical committee of the Hospital District of Southwest Finland whenever a new field study was conducted, most recently in 2017 (ETMK 68/2017). The use of the YFS-LPC data was approved by Statistics Finland (TK-53-673-13). Written informed consent for underage participants was provided by their parents or guardians. Once participants Fig 1. Flow chart of the YFS-LPC data. Participants in the YFS who were not linked to the LPC data either did not consent to the linking or were not found in the registers, due to having permanently moved abroad or having passed away before 1987. https://doi.org/10.1371/journal.pone.0297594.g001 PLOS ONE Weight and depression PLOS ONE | https://doi.org/10.1371/journal.pone.0297594 February 23, 2024 3 / 11 reached the age of 18, they themselves gave written informed consent after a full explanation of the study procedures. Measures As the dependent variable, we used the 21-item Beck’s Depression Inventory (BDI-II), which was administered in 2011 [18,19]. Items include feeling sad, feeling discouraged about the future, and feelings of failure. S1 Table documents the full inventory. Depression symptoms were self-assessed on a 4-point scale, and the total score ranges from 0 to 63. A higher value of the score indicates more severe depression symptoms. As the treatment variable, we used BMI, which is defined as weight in kilograms over height in meters squared (kg/m 2 ). BMI originated from the YFS data and was based on professional health examinations conducted in 2011. Since body measures were collected by health professionals, we avoid measurement errors and other sources of bias related to self-reported BMI [20]. Information on family background, measured by parental education, was obtained in 1980 (from LPC). The parental education indicator equaled one if the parent had completed university level education by 1980 and zero otherwise. Information on region of residence were measured in 1980, and the variable takes one of four categorical values (south, west, east, north, from YFS). Blood samples from the YFS participants, collected in 2001 or in 2007, were used for genotyping. Based on the genotyping results, SNPs associated with BMI in two GWASs were identified [21,22]. We use weighted PGSs as instruments, where the weights are determined by the strength of the SNP’s association with BMI. The sum of these weighted contributions yields the PGS. The first PGS, which has also been used in previous studies [8–10], includes 32 SNPs that are associated with BMI at p <5×10 −8 [21]. Moreover, we use five PGSs based on a more recent GWAS [22]. These PGSs include SNPs that are associated with BMI at the following significance levels: p <5×10 −8 , p <10 −5 , p <10 −4 , p <0.001, and p <0.01. The PGS based on significance level p <5×10 −8 includes 97 SNPs. In other PGSs, the number of SNPs is higher, but the exact number is unknown. Statistical analysis We first ran an OLS model that shows an association between BMI and depression symptoms. To identify a causal effect of BMI on depression symptoms, we then estimated two-stage least squares (2SLS), Mendelian randomization (MR) models, using PGSs as instruments for BMI. The motivation for the use of MR is based on the likely influence of confounding factors [5,10], and the possibility of a two-way causal relationship between body weight and depression [23,24]. The analyses were conducted using Stata, version 18.0. In all models, we controlled for age, sex, region of residence in 1980, and parental education. To account for population stratification, the MR models also include the first ten genetic principal components. To obtain unbiased MR estimates, four conditions must be met: (1) the genetic instrument must be as-if randomly assigned (independence); (2) the instrument must have a monotonic effect on BMI (monotonicity); (3) the genetic instrument must affect BMI (relevance); and (4) the genetic instrument does not affect depression except through BMI (exclusion restriction) [11]. The independence assumption is supported by Mendel’s law of segregation i.e., alleles segregate randomly in conception. To control for the possibility that the independence assumption would be violated because of population stratification, which is the case if SNP frequencies differ between population subgroups, we control for the first ten genetic principal PLOS ONE Weight and depression PLOS ONE | https://doi.org/10.1371/journal.pone.0297594 February 23, 2024 4 / 11 components. Supporting the monotonicity assumption, the SNPs in PGSs are associated with higher (not lower) BMI. To evaluate the relevance assumption, it has become standard to use the first-stage statistics F-value of 10 as a cut-off value for a sufficiently strong instrument [25]. However, a recent study suggests that this cut-off value may be too low and proposes that the first-stage F-statistics should exceed the value of 104.7 [26]. In the studies that use the 32 SNP genetic instrument, the first-stage F-statistics have been around 35 [8–10], which clearly falls below the more conservative cut-off value. In this study, we also use an instrument that exceeds the cut-off value of 104.7. The exclusion restriction assumption is discussed in the results section. Results Descriptive statistics of the variables are presented in Table 1. The OLS estimates in Fig 2 and Table 2 (Panel A) show that a one-unit increase in BMI is associated with a 0.117 [95% CI: 0.020, 0.215] and 0.238 [95% CI: 0.143, 0.333] unit increase in the depression index among males and females, respectively. Fig 3 and Table 2 (Panels B-G) summarize the MR estimates for the effect of BMI on mental health. We obtained three important results. First, the point estimates were similar regardless of the instrument used. Second, the positive link between BMI and depression symptoms applies only to females. The results using the instrument based on p <0.01 imply that among females a one-point increase in BMI is linked to a 0.302 [95% CI: 0.083, 0.521] unit increase in depression symptoms. Among males, the point estimate is negative, close to zero, and statistically insignificant (b = -0.070, 95% CI: -0.282, 0.143). Results from reduced-form models are in accordance with these results (S1 Fig). Third, the more SNPs the instrument contains, the smaller the confidence intervals. This is important pattern affects the interpretation of the Table 1. Descriptive statistics. Data source Year of measurement Mean SD Beck’s Depression Inventory score YFS 2011 4.991 5.784 BMI YFS 2011 26.275 4.677 BMI PGS, Speliotes p<5×10 −8 YFS 2001/2007 4.030 0.527 BMI PGS, Locke p<5×10 −8 YFS 2001/2007 -0.301 0.191 BMI PGS, Locke p<10 −5 YFS 2001/2007 -0.409 0.266 BMI BGS, Locke p<10 −4 YFS 2001/2007 -0.467 0.333 BMI PGS, Locke p<0.001 YFS 2001/2007 0.595 0.452 BMI PGS, Locke p<0.01 YFS 2001/2007 0.964 0.737 Female, share YFS 1980 0.550 0.498 Average age, years YFS 2011 41.861 5.039 Mother education high, share LPC 1980 0.077 0.267 Father education high, share LPC 1980 0.106 0.308 Region of residence, share Southern Finland YFS 1980 0.169 0.375 Western Finland YFS 1980 0.357 0.479 Eastern Finland YFS 1980 0.309 0.462 Northern Finland YFS 1980 0.165 0.372 YFS refers to the Cardiovascular Risk in Young Finns Study. LPC refers to the Longitudinal Population Census. SD refers to standard deviation and PGS refers to polygenic risk score. The PGSs are based on studies by Speliotes et al. (2010) and Locke et al. (2015). The indicator for high parental education equals one if the parent has obtained some university education based on the LPC data from 1980. N = 1,523. https://doi.org/10.1371/journal.pone.0297594.t001 PLOS ONE Weight and depression PLOS ONE | https://doi.org/10.1371/journal.pone.0297594 February 23, 2024 5 / 11 results, particularly for females. Using an instrument that is based on a significance level of p<10 −5 or lower, the results suggest a statistically insignificant link between BMI and depression symptoms for both females and males. Using other instruments, the link is positive and significant (p <0.10) for females. The first-stage F-statistics in the models using instruments based on significance level p <10 −5 or lower vary between 16.49 and 71.27. In the models using an instrument based on more lenient significance levels, the F-statistics vary between 39.59 and 285.84. A potential threat to the validity of the results is the violation of the exclusion restriction assumption. This assumption could be violated if SNPs, which are related to BMI, also affect depression symptoms via other pathways. To assess the validity of our instrument, we conducted a balance test comparing observable characteristics of individuals aboveand belowmedian PGS scores (Table 3). The 32 SNP PGS is not associated with covariates, but the p<0.01 PGS is associated with residing in eastern or western Finland and fathers’ education, which are controlled in all models. We also performed Sargan’s test of overidentifying restrictions using the 32 SNP PGS. The test lends support to the instrument’s validity as the test supported the null hypothesis that all 32 SNPs yielded the same MR estimate (p >0.520). As we lack information on individual SNPs for other instruments, Sargan’s test cannot be used to test the validity of those instruments. However, as the point estimates were similar across models, Sargan’s test based on 32 SNP PGS indirectly supports the conclusion that violation of the exclusion restriction does not substantially bias the results. Discussion Using genetic instruments for BMI, we found that BMI was linked to depressive symptoms among females but not among males in Finland. Although the point estimates were similar Fig 2. The association between BMI and depression symptoms based on OLS estimates. OLS point estimates (points) and the corresponding 95% confidence intervals (solid lines) based on heteroscedasticity-robust standard errors overall and stratified by sex. The models adjust for (sex), age, region of residence in 1980, and parental education. N = 1,523. https://doi.org/10.1371/journal.pone.0297594.g002 PLOS ONE Weight and depression PLOS ONE | https://doi.org/10.1371/journal.pone.0297594 February 23, 2024 6 / 11 regardless of the PGS used, the confidence intervals were substantially reduced as the instrument strength increased. The first result is consistent with previous findings that females are particularly punished for being overweight, e.g., in the labor market [27,28]. The second finding may explain why previous studies that used a genetic instrument based on relatively fewer SNPs have not found a significant relationship between BMI and depression-related traits. Consistent with prior studies [8–10], our results based on the 32 SNP PGS, did not indicate a significant link between BMI and depressive symptoms. Using a more powerful instrument, Table 2. Estimation results: BMI and depression symptoms. All Male Female Panel A: OLS results Per additional unit of BMI 0.194*** (p = 0.000) [0.123, 0.265] 0.117** (p = 0.019) [0.020, 0.215] 0.238*** (p = 0.000) [0.143, 0.333] Panel B: MR results (PGS by Speliotes, p <5×10 −8 (32 SNPs) Per additional unit of BMI 0.235 (p = 0.238) [-0.155, 0.625] 0.228 (p = 0.387) [-0.288, 0.743] 0.276 (p = 0.355) [-0.309, 0.862] First stage F-statistics 36.45 19.42 16.49 Panel C: MR results, (PGS by Locke, p <5×10 −8 (97 SNPs) Per additional unit of BMI 0.190 (p = 0.277) [-0.152, 0.531] 0.096 (p = 0.704) [-0.400, 0.592] 0.261 (p = 0.259) [-0.193, 0.714] First stage F-statistics 48.04 21.93 26.65 Panel D: MR results (PGS by Locke, p <10 −5 ) Per additional unit of BMI 0.136 (p = 0.354) [-0.151, 0.423] -0.052 (p = 0.814) [-0.484, 0.380] 0.256 (p = 0.172) [-0.112, 0.624] First stage F-statistics 71.27 35.09 37.54 Panel E: MR results (PGS by Locke, p <10 −4 ) Per additional unit of BMI 0.182 (p = 0.141) [-0.060, 0.425] -0.089 (p = 0.659) [-0.485, 0.307] 0.360** (p = 0.018) [0.063, 0.656] First stage F-statistics 101.07 39.59 60.80 Panel F: MR results (PGS by Locke, p <0.001) Per additional unit of BMI 0.127 (p = 0.242) [-0.086, 0.339] -0.116 (p = 0.450) [-0.416, 0.185] 0.289* (p = 0.050) [0.000, 0.577] First stage F-statistics 162.50 73.92 88.44 Panel G: MR results (PGS by Locke, p <0.01) Per additional unit of BMI 0.163 (p = 0.041) [0.007, 0.319] -0.070 (p = 0.520) [-0.282, 0.143] 0.302*** (p = 0.007) [0.083, 0.521] First stage F-statistics 285.84 136.42 156.62 Mean outcome 4.991 4.095 5.726 N 1,523 686 837 The table reports parameter estimates, p-values in parentheses and 95% confidence intervals based on heteroscedasticity-robust standard errors in square brackets. The dependent variable is Beck’s Depression Inventory score measured in 2011. BMI was measured in 2011. The models include unreported controls for (sex), age, the first ten principal components (Panels B-G), region of residence in 1980, and parental education. The instruments used in the MR models are the PGSs for BMI based on studies by Speliotes et al. (2010) and Locke et al. (2015) [21,22]. Boldface indicates statistical significance (*p<0.10 **p<0.05 ***p<0.001). N = 1,523. https://doi.org/10.1371/journal.pone.0297594.t002 PLOS ONE Weight and depression PLOS ONE | https://doi.org/10.1371/journal.pone.0297594 February 23, 2024 7 / 11 Fig 3. The effect of BMI on depression symptoms, MR estimates. MR point estimates (points) and the corresponding 95% confidence intervals (solid lines) based on heteroscedasticity-robust standard errors overall and stratified by sex. The instruments used in the models are the PGSs for BMI based on studies by Speliotes et al. (2010) and Locke et al. (2015) [21,22]. The PGSs include SNPs that associated with BMI at the following significance levels: p <5×10 −8 , p <10 −5 , p <10 −4 , p <0.001, and p <0.01. The models adjust for (sex), age, the first ten principal components, region of residence in 1980, and parental education. N = 1,523. https://doi.org/10.1371/journal.pone.0297594.g003 Table 3. Comparison of observable characteristics between individuals above and below the median values of PGS. Above median PGS Below median PGS Difference a (t-statistics) Female: Speliotes c , p <5×10 −8 0.548 (0.498) 0.551 (0.498) -0.003 (-0.100) Female: Locke d , p <0.01 0.550 (0.497) 0.549 (0.498) 0.001 (0.023) Age in 2011: Speliotes, p <5×10 −8 41.692 (5.013) 42.033 (5.063) -0.341 (-1.321) Age in 2011: Locke, p <0.01 41.701 (5.088) 42.022 (4.988) -0.322 (-1.245) Southern Finland b : Speliotes, p <5×10 −8 0.163 (0.370) 0.176 (0.381) -0.013 (-0.650) Southern Finland b : Locke, p <0.01 0.154 (0.361) 0.185 (0.389) -0.032 (-1.651) Western Finland b : Speliotes, p <5×10 −8 0.349 (0.477) 0.365 (0.482) -0.016 (-0.653) Western Finland b : Locke, p <0.01 0.336 (0.473) 0.377 (0.485) -0.041 (-1.678)* Eastern Finland b : Speliotes, p <5×10 −8 0.322 (0.468) 0.295 (0.456) 0.028 (1.177) Eastern Finland b : Locke, p <0.01 0.346 (0.476) 0.271 (0.445) 0.076 (3.209)*** Northern Finland: Speliotes, p <5×10 −8 0.166 (0.372) 0.165 (0.372) 0.001 (0.035) Northern Finland: Locke, p <0.01 0.164 (0.371) 0.167 (0.373) -0.003 (-0.149) High education, mother: Speliotes, p <5×10 −8 0.074 (0.263) 0.081 (0.272) -0.006 (-0.450) High education, mother: Locke, p <0.01 0.071 (0.257) 0.084 (0.278) -0.013 (-0.966) High education, father: Speliotes, p <5×10 −8 0.107 (0.309) 0.106 (0.308) 0.001 (0.087) High education, father: Locke, p <0.01 0.089 (0.285) 0.124 (0.329) -0.034 (-2.172)** The table reports means and standard deviations in parentheses. Boldface indicates statistical significance (*p<0.10 **p<0.05 ***p<0.001). N = 1,523. a A two-sample t-test was used to test the differences in means. b Regional indicators refer to the region of residence in 1980. https://doi.org/10.1371/journal.pone.0297594.t003 PLOS ONE Weight and depression PLOS ONE | https://doi.org/10.1371/journal.pone.0297594 February 23, 2024 8 / 11