Abstract
AbstractBackgroundLongitudinal weight trajectories may reflect individual health status. We examined the genetic aetiology and clinical consequences of adult weight trajectories in males and females leveraging genetic and phenotypic data in the electronic health records (EHR) of the BioMe™ Biobank.MethodsWe constructed four longitudinal weight trajectories using annual EHR-recorded weights (stable weight, weight gain, weight loss, or weight cycle) (n=21,487). After validating the accuracy of the trajectories (n=100), we conducted a hypothesis-free phenome-wide association study (PheWAS), including sex-stratified PheWAS, to identify diseases associated with each weight trajectory. We then performed a hypothesis-driven polygenic risk score (PRS) analysis on these weight trajectories, focusing on anorexia nervosa (AN) and depression—both commonly associated with weight changes.FindingsWeight trajectory classification was highly accurate (accuracy, sensitivity, and specificity > 97% for all four trajectories). Hypothesis-free PheWAS analyses identified a significant association between depression and weight cycle (OR=1.4, p≤7.7×10−16) after Bonferroni correction, but not with weight gain or loss. Compared to other weight trajectories, we also observed a significant association of osteoporosis-related phecodes with weight loss in females only (ORfemale=1.4, pfemale≤ 1.4×10−7, ORmale=0.8, pmale≥ 0.18). AN-PRS was positively associated with weight loss trajectory among individuals without eating disorder diagnoses (ORtop vs. bottom 10% PRS=1.95, p=0.00035). Consistent effect direction was observed across three ancestry groups. The AN-PRS-weight loss association was not attenuated by obesity-PRS (ORtop vs. bottom 10% PRS=1.94).InterpretationAdult weight trajectory is associated with disease both phenotypically and genetically. Our PheWAS reveals unique relationships between diseases and weight trajectory patterns, including the association of depression and weight cycle trajectory in both males and females, and osteoporosis-weight loss trajectory association in females only. In addition, our PRS analysis suggests that adults with higher AN genetic risk are more likely to have a weight loss trajectory, and this association may be independent of BMI/obesity-related genetic pathways.FundingKlarman Family Foundation, NIMH.Research in Context PanelEvidence before this studyWe used PubMed and medRxiv to search for phenome-wide association studies (PheWAS) of BMI/weight that have been published and/or are currently in preprint. For the weight PheWAS, we used search terms: “(phewas[tiab] OR phenome wide[tiab]) AND (weight[tiab] OR BMI[tiab] OR body mass index[tiab])” on PubMed, and “phewas weight”, “phewas BMI”, “phewas body mass index”, “phenome weight”, “phenome BMI”, or “phenome body mass index” for abstract or title search on medRxiv (up to March 17, 2021). The literature search identified 45 studies in total. From title screening, 13 of the studies were further reviewed, and 5 studies were ultimately included as relevant evidence of PheWAS on weight or BMI. These five PheWAS included four studies of adult populations of European ancestry, and one study conducted in children (ALSPAC). The weight-related exposure variables used in these studies were genetic variants of the obesity-associated FTO gene, BMI-associated SNPs, BMI PRS, BMI value, and obesity status. Through using BMI/obesity-related exposures, these published PheWAS identified comorbidities associated with obesity, including type 2 diabetes, sleep apnea, hypertension, edema, liver disease, asthma, bronchitis, and earlier age of puberty in at least two of the PheWAS. The childhood PheWAS found positive associations of BMI PRS with multiple biomarkers, including leptin, C-reactive protein, IL6, triglyceride, very low-density lipoprotein, and a negative association with high density lipoprotein. One BMI PheWAS published in 2020 observed that hyperlipidemia and gastroesophageal reflux disease were only significantly associated with BMI on a phenotypic level, but not on a genetic level (e.g., BMI or obesity SNPs), likely due to the small genetic effect of single genetic variants.Regarding the impact of anorexia nervosa (AN) and depression genetic risk on weight trajectory, we searched “anorexia nervosa[title] AND (weight[title] OR BMI[title] OR body mass index[title]) AND (genetic[tiab])” or “depression[title] AND (weight[title] OR BMI[title] OR body mass index[title]) AND (genetic[tiab])” on PubMed, and “anorexia polygenic weight” or “anorexia polygenic BMI” or “anorexia polygenic body mass index” or “depression polygenic weight” or “depression BMI” or “depression body mass index” for abstract or title search on medRxiv (up to March 17, 2021). The literature search identified 36 studies in total, and 21 were further reviewed through the title screening. No studies were identified that examined the effect of depression genetic risk on BMI or weight, and only two were included as relevant evidence of AN genetic risk on BMI and weight. Of these two studies, one was cross-sectional in a small adult sample (age 18-59, n=380), and the other was longitudinal in a children/young adult population in the ALSPAC cohort (age 10-24, n=8,654). BMI PRS was found to be associated with lower BMI cross-sectionally, and with weight loss over time only in females.Added value of this studyIn this study, we create a novel inflection-point based method to classify longitudinal weight trajectory using weights recorded in the EHR in a hospital-based biobank (Mount Sinai BioMe™ Biobank), with an accuracy of 98% or higher through our validation study (n=100).With this validated phenotype of weight pattern over time (i.e., weight trajectory), our PheWAS analysis afforded us the opportunity to examine comorbidity across the weight spectrum and across time. We identified 143 diseases associated with weight cycle (e.g., depression, anemias, renal failure),13 diseases positively associated with weight gain trajectory (e.g., obesity, obstructive sleep apnea, edema), and 36 with weight loss (e.g., protein-calorie malnutrition, gastrointestinal complication, end stage renal disease), after Bonferroni correction, using 5% as the cutoff for clinically relevant weight change. All diseases were negatively associated with a stable weight trajectory. Furthermore, we performed, to our knowledge, the first sex-stratified PheWAS related to weight trajectory, and identified eight sex-stratified associations with weight gain (e.g., obstructive sleep apnea), eight with weight loss (e.g., osteoporosis), and ten with weight cycle (e.g., vitamin B-complex deficiencies).On a genetic level, our study fills in the gap of the impact of AN genetic risk on longitudinal weight changes in the adult population. Unlike the finding in adolescents in the ALSPAC study, which found an AN-PRS-weight loss trajectory association only in females, we found an association of higher AN genetic risk with weight loss trajectory in both men and women, with consistent effect direction observed across individuals with European, African, and Hispanic ancestry in the BioMe™ Biobank. Additionally, this association of AN genetics with weight loss was independent of the influence of obesity/BMI related genetic variants on weight.Implications of all the available evidencePheWAS is an excellent tool for exploring comorbidities associated across the weight spectrum. Our PheWAS findings identify diseases with different weight patterns (e.g., depression and weight cycle), which may reflect characteristics of these diseases, including age of onset, progression pattern, severity, and chronicity (e.g., the episodic nature of depression with the weight cycle pattern). In addition, our sex-stratified PheWAS implicates the important role of sex in weight regulation in the presence of disease. Certain sub-populations may be at greater risk of weight loss in some disease states (e.g., women with osteoporosis) and may need targeted treatment to address nutritional needs and to prevent further weight loss.Our study also suggests that people who have high AN genetic risk are at greater risk of displaying a weight loss trajectory during adulthood. However, given the limited amount of variation in the outcome of interest (e.g., weight loss) explained by the AN-PRS, the PRS may have to be jointly modeled with other risk factors to predict weight loss more accurately, or to identify subgroups at risk of weight loss. In addition, given our finding that the effect of AN genetics on weight loss was minimally affected by the obesity-related genetics, and the previously reported low genetic correlation of −0.22 between AN and obesity in the 2019 AN GWAS, this may indicate that AN- and obesity-related weight changes might have unique genetic underpinnings. Future studies that assess the pathway-specific genetic risk on weight pattern will further our understanding of the genetic architecture of longitudinal weight trajectory.
Publisher
Cold Spring Harbor Laboratory