Exploring the clinical and genetic associations of adult weight trajectories using electronic health records in a racially diverse biobank: a phenome-wide and polygenic risk study

Abstract

Background: Weight trajectories might reflect individual health status. In this study, we aimed to examine the clinical and genetic associations of adult weight trajectories using electronic health records (EHRs) in the BioMe Biobank.

Methods: We constructed four weight trajectories based on a-priori definitions of weight changes (5% or 10%) using annual weight in EHRs (stable weight, weight gain, weight loss, and weight cycle); the final weight dataset included 21 487 participants with 162 783 annual weight measures. To confirm accurate assignment of weight trajectories, we manually reviewed weight trajectory plots for 100 random individuals. We then did a hypothesis-free phenome-wide association study (PheWAS) to identify diseases associated with each weight trajectory. Next, we estimated the single-nucleotide polymorphism-based heritability (h_SNP²) of weight trajectories using GCTA-GREML, and we did a hypothesis-driven analysis of anorexia nervosa and depression polygenic risk scores (PRS) on these weight trajectories, given both diseases are associated with weight changes. We extended our analyses to the UK Biobank to replicate findings from a patient population to a generally healthy population.

Findings: We found high concordance between manually assigned weight trajectories and those assigned by the algorithm (accuracy ≥98%). Stable weight was consistently associated with lower disease risks among those passing Bonferroni-corrected p value in our PheWAS (p≤4·4 × 10^-5). Additionally, we identified an association between depression and weight cycle (odds ratio [OR] 1·42, 95% CI 1·31-1·55, p≤7·7 × 10^-16). The adult weight trajectories were heritable (using 5% weight change as the cutoff: h_SNP² of 2·1%, 95% CI 0·9-3·3, for stable weight; 4·1%, 1·4-6·8, for weight gain; 5·5%, 2·8-8·2, for weight loss; and 4·7%, 2·3-7·1%, for weight cycle). Anorexia nervosa PRS was positively associated with weight loss trajectory among individuals without eating disorder diagnoses (OR_1SD 1·16, 95% CI 1·07-1·26, per 1 SD higher PRS, p=0·011), and the association was not attenuated by obesity PRS. No association was found between depression PRS and weight trajectories after permutation tests. All main findings were replicated in the UK Biobank (p<0·05).

Interpretation: Our findings suggest the importance of considering weight from a longitudinal aspect for its association with health and highlight a crucial role of weight management during disease development and progression.

Funding: Klarman Family Foundation, US National Institute of Mental Health (NIMH).

Figure 1:. Examples of different types of individual weight trajectories based on annual weight

(A) Stable weight trajectory. (B) Weight loss trajectory. (C) Weight gain trajectory. (D) Weight cycle trajectory. (E) Weight cycle plus weight loss trajectory. (F) Weight cycle plus weight gain trajectory.

Figure 2:. Phenome-wide association plots of weight trajectories with the BioMe Biobank phecodes with the 5% weight change cutoff

Phecodes above the blue line passed the Bonferroni-corrected p value threshold (p≤4·4 × 10⁻⁵). Phecodes are grouped into 17 different disease categories (appendix p 33). An upward triangle (Δ) denotes a positive association, whereas a downward triangle (∇) denotes a negative association. The top associations are annotated in each plot.

Figure 3:. Associations of anorexia nervosa and depression with weight trajectories in participants with European ancestry in the BioMe Biobank

The left panel shows the OR of individuals in the top versus bottom decile of PRS (anorexia nervosa, depression, and obesity class I) with different weight trajectories (defined by either 5% or 10% weight change cutoff). The right panel shows the association of anorexia nervosa and depression with different weight trajectories on a phenotypic level using phecodes as the exposure. Individuals with stable weight were used as controls for those with weight gain, weight loss, and weight cycle trajectories. Given the small sample size of individuals with diagnosed anorexia nervosa in the European ancestry samples (n=1), its parent phecode, eating disorder, was used for the phenotypic association (n=9). Obesity PRS and phecode were used as positive controls. OR=odds ratio. PRS=polygenic risk score.

Figure 4:. Associations of PRS with weight trajectory by deciles and by ancestry

(A) OR of each obesity PRS decile with weight gain trajectory, using the bottom decile as the reference group. (B) OR of each anorexia nervosa PRS decile with weight loss trajectory, using the bottom decile as the reference group. (C) OR of individuals in the top versus bottom decile of anorexia nervosa PRS with weight loss trajectory (defined by 5% cutoff) across different ancestry groups (European, African, or Hispanic Latino). OR=odds ratio. PRS=polygenic risk score.