The development of chronic disease is a long-term process involving multiple endpoints. Although multi-state Cox models can estimate state-specific survival risks over time, they are not well suited for comparing the effectiveness of treatment regimes. A discrete-time split-state framework has been proposed, which divides disease states into substates by conditioning on past history. As this framework is both "memoryless" and "memorable," the transition rates can be synthesized into summary measures, multimorbidity-adjusted life year (MALY) and substate-specific life year (SSLY). Building on this framework, we propose to investigate the causal effects of static and dynamic treatment regimes over the disease course under the assumptions of constant baseline confounders and instantaneous effects of interventions on transition rates. We identify the optimal treatment regime as the one that maximizes MALY and use SSLY to elucidate the mechanisms of how treatments influence disease progression. In the application, we identified the optimal weight targets in the ARIC study by modeling the disease course in healthy, at-metabolic-risk, coronary heart disease, heart failure, and mortality states. The estimated MALY was 1.80 years higher (95% CI: 0.62, 2.78) under regime "Normal weight initially and change to overweight if age >65 y" compared to regime "Normal weight across all states." SSLY decomposition indicates that this gain arises from increased life year in all substates except the healthy state. In summary, our method provides a framework to evaluate the health benefits of treatment regimes over the disease course and has the potential to improve the precision prevention of chronic diseases.

Infectious diseases and chronic diseases are two major fields in epidemiology that have traditionally been studied separately because of their distinct etiologies and modeling methods. Infectious disease data are typically collected at an aggregated level and analyzed using compartmental models, most commonly the susceptible (S), infectious (I), and recovered (R) (SIR) model, whereas chronic disease data are usually collected at the individual level and analyzed using multi-state survival models. Previous studies have pointed out the link between compartmental models and survival analysis by reconstructing the aggregated infection disease data into individual-level data. However, these studies have largely focused on the two-state transition from S to I state, and few studies have simultaneously modeled the three-state process, S, I, and R. In this paper, we propose to use a discrete-time multi-state framework to model the three-state progression of infectious disease. We first introduce and compare the underlying methodological foundations for modeling infectious disease and chronic disease dynamics, then show the link between compartment models and multi-state models, and finally present how infectious disease can be modeled using the multi-state framework under the two scenarios: 1) all S, I, and R states are observed, and 2) only the I state is observed, with the R state treated as latent. In the application, we applied the multi-state approach to estimate the dynamics of influenza using the data in a British boarding school in 1978, where only the infected cases were observed over time. The estimated recovery rate was 0.42 and the corresponding contact rate was0.91 (95% CI: 0.84, 0.98). The basic reproductive number was 2.17 (95% CI: 2.00, 2.33), which declined to approximately 1 by day 6, and continued to decrease thereafter. Overall, we propose a unified multi-state approach for modeling infectious and chronic disease progression, which may provide evidence to inform timely and effective infectious disease prevention.

Survival analysis has been widely used to understand the etiology of chronic diseases. However, traditional survival analysis models only a single endpoint, whereas the development of chronic diseases is a multi-state process. In this review, we examine the connections among logistic regression, the Cox model, competing risk models, and multi-state models in estimating hazards and survival risks, with findings summarized in three aspects. First, logistic regression and the Cox model are connected. The conditional likelihood of a conditional logistic regression stratified by risk set is equivalent to the partial likelihood used by the Cox model. The continuous-time survival risks can be derived from the Cox model using the Breslow estimator. Alternatively, pooled logistic regression can be used to estimate risk within small time interval that approximate discrete-time hazard, with survival risks estimated using the Kaplan-Meier (K-M) estimator. Survival risk estimated from the Cox model is more accurate, whereas discrete-time hazard is more flexible for creating complex statistics (such as counterfactual survival risks in causal inference) and provides an approach for integrating machine learning into survival analysis. Second, for nonparametric estimation of survival risks from hazards, the cumulative incidence functions (CIFs) used in competing risks and the Aalen-Johansen (A-J) estimator used in multi-state process are extensions of the K-M estimator used in single disease endpoint. Third, the cause-specific Cox model and the Markov Cox model are extensions of the Cox model to competing risk and multi-state settings, respectively. Correspondingly, multinomial pooled logistic regression and a discrete-time split-state framework extend pooled logistic regression to competing risk and multi-state settings. Our paper can serve as a tutorial to illustrate the connections between survival analysis and multi-state modeling. We anticipate that multi-state models will play an increasingly important role in understanding chronic disease dynamics and advancing precision prevention and prediction.

The course of heart disease involves multiple endpoints, and how to characterize this course needs investigation. Multi-state models are commonly used to model the disease course, estimating state occupation probabilities (survival risks) in each state over time. These transition parameters are state- and time-specific, which makes the model difficult to use for characterization of the entire course and for comparative effectiveness of exposures. A Discrete-time Split-state Framework has been proposed, which splits disease states into substates conditioned on past disease history. This framework is “memoryless” in that the newly created substates are independent of past history, thereby relaxing the Markov assumption. It is also “memorable” because the substates contain information about past history. In this paper, we leverage the “memoryless” and “memorable” features of the framework to synthesize the estimated transition parameters into two summary measures: Multimorbidity-Adjusted Life Year (MALY), and disease path. MALY takes the multimorbidity of each substate into consideration and estimates the adjusted life years in full health. Disease path describes the progression of disease and elucidates disease mechanisms. In the application, we characterized the course of heart disease using data from the Atherosclerosis Risk in Communities (ARIC) study. The disease course was modeled in five states: healthy, at metabolic risk, coronary heart disease (CHD), heart failure, and mortality. In this mid- to old-age population, the estimated MALY was 26.53 years (95

The development of chronic disease is a long-term process that involves multiple endpoints, and limited methods can assess the health benefits of a treatment regime over the disease course. Existing multi-state Cox models estimate survival risks by state over time, which are difficult to use when comparing the effectiveness of treatment regimes. We have proposed a discrete-time split-state framework [1], which divides disease states into substates by conditioning on past history. As this framework is both “memoryless” and “memorable”, the time-specific transition parameters can be synthesized into summary measures, substate-specific life year (SSLY), multimorbidity-adjusted life year (MALY), and disease path [2]. In this abstract, based on this framework, we propose to investigate the causal effects of static and dynamic treatment regimes on health benefits over the entire disease course, under the assumptions of constant confounders from baseline and instantaneous effects of interventions on transition rates. Our method can identify the optimal treatment regime that generates the most benefits using MALY, and illustrate the mechanisms of treatment regimes affecting disease progression using SSLY and disease path. In the application, we evaluated the cardiovascular benefits of smoking cessation, where the course of heart disease was modeled in healthy (S 0 ), at metabolic risk (S 1 ), coronary heart disease (S 2 ), heart failure (S 3 ), and mortality states (S 4 ). Compared to the regime “being a smoker in S 0 -S 4 ”, the MALY was 0.53 (95% CI: 0.21, 0.96), 6.10 (4.88, 7.19), and 4.34 (3.02, 5.47) years higher for the regimes “being a smoker in S 0 and S 1 and stop smoking if a person develops S 2 , S 3, or S 4 ”, “no smoking in S 0 -S 4 ”, and “being a smoker at the start of intervention and stop smoking if age>65y”, respectively. In summary, our method can evaluate the health benefits of treatment regimes over the disease course, and has the potential to improve the precision prevention of chronic disease. A preprint of this manuscript can be found online [3]. [1] Ding, Chen, Lin. BMC Med Res Methodol. 2025;25(1):54. [2] Ding, Lin, Meyer. doi: https://doi.org/10.1101/2024.09.18.24313882 (BMC Med Res Methodol. In revision) [3] Ding. medRxiv. doi: https://doi.org/10.1101/2025.07.25.25332203 (Stat Methods Med Res. In revision)

Summary-data-based multivariable Mendelian randomization (MVMR) methods, such as MVMR-Egger, MVMR-IVW, MVMR median-based, and MVMR-PRESSO, are used to assess the causal effects of multiple risk factors on disease. However, accounting for variances in the summary statistics of risk factors remains a challenge. We propose a linear mixed model with measurement error correction (LMM-MEC) that accounts for the variance in summary statistics for both disease outcomes and risk factors. First, under the NOME assumption, we apply a linear mixed model to account for variance in disease summary statistics by treating it as fixed- or random-effects, depending on whether there is heterogeneity in the effect sizes of the genetic variants on the disease outcome. Next, we relax the NOME assumption and further take the estimation error (or variance) in the summary statistics of risk factors into consideration by measurement models through a regression calibration approach. In a simulation study, using independent genetic variants as instrumental variables (IV), our method showed comparable performance to existing MVMR methods under conditions of no pleiotropy or with balanced pleiotropy on the disease outcome, and it achieved slightly improved coverage rates and power under directional pleiotropy. When genetic variants are in low to moderate linkage disequilibrium (LD) (0 <  ρ $\rho $ 2 ≤ 0.3), our method showed comparable performance to MVMR-Egger, although both methods showed reduced coverage rates and power compared to situations where genetic variants as IVs are in LD. In the application study, we examined causal associations between correlated cholesterol biomarkers and longevity. By including 739 genetic variants selected based on p values < 5 × 10-5 from GWAS and allowing for low LD ( ρ $\rho $ 2 ≤ 0.1), our method identified that large LDL-c levels were causally associated with a lower likelihood of achieving longevity.

Background: In cardiovascular epidemiology, heart disease has been largely investigated by focusing on a single (or composite) endpoint. A few studies examined the disease course using multi-state Markov models, assuming the current disease state is independent of past states. Markov models are unsuitable to heart disease due to three reasons: First, risk factors interplay and affect future disease risk; second, Markov assumption does not allow for multimorbidity, a common condition for heart disease; and third, the estimated transition probabilities are time-specific and indirect for public health use. Methods: We have developed a multi-state non-Markov framework that splits disease state into substates by conditioning on past states (Ding et al. BMC Medical Research Methodology. 2024. In revision. doi: https://doi.org/10.1101/2024.09.18.24313882). As the substates track history of past states, transition rates between substates can be synthesized into a summary estimate, multimorbidity-adjusted life year (MALY) that represents the adjusted life year in full health. The derivation of MALY can be found in our second paper (Ding et al. medRxiv. 2024.doi: https://doi.org/10.1101/2024.09.18.24313882). In this study, we applied our framework to characterize the course of heart disease among participants enrolled in the Atherosclerosis Risk in Communities Study (ARIC). ARIC data was obtained from NHLBI BioLINCC. Results: We modeled the course of heart disease in five states: healthy, at metabolic risk, coronary heart disease (CHD), heart failure, and mortality ( Figure 1 ). The disability weights assigned to each disease state is shown in Figure 2 . In this mid- to old-age population, the estimated MALY was 24.13 (95% CI: 16.55, 32.06) years ( Figure 3 ). The multimorbidity-adjusted life expectancy was higher among women (79.29 [95% CI: 77.42, 81.57] years) than men (77.07 [95% CI: 74.37, 80.05] years), and was higher among Whites (79.04 [95% CI: 77.05, 81.31] years) than Blacks (75.49 [95% CI: 73.19, 79.30] years). Impact: MALY summarizes the course of heart disease into one estimate and can be used in comparative effectiveness of disease intervention in the near future.

Arthritis, a chronic inflammatory condition linked to cardiovascular disease (CVD) and bone fracture, is more frequent among military veterans and postmenopausal women. This study examined correlates of arthritis and relationships of arthritis with risks of developing CVD, bone fractures, and mortality among postmenopausal veteran and non-veteran women. We analyzed longitudinal data on 135,790 (3,436 veteran and 132,354 non-veteran) postmenopausal women from the Women’s Health Initiative who were followed-up for an average of 16 years between enrollment (1993–1998) and February 17, 2024. Regression and multistate Markov modeling were applied to meet study objectives. The prevalence of arthritis at enrollment (1993–1998) did not differ by veteran status in a fully adjusted logistic model. Variable selection yielded 5 key predictors of prevalent arthritis among veterans and 15 key predictors among non-veterans. In fully-adjusted Cox models, prevalent arthritis was associated with CVD (hazard ratio [HR] = 1.08, 95

Summary-data-based multivariable Mendelian randomization (MVMR) methods, such as MVMR-Egger, MVMR-IVW, MVMR median-based, and MVMR-PRESSO, assess the causal effects of multiple risk factors on disease. However, accounting for variances in summary statistics related to risk factors remains a challenge. We propose a linear mixed model with measurement error correction (LMM-MEC) that accounts for the variance of summary statistics for both disease outcomes and risk factors. In step I, a linear mixed model is applied to account for the variance in disease summary statistics. Specifically, if heterogeneity is present in disease summary statistics, we treat it as a random effect and adopt an iteratively re-weighted least squares algorithm to estimate causal effects. In step II, we treat the variance in the summary statistics of risk factors as multiple measurement errors and apply a regression calibration method for simultaneous multiple measurement error correction. In a simulation study, when using independent genetic variants as instrumental variables (IV), our method showed comparable performance to existing MVMR methods under conditions of no pleiotropy or balanced pleiotropy with the outcome, and it exhibited higher coverage rates and power under directional pleiotropy. Similar findings were observed when using genetic variants with low to moderate linkage disequilibrium (LD) (0 < ρ 2 ≤ 0.3) as IVs, although coverage rates reduced for all methods compared to using independent genetic variants as IVs. In the application study, we examined causal associations between correlated cholesterol biomarkers and longevity. By including 739 genetic variants selected based on P values <5×10 -5 from GWAS and allowing for low LD ( ρ 2 ≤ 0.1), our method identified that large LDL-c were causally associated with lower likelihood of achieving longevity.

In chronic disease epidemiology, the investigation of disease etiology has largely focused on an endpoint, while the course of chronic disease is understudied, representing a knowledge gap. Multi-state models can be used to describe the course of chronic disease, such as Markov models which assume that the future state depends only on the present state, and semi-Markov models which allow transition rates to depend on the duration in the current state. However, these models are unsuitable for chronic diseases that are largely non-memoryless. We propose a Discrete-Time Split-State Framework that generates a process of substates by conditioning on past disease history and estimates discrete-time transition rates between substates as a function of duration in a (sub)state. Specifically, as the substates are created by conditioning on past history, they satisfy the Markov assumption, regardless of whether the original disease process is Markovian; and the transition rates are approximated by competing risks in a short time interval estimated from cause-specific Cox models. In the simulation study, we simulated a Markov process with an exponential distribution, a semi-Markov process with a Weibull distribution, and a non-Markov process with an exponential distribution. The coverage rate of transition rates estimated using our framework was 94% for the Markov process and 93% for the non-Markov process. However, the estimated transition rates were under coverage (72%) for the semi-Markov process, which is likely due to the approximation of transition rates in discrete time. In the application, we applied the framework to describe the course of heart disease in a large cohort study. In summary, the framework we proposed can be applied to both Markov and non-Markov processes and has potential to be applied to semi-Markov processes. For future research, as substates created using our framework track past disease history, the transition rates between substates have the potential to be used to derive summary estimates that characterize the disease course.

Background: In existing literature, prevention and prediction of heart disease have largely focused on one endpoint. However, the development of heart disease is a multi-state process that is biologically inseparable. Only a few studies investigated the course of heart disease using multi-state Markov models, which are unsuitable to the course of heart disease due to several reasons. First, by assuming present state is independent of past history, Markov models do not allow for multimorbidity, a common condition for heart disease. Second, risk factors of heart disease interplay and affect future states, which violates the Markov assumption. Finally, the transition probabilities estimated from Markov modelsare time-specific, and these high-dimensional matrices are not straightforward for public health interpretation. Method: We have developed a multi-state non-Markov framework that splits disease state into substates by conditioning on past states (Ding et al. BMC Medical Research Methodology. 2024. In revision. doi: https://doi.org/10.1101/2024.09.18.24313882). As the substates track history of past states, transition rates between substates can be synthesized into one summary estimate, disease path. The derivation of the disease path can be found in our second paper (Ding et al. medRxiv. 2024. doi: https://doi.org/10.1101/2024.09.18.24313882). We applied our method to identify disease paths over the course of heart disease using the Atherosclerosis Risk in Communities Study (ARIC). We obtained the ARIC data from NHLBI BioLINCC. Results: We modeled the course of heart disease in five states: healthy, at metabolic risk, coronary heart disease (CHD), heart failure, and mortality ( Figure 1a ). By applying our framework, the disease states were divided into substates by conditioning on past disease history ( Figure 1b ), where transition rates between substates were estimated and used to project disease paths in the whole population. In this mid- to old-age population, the most likely disease path was “Healthy → at metabolic risk → mortality” (37%) for healthy participants at baseline, and the most likely disease path was “At metabolic risk → mortality” (51%) for at-metabolic-risk participants at baseline ( Figure 2 ). The distribution of disease path was similar across sex and race subgroups. Impact: The path of heart disease characterizes the disease course in a summary manner and may elucidate pathophysiology of heart disease at the population level.

OBJECTIVE:Multiple plant-based dietary patterns are inversely associated with gout, although the individual constituents driving this association remain unclear. Dietary lignans, a major group of phytoestrogens abundant in plant foods, are metabolized by the gut microflora and may modulate gout risk. We examined the associations between dietary lignan intake, certain whole grain foods rich in lignans, and incident gout. METHODS:We analyzed data from 122,680 individuals in the Health Professionals Follow-up Study and Nurses' Health Study. We administered a food frequency questionnaire every two to four years. We used Cox models to evaluate associations between dietary lignans, whole grain foods, and confirmed gout. RESULTS:Higher intakes of matairesinol (hazard ratio [HR] comparing extreme quintiles, 0.78; 95% confidence interval [CI], 0.69-0.90; P trend = 0.002) and secoisolariciresinol (HR, 0.78; 95% CI, 0.68-0.89; P trend = 0.002) were both associated with lower gout risk, whereas pinoresinol and lariciresinol were not associated with gout. We found inverse associations of whole grain cold breakfast cereals (HR for those consuming ≥1 serving per day, 0.62; 95% CI, 0.53-0.73), cooked oatmeal/oat bran (HR for those consuming ≥2 servings per week, 0.78; 95% CI, 0.70-0.86), and bran added to food (HR for those consuming ≥2 servings per week, 0.84; 95% CI, 0.74-0.95), but not dark breads or other cooked breakfast cereals, with gout. CONCLUSION:Higher intakes of matairesinol and secoisolariciresinol, as well as whole grain cold breakfast cereals, oatmeal, and added bran, were each significantly associated with lower gout risk. These findings support adherence to healthful plant-based diets for gout and support a potential role of the gut microbiome in gout pathogenesis.

Multi-state Markov models have been used to model the course of chronic disease. However, they are unsuitable to chronic disease where past and present states interplay and affect future states, and the estimated transition probabilities are time-specific which are not straightforward for public health interpretation. We have proposed a multi-state non-Markov framework that splits disease states into substates conditioning on past states. As the substates track past states and indicate multimorbidity, the estimated transition rates can be used to derive two summary estimates: Disease path which shows path of state transition, and multimorbidity-adjusted life year (MALY) which represents the adjusted life year in full health. In this paper, we showed the derivation of the two summary estimates and applied them to characterize the course of heart disease using data from the Atherosclerosis Risk in Communities Study (ARIC) study. The course of heart disease was modeled in five states, namely, healthy, at metabolic risk, coronary heart disease (CHD), heart failure, and mortality. In this mid- to old-age population, the estimated MALY was 24.13 (95% CI: 16.55, 32.06) years. For healthy participants at baseline, the most likely disease paths were: Healthy, at metabolic risk, mortality (37%), Healthy, mortality (21%), Healthy, at metabolic risk, heart failure, mortality (19%), and Healthy, at metabolic risk, CHD, mortality(8%). The MALY was higher among women than men and higher among Whites than Blacks. The distribution of disease path was similar across sex and race subgroups. In summary, MALY and disease path characterize the disease course in a summary manner and have potential use in chronic disease prevention.

Background: HDL cholesterol, together with LDL-c and triglycerides, are biomarkers widely used to assess risk of myocardial infarction (MI). Thereafter, apoA1 and HDL particle number (HDL-P) were identified as biomarkers inversely associated with risk of MI. State-of-the-art nuclear magnetic resonance spectroscopy allows for profiling of lipidomic markers within HDL, providing an unprecedented opportunity to identify novel biomarkers of MI. Methods: We examined associations between HDL lipid components and risk of MI in the UK Biobank. We leveraged the most up to date lipidomic data measured by NMR. A total of 8229 MI incidents were documented among 249,140 participants over a median 17.7 years of follow-up. Results: Independent of existing risk factors including LDL-c, triglycerides, apo-A1, HDL-P, the percentage of cholesterol over total lipids in HDL was inversely associated with the risk of MI with Hazard Ratio (HR) of 0.85 (0.76, 0.95), when comparing the highest quintile to the lowest. The percentage of triglycerides in HDL was positively associated with 1.13 (1.01, 1.26). We created the ratio of cholesterol to triglyceride within HDL and found it to be inversely associated with risk of MI (HR=0.89 [0.79, 0.99]). This finding is robust and consistently supported across subclasses of HDL. Conclusions: The ratio of cholesterol to triglycerides within HDL particles may be a new marker of risk for MI, independent of existing biomarkers. Table 1 . Hazard Ratios for Myocardial Infarction (MI) across Percentages of Lipids in HDL. Model 1: Cox model adjusted for conventional covariates: Age, Sex, Race, Body Mass Index, Hypertension status, Statin Use, Type 2 Diabetes status, Smoking, Education, Alcohol intake, and Townsend Deprivation Index, LDL-C, and TriglyceridesModel 2: Cox model adjusted for Model 1 covariates plus Apolipoprotein A1 and HDL particle number.

In chronic disease epidemiology, investigation of disease etiology has largely focused on one single endpoint, and progression of chronic disease as a multi-state process is understudied, representing a knowledge gap. Most of existing multi-state regression models require Markov assumption and are unsuitable to estimate progression of chronic diseases that is largely non-memoryless. We propose a new non-Markov framework that allows past states to affect transition rates of current states. The key innovation is that we convert a non-Markov to Markov process by dividing disease states into substates through conditioning on past disease history. Specifically, we apply cause-specific Cox models (CSC) including past states as covariates to obtain transition rates (TR) of substates, which were used to obtain transition probabilities (TP) and state occupational probabilities (SOP) of substates. We applied our model to describe progression of coronary heart disease (CHD) in the ARIC study, where CHD was modeled in healthy, in risk, CHD, heart failure, and mortality states. We presented transition rates, transition probabilities, and state occupational probabilities between states from age 45 to 95 years. In summary, the significance of our framework lies in that transition parameters between disease substates may shed light on new mechanistic insight of chronic disease and may provide more accurate description of non-Markov process than Markov regression models. Our method has potential of wide application in chronic disease epidemiology.

As the most commonly consumed dietary supplement, how multivitamin use associates with risk of mortality remains inconclusive. Methods: We used data from three cohort studies to examine the association of time-varying multivitamin use with risk of mortality. We also emulated a target trial to estimate the association by applying the parametric g-formula. Results: Of 241,068 eligible participants (mean age at baseline: 50 for NHS, 37 for NHSII, and 53 for HPFS), there were 44,415 deaths over 30-years of follow up. Compared to non-users, the pooled age-adjusted harzard ratio (HR) of mortality was 0.92 (95% CI: 0.89, 0.94) and 0.89 (95% CI: 0.88, 0.91) for users who used <1 and ≥1 pill/day, respectively. These results were attenuated after multivariable adjustment (0.97 [0.95, 1.00] and 0.98 [0.96, 0.99], respectively), and the inverse association was primarily due to lower risk of CVD mortality ( table 1 ). In the emulated target trial, the estimated 30-y all-cause mortality under a multivitamin intervention was slightly lower than no intervention ( table 2 ). Conclusions: Multivitamin use may relate to a small reduction in mortality among middle-aged adults.

Objectives: Studies suggest that endogenous estrogen may be involved in regulating body weight. Dietary lignans can act as estrogen receptor agonists and may be beneficial for weight control. We prospectively examined the association between changes in dietary lignan intake with weight change. Methods: We analyzed data from 124,875 men and women in the Nurses’ Health Study (1986-2010), Nurses’ Health Study II (1991-2011), and the Health Professionals Follow-up Study (1986-2010). We calculated 4-year changes in total lignan intake as well as in four major individual lignans (matairesinol, secoisolariciresinol, pinoresinol, and lariciresinol) using data from a semi-quantitative food frequency questionnaire. Participants reported their height and weight at baseline and updated their current weight every 2 years thereafter, from which we calculated 4-year changes in weight as the primary study outcome. We used multivariable generalized linear regression models to examine the association between 4-year changes in lignan intake with 4-year weight changes over the same period, and these models were adjusted for age, baseline BMI in each period, and relevant lifestyle factors. Results: Each 1-standard deviation (SD) increase in total lignan intake was associated with a 4-year weight change of -0.36 lbs (95% confidence interval [CI] -0.40 to -0.33 lbs; P for trend <0.0001). This inverse association persisted among all four individual lignans, although the magnitude of this association was the strongest for lariciresinol (-0.63 lbs [95% CI, -0.67 to -0.59 lbs]; P for trend <0.0001) (Figure) . The observed associations were stronger among certain subgroups, including those with BMI ≥25 kg/m 2 , below-median diet quality, or below-median physical activity level. Conclusions: Increases in dietary lignan intake were associated with less weight gain over the same 4-year period. Our findings support existing dietary recommendations to consume a healthy plant-based diet.

Importance: Consumption of energy drinks has increased drastically in recent years, particularly among young people. It is unknown whether intake of energy drinks is associated with health during pregnancy.Objective: To examine associations of energy drink intake before and during pregnancy with risk of adverse pregnancy outcomes (APOs).Design, setting, and participants: This prospective cohort study included data from women enrolled in the Nurses' Health Study 3 (NHS3) between June 1, 2010, and September 27, 2021, and the Growing Up Today Study (GUTS) who reported 1 or more singleton pregnancy from January 1, 2011, to June 1, 2019. Data were analyzed from October 1, 2021, to September 28, 2023.Exposure: Intake of energy drinks, assessed by food frequency questionnaire.Main outcomes and measures: The main outcomes were self-reported APOs, including pregnancy loss, gestational diabetes, gestational hypertension, preeclampsia, or preterm birth, and a composite APO, defined as development of any of the APOs. Risk of APOs was compared between consumers and nonconsumers of energy drinks.Results: This study included 7304 pregnancies in 4736 participants with information on prepregnancy energy drink intake and 4559 pregnancies in 4559 participants with information on energy drink intake during pregnancy. There were 1691 GUTS participants (mean [SD] age, 25.7 [2.9] years) and 3045 NHS3 participants (mean [SD] age, 30.2 [4.1] years). At baseline, 230 GUTS participants (14%) and 283 NHS3 participants (9%) reported any intake of energy drinks. While no associations were found for pregnancy loss (odds ratio [OR], 0.89; 95% CI, 0.71-1.11), preterm birth (OR, 1.07; 95% CI, 0.71-1.61), gestational diabetes (OR, 0.89; 95% CI, 0.58-1.35), preeclampsia (OR, 0.73; 95% CI, 0.41-1.30), or the composite APO (OR, 1.05; 95% CI, 0.87-1.26), prepregnancy energy drink use was associated with a higher risk of gestational hypertension (OR, 1.60; 95% CI, 1.12-2.29). A significant interaction was found between age and energy drink intake in relation to hypertensive disorders (P = .02 for interaction for gestational hypertension; P = .04 for interaction for any hypertensive disorders), with stronger associations for participants above the median age. No associations of energy drink intake during pregnancy with any of the APOs were found in NHS3 (eg, any APO: OR, 0.86; 95% CI, 0.41-1.79).Conclusions and relevance: In this study, energy drink intake before pregnancy was associated with an elevated risk of gestational hypertension. Given the low prevalence of energy drink intake and low consumption levels among users, the results should be interpreted cautiously.