Comparison of modeling strategies to study the effect of an exposure on changes in continuous outcomes: evaluating the impact of alcohol consumption on changes in cardiometabolic health markers based on the German National Cohort (NAKO).
20 affiliations
- 1Molecular Epidemiology Research Group, Max-Delbrück-Center for Molecular Medicine in the Helmholtz Association (MDC), Berlin, Germany.
- 2Institute for Public Health, Charité - Universitätsmedizin Berlin, Corporate Member of Freie Universität Berlin and Humboldt-Universität zu Berlin, Berlin, Germany.
- 3Epidemiology Research Group, Faculty of Health Sciences Brandenburg, University of Potsdam, Potsdam, Germany.
- 4Department for Biostatistics and Clinical Epidemiology, Centre for Stroke Research, Charité - Universitätsmedizin Berlin, Berlin, Germany.
- 5Population Sciences in the Pacific, University of Hawai'i Cancer Center, Honolulu, HI, United States.
- 6Institute of Epidemiology, Helmholtz Zentrum München-German Research Center for Environmental Health (GmbH), Neuherberg, Germany.
- 7Chair of Epidemiology, Institute for Medical Information Processing, Biometry and Epidemiology, Medical Faculty, Ludwig-Maximilians-Universität München, Munich, Germany.
- 8DZHK (German Centre for Cardiovascular Research), partner site Munich Heart Alliance, Munich, Germany.
- 9German Center for Diabetes Research (DZD e. V.), Neuherberg, Germany.
- 10Institute for Community Medicine, University Medicine Greifswald, Greifswald, Germany.
- 11Department of Epidemiology and Preventive Medicine, University of Regensburg, Regensburg, Germany.
- 12Heidelberg Institute of Global Health, University Hospital Heidelberg, Heidelbrerg, Germany.
- 13Cancer Prevention Graduate School, German Cancer Research Center (DKFZ), Heidelberg, Germany.
- 14Department of Epidemiology, Helmholtz Zentrum für Infektionsforschung, Brausnchweig, Germany.
- 15Institute of Social Medicine, Epidemiology and Health Economics, Charité - Universitätsmedizin Berlin, Berlin, Germany.
- 16Institute for Occupational and Maritime Medicine (ZfAM), University Medical Center Hamburg-Eppendorf, Hamburg, Germany.
- 17Institute of Epidemiology and Prevention, Faculty of Medicine and Medical Center-University of Freiburg, Freiburg, Germany.
- 18Institute for Medical Epidemiology, Biometrics, and Informatics, Interdisciplinary Center for Health Sciences, Medical Faculty of the Martin Luther University Halle-Wittenberg, Halle, Germany.
- 19Biobank Technology Platform, Max-Delbrück-Center for Molecular Medicine in the Helmholtz Association (MDC), Berlin, Germany.
- 20Charité - Universitätsmedizin Berlin, Corporate Member of Freie Universität Berlin and Humboldt-Universität zu Berlin, Berlin, Germany.
Abstract
Background
When investigating effects of baseline exposures on changes in continuous outcomes in longitudinal studies, the change-score model is increasingly recommended, while baseline adjustment is debated. Limited empirical evidence exists on the extent to which these choices matter. We examined how different modeling strategies influence estimated effects using alcohol and prospective changes in cardiometabolic health markers as an example.Methods
We used data from the German National Cohort (NAKO; ∼205 000 participants at baseline, ∼140 000 at 4-5-year follow-up). Participants reported alcohol consumption at baseline (2014-19) and were classified as light, moderate, heavy, former, or never drinkers. Triglycerides (TG), systolic and diastolic blood pressure (SBP and DBP), high-density lipoprotein cholesterol (HDL-C), and glycated hemoglobin were measured at baseline and 5 years later. Longitudinal effects were estimated using a change-score model informed by directed acyclic graphs without baseline adjustment, and compared to a baseline-adjusted model and a model that adjusted for the average of baseline and follow-up (Oldham's). Statistical tests assessed whether changes in the outcome depended on baseline levels.Results
Change-score and Oldham's models produced highly consistent results across all five cardiometabolic markers, whereas baseline-adjusted change-score models yielded different results for all. Statistical tests assessing the relationship between baseline levels and subsequent change did not indicate whether baseline adjustment would materially alter the results. Based on the change-score models, former drinkers had lower 5-year TG change and higher SBP, DBP, and HDL-C changes compared with light drinkers, while heavy drinkers had lower DBP change.Conclusion
When exposure precedes the baseline outcome, baseline adjustment in change-score analyses may substantially alter effect estimates.
Related
- Fish Consumption and Sustainability: A Modeling Study Evaluating the Effects of Fish and Canned Fish Consumption on Nutritional Adequacy and Selected Metrics of Environmental Impact
- Age-Standardized Comorbidity Prevalence of Cardiometabolic Conditions and Socioeconomic Inequalities in Hong Kong: A District Comparison
- Implementation strategies, outcomes and determinants of integrated care for cardiometabolic conditions and mental, neurological and substance use disorders: protocol for a mixed-methods systematic review
- Diet, Genes, or Both? Coffee Consumption, Snacking, and BDNF and TCF7L2 Genotypes in Relation to Cardiometabolic Traits: A Cross-Sectional Study in Polish Adults
- Understanding consumption patterns of sugar-sweetened beverages across stages of adolescence in Mumbai city, India: an exploratory cross-sectional study for insights to develop innovative sugar reduction strategies
BibTeX
@article{bouka2026comparison,
title = {{Comparison of modeling strategies to study the effect of an exposure on changes in continuous outcomes: evaluating the impact of alcohol consumption on changes in cardiometabolic health markers based on the German National Cohort (NAKO).}},
author = {Martina Bouka and Katharina Nimptsch and Ulrike Grittner and Solange Gonzalez Chiappe and Gertraud Maskarinec and Annette Peters and Claudia Meinke-Franze and Hansjörg Baurecht and Heiko Becher and Henry Völzke and Hermann Brenner and Isti Rodiah and Lilian Krist and Michael Leitzmann and Nadia Obi and Peggy Sekula and Stefan N Willich and Rafael Mikolajczyk and Tobias Pischon},
journal = {International journal of epidemiology},
year = {2026},
month = oct,
doi = {10.1093/ije/dyag245},
url = {https://europepmc.org/article/MED/42974917},
note = {CC BY-NC-ND},
}
RIS
TY - JOUR
TI - Comparison of modeling strategies to study the effect of an exposure on changes in continuous outcomes: evaluating the impact of alcohol consumption on changes in cardiometabolic health markers based on the German National Cohort (NAKO).
AU - Martina Bouka
AU - Katharina Nimptsch
AU - Ulrike Grittner
AU - Solange Gonzalez Chiappe
AU - Gertraud Maskarinec
AU - Annette Peters
AU - Claudia Meinke-Franze
AU - Hansjörg Baurecht
AU - Heiko Becher
AU - Henry Völzke
AU - Hermann Brenner
AU - Isti Rodiah
AU - Lilian Krist
AU - Michael Leitzmann
AU - Nadia Obi
AU - Peggy Sekula
AU - Stefan N Willich
AU - Rafael Mikolajczyk
AU - Tobias Pischon
PY - 2026
DA - 2026/10/01
JO - International journal of epidemiology
DO - 10.1093/ije/dyag245
UR - https://europepmc.org/article/MED/42974917
AB - Background When investigating effects of baseline exposures on changes in continuous outcomes in longitudinal studies, the change-score model is increasingly recommended, while baseline adjustment is debated. Limited empirical evidence exists on the extent to which these choices matter. We examined how different modeling strategies influence estimated effects using alcohol and prospective changes in cardiometabolic health markers as an example.Methods We used data from the German National Cohort (NAKO; ∼205 000 participants at baseline, ∼140 000 at 4-5-year follow-up). Participants reported alcohol consumption at baseline (2014-19) and were classified as light, moderate, heavy, former, or never drinkers. Triglycerides (TG), systolic and diastolic blood pressure (SBP and DBP), high-density lipoprotein cholesterol (HDL-C), and glycated hemoglobin were measured at baseline and 5 years later. Longitudinal effects were estimated using a change-score model informed by directed acyclic graphs without baseline adjustment, and compared to a baseline-adjusted model and a model that adjusted for the average of baseline and follow-up (Oldham's). Statistical tests assessed whether changes in the outcome depended on baseline levels.Results Change-score and Oldham's models produced highly consistent results across all five cardiometabolic markers, whereas baseline-adjusted change-score models yielded different results for all. Statistical tests assessing the relationship between baseline levels and subsequent change did not indicate whether baseline adjustment would materially alter the results. Based on the change-score models, former drinkers had lower 5-year TG change and higher SBP, DBP, and HDL-C changes compared with light drinkers, while heavy drinkers had lower DBP change.Conclusion When exposure precedes the baseline outcome, baseline adjustment in change-score analyses may substantially alter effect estimates.
SN - 0300-5771
SN - 1464-3685
ER -