Babies’ Blood Proteins and Gut Microbes Predicted Obesity 26 Years Later

A study following 16,683 newborns in Sweden for more than 26 years has identified specific proteins in cord blood and gut bacteria present at age one that together predict obesity risk with an accuracy of 83%. The findings, published in mSystems, suggest that the biological groundwork for adult obesity may be laid in the first months of life, well before any weight gain becomes visible.

Cord Blood Proteins and Gut Bacteria Forecast Decades of Weight Gain

The researchers, led by Angelica P. Ahrens at the University of Florida and Johnny Ludvigsson at Linköping University in Sweden, drew on the All Babies in Southeast Sweden (ABIS) cohort, which recruited births between 1997 and 1999. Infants who went on to develop obesity showed measurable biological differences from the very beginning. Elevated cord blood levels of three proteins stood out: angiopoietin-like protein 4 (ANGPTL4), follistatin (FST), and hepatocyte growth factor (HGF). These elevations were independent of maternal BMI, meaning they were not simply a reflection of having an overweight mother. At the same time, infants who later became obese had lower cord blood concentrations of isocaproic acid, tryptophan, and oleic acid. By age one, their gut microbiomes already looked different: five bacterial genera, including Akkermansia, Phascolarctobacterium, and Slackia, were less abundant compared to children who maintained healthy weights. Research on early-life gut microbiome development and long-term health has shown that microbial communities established in infancy shape immune and metabolic programming for years afterward, lending biological plausibility to these early divergences.

Why a 26-Year Follow-Up Changes What We Know About Childhood Obesity

Most studies connecting the infant microbiome to obesity have measured outcomes in early or middle childhood, leaving open the question of whether early signals persist into adulthood. This cohort's mean follow-up of 25.3 years closes that gap. Obesity diagnoses were pulled from the Swedish National Patient Register using ICD-10 codes and supplemented with self-reported BMI, giving the team a robust outcome measure across a large population. When researchers combined birth proteomics with one-year gut bacteria data in machine learning models, the area under the curve (AUC) reached 0.83, indicating that the models correctly ranked obese versus non-obese individuals 83% of the time. Adding parental BMI and the child's own BMI before age eight pushed that figure to 0.89. That improvement from clinical variables is meaningful, but so is the 0.83 baseline: it means biology measurable at birth already carries most of the predictive signal, before any family history information is included.

How a Swedish Cohort of 16,000 Births Made the Science Possible

The scale and design of ABIS are what allow these conclusions to stand up. Cord blood metabolomics were performed on 290 samples using liquid chromatography-tandem mass spectrometry; targeted proteomics via the Olink platform covered 358 samples; and 16S rRNA gene sequencing of one-year fecal samples covered 1,743 children. Early-life factors beyond biology were also recorded: maternal smoking, infections, psychosocial stressors, and diet in the first year were all linked to obesity risk by age 26. Four protein biomarkers recurred across multiple analyses: fibroblast growth factor 19 (FGF19), ANGPTL4, sulfotransferase family 2A member 1 (SULT2A1), and interleukin 20 (IL-20). FGF19 is secreted by the ileum and regulates bile acid synthesis and glucose homeostasis. SULT2A1 handles bile acid detoxification. ANGPTL4 inhibits intestinal lipase activity and is itself regulated by microbiota, particularly short-chain fatty acid-producing bacteria like Roseburia and Blautia. Taken together, these markers converge on bile acid metabolism as a plausible early-life mechanism connecting gut microbial composition to long-term fat accumulation and insulin regulation.

Open Questions: Sample Sizes, Causality, and Functional Gaps

The study has real limitations. Some omics subgroups were small; the proteomics and metabolomics samples numbered in the hundreds rather than the thousands covered by microbiome sequencing. Clinical obesity diagnoses drawn from a national register likely undercount true prevalence, since many individuals with obesity never receive a formal code. There was no external validation cohort, so the machine learning models have not yet been tested in a different population. The microbiome data relied on 16S rRNA gene sequencing, which identifies bacteria by taxon but does not reveal what those bacteria are actually doing, because functional metagenomic annotation was not performed. The broader challenge of establishing causality between early microbiome composition and obesity outcomes remains unresolved; as work on antibiotic exposure and obesity has shown, identifying associations is considerably easier than tracing the mechanisms that drive them.

Personalized Prevention Strategies Could Start at Birth

The authors frame the clinical implication carefully: these markers identify measurable targets for preventing obesity from early life, not a diagnostic test ready for deployment. Replication in an independent birth cohort, ideally with functional microbiome profiling added, would be the logical next step. If the markers hold up, the practical question becomes what to do about them. Identifying high-risk infants before weight gain begins opens a window for early interventions, whether dietary, microbial, or behavioral. Research into probiotic strategies targeting early-life metabolic risk suggests one possible avenue, though evidence for specific interventions in this age group remains preliminary. What the ABIS data establish, at minimum, is that the 26-year arc of obesity risk is already being written in cord blood and infant gut flora, and that reading it accurately may one day allow clinicians to intervene before the story is set.