Metabolic syndrome is one of those conditions where the public description and the research description have drifted apart. The public version is a shopping list. The research version is a set of correlated lab values, and the interesting question is not whether they cluster but whether the cluster predicts anything you would have predicted anyway.
A 2025 cohort gives a clean answer to that, and it also shows how far a large cohort can go and where it stops.
What the study is
A prospective analysis built on the China Health and Retirement Longitudinal Study, which enrolled 17,705 middle-aged and older people. The subject is cardiovascular-kidney-metabolic syndrome, CKM stages 0 to 3 — stages 0 through 3 means the analysis deliberately includes people with no CKM diagnosis at all, which is the point of staging it that way.
The analysis used Cox proportional hazard models, Kaplan-Meier curves and restricted cubic spline analysis, plus stratified analysis across socio-demographic groups.
The index itself combines three blood markers — C-reactive protein, triglycerides and fasting plasma glucose — into one number. The stated rationale is that it may represent insulin resistance and inflammatory response together, which is why it sits inside a combined cardiovascular-kidney-metabolic framing rather than a metabolic-only one.
The three numbers that matter most
The paper reports its analysed samples, and they are not the same:
- 17,705 enrolled
- 5,723 analysed for cardiovascular disease
- 5,847 analysed for all-cause mortality
That gap is the most honest number in the study. Roughly two-thirds of the enrolled cohort did not make it into the analysed samples. Cohorts do this for ordinary reasons — blood markers not drawn, incomplete linkage to mortality records, participants lost to follow-up — and any of those mechanisms is correlated with health in ways that are not random. When the analysed subset is two-thirds of the enrolled one, the direction of that selection is the first question, and it is rarely the first question.
The findings
- Restricted cubic spline analysis showed a notable non-linear association between the index and cardiovascular disease occurrence.
- The association with all-cause mortality was linear.
- After comprehensive multivariable adjustment, the data showed a 111 per cent increase in overall mortality risk for every one-unit rise in continuous index measurements.
Three things are worth separating in that list.
The shape differs by outcome. Non-linear for cardiovascular disease, linear for all-cause death. A non-linear relationship means there is a threshold region where risk climbs faster than elsewhere — which is clinically the more interesting of the two findings and the one that gets quoted least, because thresholds are harder to put on a poster than a percentage.
111 per cent is a relative figure on a continuous scale. A one-unit rise in a composite index is not a clinically actionable event. It is a step on a scale designed by the authors. The number tells you the marker carries information about mortality risk across its range. It does not tell you what happens to any individual who moves one unit, and it does not give a threshold a clinician could act on — that is a separate piece of work that this study did not do.
Multivariable adjustment is not the same as causation. The authors adjusted for a broad set of covariates. Adjustment removes confounding by what you measured. It cannot remove confounding by what you did not.
Where this belongs relative to the GH axis
There is a reason metabolic health shows up in this literature at all: the growth-hormone axis and insulin sensitivity are linked in both directions, and the peptide research pages track the human data on that class specifically. The CJC-1295 DAC human data page covers what was measured in people rather than in animals — and the honest framing there is the same framing this cohort requires: an association measured in humans is not an intervention.
What this study does not show
It does not show that the index causes cardiovascular disease or death. It is observational. The design supports association, and the authors’ own conclusion is carefully worded as association.
It does not show that lowering the index lowers risk. No one in this study had their index lowered on purpose and was then followed. That is the missing half of the claim, and it is a different study design.
It does not show that the index beats the individual markers. The three components are all established risk markers on their own. The composite is a convenience, and this study does not establish that it carries information the components did not already carry.
It does not generalise beyond the cohort. One national dataset, one country, one age range, and staging that runs to CKM 3. The composition of a Chinese middle-aged and older cohort is not the composition of any other population.
It does not speak to the analysed subset. 5,723 and 5,847 out of 17,705. The paper reports the numbers; it does not establish that the excluded third would have produced the same association.
It does not establish the index as a clinical instrument. The authors call it a potential biomarker. A biomarker that predicts risk in a cohort is a long way from a biomarker that changes decisions.
The question this cohort actually answers
Every observational cohort of this size answers the same question: does this marker carry information beyond what we already measure?
For this index, the answer the paper gives is yes — a substantial association survived full adjustment, and the cardiovascular-disease relationship was non-linear.
The question it does not answer is the one that matters to a person holding a lab result: if this number goes down, does anything change? That is a different study, with different participants and different ethics, and it has not been done here.
Knowing which question you are looking at is most of what separates reading a cohort paper from being recruited by one.
