Changing diagnostic criteria and their impact on metabolic syndrome: a systematic review

Metabolic syndrome was classified at markedly different rates when the same adult populations were assessed with alternative diagnostic definitions, according to a systematic review of nine observational studies from seven countries. Within individual populations, reported prevalence ranged from 11.5% to 70.1%, and differences between definitions reached 45.2 percentage points. The review, titled “Changing diagnostic criteria and their impact on metabolic syndrome: a systematic review,” compared definitions within the same participants, avoiding the confounding that can occur when prevalence estimates from unrelated populations are placed side by side.

Researchers searched PubMed/MEDLINE, Scopus, Web of Science Core Collection and Embase through 31 July 2026, then screened records and assessed full texts. Of 2,696 database records identified, 629 duplicates were removed and 2,067 records underwent title and abstract screening. Sixty-six reports were assessed in full text, 54 were excluded, and three additional candidates were removed because of overlapping cohorts or insufficient independently verifiable comparative data.

Nine studies, published from 2007 to 2024, were retained for qualitative synthesis. The included populations came from Turkey, Seychelles, Iran, Finland, China, Ethiopia and Pakistan. Sample sizes ranged from 314 adults with type 2 diabetes in Ethiopia to 18,414 adults in a nationwide Iranian survey.

The frameworks compared included World Health Organization criteria, the National Cholesterol Education Program Adult Treatment Panel III, also called NCEP-ATP III, the International Diabetes Federation, the American Heart Association/National Heart, Lung, and Blood Institute, the Joint Interim Statement, the European Group for the Study of Insulin Resistance, American Association of Clinical Endocrinology criteria, and harmonized or region-specific modifications. Because populations, settings, diagnostic structures and component thresholds differed substantially, the investigators used descriptive synthesis rather than pooling prevalence into a single meta-analysis. The definitions differed not only in their numerical cutoffs but also in their architecture.

WHO and EGIR explicitly incorporated insulin resistance, while NCEP-ATP III, IDF, AHA/NHLBI and JIS relied more heavily on combinations of routinely measured abnormalities such as central obesity, raised triglycerides, reduced high-density lipoprotein cholesterol, high blood pressure and abnormal glucose regulation. Some frameworks required two abnormalities and others three; classical IDF criteria made central obesity mandatory, whereas several component-based definitions treated waist circumference as one feature among several. The number of abnormalities, the components given special weight and the treatment of insulin resistance therefore helped determine which people met the definition, even when the same measurements were used.

Prevalence differences were substantial even when the population remained unchanged. In the Finnish study, estimates ranged from 17.7% under WHO criteria to 43.0% under JIS, a difference of 25.3 percentage points. The nationwide Iranian survey produced estimates of 38.3% with NCEP-ATP III, 43.5% with international IDF, 40.9% with AHA/NHLBI and 47.6% with JIS.

Regional IDF criteria yielded 32.0%, compared with 43.5% under international IDF, while regional JIS produced 40.8% rather than 47.6%. In the rural Chinese population, estimates for men ranged from 11.5% to 39.7%, while estimates for women ranged from 15.7% to 54.2% across the frameworks evaluated. In the Ethiopian clinical population, prevalence ranged from 31.2% under WHO criteria to 70.1% under NCEP-ATP III.

Concordance statistics showed a similar dependence on diagnostic structure. The highest reported agreement in the evidence base was between IDF and JIS criteria in Finland, with Cohen’s κ of 0.97. Other close pairings included modified NCEP-ATP III with IDF in Pakistan, with κ of 0.960, and AHA with NCEP-ATP III in an Iranian cohort, with κ of 0.909.

Agreement was generally weaker when an insulin-resistance-centered framework was compared with a component-based one. In the Turkish population, κ values ranged from 0.77 to 0.84 among NCEP, AACE and IDF comparisons but from 0.32 to 0.37 when those frameworks were compared with WHO and EGIR. These values measure agreement between classifications, not the clinical superiority of either system, because the review identified no biological gold standard for metabolic syndrome.

High agreement between two definitions therefore cannot by itself establish which one is better at predicting clinical outcomes. Central obesity and waist circumference were the components most consistently associated with disagreement. In Seychelles, the adiposity component was present in 35.2% of participants under ATP III, 61.5% under WHO and 55.8% under IDF, with only 48% meeting the criterion under all three.

The clearest threshold effects appeared in Iran. In the Tabari cohort, prevalence declined from 44.6% with international IDF criteria to 30.8% with Iranian-specific criteria, an absolute difference of 13.8 percentage points. In the nationwide survey, the corresponding decline was from 43.5% to 32.0%, or 11.5 percentage points.

Other components also shifted in prominence: elevated blood pressure was most common among Finnish men, central obesity among Finnish women, and different lipid and waist components emerged most often under different criteria in the Iranian Shiraz cohort. These were classification differences, not evidence that the underlying biology of the population changed. Findings for age and sex were descriptive rather than directly comparable across studies.

In the Finnish and Iranian populations, older age groups generally had higher measured prevalence, but the studies used inconsistent age categories and reporting procedures. Sex-specific estimates were available for rural China, where the range across definitions was 28.2 percentage points for men and 38.5 points for women. The evidence did not show all ethnic populations moving in a consistent direction under localized criteria; the directly documented ethnicity-specific effect involved waist-circumference thresholds.

Waist circumference is affected by ethnic group, sex, age and body-fat distribution, which can make a single threshold classify metabolic risk differently across populations. The review could evaluate classification changes, however, but it could not determine whether any definition better predicted future diabetes, cardiovascular disease or mortality. The evidence base has important limitations.

Only nine studies qualified, several lacked complete participant-level cross-classification tables, and some reported ethnicity-specific cutoffs inconsistently. The risk-of-bias review was not conducted independently and was described by the authors as provisional. Three studies were provisionally rated at moderate risk of bias and the remainder at low to low-moderate risk; none was judged high risk.

Some populations were clinical rather than general, including 314 Ethiopian adults with type 2 diabetes, which limits generalization. The researchers also reported incomplete database-level retrieval counts for all planned information sources. Reporting was incomplete in other ways: prevalence estimates and Cohen’s κ values were generally more available than full participant-level cross-tabulations, making the size and direction of individual reclassification difficult to quantify in some studies.

Accordingly, the synthesis is not a worldwide estimate of metabolic syndrome prevalence, but a structured comparison showing how the same individuals can be classified differently under competing frameworks. Overall, the review shows that metabolic syndrome classification is strongly definition-dependent, with waist-circumference thresholds a particularly important source of reclassification. The authors call for future studies to report exact thresholds, provide cross-classification data and compare competing definitions against clinically relevant outcomes.

The present evidence supports the narrower conclusion that diagnostic definitions materially shape who is counted as having metabolic syndrome, while further outcome-based research is needed to establish which definitions identify meaningfully different risk profiles.

Subject of Research: Systematic review of how metabolic syndrome diagnostic definitions affect adult prevalence, classification and concordance

Article Title: Changing diagnostic criteria and their impact on metabolic syndrome: a systematic review

Article References: Obeidat, A. A., Abu Al-Haijaa, D. M., Rahahleh, R. J., Ahmad, M. N., Jaradat, S., Nassar, N. A., Khamis, N. A., & Azzeh, F. S. (2026). Changing diagnostic criteria and their impact on metabolic syndrome: a systematic review. Frontiers in Endocrinology, 17, Article 1976380. https://doi.org/10.3389/fendo.2026.1976380

Image Credits: AI Generated

DOI: 10.3389/fendo.2026.1976380

Keywords: metabolic syndrome, diagnostic criteria, systematic review, prevalence, diagnostic concordance, waist circumference, ethnic thresholds, WHO criteria, IDF criteria, NCEP-ATP III, Changing, diagnostic