🧬 Aging Wiki

skin-autofluorescence-age-reader

Properties1
aliasesSAF, AGE Reader, skin AGEs, skin autofluorescence, DiagnOptics AGE Reader, fluorescent advanced glycation end-products, tissue AGE biomarker

21 min read

Skin Autofluorescence — AGE Reader

Skin autofluorescence (SAF), measured via the AGE Reader device (DiagnOptics Technologies, Groningen, Netherlands), is a non-invasive fluorescence spectroscopy method for in-vivo quantification of fluorescent advanced glycation end-product (AGE) accumulation in dermal collagen. A handheld UV light source excites the forearm skin at ~370 nm; emission integrated over 420–600 nm (blue-green fluorescence) is ratioed against the excitation intensity to produce a dimensionless reading in arbitrary units (AU). The device is FDA-cleared (Class II), CE-marked, and commercially deployed in diabetes clinics and cardiovascular risk stratification programs across Europe and North America.

SAF is conceptually distinct from DNA-methylation epigenetic clocks: it measures an accumulation of structural molecular damage (covalent AGE crosslinks on long-lived ECM proteins) rather than an epigenetic state change. Because dermal collagen-I turns over on a ~15-year half-life, the SAF signal reflects cumulative lifetime AGE burden — an irreversible structural record of metabolic insults integrated over decades. This makes SAF useful as a “tissue memory” biomarker that is difficult to transiently manipulate but provides information orthogonal to blood-based biomarkers.

Identity and Device Specifications

  • Manufacturer: DiagnOptics Technologies B.V., Groningen, Netherlands
  • Device name: AGE Reader (also “AGE Reader SU” for clinical/standalone version); second-generation device launched ~2010
  • Regulatory status: FDA-cleared Class II medical device; CE-marked (EU)
  • Primary citation (device validation): Meerwaldt R et al. (2004) Diabetologia 47:2558–2565 — doi:10.1007/s00125-004-1451-2
  • Measurement principle: UV-A excitation (~370 nm); emission read 420–600 nm (blue-green fluorescence window); autofluorescence ratio = emission/excitation. Reading standardized against a white ceramic tile reference on every measurement.
  • Site of measurement: Volar (inner) forearm — preferred because skin is relatively thin, unpigmented in most individuals, and minimally sun-damaged. Forehead and other sites have been studied but volar forearm is the clinical standard.
  • Output units: AU (arbitrary fluorescence units) — typical range ~1.0–4.0 in adults; no absolute physical unit equivalent
  • Measurement duration: ~30 seconds per reading; 2–3 readings are typically averaged
  • Archive status: Meerwaldt 2004 — doi confirmed (bronze OA; download failed — Springer paywall); Meerwaldt 2005 — doi confirmed (not_oa); van Waateringe 2019 — doi confirmed (downloaded); Boersma 2024 — doi confirmed (downloaded); Reurean-Pintilei 2024 — doi confirmed (downloaded); Koetsier 2010 — doi confirmed (downloaded); Chen 2022 — doi confirmed (not_oa)

Measurement Principle and Molecular Basis

Fluorescent AGEs detected

The SAF reading captures only the fluorescent AGE subset of total tissue AGEs. This is an important caveat: the most abundant AGE in human tissue (CML, carboxymethyl-lysine) is not fluorescent and is not detected by SAF. The major fluorescent AGEs contributing to the SAF signal are:

Fluorescent AGE speciesChemistryContribution to SAF
PentosidineLys–Arg crosslink; oxidative; pentose-derivedStrong; r=0.75 correlation with SAF in Meerwaldt 2005 1
Collagen-linked fluorescence (CLF)Heterogeneous; multiple crosslinks at ~370/440 nmStrong; r=0.71 with SAF in Meerwaldt 2005 1
Glucosepane-derived fluorescencePartial; see glucosepanePartial contributor; exact fraction unclear
CML (carboxymethyl-lysine)Non-crosslink adduct; glycoxidationNOT fluorescent at this wavelength — NOT detected
CEL (carboxyethyl-lysine)Methylglyoxal-derivedNOT fluorescent at this wavelength — NOT detected

The non-detection of CML is a fundamental analytical limitation. CML is the dominant non-crosslink AGE by mass in tissues and plasma, and many downstream effects of AGEs (RAGE signaling, inflammation) are driven by CML and other non-fluorescent species. SAF measures cumulative crosslink burden, not total AGE burden. See advanced-glycation-end-products for the full AGE taxonomy.

Why dermal collagen?

Dermal collagen-I is one of the body’s longest-lived structural proteins, with an estimated half-life of ~15 years in adult humans [^dermis-page]. Because AGE formation is essentially irreversible under physiological conditions and accumulates proportionally to glucose exposure × time, dermal collagen acts as a metabolic integration register — its current AGE content reflects the individual’s cumulative glycemic and oxidative stress history rather than current conditions. This is why SAF responds slowly to short-term interventions but discriminates long-term metabolic exposure differences between individuals.

The physics: UV-A photons (~370 nm) penetrate the epidermis and excite fluorescent crosslinks in the papillary and reticular dermis. Emitted photons (420–600 nm blue-green fluorescence) are detected by a photodiode array. The ratio method (emission/excitation) partially corrects for inter-individual differences in optical path length and tissue absorbance. See dermis for collagen-I biology and AGE substrate context.

Skin pigmentation interference

The most significant technical limitation of the AGE Reader is interference from melanin pigmentation. Melanin absorbs both the excitation UV-A and the emission blue-green light, attenuating the fluorescence signal in proportion to pigmentation. Darker-skinned individuals (Fitzpatrick phototypes IV–VI) show artificially lowered SAF readings, potentially masking true AGE accumulation 2.

Koetsier et al. (2010) developed a correction algorithm achieving adjusted R² = 80% (three-parameter model) for predicting SAF deviation in individuals with darker skin (UV-reflectance below 12%; cohorts included Afro-Caribbean and southern Chinese subjects) 2. The paper characterises skin color via UV-reflectance rather than Fitzpatrick phototype directly, but the correction addresses the same population (approximately Fitzpatrick IV–VI). Not all AGE Reader firmware versions implement this correction; DiagnOptics documentation should be consulted to confirm whether a given device version applies the Koetsier correction. Clinical interpretation of SAF in patients with darker skin phototypes should be done with caution, and current population-norm reference ranges (derived predominantly from European-ancestry cohorts) may not apply. needs-replication — large validation of SAF in African-ancestry and South Asian populations using the pigmentation correction algorithm has not been published.

Clinical Validation

Original validation: Meerwaldt 2004 — diabetic vs healthy controls

The first in-vivo validation study compared SAF against biochemical analysis of skin punch biopsies in 46 diabetic patients and 46 healthy controls 3:

  • SAF correlated with biopsy-derived AGE markers (CLF, pentosidine, CML, CEL) with r = 0.47–0.62 across all markers (p≤0.002)
  • 70% of T2DM patients showed SAF values above the 95th percentile of healthy controls
  • SAF correlated with diabetes duration, mean HbA1c, and creatinine
  • Procedure: non-invasive, ~30 seconds, no patient discomfort

The differential correlations between fluorescent and non-fluorescent AGEs were further characterised in the Meerwaldt 2005 hemodialysis study 1, which reported r=0.75 for pentosidine and r=0.71 for CLF (both p<0.001) with r=0.45 for CML (p<0.01). This is the paper establishing SAF as a valid proxy for tissue AGE burden — but it also establishes the detection-subset limitation.

DimensionStatus
Pathway conserved in humans?yes — AGE accumulation is a human-specific and universal aging mechanism
Phenotype conserved in humans?yes — the measurement is performed in humans
Replicated in humans?yes — multiple large prospective cohorts

Mortality prediction: Meerwaldt 2005 — hemodialysis cohort

In 109 hemodialysis patients followed for 3 years (38.5% died), SAF independently predicted all-cause mortality with an odds ratio of 3.9 1. This was the first prospective mortality endpoint study and established SAF as clinically predictive beyond standard markers in a high-AGE-burden population. Dialysis patients are particularly informative for AGE biology because impaired renal clearance dramatically accelerates AGE accumulation.

A notable finding in this study: kidney transplant recipients showed intermediate SAF levels (2.5 AU) between dialysis patients (2.8 AU) and healthy controls (2.0 AU), with incomplete reversal of AGE accumulation after restored renal function — consistent with the irreversibility of collagen-bound AGEs 4.

Large prospective cohort: van Waateringe 2019 — LifeLines (n=72,880)

The largest prospective validation study published a LifeLines Cohort analysis (n=72,880 general population participants, 4-year follow-up) 5:

  • Incident T2DM: higher baseline SAF significantly predicted new-onset T2DM (~1.4% 4-year incidence)
  • Incident CVD: higher SAF predicted incident cardiovascular events (~1.7% 4-year incidence)
  • All-cause mortality: higher SAF predicted mortality (~1.3% 4-year incidence)
  • Effects were independent of traditional risk factors including metabolic syndrome components and HbA1c

This established SAF as an independent predictor of future disease across all three major age-related morbidity domains in a general (non-diabetic, non-CVD) population. A non-invasive risk score combining age, BMI, and SAF (AUC = 0.783) was non-inferior to the FINDRISC clinical diabetes-risk calculator (AUC = 0.797) in the same cohort 6.

T2DM cohort prediction: Boersma 2021 (n=2,349)

In 2,349 T2DM patients (median follow-up 3.7 years), SAF predicted the combined CVD/mortality endpoint with OR 2.59 (95% CI 2.10–3.20, p<0.001), incident CVD with OR 2.05 (95% CI 1.61–2.61), and mortality with OR 2.98 (95% CI 2.25–3.94) in multivariate analysis 7.

Cause-specific mortality in the general population: Boersma 2024 (n=77,143)

The most recent large LifeLines analysis (n=77,143 participants without known diabetes or CVD at baseline; median follow-up 115 months; 1,447 deaths) found 8:

  • Cancer mortality: highest SAF quartile HR = 1.7 (95% CI 1.3–2.4) vs lowest, fully adjusted
  • Cardiovascular mortality: fully-adjusted HR 1.4 (95% CI 0.8–2.5) — attenuated and marginally significant after full adjustment (unadjusted HR was 12.6)
  • The attenuation of the CVD mortality signal upon full adjustment (vs the cancer mortality signal which persisted) suggests SAF’s cardiovascular association is partially confounded by shared metabolic risk factors — a finding that moderates the earlier observational signals

contradictory-evidence: the full-adjustment attenuation in Boersma 2024 for cardiovascular mortality (unadjusted HR 12.6 → fully adjusted HR 1.4, CI crossing 1.0) contrasts with the Boersma 2021 T2DM-cohort result where SAF retained independent significance for mortality after multivariable adjustment. This likely reflects the difference between a healthy general population (Boersma 2024) vs a T2DM-enriched cohort (Boersma 2021) and the shared confounding between SAF and conventional CVD risk factors in low-risk individuals.

Cancer incidence: Boersma 2025 (n=77,961)

The cancer-incidence companion analysis from the same LifeLines cohort (n=77,961 enrolled; n=72,038 in fully-adjusted Model 3 due to missing covariate data; median 11.5-year follow-up, IQR 10.5–12.6 years, through March 2023; cancers ascertained via the Dutch Nationwide Pathology Databank [Palga]) found that elevated baseline SAF predicted incident cancer across all participants 9:

  • Unadjusted HR 2.36 (95% CI 2.26–2.45)
  • Age + sex (Model 2) HR 1.15 (95% CI 1.09–1.21)
  • Fully-adjusted HR 1.11 (95% CI 1.05–1.17, p<0.001) — Model 3 covariates: age, sex, waist circumference, BMI, pack-years of smoking (log-transformed), alcohol intake (g/day, log-transformed), diabetes, metabolic syndrome
  • Site-specific signals (fully adjusted): lung, oesophageal, urinary tract p<0.001; vagina/vulva, ovary p<0.05; liver p<0.01

The substantial attenuation from unadjusted HR 2.36 to fully-adjusted HR 1.11 reflects high collinearity between SAF and conventional cancer risk factors — particularly smoking and diabetes, which independently drive both AGE accumulation and cancer risk. The residual HR 1.11 after full adjustment indicates a modest but statistically robust independent association in this large general-population cohort. The cancer-incidence HR 1.11 here is lower than the cancer-mortality HR 1.7 from Boersma 2024 — expected, because cancer incidence captures many low-fatality cancers, whereas cancer mortality stratifies higher-severity outcomes where cumulative AGE burden may matter more. no-mechanism — the mechanistic pathway from tissue-AGE accumulation to cancer initiation is not established.

Meta-analysis: Chen 2022 — MACE risk (n=79,389)

A 2022 meta-analysis of 14 prospective studies (n=79,389 participants) found that higher SAF was associated with increased major adverse cardiovascular events (MACE) with a pooled HR of 1.54 (95% CI 1.31–1.81) 10. This is the strongest available summary effect estimate for SAF as a CV risk predictor.

Outcome domainPooled effect (Chen 2022 meta-analysis)Source
Major adverse cardiovascular events (MACE)HR 1.54 (95% CI 1.31–1.81)10
All-cause mortality (T2DM cohort, OR)OR 2.98 (95% CI 2.25–3.94)7
Incident T2DM (general population)Significant; exact HR not extracted5
Cancer mortality (general population, fully adjusted)HR 1.7 (95% CI 1.3–2.4)8
Cancer incidence (general population, fully adjusted)HR 1.11 (95% CI 1.05–1.17)9

CKD progression: Jin 2022

In T2DM patients, SAF predicted kidney disease progression (HR 1.15 per SD; renal markers accounted for 38.7% of the SAF effect) 11. This positions SAF as a potential marker of the cumulative renal-glycemic damage interaction, independent from — but co-linear with — standard renal biomarkers (eGFR, albuminuria).

Systematic review — CV risk in T2DM: Reurean-Pintilei 2024

A 2024 systematic review of 7 studies (n=8,934 T2DM patients, mean age 63) confirmed SAF as “a valuable, non-invasive marker for evaluating CV risk in T2DM patients” 12. The review noted consistent directionality across all included studies but heterogeneous effect sizes, partly attributable to variation in SAF cutoffs applied and variation in the comparator risk score.

Reference Values and Interpretation

Population norms were established primarily in Dutch/European cohorts. Age-stratified reference ranges:

Age groupApproximate mean AU (healthy individuals)Notes
20–30 years~1.4–1.7 AUYoung adults without metabolic disease
30–50 years~1.7–2.1 AUModerate accumulation; diabetes-dependent accelerated
50–65 years~2.1–2.6 AUSubclinical disease burden increasing
65–80 years~2.6–3.0 AUCommunity-dwelling elderly
80+ yearsoften >3.0 AUHigh disease/medication burden

In T2DM patients: mean SAF ~2.44 AU vs non-diabetics ~2.04 AU in a LifeLines subcohort (n=9,009) 13. The T2DM mean is ~0.4 AU above age-matched controls, reflecting accelerated glycation.

Practical interpretation cutoff: SAF >2.35 AU has been used as an elevated-risk cutoff in some clinical studies (Reurean-Pintilei 2024: sensitivity 67.7%, specificity 56.2%, AUC 0.634 for cardiovascular complications in T2DM), but no single authoritative cutoff has been universally adopted. dose-response-unclear — no standardized clinical decision threshold exists; existing cutoffs derive from individual study populations and may not generalize.

Skin pigmentation caveat: reference values above apply to Fitzpatrick I–IV skin types. Darker skin requires the Koetsier (2010) correction algorithm and the corrected norms; uncorrected SAF in Fitzpatrick V/VI individuals is systematically underestimated. See technical caveat in Measurement Principle section above.

Comparison: T2DM patients vs hemodialysis: dialysis patients accumulate dramatically higher SAF (mean ~2.8 AU in Hartog 2005 vs ~2.0 for controls) because impaired renal clearance prevents elimination of soluble AGE precursors, accelerating tissue deposition beyond the diabetes alone.

Intervention Responsiveness

SAF reflects AGE accumulation on long-lived collagen proteins with ~15-year half-life — this sets a fundamental thermodynamic constraint on how rapidly SAF can change with any intervention.

What does and does not reduce SAF

Intervention typeSAF responseEvidence quality
Intensive long-term glycemic control (T2DM)Modest reduction; slow timecourse (years)Multiple observational studies; 1 RCT subgroup
Bariatric surgery / major sustained weight lossReduction reported in observational follow-upObservational; needs-replication
Hemodialysis (high-flux vs low-flux)SAF unchanged; only plasma AF reducedRamsauer 2015 14; n=?
6-month vitamin D supplementationNo effect on SAFKrul-Poel 2015 15; RCT
Mediterranean diet adherence ≥4 weeksSerum AGE reduction; SAF appears to improve in clinical trialsOliveira 2025 systematic review 16; 6 trials n=2,935 — note: SAF reduction likely reflects serum AGEs and short-timecourse changes; biological plausibility for dermal collagen SAF improvement in 4 weeks is low; interpretation requires caution
Anthocyanin supplementation 12 weeks (impaired glucose tolerance)SAF decreased significantly (anthocyanin group only)Yu 2026 RCT 17; n=68; short-term; needs-replication
Equol supplementation 3 months (postmenopausal women)Significant improvement in equol-producers onlyYoshikata 2021 RCT 18; n=57; subgroup; needs-replication
Topical interventions / short-term anti-aging supplementsNo credible evidence of SAF reductionCollagen turnover timescale prohibits short-term response
Senolytics / rapamycin / NAD precursorsNo published SAF endpoint dataneeds-human-replication

Bottom line for intervention trials: SAF is a poor endpoint for short-duration (< 6–12 months) intervention trials because collagen-AGE half-life exceeds measurement windows. For trials seeking a structural-damage endpoint with AGE readout, plasma or urinary AGE markers (serum CML, urinary glucosepane) may be more responsive on shorter timescales (see glucosepane for the emerging urinary glucosepane assay). SAF is best suited as a prognostic stratifier in trials (i.e., high SAF = high-risk subgroup) rather than as a primary intervention endpoint.

Comparison with Other AGE Measurement Methods

MethodSpecies measuredInvasivenessProsCons
SAF (AGE Reader)Fluorescent AGEs (pentosidine + CLF; partial glucosepane)Non-invasiveFast, cheap, reproducible, predicts outcomesMisses CML (most abundant AGE); dark-skin confound
Serum CML (ELISA)CML (circulating)Blood drawCovers most abundant AGE class; sensitiveReflects recent exposure + clearance, not cumulative tissue burden
Urinary glucosepane (HPLC/MS)Glucosepane (excreted crosslink fragments)Non-invasive (urine)Covers dominant tissue crosslink; emerging technologyLow-throughput; not commercially available; see glucosepane
Skin biopsy + HPLC-AGE analysisMultiple AGE species (pentosidine, CML, CEL, CLF)Invasive (punch biopsy)Gold standard for tissue AGE contentImpractical for large studies; site-specific sampling
Plasma pentosidine (HPLC)Circulating pentosidineBlood drawValidated mortality predictor in CKDCirculating levels don’t directly reflect tissue crosslink burden

SAF occupies the optimal niche for large-scale population studies: non-invasive, fast, reproducible, and predictive of clinical outcomes. Its limitation — missing the non-fluorescent AGE majority — is the price of non-invasiveness.

Aging Context

Position in the biological-age measurement landscape

SAF is a structural biomarker rather than an epigenetic or proteomic clock. Its relationship to epigenetic clocks:

  • SAF measures accumulated molecular damage (irreversible covalent modifications on structural proteins) — the GlycoSENS category in sens-damage-categories
  • Epigenetic clocks (e.g., horvath-clock-2013, grimage-2019, dunedinpace-2022) measure information-state changes in the methylome
  • The two measurement dimensions are conceptually orthogonal and may provide complementary information; head-to-head comparisons in the same cohort are rare as of 2026 needs-replication

SAF is most directly related to the advanced-glycation-end-products process and the altered-intercellular-communication hallmark (downstream RAGE signaling + ECM stiffening). One of the most prominent macroscopic phenotypes downstream of AGE accumulation in the large-artery ECM is arterial-stiffening — the increase in pulse wave velocity and pulse pressure that underlies isolated systolic hypertension in older adults. SAF does not directly measure the dozen hallmarks captured by the epigenetic clocks.

NAT2 acetylator status

In the LifeLines 2016 analysis (n=9,009), a NAT2 slow-acetylator polymorphism was independently associated with higher SAF, suggesting genetic variation in AGE-processing pathways modulates tissue AGE accumulation 13. This is one of the few identified genetic modifiers of SAF and is consistent with NAT2’s role in processing reactive carbonyl species. It also indicates that some inter-individual SAF variation is heritable — a partial genetic confound for epidemiological causal inference.

Limitations and Knowledge Gaps

  1. Fluorescent-AGE subset only. SAF detects ~20–30% of total AGE burden by species diversity. CML, the most abundant AGE and primary RAGE ligand, is undetected. Effect sizes in SAF studies should be interpreted as lower bounds of what full-AGE burden would predict. no-mechanism — the relative clinical contribution of fluorescent vs non-fluorescent AGEs to cardiovascular outcomes is not established.

  2. Skin pigmentation interference. Individuals with darker skin (UV-reflectance below ~12%, approximately Fitzpatrick IV–VI) require the Koetsier 2010 correction algorithm; clinical deployments without this correction produce systematically biased (underestimated) readings. The correction algorithm (adjusted R²=80%) was developed in Afro-Caribbean and Chinese subjects (N=99 model, N=141 validation); its performance across more diverse ancestry groups has not been independently confirmed. needs-replication

  3. Single-device manufacturer. All published SAF data uses the DiagnOptics AGE Reader or its predecessor. No head-to-head comparison against alternative UV-fluorescence spectroscopy platforms exists; inter-platform agreement is unknown. This is also a commercial dependency risk for the field.

  4. Reference ranges derived from European cohorts. The major normative databases (LifeLines, Dutch clinical cohorts) are predominantly of Northern European ancestry and cannot be directly applied to other populations without recalibration.

  5. Confounding by traditional risk factors. The Boersma 2024 LifeLines analysis showed that full adjustment for conventional CVD risk factors substantially attenuates the cardiovascular mortality HR (12.6 → 1.4; CI now crossing 1.0). SAF’s value beyond existing risk scores is modest in fully-adjusted models in a low-risk general population, though it retains significance for cancer mortality and retains strong predictive value in high-risk populations (T2DM, CKD, dialysis). contradictory-evidence — the magnitude of incremental prognostic value above standard risk scores remains contested.

  6. Collagen-AGE irreversibility limits intervention utility. The slow timecourse of change (years to decades) makes SAF unsuitable as a short-term intervention endpoint. The currently positive intervention signals (Mediterranean diet, anthocyanins, equol) in the recency literature are from short-duration trials where the biological plausibility of dermal SAF reduction is low; these results require cautious interpretation and replication.

  7. No MR evidence. Mendelian randomization to test whether genetically higher SAF (via NAT2 or other instruments) causally causes cardiovascular outcomes has not been published. The observational associations may be confounded or reverse-causal. mendelian-randomization: not-tested is assigned because SAF is itself a measurement construct rather than a heritable molecular trait — but genetic modifiers of AGE accumulation could in principle be used as instruments. needs-replication

Cross-references

Footnotes

Footnotes

  1. doi:10.1681/ASN.2005020144 · Meerwaldt R et al. · JASN 2005 · n=109 hemodialysis patients; 38.5% (42/109) died over 3-year follow-up · prospective cohort · OR for all-cause mortality 3.9 (independent predictor); SAF–pentosidine r=0.75 (p<0.001); SAF–CLF r=0.71 (p<0.001); SAF–CML r=0.45 (p<0.01) · archive: confirmed (not_oa; not downloadable) ↩ ↩2 ↩3 ↩4

  2. doi:10.1364/OE.18.014416 · Koetsier M et al. · Optics Express 2010 · skin-color-independent correction algorithm for SAF; adjusted R²=80% for predicting Fitzpatrick V/VI deviations from Caucasian reference values · archive: confirmed (pending download) ↩ ↩2

  3. doi:10.1007/s00125-004-1451-2 · Meerwaldt R et al. · Diabetologia 2004 · n=46 diabetics + 46 healthy controls · cross-sectional validation vs skin biopsy HPLC-AGE reference standard · SAF correlated r=0.47–0.62 (p≤0.002) across CLF, pentosidine, CML, CEL; 70% of T2DM above 95th percentile of controls · archive: confirmed (not_oa; download failed — bronze OA URL not accessible) ↩

  4. doi:10.1196/annals.1329.089 · Hartog JWL et al. · Ann N Y Acad Sci 2005 · dialysis patients 2.8 AU vs controls 2.0 AU (p<0.0001); kidney transplant recipients 2.5 AU (partial reversal) · cross-sectional · needs-replication — full-text not independently verified; DOI not yet confirmed in archive ↩

  5. doi:10.1007/s00125-018-4769-x · van Waateringe RP et al. · Diabetologia 2019 · n=72,880 LifeLines participants; 4-year follow-up · prospective cohort · baseline SAF independently predicted incident T2DM (~1.4%), CVD (~1.7%), mortality (~1.3%); independent of metabolic syndrome and HbA1c · archive: confirmed (pending download) ↩ ↩2

  6. doi:10.1038/s41598-022-26313-9 · Boersma HE et al. · Scientific Reports 2022 · n=58,377 LifeLines participants · prospective cohort · age + BMI + SAF risk score: AUC 0.783 for undiagnosed/incident T2DM; non-inferior to FINDRISC (AUC 0.797) · archive: confirmed (pending download) ↩

  7. doi:10.1186/s12902-020-00676-4 · Boersma HE et al. · BMC Endocrine Disorders 2021 · n=2,349 T2DM patients; median 3.7-year follow-up · prospective cohort · combined CVD/mortality OR 2.59 (95% CI 2.10–3.20, p<0.001); mortality OR 2.98 (95% CI 2.25–3.94); multivariate analysis · archive: confirmed (pending download) ↩ ↩2

  8. doi:10.1038/s41598-024-71037-7 · Boersma HE et al. · Scientific Reports 2024 · n=77,143 participants without diabetes or CVD; 115-month median follow-up; 1,447 deaths · prospective cohort · cancer mortality fully-adjusted HR 1.7 (95% CI 1.3–2.4); CVD mortality fully-adjusted HR 1.4 (95% CI 0.8–2.5) · archive: confirmed (pending download) ↩ ↩2

  9. boersma-2025-saf-cancer-incidence · doi:10.1186/s12885-025-14801-w · PMID 40859200 · PMC12379543 · Boersma HE et al. · BMC Cancer 2025 · n=77,961 enrolled / n=72,038 fully-adjusted Model 3; 11.5-year median follow-up (IQR 10.5–12.6), through March 2023; cancers via Dutch Nationwide Pathology Databank (Palga) · prospective cohort · cancer-incidence unadjusted HR 2.36 (95% CI 2.26–2.45); Model 2 (age+sex) HR 1.15 (1.09–1.21); fully-adjusted HR 1.11 (95% CI 1.05–1.17, p<0.001); site-specific signals lung/oesophageal/urinary-tract p<0.001, liver p<0.01, vagina/vulva/ovary p<0.05; Model 3 covariates: age, sex, waist, BMI, pack-years smoking (log), alcohol g/day (log), diabetes, metabolic syndrome · archive: downloaded (gold OA via BMC / PMC) ↩ ↩2

  10. doi:10.1016/j.numecd.2022.01.032 · Chen Q et al. · Nutr Metab Cardiovasc Dis 2022 · meta-analysis of 14 prospective studies; n=79,389 participants · MACE pooled HR 1.54 (95% CI 1.31–1.81) · archive: confirmed (not_oa; pending download) ↩ ↩2

  11. doi:10.1016/j.numecd.2021.10.007 · Jin Q et al. · Nutr Metab Cardiovasc Dis 2022 · T2DM patients · prospective cohort · SAF independently associated with kidney disease progression HR 1.15 per SD; renal markers accounted for 38.7% of SAF effect · archive: confirmed (pending download) ↩

  12. doi:10.3390/ijms25073889 · Reurean-Pintilei D et al. · Int J Mol Sci 2024 · systematic review of 7 studies; n=8,934 T2DM patients; mean age 63 · SAF as CV risk adjunctive marker in T2DM; SAF cutoff 2.35 AU: sensitivity 67.7%, specificity 56.2%, AUC 0.634 for CV complications · archive: confirmed (pending download) ↩

  13. doi:10.1111/eci.12627 · van Waateringe RP et al. · Eur J Clin Invest 2016 · n=9,009 LifeLines participants · cross-sectional · T2DM mean SAF 2.44 AU vs non-diabetics 2.04 AU; NAT2 slow-acetylator polymorphism independently associated with higher SAF · archive: confirmed (pending download) ↩ ↩2

  14. doi:10.5301/ijao.5000434 · Ramsauer B et al. · Int J Artif Organs 2015 · hemodialysis intervention (high-flux vs low-flux) · skin-AF unchanged by dialysis type; total plasma-AF reduced 21% (low-flux) and 5% (high-flux) · archive: not yet confirmed — DOI only abstracted from search result needs-replication ↩

  15. doi:10.1186/s12933-015-0250-z · Krul-Poel YHM et al. · Cardiovasc Diabetol 2015 · n=not stated; 6-month vitamin D supplementation in T2DM patients · rct · no effect on SAF (skin AGE accumulation) · archive: confirmed (pending download) ↩

  16. doi:10.1093/nutrit/nuae045 · Oliveira JS et al. · Nutrition Reviews 2025 · systematic review of 6 Mediterranean diet clinical trials; n=2,935 participants · Mediterranean diet adherence ≥4 weeks associated with reduced serum AGE concentrations and SAF levels · archive: confirmed (pending download) · caution: short-duration trials; collagen-AGE biological plausibility for SAF reduction at 4-week timescale is low; may reflect serum AGE changes misattributed to SAF ↩

  17. doi:10.1016/j.ajcnt.2026.101210 · Yu L et al. · Am J Clin Nutr 2026 · n=68 individuals with impaired glucose tolerance · rct (randomized double-blind placebo-controlled) · 12-week anthocyanin supplementation 160 mg/day; skin AGEs decreased significantly in anthocyanin group; 55.9% vs 29.4% glucose tolerance reversion (P=0.013) · archive: not yet confirmed — abstract-level only; full PDF not verified needs-replication ↩

  18. doi:10.1371/journal.pone.0257332 · Yoshikata R et al. · PLoS One 2021 · n=57 postmenopausal women · rct · 3-month equol supplementation; significant SAF improvement in equol-producers only; subgroup effect needs-replication ↩

Graph View

Backlinks