Skip to main content
Dryad logo

Lens fluorescence and skin fluorescence in the Copenhagen Twin Cohort Eye Study: Covariates and heritability


Bjerager, Jakob (2021), Lens fluorescence and skin fluorescence in the Copenhagen Twin Cohort Eye Study: Covariates and heritability, Dryad, Dataset,


Lens and skin fluorescence are related to the systemic accumulation of advanced glycation end products, which is accelerated in diabetes. We have examined lens fluorescence and skin fluorescence in healthy adult twins. The study enrolled twins aged median 59 years from a national population-based registry. Diabetic individuals were excluded from analysis. The interrelatedness between fluorescence parameters and relations between fluorescence and age, current HbA1c and smoking pack years were examined using correlation tests and mixed model linear regression analyses. Broad-sense heritability was analyzed and compared for lens fluorescence, skin fluorescence and HbA1c. Lens fluorescence and skin fluorescence were crudely interrelated (R = 0.38). In linear regression analyses, age explained a larger fraction of the variance in lens fluorescence (R2 = 32 %) than in skin fluorescence (R2 = 20 %), whereas HbA1c explained smaller variance fractions (R2 = 3 % and 8 %, respectively) followed by smoking pack years (4 % and 3 %, respectively). In multivariate analyses, age, HbA1c and smoking pack years combined explained more of the variance in lens fluorescence (R2 = 35 %) than in skin fluorescence (R2 = 21 %), but the influence of HbA1c on lens fluorescence was not statistically significant (p = .2). Age-adjusted broad-sense heritability was 85 % for lens fluorescence, 53 % for skin fluorescence and 71 % for HbA1c in best fitting heritability models. Both fluorescence parameters increased with age, current glycemia and cumulative smoking. Lens fluorescence was found to be a predominantly heritable trait, whereas skin fluorescence was more influenced by environmental factors and closer related to current glycemia. The results suggest that skin fluorophores have a faster turn-over than lens fluorophores.


Data sources:

Lens fluorescence was measured in the right eye of phakic subjects using a commercial ocular fluorometer (Fluortron Master TM-2 with Windows software, revision B.17, OcuMetrics, Mountain View, California, USA), approximately 1 hour after dilation with tropicamide 1% eye drops. The device measures blue-green fluorescence at incremental steps of 0.125 mm along the optical axis of the eye using excitation light at 430-490 nm and detection at 530-630 nm with results reported in units of equivalent fluorescein concentration in water. Measurements were performed under scotopic lighting conditions. Absorption-corrected anterior lens peak fluorescence was calculated using the manufacturer’s software. Lens fluorescence peak values were corrected for ambient light in the examination room by subtraction of the lowest fluorescence readings in each scan (averaged from the fluorescence intensities of the 15/148 steps with the lowest intensities). Three successful scans were attempted for each subject. Unsuccessful scans counted scans with ambient background light values above 30% of the posterior absorption-corrected lens peak fluorescence, as recommended by the manufacturer, and if blinking had occurred at critical points during the scan. Lens fluorescence values used for data analysis were based on the average absorption-corrected anterior peak value of three successful scans. Study subjects in whom three successful scans could not be obtained were omitted from analysis. Skin fluorescence was measured on the anterior forearm with a designated commercial device (Diagnoptics AGE Reader, Diagnoptics Technologies B.V., Groningen, Netherlands). The instrument emits light at 300-420 nm, with peak intensity at 370 nm, on a 4 cm2 skin area and measures emission at 300-600 nm. Data output is a double-digit arbitrary unit (AU) index of 420-600 nm fluorescence relative to reflected 300-420 nm emission multiplied by 100. The average of three readings per subject was used for analyses. According to the manufacturer’s guidelines, subjects with excessive sweating, tattoos, recently applied skin cream or recent intensive sunbathing affecting the skin region of interest were excluded from analysis. The first measurement after device start-up was routinely discarded, as recommended by the manufacturer.

We choose to include three lens fluorescence and three skin fluorescence readings per subject since three are enough to evaluate reproducibility of measurements and the maximum to which one can reasonably expose a study participant in a study that also includes other procedures. Also, unilateral measurements only do not require statistical adjustments due to paired organ data dependencies.

Blood samples obtained during the examinations were analyzed for HbA1c . Data on accumulated smoking pack years were obtained by interview. Participants who reported <1 pack year were categorized as non-smokers.

Statistical analyses:

Microsoft Excel 360 for Windows 10 was used for demographic statistics, GraphPad Prism v9.0.0.121 for reproducibility of measurements analyses and R-Studio v1.2.5001 for Windows 10 was used for all other statistical analyses. Normality was tested by Shapiro-Wilk normality tests. Fluorescence parameters were transformed by log10 to obtain normal distributions. All fluorescence values reported have been back-transformed to geometric mean values with 95% confidence interval. Non-parametric parameters were reported in medians and inter-quantile ranges. Reproducibility of fluorescence measurements were assessed by three-group one-way ANOVA analyses of log10-transformed data. Pearson’s correlation tests were used in case of normal distributions and Spearman’s rank correlation tests were used for non-normally distributed parameters. Univariate and multivariate log-level linear mixed model regression analyses adjusted for twin-pair dependencies were performed with the R functions ‘lmer()’ (lme4 v.1.1.26 package) and ‘modelTest ()’ (JWileymisc v. 1.2.0 package). Reported coefficient estimates from regression analyses were transformed by antilog to designate percentage change in fluorescence per unit change in either age (years), smoking (pack years) or HbA1c (mmol/mol),. Broad-sense heritability was calculated for lens fluorescence, skin fluorescence and HbA1c using a linear regression model of each parameter as a function of age with the R function ‘twinlm()’ (mets v. package). Best-fit models were found by Akaike’s information criterion (AIC). The lowest AIC-value defined the best fitting model for lens fluorescence, skin fluorescence and HbA1c, but models with AIC-values between the value of the best fitting model and the value of the best model plus two units were considered non-inferior to true best fitting models.

Usage Notes

Please review the spreadsheet tab "ReadMe_tab" for usage notes.