Correlation of retino-choroidal thickness and vascular metrics with drusen volume as a severity marker of age-related macular degeneration

Abstract

Purpose

To assess retinal vascular perfusion and choroidal vascularity biomarkers correlated with drusen volume and severity of age-related macular degeneration (AMD).

Methods

Patients underwent swept-source optical coherence tomography angiography (SS-OCTA) (PlexElite-9000). Eyes with geographic atrophy or neovascular AMD were excluded. Retinal thickness, retinal perfusion including superficial (SCP) and deep capillary plexuses (DCP), foveal avascular zone (FAZ), drusen volume, choroidal thickness (ChT) and choroidal vascularity index (CVI) were assessed through the Advanced Research and Innovation Network. Linear mixed model and Spearman test were used for statistical analysis.

Results

We assessed 81 eyes from 57 subjects (34 early-stage, 47 intermediate-stage AMD). The mean age was 74.95 ± 8.79 years. The mean LogMar visual acuity (VA) was 0.16 ± 0.18 (early-stage: 0.12 ± 0.17, intermediate-stage: 0.19 ± 0.18, P = 0.122). Between early and intermediate AMD, no significant differences were seen in SCP and DCP vascular perfusion (P = 0.368, 0.859, respectively), FAZ (p = 0.836) and retinal thickness within the 6-mm area (P = 0.680). Drusen volume showed a significant difference (early-stage: 0.0706 ± 0.1272, intermediate-stage: 0.2102 ± 0.2211mm³, P < 0.01). Intermediate-stage AMD had significantly lower mean ChT (266.40 ± 115.55 vs. 204.97 ± 70.69 µm, P = 0.038) and CVI (0.605 ± 0.021 vs. 0.591 ± 0.015, P = 0.004) within the 5-mm area. Drusen volume was negatively correlated with ChT (r = −0.198, P = 0.017) and CVI (r = −0.209, P = 0.029). No significant correlation was found between drusen volume and VA (r = 0.051, P = 0.143), retinal thickness (−0.03, P = 0.393), FAZ (r = −0.023, P = 0.150), SCP (r = −0.011, P = 0.307), and DCP (r = −0.022, P = 0.190).

Conclusion

Drusen volume, a key AMD severity marker, correlates more strongly with choroidal parameters like ChT and CVI than retinal thickness and perfusion. It may serve as a biomarker for dry AMD severity, with choroidal biomarkers showing earlier disease changes.

Introduction

Age-related macular degeneration (AMD) is a common and progressive degenerative disease among the elderly that leads to central vision loss [1]. It starts with retinal pigment epithelium (RPE) abnormality and accumulation of drusen material between the Bruch’s membrane and the RPE, leading to geographic atrophy (GA) and/or neovascular AMD at the late stage [2]. The exact pathogenesis of AMD remains incompletely understood, but it appears that the most involved layers are RPE, choriocapillaris, and the outer retina [3].

Recent research has highlighted a growing interest in the role of retinal and choroidal vasculature in the pathogenesis of AMD to show how variations in blood supply to retinal tissues may influence tissue damage [1, 4]. The evaluation of retinal blood flow in AMD has utilized various imaging techniques, including colour fundus photographs, colour Doppler imaging, and optical coherence tomography angiography (OCTA) [5–10]. Colour Doppler imaging studies on eyes with AMD investigated increased pulsatility in the central retinal artery [5], with lower end-diastolic velocity with a higher resistance index [6]. However, colour fundus photograph studies showed mixed results, including no differences in eyes with AMD [7, 8], wider venule and arterial calibre in early AMD [11–13], and a smaller arteriole-to-venule diameter ratio in advanced AMD [14].

Studies on OCTA provided a better assessment of blood flow in both choroid and retina [9, 10, 15]. It has been shown that both superficial and deep capillary plexuses are altered compared to healthy controls [9, 10]. However, some studies found no difference in retinal vessel density between eyes with intermediate-stage AMD and healthy controls [16, 17], and a longitudinal study on intermediate-stage AMD over 12 months reported no significant differences in vessel density of the superficial and deep capillary plexuses [18]. Reports on the foveal avascular zone (FAZ) among AMD and healthy people did not show significant differences [19, 20]. Studies on choroidal vascular metrics, including choroidal vascularity index (CVI), showed that choroidal vascularity is altered in AMD [1, 21]. Choroidal blood flow parameters, such as choroidal volume, blood flow, and velocity, measured using laser Doppler flowmetry, have shown a progressive decline corresponding to the increasing severity of AMD [21].

The Beckman classification uses clinical examination to evaluate lesions within two-disc diameters of the fovea, categorizing non-atrophic dry AMD into two stages. Early-stage is identified by medium-sized drusen (63–124 μm) without pigmentary abnormalities, while intermediate-stage is characterised by either large drusen (125 μm or larger) or the presence of medium-sized drusen accompanied by pigmentary changes [22]. Studies examining the correlation of retinal and choroidal vascularity parameters across different stages of AMD are limited, with most studies using the Beckman classification. There is a lack of studies correlating vascular metrics of the retina and choroid with quantitative biomarkers of the severity of AMD. Our study aims to assess retinal vascular perfusion and choroidal vascularity biomarkers in relation to drusen volume as a marker of AMD severity, focusing on vascular changes based on drusen volumes rather than solely on clinical staging.

Methods

Patients’ selection

This retrospective, cross-sectional study examined individuals with early and intermediate AMD stages. Conducted at the ophthalmology department of the University of Pittsburgh School of Medicine from July 2023 to October 2024, our study strictly adhered to the principles of the Declaration of Helsinki. Due to its retrospective nature, we obtained a “waiver of informed consent”.

This study included patients over 55 years old with a confirmed diagnosis of early and intermediate AMD as per the Beckman classification. The early stage is identified by medium-sized drusen (63–124 μm) without pigmentary abnormalities, while the intermediate stage is characterised by either large drusen (125 μm or larger) or the presence of medium-sized drusen accompanied by pigmentary changes [22]. An expert retina specialist (JC) confirmed the diagnosis.

Eyes exhibiting geographic atrophy and neovascular AMD or any other vitreoretinal diseases, including glaucoma, diabetic retinopathy, retinal vascular occlusion, uveitis, or high myopia, were excluded from the study. Additionally, patients with previous ocular surgeries, except uncomplicated phacoemulsification and intraocular lens insertion surgeries, were excluded. Factors that impaired the quality of OCTA scans, including corneal opacity, dense cataracts, and vitreous opacities, were considered exclusion criteria. Best-corrected visual acuity (BCVA) was assessed using the Snellen chart and subsequently converted to LogMAR (Logarithm of the Minimum Angle of Resolution).

OCTA imaging acquisition

Images centred on the fovea were obtained using the Plex Elite 9000 device (Carl Zeiss Meditec, Dublin, CA). The scan quality was assessed using the SS-OCTA software’s built-in scoring system, and scans scoring 6 or higher out of 10 were included.

Automated choroidal vessel segmentation

The OCT angio and structural volumes were imported into the Advanced Research and Innovation (ARI) network (Zeiss Portal v5.4) to evaluate drusen volume, retina thickness, retinal perfusion density, choroidal thickness (ChT), and CVI. All qualitative and quantitative data were thoroughly reviewed by an expert ophthalmologist (E.S.). Scans exhibiting incorrect segmentation or inaccurate en-face images were excluded from the study to ensure the highest accuracy and reliability. See Figs. 1, 2. (https://arinetworkhub.com/tutorials-and-help/content/algorithm-info).

Fig. 1. Advanced research and innovation (ARI) network algorithms in an eye with early-stage age-related macular degeneration (AMD).

A Swept-source optical coherence tomography (SS-OCT) segmentation. B Retinal thickness colour map. The mean retinal thickness in the 6-mm centre of the macular is 278.037 microns. C Choroidal thickness colour map. The mean choroidal thickness in the 5-mm centre of the macula is 285.294 microns. D The elevation map shows 0.029 mm³ drusen volume. E, F Superficial and deep capillary retinal plexuses show respectively 0.445 and 0.271 perfusion density. G Choroidal vascularity index (CVI) colour map. The CVI was 0.620.

Fig. 2. Advanced research and innovation (ARI) network algorithms in an eye with intermediate-stage age-related macular degeneration (AMD).

A Swept-source optical coherence tomography (SS-OCT) segmentation. B Retinal thickness colour map. The mean retinal thickness in the 6-mm centre of the macular is 325.820 microns. C Choroidal thickness colour map. The mean choroidal thickness in the 5-mm centre of the macula is 147.079 microns. D The elevation map shows 0.236 mm³ drusen volume. E, F superficial and deep capillary retinal plexuses show respectively 0.415 and 0.264 perfusion density. G Choroidal vascularity index (CVI) colour map. The CVI was 0.589.

Drusen volume

We employed the Advanced RPE Analysis v0.10 algorithm to evaluate RPE elevation relative to Bruch’s membrane. This algorithm generates an en-face elevation heat map, highlighting the elevation degree using a colour scale ranging from 0 microns (blue) to a maximum of 100 microns (bright red). The algorithm also produces quantitative data, utilizing Bruch’s membrane segmentation as a baseline created through multilayer segmentation. We analysed the area within a 5 mm radius circle centred on the fovea to quantify RPE elevation and drusen volume.

Retina thickness

We utilized the ETDRS Retina Thickness v0.3 algorithm from the ARI portal to assess retinal thickness. This algorithm measures the total retinal thickness between the internal limiting membrane (ILM) and the RPE. The output includes retinal thickness topography maps and retinal thickness values in the ETDRS sectors. For the retinal thickness measurement, we focused on the area within a 6 mm radius circle centred at the fovea.

Perfusion density

The Macular Density v0.7.3.3 algorithm on the ARI portal was used to evaluate retinal perfusion density in the superficial (SCP) and deep capillary plexuses (DCP) as well as the whole retina. Perfusion density refers to the proportion of the area within a specific region that is occupied by perfused blood vessels. The algorithm reconstructs slab images of the whole retina (top ILM, bottom Bruch’s minus 41 microns), superficial layer (top ILM, bottom IPL), and deep layer (top IPL, bottom OPL) to binarized to form black-and-white illustrations of the vessels. The corresponding binarized images maintain the individual vessel thicknesses representing the ‘perfusion density’ metric. To assess perfusion density in the SCP, DCP, and the entire retina, we analysed the area within a 6 mm radius circle centred on the fovea. Perfusion density was quantified on a scale from 0 (no perfusion) to 1 (fully perfused), with pixel values ranging from 0 to 255. The FAZ area was measured using ImageJ 1.51 s (National Institutes of Health, Bethesda, MD) on the superficial slab angio image.

Choroidal thickness and CVI

ChT and CVI were measured using the Choroid Quantification v20220224-B algorithm. The choroid was segmented from Bruch’s membrane to the choroidal-scleral interface. A thickness colour map image was generated to support qualitative analysis and verify numerical output values, with thickness ranging from 0 microns (blue) to a maximum of 600 microns (bright red). The quantitative ChT was also provided. This algorithm visualized the choroidal vasculature based on structural scans information, which was also used to calculate the CVI, representing the ratio of choroidal vessels (lumen volume) to the entire choroidal volume. The CVI map, shown in a colour map, ranges from 0.00 (blue) to a maximum of 1.00 (bright red). The algorithm also provided quantitative CVI data [23, 24]. We assessed the ChT and CVI for each volume within a 5 mm radius circle centred at the fovea.

Statistical analysis

The Shapiro–Wilk test was used to assess data normality, followed by parametric tests. The Chi-square test was utilized for categorical data analysis. Demographic data, drusen volume, retina perfusion, retina and ChT, and CVI were compared across groups, including early and intermediate-stage AMD, using linear mixed models. We included patients as a random effect to account for cases where one or both eyes were included from the same individual. The dependent variables in our model were LogMar BVCA, Drusen volume, Average retina thickness, Central retinal thickness, Average choroidal thickness, Superficial capillary plexus perfusion density, Deep capillary plexus perfusion density, Retinal capillary plexus perfusion density, FAZ, and CVI. The explanatory variables included age and sex. Random effects were incorporated to adjust for intra-patient correlations when both eyes were analysed. The Spearman test was used to analyse correlations between all parameters. A P value of less than 0.05 was used to indicate statistical significance. All statistical analyses were conducted using IBM’s Statistical Package for Social Sciences (SPSS) version 26. Correlation heatmap was designed by Python (version 3.11).

Results

Demographic data and BCVA

This study analysed 81 eyes from 57 patients with non-atrophic dry AMD, including 34 eyes with early-stage from 23 patients, and 47 eyes with intermediate-stage from 34 patients. The mean age of all patients was 74.95 ± 8.79 years (early-stage: 72.56 ± 8.93, intermediate-stage: 77.82 ± 8.11, P = 0.025). Among the subjects, 34 (59.64%) were female (female with early-stage: 65.21%, female with intermediate-stage: 57.57%, P = 0.634). The mean LogMar BCVA was 0.12 ± 0.17 in early-stage and 0.19 ± 0.18 in intermediate-stage (P = 0.122).

Retino-choroidal thickness, vascular metrics, and drusen volume between early and intermediate-stage AMD

The mean drusen volume within a 5 mm radius circle centred at the fovea was 0.0706 ± 0.1272 mm³ in early-stage and 0.2102 ± 0.2211 mm³ in intermediate-stage AMD (P = 0.009). The average retinal thickness within a 6 mm radius circle was 280.67 ± 12.57 microns in the early-stage and 282.36 ± 14.08 microns in the intermediate-stage AMD, which showed no difference (P = 0.680); however, the ChT was significantly higher in the early-stage (266.40 ± 115.55 microns vs. 204.97 ± 70.69, P = 0.038). Additionally, FAZ was slightly larger but insignificant in eyes with intermediate-stage AMD (0.314 ± 0.266 vs. 0.384 ± 0.677 mm², P = 0.836).

In evaluating retinal perfusion density within a 6 mm radius circle, the analysis showed no significant difference between early and intermediate-stage AMD in SCP perfusion density (0.3609 ± 0.1033 vs. 0.3742 ± 0.0762, P = 0.368), and DCP perfusion density (0.1900 ± 0.0981 vs. 0.1897 ± 0.0784, P = 0.859). The CVI within a 5 mm radius circle showed a significant reduction in eyes with intermediate-stage AMD compared to early-stage (0.605 ± 0.021 vs. 0.591 ± 0.015, P = 0.004). See Table 1.

Table 1.

Structural and vascular metrics and drusen volume between early and intermediate-stage AMD.

	Early-stage AMD (34 eyes)	Intermediate-stage AMD (47 eyes)	P value
Best-corrected visual acuity, LogMar	0.12 ± 0.17	0.19 ± 0.18	0.122
Drusen volume (5 mm), mm3	0.0706 ± 0.1272	0.2102 ± 0.2211	0.009
Average retina thickness (6 mm), micron	280.67 ± 12.57	282.36 ± 14.08	0.680
Central retinal thickness, micron	266.49 ± 12.71	264.61 ± 30.47	0.697
Average choroidal thickness (5 mm), micron	266.40 ± 115.55	204.97 ± 70.69	0.038
Superficial capillary plexus perfusion density (6 mm)	0.3609 ± 0.1033	0.3742 ± 0.0762	0.368
Deep capillary plexus perfusion density (6 mm)	0.1900 ± 0.0981	0.1897 ± 0.0784	0.859
Retinal capillary plexus perfusion density (6 mm)	0.3794 ± 0.1117	0.4028 ± 0.0808	0.217
Foveal avascular zone (FAZ), mm2	0.314 ± 0.266	0.384 ± 0.677	0.836
Choroidal vascularity index (5 mm)	0.605 ± 0.021	0.591 ± 0.015	0.004

A P value less than 0.05 is significant and is bold in the table.

In correlation analysis by Spearman test, drusen volume was negatively correlated with mean ChT (r = −0.198, P = 0.017) and CVI (r = −0.209, P = 0.029). There was no significant correlation between drusen volume and BCVA (r = 0.051, P = 0.143), retina thickness (r = −0.03, P = 0.393), FAZ (r = −0.023, P = 0.150), SCP (r = −0.011, P = 0.307), and DCP (r = −0.022, P = 0.190). The correlation heatmap is shown in Fig. 3.

Fig. 3.

Correlation Heatmap Analysis, which illustrates the relationships between various factors, including best corrected visual acuity (BCVA) LogMar, retinal pigment epithelium (RPE) elevation, retinal parameters including average retinal thickness, fovea avascular zone (FAZ), whole retina, and superficial and deep capillary plexuses perfusion density, and choroidal parameters, including average choroidal thickness and choroidal vascularity index (CVI).

Discussion

This study investigated the correlation between retinal and ChT and vascularity changes with AMD stage and drusen volume. Results demonstrate that retinal perfusion markers, such as FAZ, vascular perfusion in the whole retina, SCP, and DCP, do not associate with the AMD stage or drusen volume. Conversely, CVI shows a negative correlation with AMD severity and drusen volume. Changes in retinal structure, including retinal thickness, are not linked to AMD stage or drusen volume. However, ChT is inversely related to AMD severity and drusen volume. Drusen volume, indicative of AMD severity, negatively correlates with choroidal changes, such as ChT and CVI, but shows no correlation with VA or retinal changes, including retinal thickness and retinal perfusion in both superficial and deep plexuses. BCVA is not associated with AMD stage or drusen volume.

Most studies on AMD severity utilize the Beckman classification (21), which assesses lesions based on drusen size through fundus photography [25, 26]. However, this classification has two main limitations: it relies solely on clinical examination, which may fail to detect smaller drusen, and it only considers drusen larger than 63 microns. Some studies have used other RPE characteristics, such as RPE curvature [27], RPE thickness [28], or drusen volume in the analysis, which indicates drusen volume [29]. It has been shown that drusen volume increased during AMD progression, and spontaneous regression in drusen volume is associated with conversion to advanced AMD [29]. Abdelfattah et al. have indicated that individuals with drusen volumes exceeding 0.03 mm³ are at more than four times the risk of progressing to late-stage AMD [30]. In this study, we quantified drusen volume within a 5 mm radius circle centred on the fovea and evaluated its correlation with retinal and choroidal structural and vascularity metrics. Our findings indicated that the drusen volume negatively correlates with ChT and CVI but not retinal thickness and retinal vessel perfusion. Our findings showed no significant correlation between drusen volume within the central 5 mm of the macula and VA. This result is consistent with the study by Ou et al., which reported that the amount of drusen within the central 3 mm of the macula does not appear to be related to low-luminance deficit (LLD) VA in intermediate AMD [31].

OCTA has gained popularity over the last decade due to its ability to evaluate retinal and choroidal vasculature, aiding in understanding disease pathogenesis. By using the XR Avanti SD-OCT (Optovue Inc., Fremont, CA), Toro et al. showed that the SCP and DCP are altered in dry AMD, with more pronounced alterations in the SCP in the intermediate stage [9]. In another study by the same team, they compared intermediate-stage AMD patients with and without pre-atrophic changes. They demonstrated that the flow density of the parafoveal SCP was significantly altered in intermediate-stage AMD with pre-atrophic changes compared to those without pre-atrophic changes [32]. Using Zeiss Cirrus Angioplex OCTA (Carl Zeiss Meditec, Jena, Germany), Trinh et al. reported that vessel density in SCP, along with vessel length, diameter, and complexity, are significantly altered in intermediate-stage AMD [33]. Another study by the same group revealed that perfusion density reduction in the SCP was more pronounced in the temporal area and FAZ in eyes with intermediate-stage AMD. In the DCP, the decrease was mainly observed in the FAZ and diffusely in other regions [34]. Conversely, two studies by Parisi et al. [17]. and Vaghefi et al. [16] identified no significant differences in retinal vessel density between eyes with intermediate-stage AMD and healthy controls. In a 12-month follow-up study of intermediate-stage AMD by using RTVue XR Avanti (Optovue), Reiter et al. found no association between changes in drusen volume and vessel density or flow area in either the SCP or DCP, except for a significant increase in the foveal flow area of the SCP. They noted that changes in retinal vessels might occur more slowly than anticipated despite increased drusen volume and VA reduction [18]. Additionally, two other studies found no significant differences in FAZ between AMD patients and healthy controls [19, 20]. Our study revealed no significant differences in retinal vascularity metrics, such as vessel perfusion in the SCP, DCP, and FAZ, between early and intermediate-stage AMD. Additionally, we found no significant correlation between drusen volume and retinal vascularity metrics. Our findings suggest that retinal vascularity changes might lag behind the progression of AMD stages and the increase in drusen volume.

Multiple studies documented decreased ChT, altered choroidal blood flow, and reduced CVI during AMD progression, suggesting that choroidal ischaemia might play a role in pathogenesis and could serve as a biomarker of disease progression [15, 35–37]. A recent study by our team evaluating three-dimensional choroidal vessel assessment in dry AMD showed that eyes with dry AMD have lower ChT, reduced CVI, and dilated choroidal vessels compared to healthy age-match controls [1]. A longitudinal study by Vidal-Oliver reported that ChT and CVI are reduced in eyes with AMD, and these two biomarkers may aid in assessing the risk of AMD progression [38]. This study demonstrated a significant decrease in ChT and CVI in intermediate-stage AMD compared to early-stage AMD. Furthermore, a significant negative correlation was found between drusen volume and choroidal changes. These findings suggest that alterations in choroidal structure and choroidal vascularity are more pronounced as AMD progresses, making them more accurate biomarkers for evaluating disease progression in these patients.

Our study has certain limitations stemming from its retrospective design and the relatively small sample size. We used drusen volume as a quantitative measure of the severity of the disease. However, we understand that drusen volume can start decreasing in advanced stages. Evaluating late-stage AMD, including neovascular AMD and geographic atrophy, and including incomplete and complete RPE and outer retinal atrophy (iRORA/cRORA) as severity markers would offer a more comprehensive understanding of disease progression. Another limitation of our study was that ARI measures drusen volume and choroidal metrics within a 5 mm central circle, while retinal metrics are assessed within a 6 mm central circle. The cross-sectional design limits our ability to monitor progressive changes. Longitudinal data is essential for comprehending the evolution of choroidal changes in conjunction with the increase in drusen volume and exploring the pathophysiological mechanisms underlying these correlations.

In conclusion, we report that ChT and CVI are early biomarkers influenced by AMD progression from early to intermediate stages, showing a negative correlation with drusen volume. Retinal thickness and retinal vessel perfusion do not change with the AMD stage, including early and intermediate stages, nor drusen volume. Alterations in choroidal parameters, such as ChT and CVI, are more closely linked to AMD severity in early/intermediate stages than retinal thickness, retinal vessel perfusion, and VA. Drusen volume, in particular, helps quantify AMD severity and addresses the limitations of the Beckman criteria, and it could serve as a marker of AMD progression.

Summary

What was known before

• Research on AMD severity has traditionally relied on the Beckman classification, which assesses lesions based on drusen size via fundus photography.

• However, this method has limitations, including the inability to detect smaller drusen and its focus only on drusen larger than 63 microns

• Previous studies have also reported controversial results regarding retinal vascular perfusion in AMD.

What this study adds

• Drusen volume is a valuable marker for quantifying AMD severity and addressing the limitations of the Beckman criteria, potentially serving as a marker for AMD progression.

• Our findings show that ChT and CVI are early biomarkers impacted by AMD progression from early to intermediate stages, demonstrating a negative correlation with drusen volume

• Retinal thickness and retinal vessel perfusion, on the other hand, do not change with the AMD stage, including early and intermediate stages, nor drusen volume

• Alterations in choroidal parameters, such as ChT and CVI, are more closely linked to AMD severity in early and intermediate stages than retinal thickness, retinal vessel perfusion, and VA.

Author contributions

ES: conceptualization, data gathering, data analysis, drafting, revision, final approval. AS: data gathering, final approval. SCV: data analysis, final approval. SRS: revision, final approval. NH: revision, final approval. SCB: revision, final approval. KKV: revision, final approval. JAS: revision, final approval. AWE: revision, final approval. JC: conceptualization, data preparation, revision, final approval.

Funding

NIH CORE Grant P30 EY08098 supported this work to the Department of Ophthalmology, the Eye and Ear Foundation of Pittsburgh, and from an unrestricted grant from Research to Prevent Blindness, New York, NY.

Data availability

The datasets generated during and/or analysed during the current study are available from the corresponding author on reasonable request.

Competing interests

JAS: Avista Therapeutics, Tenpoint, Code C (Consultant/Contractor), Clinical Trials: Gensight, SparingVision, Meira, Code F (Financial Support), Netramind Innovations, Gensight, Sparing Vision, Avista, Tenpoint, Prophesee, Chronolife, Tilak Healthcare, SharpEye, Cilensee, Vegavect, Code O (Owner), Allotopic Expression, Rod-derived Cone Viability Factor and related patents., Code P (Patent), Patent Royalties, Gensight, Code R (Recipient), Observer: Gensight, SparingVision, Avista, Vegavect. President: Fondation Voir et Entendre, Paris; President: StreetLab, Paris., Code S (non-remunerative); JC: Netramind Innovations, Code O (Owner). KKV: Netramind Innovations, Code O (Owner). SCB: Netramind Innovations, Code O (Owner).

Supplementary information

The online version contains supplementary material available at 10.1038/s41433-025-03847-6.